Temporal molecular changes in ageing mammalian organs are of relevance to disease aetiology because many age-related diseases are linked to changes in the transcriptional and epigenetic machinery that regulate gene expression. We performed quantitative proteome analysis of chromatin-enriched protein extracts to investigate the dynamics of the chromatin proteomes of the mouse brain, heart, lung, kidney, liver, and spleen at 3, 5, 10, and 15 months of age. Each organ exhibited a distinct chromatin proteome and sets of unique proteins. The brain and spleen chromatin proteomes were the most extensive, diverse, and heterogenous among the six organs. The spleen chromatin proteome appeared static during the lifespan, presenting a young phenotype that reflects the permanent alertness state and important role of this organ in physiological defence and immunity. We identified a total of 5928 proteins, including 2472 nuclear or chromatin-associated proteins across the six mouse organs. Up to 3125 proteins were quantified in each organ, demonstrating distinct and organ-specific temporal protein expression timelines and regulation at the post-translational level. Bioinformatics meta-analysis of these chromatin proteomes revealed distinct physiological and ageing-related features for each organ. Our results demonstrate the efficiency of organelle-specific proteomics for in vivo studies of a model organism and consolidate the hypothesis that chromatin-associated proteins are involved in distinct and specific physiological functions in ageing organs.
The authors have performed an extensive analysis of chromatin enriched proteins as a function of age in mice. Time-course quantitative proteomics reveals the molecular complexity and diversity of mammalian organs and identified aging-related molecular features of chromatin.https://doi.org/10.7554/eLife.73524.sa0
Ageing is a natural process resulting in progressive changes of most, if not all, cellular components. Ageing is generally associated with declining biological performance and increased incidence of disease (Oberdoerffer and Sinclair, 2007). The gene expression apparatus, comprising the DNA itself, the chromatin environment it is housed in, and the machinery of transcription and translation, is profoundly affected by ageing (Busuttil et al., 2007; Edifizi and Schumacher, 2015). Models of ageing often display similar phenotypes to those undergoing senescence or genome instability, highlighting how integrated the ageing process is with these phenomena (Oberdoerffer and Sinclair, 2007). Diseases that mimic or accelerate the ageing process, including Hutchinson–Gilford progeria and Werner syndromes, result in molecular changes in nucleosomes and chromatin (Arancio et al., 2014; Burtner and Kennedy, 2010; Feser and Tyler, 2011).
In eukaryotes, chromatin includes histone molecules that package the DNA, locally controlling access to the underlying genes by facilitating ‘open’ or ‘closed’ states associated with transcriptional activation or repression, respectively (Laugesen and Helin, 2014). In this way, transcription of individual gene products may be regulated in temporal or location-specific manners.
One of the main routes of proteome expansion is dedicated to enzymes that carry out post-translational modifications (PTMs) of proteins. Enzyme-catalysed PTMs at distinct amino acid residues regulate or modulate protein structure, interactions, and functions (Santos and Lindner, 2017).
The mouse, Mus musculus, is the most commonly used experimental animal in biomedical research and serves as a model system for studying human health and disease (Phifer-Rixey and Nachman, 2015; Fontana and Partridge, 2015). Mice have a relatively short lifespan, with one adult mouse month equivalent to approximately three human years (Sengupta, 2013; Dutta and Sengupta, 2016). This allows for maximum lifespan studies to proceed within the timelines of typical research projects, while environmental factors that affect ageing can be controlled (Shoji and Miyakawa, 2019; Delgado-Morales, 2017; Shoji et al., 2016). Lifespan and health span are mutually influenced by many genes that can either predispose to age-related diseases or slow the ageing process itself (Murabito et al., 2012).
We recently applied a middle-down proteomics strategy to demonstrate that mouse chromatin undergoes major changes during ageing, specifically that histone H3.3 replaces H3.1 and that the extent of H3 methylation marks at multiple sites is profoundly altered during ageing (Tvardovskiy et al., 2017). We here extend these proteomics studies of mouse chromatin to investigate the protein composition of chromatin in multiple mouse organs during ageing.
We hypothesised that a time-course investigation of the dynamic chromatin proteome could reveal distinct molecular differences of mammalian organs and provide new insights into the regulatory mechanisms in different organs during ageing.
We studied the progressive chromatin protein expression changes in six mouse organs during ageing by quantitative proteomics by mass spectrometry (graphical abstract).
Among almost 6000 proteins identified, organ-specific patterns predominated, with age-responsive subsets identified for each organ. We mapped pathway-level molecular changes specific to individual organ over time.
Our results demonstrate that the ageing process affects each mouse organ in a distinct manner illustrated by the diversity and heterogeneity of the temporal chromatin proteomes of each organ.
We aimed to provide a comprehensive overview of the chromatin-enriched proteome in mouse organs and obtain insights into the molecular processes involved in growth and ageing.
We initially applied a nuclear protein extraction protocol to the mouse embryonic stem cell (mESC) model, which is a ‘gold standard’ for epigenetics research (Supplementary file 1; Tobin and Kim, 2012; Takahashi et al., 2018). Briefly, mESCs were lysed and cellular compartments were isolated by mechanical disruption followed by a high salt gradient separation to obtain cytosolic, nuclear, chromatin, and histone fractions (Streubel et al., 2017; Herrmann et al., 2017). These lysates were initially analysed by Western blotting using specific protein markers for each cellular compartment to assess the degree of enrichment of chromatin-associated proteins (Figure 1A).
We performed quantitative proteomics by triplicate high-mass-accuracy mass spectrometry analysis of the mESC proteome and the mESC chromatin proteome to assess the enrichment of chromatin-associated proteins (Figure 1B and C). We present the detected chromatin enzyme/protein complex components in Table 1. We annotated all nuclear or chromatin-associated proteins using available database resources (Figure 1—figure supplement 1; Medvedeva et al., 2015; van Mierlo et al., 2019; Christoforou et al., 2016; Xu et al., 2016).
The dynamic range plot was used to assess the measurements of protein expression across these proteomes. The mESC chromatin fraction was indeed highly enriched for chromatin-associated proteins compared to the whole-cell lysate (Figure 1B and C).
We detected major chromatin-associated protein complexes, including the polycomb repressive complex 2 (PRC2), nucleosome remodelling and deacetylase (NuRD), BRAF-HDAC complex (BHC), and mixed-lineage leukaemia (MLL) complex (Figure 1B; Medvedeva et al., 2015; van Mierlo et al., 2019; Xu et al., 2016).
In general, we observed an overall increase in the proportion of proteins classified as either ‘nuclear’ (20%) or ‘chromatin-associated’ (35%) within the mESC chromatin sample (Figure 1).
Gene Ontology (GO) analysis showed a distinct enrichment of proteins associated with ‘DNA-protein binding’, ‘histone binding’, and ‘chromatin and nucleosome organisation’ within the chromatin sample (Figure 1D). The mESC total lysate sample mainly contained proteins involved in ‘translational protein’, ‘cell-cell structure organisation’, and ‘ribosomal and ATP processes’ (Figure 1D).
We next demonstrated that the chromatin enrichment protocol used for mESCs is applicable to mouse organs. We first extracted chromatin proteins from mouse brain and assessed chromatin enrichment by Western blotting using specific protein markers (Figure 1—figure supplement 2). We observed enrichment of the chromatin marker histone H3 and reduced level of the cytosolic marker GAPDH expression within the ‘chromatin fraction’ lysate.
Next, we performed quantitative mass spectrometry profiling of the mouse brain proteome and the mouse brain chromatin-enriched proteome. The dynamic range plots demonstrated that the major chromatin-binding protein complexes were enriched in the brain chromatin sample, including MLL, NuRD, Polycomb, and BHC complexes (Figure 2A, Supplementary file 2, Table S2).
The mouse brain chromatin-enriched fraction contained more than 30% chromatin-associated proteins, up from ~8% in the total brain lysate (Figure 2A, bottom). This was accompanied by a large increase in the content of ‘nuclear proteins’ (~20%) (Figure 2A).
We conclude that our chromatin proteome fraction of mouse brain tissue was highly enriched in chromatin-associated and nuclear proteins compared to the whole brain lysate (Figure 2A).
We performed quantitative chromatin proteomics of six mouse organs to investigate the in vivo dynamics of chromatin during ageing. We isolated chromatin-associated proteins from mouse brain, heart, liver, kidney, lung, and spleen at time points 3, 5, 10, and 15 months representing the ‘mature adult mouse lifespan’, from the early adult stage (3 months), middle-aged adult (5–10 months), and mature adult (15 months) (graphical abstract) (Dutta and Sengupta, 2016; Shoji and Miyakawa, 2019; Delgado-Morales, 2017; Shoji et al., 2016).
We excluded mice older than 18 months to minimise any age-related changes that might be due to social behaviour, physical characteristics of motor function, and locomotor activity.
Proteins were identified and quantified by high-mass-accuracy LC-MS/MS by hybrid quadrupole-orbitrap technology using a peptide intensity-based (label-free) protein quantification strategy (see Materials and methods, Supplementary file 3, Table S3).
The proportion of nuclear proteins or chromatin-associated proteins ranged from 30% to 60% of all detected proteins, and it was similar across all time points for each organ (Figure 2B).
Subsequent data analysis included all identified proteins of each organ and all time points to avoid loss of essential information and achieve a more detailed characterisation of the chromatin proteomes of mouse organs.
We identified a total of 5928 proteins in the chromatin-enriched protein samples across all six mouse organs over time (Figure 2C). Most proteins were identified in the chromatin fractions of mouse brain (3110) and spleen (3125), whereas the lowest number of proteins were identified in the chromatin fraction of mouse heart organ (2051) (Figure 2C). The lung and spleen samples were highly enriched in nuclear proteins and chromatin-associated proteins (55–65%), whereas the chromatin-enriched heart sample contained 30–35% nuclear/chromatin-associated proteins.
The very different morphology and cell-type compositions of the mouse organs likely influence the efficiency of the chromatin protein extraction protocol and thereby the detected proteome compositions. Cell-type and cell cycle-specific transcriptional activities likely explain some of the observed variation in proteome composition.
Nevertheless, the fraction of proteins classified as either ‘nuclear’ or ‘chromatin-associated’ was similar among all mouse organs and time points, demonstrating the high reproducibly and reliability of the experimental approach with a coefficient of variation estimated less than 10% among all the samples (Figure 2B).
Overall, we identified a total of 4581 different chromatin-associated proteins across all organs, including 2717 different nuclear proteins (Table 2).
We assessed the entire mouse organ chromatin-enriched proteome dataset using principal component analysis (PCA) (Figure 3A). The time points (age of mouse at organ harvest) were well separated from one another for each organ, such that the replicates for each time point were more closely clustered to one another than to replicates of other time points. More striking was the observation that the origin of organ was the fundamental discriminating factor for the overall clustering of samples: each organ formed its distinct cluster made up of subclusters comprising the different ages of the organ samples (Figure 3A).
Pearson correlation analysis (Figure 3—figure supplement 1) demonstrated the reproducibility of biological replicates and confirms the robustness of our biochemical and proteomics methodology. To further characterise similarities between ageing organ and proteome expression profiles, we performed clustering of Pearson correlation coefficients (Figure 3B). This showed the consistency and reproducibility of analysis of three biological replicates at all time points and revealed the distinct organ proteome profiles. The brain-derived proteome exhibited poor correlation to all other organs. Kidney, liver, heart, and lung samples exhibited protein expression profiles, which were slightly positively correlated. Spleen displayed a slightly positive correlation with kidney and lung and poor correlation to the brain, liver, and heart.
Overall, our PCA and Pearson correlation analyses demonstrate that each mouse tissue exhibits a distinct and ageing-related chromatin proteome profile (Figure 3).
To define ageing signatures across mouse lifespan in each organ, we performed a comprehensive functional analysis of the proteomics datasets using an UpSetR plot to investigate ageing markers (Figure 4A; Conway et al., 2017).
Briefly, comparative organ proteome analysis was performed to estimate the number of shared proteins and to determine the degree of similarity between datasets and, subsequently, to identify unique organ-specific feature. Then, the datasets were investigated by GO analysis to explore the relationship between chromatin protein composition and ageing.
A ‘core’ proteome of 863 proteins was identified across all six organs during mouse lifespan across four time points, including 289 shared chromatin-binding proteins and 157 shared nuclear proteins (Figure 4A, Table 3).
The core chromatin proteome contained proteins shared across all tissue during the mouse lifespan. These proteins were associated with the major transcriptional/epigenetic chromatin complexes such as BHC (GTF2I), MLL (ACTB, RUVBL2, DPY30, WDR82, PPP1CA, SEPT9, PPP1CC, PPP1CB, SNX2, LPP, DDX6), NuA4HAT (RUVBL1), NuRD (CHD4, TRIM28, CSNK2A1), NURF (SMARCA5), PcG (RBBP4, RBBP7), SAGA (FGG), SIN3A (MECP2, SFPQ, PA2G4), and SWI/SNF (SMARCC2, YWHAB, H2AFY, RAC1, EIF4B) (Medvedeva et al., 2015; van Mierlo et al., 2019; Xu et al., 2016; Spruijt et al., 2016; Varier et al., 2016). Many of these proteins are expressed during cell fate commitment (Medvedeva et al., 2015; Signolet and Hendrich, 2015).
The mouse brain chromatin proteome contained 793 unique proteins, constituting the largest set of unique proteins among the six organs (Figure 4A). Approximately 24% of these proteins (188) were classified as chromatin-associated proteins or nuclear proteins (Figure 4A, Table 3). They included protein and histone methyltransferase enzymes (HNMT, SAP30L, CARM1, SETD7, SUV39H2) and histone deacetylase enzymes (CHD5, IRF2BP1, MAPK10, MEF2D, MACROD2) that are mainly involved in transcriptional gene silencing (Medvedeva et al., 2015; van Mierlo et al., 2019; Xu et al., 2016; Hyun et al., 2017).
The spleen chromatin preparation contained 582 unique proteins, ~50% of which are chromatin-associated proteins or nuclear proteins, strongly suggesting a distinctive chromatin proteome profile for this organ (Figure 4A, Table 3).
The spleen is an organ with the innate capacity to regenerate (Holdsworth, 1991). It acts as a filter for blood and controls the blood-borne immune response (Tan and Watanabe, 2018). We detected several epigenetic complexes such as BHC, ING, MLL, NuRD, ORC, PCG, SIN3A, SAGA, and SWI/SNF. Also, we detected chromatin ‘reader’ enzymes not yet assigned to a specific chromatin remodelling complex (KDM3B, KDM2A, PHF23, CHD1L, UHRF2, MORC3, BRD9, ZMYND11, BAZ1A). These proteins recognise single post-translational histone marks or combinations of histone marks and histone variants to direct a particular transcriptional outcome (Medvedeva et al., 2015; van Mierlo et al., 2019; Xu et al., 2016; Hyun et al., 2017).
This result is in line with our above observations, suggesting that spleens were highly enriched in epigenetic markers and therefore are a good model to study chromatin remodelling complexes. Ageing effects are difficult to distinguish in spleen tissue.
The relatively large numbers of unique proteins of each organ likely reflect the inherent features of the individual organs, the diversity of cell types, and physiology (Figure 4A).
GO analysis of the ‘core’ 863 proteins revealed a large number of ‘chromatin-associated’ or ‘nuclear’ proteins. The GO output was enriched for categories related to mouse ageing, such as ‘oxidation-reduction process’, ‘ageing’, ‘regulator of cell cycle’, and ‘stress response’ (Figure 4B; Go and Jones, 2017; Haigis and Yankner, 2010; Epel and Lithgow, 2014; Postnikoff and Harkness, 2012; Chandler and Peters, 2013).
Subsequently, each category of the core proteome was further broken down into its constituent parts to create a map of the shared ageing-related molecular network of the six mouse organs (Figure 4B). Our aim was to identify novel biological features and associate them with ageing-related pathways and annotations.
GO classification of the unique proteins present in each organ proteome suggested distinctive molecular signs of ageing in each organ (Figure 4C). We observed distinctive organ-specific categories, ‘age classes and age development’, and categories that reflected their organ source (Figure 4C). For instance, unique GO term categories were associated with each organ, such as ‘nervous system development’ and ‘chemical synaptic transmission’ related to the brain; ‘cardiac myofibril assembly’ and ‘adult heart development’ were attributed to the heart; ‘steroid metabolic process‘ and ‘liver development’ were distinctive to the liver; ‘Transport’ and ‘sodium ion transport’ categories related to the kidney; and ‘angiogenesis’ and ‘respiratory gaseous exchange’ were present in the lung.
These results confirm that many chromatin proteins found in individual organs likely confer organ-specific functions (Figure 4C).
Taken together, our proteomics analysis showed a robust enrichment of chromatin-associated proteins in mouse organs as confirmed by GO term analysis. We reported a significant enrichment of age-related proteome features, including a large class of protein annotations associated with the core chromatin environment present in all organs.
We hypothesised that different mammalian organs have ageing-dependent and distinct expression profiles of characteristic chromatin-associated proteins.
We employed a temporal analysis of the overall dataset which included chromatin-associated proteins, nuclear proteins, and unassigned proteins to uncover the common features that may contribute to the chromatin environment during ageing.
We used the rank products test to identify the proteins that exhibited significant abundance changes in one direction during the ageing of the organ or linearly throughout the lifespan of the animal (Figure 5—figure supplement 1A and C; Koziol, 2010; Breitling et al., 2004). We call these ‘differentially regulated proteins’ (Figure 5, Figure 5—figure supplement 2A).
We then retrieved those unique regulated proteins that were specifically detected in only one organ, two organs, or three organs (UpSet plot) (Figure 5A). The majority of uniquely regulated proteins were indeed specific to one organ.
We subjected the organ-specific uniquely regulated proteins to hierarchical clustering based on their expression changes and depicted them as heatmaps for each organ (three replicates per time point) (Figure 5B). This allowed us to compare how the relative expression of unique chromatin-associated proteins changed over time in each organ, from early to late time points (3–15 months) (Figure 5B, Supplementary file 4, Table S4).
The hierarchical clustering shows that the number of up- and downregulated proteins is similar in each organ. Further, assigned cell compartments of the ‘unique differentially regulated proteins’ are shown as sidebars (Figure 5B, Supplementary file 4, Table S4). A large number of non-annotated ‘unique differentially regulated proteins’ show a similar quantitative behaviour across all organs. We hypothesised that these proteins may represent a useful list of candidates that may regulate gene expression by being transiently recruited to chromatin at distinct time points during ageing.
Brain, heart, and kidney showed mainly gradual changes of protein levels over time. Lung and liver show a more dramatic change of protein expression between 5 and 10 months of age (Figure 5B, Figure 5—figure supplement 2B). We detected a few significant protein abundance changes in the spleen. The spleen fraction was highly enriched in chromatin-associated and nuclear proteins (~60%) and contained many unique proteins (Figures 3 and 4). Thus, the spleen seems to continuously exhibit a young phenotype that may be due to the constitutively active role of the spleen in maintaining immune functions, red blood cell turnover, and microbial defence (Turner and Mabbott, 2017). The spleen contains multiple cell populations capable of supporting immune responses, which may indicate the presence of self-renewal cell types that are ‘age-less’ (Holdsworth, 1991).
In summary, we identified a large number of unique differentially regulated proteins in the chromatin-enriched proteomes of mouse organs. The abundance of these proteins changes dramatically during ageing from month 3 to month 15, across all organs, except for spleen.
Next, we explored the potential functional links between chromatin proteome dynamics and ageing. We investigated all chromatin-associated proteins that are part of known well-defined chromatin complexes that were identified among the ‘unique differentially regulated proteins’.
The majority of chromatin-modifying enzymes belonging to a given multiprotein complex exhibited similar expression profiles over time within a specific organ (Figure 5C). For instance, the MLL subunits WDR82, CTR9, and WDR61 were downregulated in the liver during ageing. Components of the same chromatin-modifying complexes were detected in several organs, albeit not by the same subunits and with opposite temporal expression profiles. NuRD subunit HDAC2 was downregulated in the liver, whereas NuRD subunits GATA2AD and TRIM28 were upregulated in the kidney.
This is consistent with the highly dynamic nature and spatio-temporal regulation of chromatin remodelling complexes. Some protein subunits are only present in a complex at distinct time points to provide a unique function or feature (Oliviero et al., 2016).
A series of ‘Reader’ enzymes were upregulated in the brain (GLYR1, BAZ1B) and spleen (BRD3, BRD7), whereas other ‘Reader’ enzymes were downregulated in the liver (DPF2, BRD2, CHD2) (Figure 4C).
Next, we queried the ‘Human Ageing Genomic Resources’ and ‘GenAge machine learning databank’ using our complete list of ‘unique regulated proteins’ that are not yet assigned to chromatin or nuclear environment to demonstrate the ability of our mouse organ proteomics approach to detect known human ageing biomarkers (Figure 5—figure supplement 2C; Tacutu et al., 2013; Kerepesi et al., 2018).
We identified a series of human protein biomarker candidates for ageing. The brain protein IREB2 is associated with Alzheimer’s disease, whereas the brain protein MAOB is associated with both Alzheimer’s and Parkinson’s diseases. The heart proteins ADD3, PTGIS, and COL1A2 are candidates for hypertension and myocardial infarction. The liver proteins INSR, PTPN1, and ENPP1 are associated with diabetes mellitus type 2 and obesity. Lung protein CYP2E1 is related to lung adenocarcinoma, and MMP9 is associated with lung neoplasms. In the spleen, KLK1 protein is a biomarker candidate for hypertension.
Next, we applied GO analysis to characterise all ‘unique differentially regulated proteins’ detected in each organ (Figure 6).
We listed the overall common pathways and processes that were found by quantitative chromatin proteomics to be differentially regulated during ageing across all the organs. We sorted the annotated features by their relative GO term category and separated them by their main family source (Figure 6, left panel).
During adult mouse lifespan, we observed several ageing stress response pathways and GO categories associated with the regulation of chromatin architecture that may affect cellular structures and morphology.
By listing every single category, we could describe the biological profile and pathways affected by the age-related protein expression responses present in all organs (Figure 6).
Subsequently, the differentially regulated proteins were sorted by their organ source and subject to further GO term analysis to distinguish unique organ-related processes from those pathways associated with ageing (Figure 6, centre panel).
We report a high proportion of uniquely annotated categories for each mouse organ. For instance, the highest unique changes observed in the brain were relative to ‘gene expression’ and ‘ageing/development’; in the heart and kidney, significant changes were observed relative to ‘structure organisation and biogenesis’; the liver showed changes across the ‘gene expression’ and ‘structure organisation and biogenesis’ categories; the lung showed the highest unique changes in the ‘ageing/development’ category, and relative changes in the spleen were detected at the ‘gene expression’ level.
These results are in line with our above observations, suggesting a unique ageing response from each organ as evidenced by distinct dynamic changes of chromatin-associated proteins.
We further interrogated the list of unique annotated organ categories to highlight the distinct and significant temporal pathway profiles among the up- and downregulated proteins in each organ to reveal the most distinctive regulated features (Figure 6, right panel).
In the mouse brain tissue, proteins involved in pathways such as ‘chromosome organisation’ and ‘histone modification’ were strongly upregulated, while those involved in the regulation of ‘neuron projection regeneration’ were downregulated. In mouse heart tissue, upregulated protein pathways included ‘oxidoreductase activity’ and ‘regulation of response to stress’, while downregulated pathways included ‘cardiovascular system development’ and ‘chromosome organisation’. In the kidneys, proteins involved in pathways related to changes in chromatin conformation were both going up and going down during the mouse lifespan. In the liver, we observed that downregulated proteins were associated with ‘chromosome organisation’ and ‘histone modification’, while ‘oxidation-reduction process’ pathways were upregulated. In the lungs, we noticed that proteins associated with ‘apoptotic process’ and ‘programmed cell death’ were upregulated, while pathways related to ‘muscle organisation and reassemble’ were downregulated.
Similarly, several different processes were altered in the other organs, except for the spleen that did not show many significant changes during ageing. In line with previous data, little pathway-level changes were observed in the spleen, especially in the downregulated proteins, possibly indicating the important role of this organ in removing old red blood cells and microbes which seem, from our data, to not be affected by ageing. For this reason, the spleen was not considered for further analysis.
Overall, using GO term analysis we dissected the biological features of the chromatin proteomes of organs in the context of mouse lifespan. By breaking down common and unique regulated functional categories, we surveyed ageing-related pathways and improved gene-annotation enrichment analyses.
In conclusion, we identified and measured distinct and extensive protein abundance changes during ageing, specifically in early to mature adult mouse lifespan. A large number of differentially expressed proteins are unique for each organ as defined by specific GO term categories. This demonstrated that the ageing process affects each mouse organ differently.
We looked in more detail at the most significantly regulated organ-specific upregulated and downregulated protein categories during mouse ageing. To validate the most significant organ-specific biological processes changing during mouse ageing, the selected GO terms related to chromatin environment were further investigated (Figure 6, right panel, dashed box). Additionally, we selected GO terms known to be associated with ageing effects to see whether chromatin-associated proteins fell into these categories.
We used protein-protein interaction (PPI) data from the STRING database to map the network of chromatin-associated protein belonging to the most significant GO categories of each organ (Szklarczyk et al., 2015; Szklarczyk et al., 2017; Figure 7A).
Figure 7—source data 1
Figure 7—source data 2
Figure 7—source data 3
Figure 7—source data 4
Subsequently, we combined the protein interaction networks with the quantitative dataset (Figure 7A). By integrating PPIs and protein expression, we derived co-interaction and co-expression networks to improve our understanding of biological mechanisms involved in ageing.
Finally, we attempted to confirm independently, by Western blot, the observations noted in our wider dataset, specifically to the co-expression network generated (Figure 7B).
The majority of upregulated proteins of the ageing brain belonged to the category ‘chromosome organisation’, including histones and histone-binding enzymes (Figure 7A). We noticed a strong subnetwork of PPIs between two histone variants and histone-binding enzymes: macro-H2A2.1, macro-H2A2.2, and HP1BP3.
Macro-H2A2.1, macro-H2A2.2, and HP1BP3 accumulated during mouse brain ageing (Figure 7B). Macro-H2A2.1, macro-H2A2.2, and HP1BP3 are mammalian heterochromatin components that are highly expressed in adult mouse brains (Douet et al., 2017; Barrero et al., 2013; Garfinkel et al., 2015; Garfinkel et al., 2016). All three proteins increased in expression during organ development and ageing. These proteins were not detected in any other mouse organ in this proteomics study (data not shown). Macro-H2A2.1 is an epigenetic marker whose major function is to maintain nuclear organisation and heterochromatin architecture (Douet et al., 2017). HP1BP3 is a heterochromatin marker protein that recognises the histone mark H3K9me3 and promotes transcriptional repression (Garfinkel et al., 2015). HP1BP3 loss of function is associated with cognitive impairment, suggesting a role for this protein in establishing or maintaining cognitive functions (Garfinkel et al., 2016). We also confirmed the expression of other heterochromatin markers, including RING1b (Saksouk et al., 2015) and RNF20 (Kim et al., 2005), used as protein controls to monitor our strategy (Figure 7B). These results suggest a link between regulation of heterochromatin components and accumulation of histone variants during the ageing of brain tissue in adult mammals.
GO analysis of the regulated proteins of ageing heart tissue indicated upregulation of ‘oxidoreductase activity’ (Figure 7A). A chromatin-associated protein of the sirtuin family (SIRT5) is a member of this category. Sirtuins are histone deacylases that play an important role in age-related pathological conditions such as cancer and the deregulation of metabolism (Gillette and Hill, 2015; Mei et al., 2016; Masse et al., 1991). Sirtuin proteins are mostly annotated as mitochondrial proteins, but they can translocate further into the nucleus or other cell compartment (Kupis et al., 2016; Vaquero, 2009; Li et al., 2013). Our results suggest that the activation of specific members of the sirtuin family and their translocation to the nucleus is involved in the ageing process (Kupis et al., 2016; Rajendran et al., 2011).
SIRT5 and SIRT3 were detected in our chromatin-enriched proteome dataset of ageing mouse heart. Both SIRT3 and SIRT5 expression levels increased during the mouse lifespan, with SIRT5 being upregulated from 3 to 15 months (Figure 7B).
The abundance change of SIRT3 levels was less pronounced. This data was confirmed by Western blotting (Figure 7B). SIRT5 is a histone deacylase that removes malonyl, succinyl, and glutaryl groups from histones. The ageing-dependent increase in histone H3 acetylation observed in our proteomics study and by Western blotting (Figure 7B) is consistent with the fact that SIRT5 has no deacylase activity towards histone H3.
We observed drastically downregulated ‘chromosome organisation’ in the liver during ageing (Figure 7A). This category contained histone acylation/acetylation-related chromatin remodelling enzymes associated with different chromatin-modifying complexes. Examples include KAT2A (either ATAC or SAGA complex), CHD2 (NuRD), HDAC2 (CoREST, NuRD, SWI/SNF, Sin3A-like), WDR82 (COMPASS), and DNMT1 (ACF) (Medvedeva et al., 2015; Xu et al., 2016; Figure 7).
Western blot analysis confirmed that the protein expression levels of CHD2 and KAT2A are significantly reduced at 15 months (Figure 7B), which leads to lower histone acylation levels. We indeed observed decreased levels of global histone H3 acetylation at this time point (Figure 7B), which contrasts with what we observe in the heart tissue. These results confirm our previously published data on decreased H3 acetylation (H3K14, K23, K27) in the liver tissue during ageing (Tvardovskiy et al., 2017), and that there is a decreased activity of histone acylation in the liver at late stages of ageing.
The upregulated ‘chromosomal part’ category in ageing kidney tissue included components of the SCF-type E3 ubiquitin-protein ligase family (FBXO41, RNF13, NEDD4, TRIM28) and the proteasome subunits PSMD14 and PSME2. (Figure 7A). The mechanistic links between proteasome activity and ageing are well established (Saez and Vilchez, 2014; McCann and Tansey, 2014).
The proteasome is a large self-compartmentalised protease complex that recognises, unfolds, and destroys ubiquitylated substrates (McCann and Tansey, 2014).
The protein expression levels of FBX041 and NEDD4 increased gradually during kidney ageing, while the signal intensity for PSMD14 and PSME2 was more pronounced at the latest time point (15 months) (Figure 7B). Thus, ageing kidney increases E3 ubiquitin-protein ligases, enhances the ubiquitylated substrates, and stimulates proteasome abundance and activity.
Many downregulated proteins of ageing lungs were involved in processes such as ‘muscle organisation and reassembly’, and the top GO term encompasses an array of myosin motor proteins (Figure 7A). The downregulated ‘myosin complex’ GO category included several proteins belonging to the myosin family, following the observations that human muscle ageing is accompanied by changes in the expression of myosins (Murgia et al., 2017). We did not perform any immunoblotting validation of these myosin proteins due to the high sequence similarity among myosin subunits and the lack of highly specific antibodies against them.
These observations suggested that ageing might influence the lung cell structure through alteration of the higher-order chromatin architecture.
Only a few differentially expressed proteins were observed in the spleen, and they did not allow for useful GO analysis and protein network analysis.
We implemented a comprehensive high-mass-accuracy mass spectrometry-based proteomics strategy to monitor changes in the chromatin-enriched proteomes of six mouse organs over a time course that mimics adult development and ageing, from the early adult to the mature adult stage.
Many age-related diseases are linked to changes in the transcriptional and epigenetic machinery that regulate gene expression.
We focused on the changes in the expression of proteins that mediate transcription, including DNA-binding proteins and chromatin modifiers such as ‘writers’, ‘erasers’, and ‘readers’ (Signolet and Hendrich, 2015; Hyun et al., 2017). Chromatin modifiers add, remove, or recognise particular PTMs of proteins associated with the alteration of chromatin architecture, and ultimately involved in the regulation of gene expression (Santos and Lindner, 2017).
Over 2000 proteins were quantified in each organ, generating a useful resource for researchers investigating mammalian development and ageing.
We identified distinct and organ-specific unique ageing features associated with each organ. We observed unique chromatin modifiers that were expressed and accumulated differently during ageing, leading to changes in the chromatin architecture, including changes in the expression of heterochromatin markers, histone deacetylases, ubiquitin-protein ligase, histone acetylation enzymes, and myosin complex in the brain, heart, kidney, liver, and lung, respectively.
Brain and spleen displayed the largest and most diverse and heterogenous chromatin proteomes. The brain is arguably the most complex organ of a mammal. Brain chromatin structure and function is sustained by a large set of chromatin-associated and nuclear proteins, which also exhibit temporal dynamics of expression during the mouse lifespan as demonstrated here. The spleen chromatin proteome was rather constant during the lifespan whilst large and diverse, which may reflect the physiological role of the spleen for continuous maintenance of important immune and defence activities of the organism.
We demonstrated progressive changes of chromatin-associated protein expression in response to ageing. Also, the specific nature of the organ was more of a significant discrimination factor, and subsequently distinct proteome profiles in response to ageing were observed. For instance, we noticed over the mouse lifespan strongly upregulated chromatin-associated proteins relate to distinctive pathways involved in ‘oxidation-reduction response’, ‘response to oxidation stress’ and ‘nucleosome assembly’, as well as signals that promote apoptosis processes. Conversely, chromatin-associated proteins strongly downregulated are related to “muscle organisation and reassembly” and “histone-modifying enzymes” associated with chromatin assembly and organisation.
Our study of chromatin-enriched proteomes demonstrated that macro-H2A2 accumulates in the mouse brain during ageing. The epigenetic regulator HP1BP3 accumulates at a similar rate and very likely interacts with macro-H2A2. Both macro-H2A2 and HP1BP3 are highly expressed in the adult mouse brain, and we suggest that a complex involving these two proteins is implicated in maintaining heterochromatin integrity and promote gene silencing during the mouse lifespan.
Reversible acetylation of histones plays a critical role in transcriptional regulation in eukaryotic cells. We detected reduced levels of histone H3 acetylation during ageing in mouse liver. The opposite trend was observed in ageing heart, that is, an increase in histone H3 acetylation. Two families of deacylase enzymes were identified: the histone deacetylases, HDACs, and the Sir family protein (Silent Information Regulator)-like family of NAD-dependent deacylases, or sirtuins (Grozinger et al., 2002). Both enzyme families play a major role in gene regulation by modifying the histone acetylation/acylation landscape in response to external stimuli and specific environmental stress conditions, such as oxidative stress (Drazic et al., 2016). In mammals, the brain and heart have the greatest oxygen demand for their ATP-dependent processes. Nearly all cellular processes of cardiomyocytes are driven by ATP-dependent pathways (Hu et al., 2016). SIRT3 plays an important role in maintaining basal ATP levels and regulated energy production in mouse embryonic fibroblasts (Hu et al., 2016; Ahn et al., 2008).
Sirtuin proteins are mostly annotated as mitochondrial proteins, but they can translocate into the nucleus or other cell compartments (Kupis et al., 2016; Vaquero, 2009; Li et al., 2013). The translocation of sirtuins through different cellular compartments is poorly described. Here, we speculate that the roles of SIRT3 and SIRT5 in the heart are essential as they compensate for age-related cellular dysfunction by controlling the levels of histone acylation marks. They may thereby promote the expression of proteins required for DNA repair to prevent cardiac hypertrophy in response to oxidative stress (Hu et al., 2016).
The SAGA complex is a multi-subunit histone-modifying complex. KAT2A is a SAGA component in mammals, containing both a HAT domain and a bromodomain and extended N-terminal domain which confers the ability to acetylate mononucleosomal H3 (Gamper et al., 2009). KAT2A is required for normal development in mice (Xu et al., 2000; Lin et al., 2007). Furthermore, KAT2A-level expression decreases during cell differentiation (Xu et al., 2000), suggesting that the downregulation of histone acetylation is connected with reducing activation of gene expression of target genes which promote self-renewal and pluripotency state during ageing. Overall, our experiments suggest a connection between the roles of two epigenetic enzymes CHD2 and KAT2A, whereby their mutual protein expression is associated with liver differentiation during the mouse lifespan.
The proteasome is a complex proteolytic machine formed by the assembly of several subunits (Finley, 2009). The ubiquitin-proteasome system (UPS) is the primary selective degradation system in eukaryotic cells, localised both in the nuclei and cytoplasm compartment, which is required for the turnover of soluble proteins (Schmidt and Finley, 2014). The UPS is mainly implicated in protein degradation in response to the regulation of several processes, including the maintenance of cellular quality control, transcription, cell cycle progression, DNA repair, receptor-mediated endocytosis, cell stress response, and apoptosis (Lecker et al., 2006). Before a protein is degraded, it is first flagged for destruction by the ubiquitin conjugation system, which ultimately results in the attachment of a polyubiquitin chain on the target protein (Tanaka, 2009; Adams, 2003).
Ubiquitin and the proteasome have been implicated in processes as diverse as the control of transcription, the response to DNA damage, the regulation of chromatin structure and function, and the export of RNAs from the nucleus (McCann and Tansey, 2014).
We found an increase in the levels of two proteasome subunits in ageing kidney organ consistent with increased E3-ubiquitin ligase activity. Increased expression of FBXO41, a subunit of the SCF E3 ubiquitin ligase complex, correlates with NEDD4 expression. Recent reports suggested a decline of proteasome function related to senescence observed in several mammalian tissues and human cells (Bulteau et al., 2000; Carrard et al., 2003; Petropoulos et al., 2000; Bardag-Gorce et al., 1999). During the ageing process, dysfunction of the ubiquitination machinery or the proteolytic activity may occur, leading to proteasome failure, which is linked to several age-related human diseases (Schmidt and Finley, 2014; Chondrogianni and Gonos, 2010). We, therefore, speculate that an increased expression of E3-ubiquitination ligase activity may compensate for the proteasome activity during ageing in the kidneys.
Not all functions of actin and actin-related proteins in complexes are yet clear: it is known that they play important roles in maintaining the stability of the proteins, possibly by bridging subunits and recruiting the complexes to chromatin (Farrants, 2008). In line with previous analysis, the majority of downregulated ‘chromatin-associated proteins’ in lung tissue were assigned to the categories ‘myosin complex’, ‘chromatin organisation’ and ‘histone modification’. The most significant subnetwork corresponds to the ‘muscle organisation and reassembly’ category and is related to the myosin family. This is in accordance with recent reports where the change of protein expression of particular myosin subsets implied a human ageing response (Murgia et al., 2017; Lang et al., 2018) The presence of actomyosin-like protein in the chromatin environment raises questions about the role of actin-like protein such as myosin in nuclear and chromatin processes (Walther and Mann, 2011).
In the spleen, we detected a large number of chromatin-associated proteins but only a fraction of these proteins were differentially regulated during ageing. Thus, the spleen chromatin proteome exhibits a ‘young’ or ‘age-less’ phenotype. Consequently, only a few pathway-level changes were observed, possibly indicating the important role of this organ in maintaining immune functions, removing old red blood cells and microbes which seem, from our data, to be not affected by ageing and consistent with our hypothesis that the time-course changes in protein expression distinctly affect each organ. We observed the largest numbers of chromatin remodelling proteins in the spleen, which suggests that this organ is may provide a good model for epigenetic studies.
Overall, our findings using high mass resolution LC-MS/MS suggest a new approach to investigate the dynamic chromatin protein environment during the lifespan of an organism. We provide a high-quality and robust dataset of protein expression changes in mouse organs during the ageing process. The dataset shows that in vivo models can describe how the dynamic changes of chromatin-associated protein may alternatively promote or repress gene expression during ageing, also reflecting some physiological features of the organ.
Our study adds novel details to mouse biology and chromatin dynamics of organs, and it complements previous attempts to identify biomarkers for mouse lifespan (Shavlakadze et al., 2019; Rappsilber et al., 2003). Walther et al. reported that the bulk proteins’ abundance is less prone to changes in organs such as the brain, heart, and kidney obtained from mice aged 5 or 26 months. They reported only a few proteins that exhibited statistically significant expression changes during ageing (Shavlakadze et al., 2019). Thus, bulk proteome analysis of mammalian organs has limitations, whereas organelle-specific proteomics, as presented here for chromatin, is a more viable strategy to reveal the molecular details of important biological processes, such as ageing and chromatin regulation.
Using a rat model (Rappsilber et al., 2003), Glass et al. studied the alterations of gene expression to identify a putative mammalian ageing signature. Unfortunately, our chromatin proteomics strategy does not readily compare to the study by Glass et al. as only three out of the seven time points are in common between the studies. We observed common ageing signatures (trends) such as cell stress response and transcriptional alterations changing at the late stage of adult rodent lifespan.
We see a consistent overlap between our results and a list of human ageing-related biomarker candidates of the ‘Human Ageing Genomic Resources’ and ‘GenAge machine learning databank’ (Figure 5; Tacutu et al., 2013; Kerepesi et al., 2018). Our differentially expressed candidate mouse proteins behaved just as human ageing biomarkers or ageing-related human proteins that promote disease.
Whereas our study provides novel features and details of molecular ageing processes in mammals, it does not provide the mechanistic details of the protein-mediated ageing process in chromatin. Quantitative proteomics is an important tool for further studies of chromatin dynamics, and the emerging field of high-sensitivity single-cell proteomics will assist in revealing the features of organ function in health, ageing, and disease using limited sample amounts. The experimental protocols used in this study provide a foundation for a more detailed interrogation of chromatin biology by functional proteomics. The data resource associated with this study provides a framework for generating novel hypotheses aimed at revealing the molecular features of ageing and developing novel approaches to mitigate age-related ailments.
Male C57BL/6J mice were obtained from a study approved by the Danish Animal Ethics Inspectorate (J.nr. 2011/561-1950). Wild-type mice were bred in the Biomedical Laboratory, University of Southern Denmark, under a 12 hr/12 hr light/dark cycle (lights on at 6:30 am). Food and water were available ad libitum. Mice were sacrificed by cervical dislocation at the ages of 3, 5, 10, and 15 months. Liver, kidneys, brain, heart, spleen, and lungs were excised, rinsed in ice-cold phosphate-buffered saline (PBS), and immediately snap frozen. Organs were stored at –80°C until further processing.
Mouse embryonic cell pellets were washed in PBS and resuspended in lysis buffer (25 mM Tris·HCl pH 7.6, 150 mM NaCl, 1% NP40, 1% sodium deoxycholate, 0.1% SDS, cOmplete Protease Inhibitor w/o EDTA [Roche], 0.5 mM DTT). The lysates were incubated for 15 min on ice and cell membranes disrupted mechanically by syringing five times with 21G narrow gauge needle and sonicating 3 Å approximately for 2 s at high power. Lysates were incubated on ice for another 15 min and cleared by centrifugation at 14,000 rpm 4°C 30 min. To harvest the nuclear fraction, lysates were resuspended in an equal volume of Nuclear Buffer (120 mM NaCl, 20 mM HEPES pH 7.9, 0.2 mM EDTA, 1.5 mM MgCl2, 20% glycerol, cOmplete Protease Inhibitor w/o EDTA [Roche], 0.5 mM DTT) and dounced 20 times with tight pestle type B. Lysates were incubated for 45 min rotating to dissociated chromatin-bound proteins and pre-cleared by centrifugation at 14,000 rpm 4°C for 30 min. Subsequently, nuclear pellets were lysed in buffer C containing protease inhibitors (420 mM NaCl, 20 mM HEPES pH 7.9, 20% [v/v] glycerol, 0.42 M NaCl, 1.5 mM MgCl2, 0.2 mM EDTA, cOmplete Protease Inhibitor w/o EDTA [Roche], 0.5 mM DTT). Lysates were incubated for 1 hr rotating at 4°C, in the presence of 250 U/mL benzonase nuclease, to form dissociated chromatin-bound proteins and pre-cleared by centrifugation (20,000 × g, 1 hr at 4°C). After centrifugation, the supernatant was snap frozen.
Organ samples were homogenised on ice in a homogenisation buffer (2.2 M sucrose, 10 mM HEPES/KOH pH 7.6, 15 mM KCl, 2 mM EDTA, 0.15 mM spermine, 0.5 mM spermidine, 1 mM DTT, and cOmplete Protease Inhibitor w/o EDTA [Roche]), 0.5 mM PMSF, and phosphatase inhibitor (PhosSTOP, Roche) using a ‘loose’-type pestle tissue. The solution was stacked over with a cushion buffer (2.05 M sucrose, 10 mM HEPES/KOH pH 7.6, 15 mM KCl, 2 mM EDTA, 0.15 mM spermine, 0.5 mM spermidine, 1 mM DTT, cOmplete Protease Inhibitor w/o EDTA [Roche], 0.5 mM PMSF) in an ultracentrifuge tube and cleared by centrifugation (20,000 × g, 1 hr at 4°C). Pellet containing nuclei was washed twice with 1 mL Dulbecco PBS (3000 × g, 5 min at 4°C). To harvest the nuclear fraction, lysates were subsequently resuspended in buffer C (420 mM NaCl, 20 mM HEPES pH 7.9, 20% v/v glycerol, 2 mM MgCl2, 0.2 mM EDTA, 0.1% NP40, cOmplete Protease Inhibitor w/o EDTA [Roche], 0.5 mM DTT) and dounced 20 times with a ‘tight’ pestle. Lysates were incubated for 1 hr rotating at 4°C, in the presence of 250 U/mL benzonase nuclease, to form dissociated chromatin-bound proteins and pre-cleared by centrifugation (20,000 × g, 1 hr at 4°C). After centrifugation, the supernatant was snap frozen.
Protein lysate was quantified by using the Bradford assay. Approximately 30 μg protein lysates were separated on SDS–PAGE gels and transferred to nitrocellulose membranes. Membranes were blocked with 5% non-fat milk or 5% BSA at room temperature for 1 hr and incubated overnight with diluted primary antibody at 4°C. Membranes were then washed and incubated with HRP-conjugated goat-anti-rabbit or mouse IgG secondary antibody for 1 hr at room temperature. Membrane was incubated with enhanced chemiluminescence reagents (Thermo Scientific) followed by exposure to X-ray films. Immunoblotting was performed using the antibodies and conditions listed in the Key resources table.
Protein concentration was measured by using the Bradford assay. 50 µg of chromatin lysate was precipitated with 4 vol of acetone and resuspended in 20 µL 6 M urea/2 M thiourea in 50 mM ABC. Cysteines were reduced with 1 mM DTT for 30 min and alkylated with 5 mM iodoacetamide for 30 min. 500 ng LysC (Wako) was added, and proteins were digested at room temperature for 3 hr. Samples were diluted 1:4 with 50 mM ABC and 1 µg trypsin (Sigma) and incubating overnight at 37°C. Digestion was quenched by addition of formic acid (FA) to a final concentration of ~1%, and the resulting peptide mixture was centrifuged at 5000 × g for 15 min. Peptides in the supernatant were desalted by stage tipping with C18 material (Silva et al., 2006) and eluted by 50% ACN, 0.1% FA in water, and dried in a SpeedVac. Samples were stored at −80°C until further use.
Samples were redissolved in 50 μl of trifluoroacetic acid 0.1% (vol/vol) in water, as buffer A, and sonicated for 1 min and centrifuged for 15 min at 15,000 × g. Analysis was carried out on an Ultimate 3000 RSLCnano HPLC system connected to a mass accuracy high-resolution mass spectrometry, Q Exactive HF (Thermo Fisher). The MS instrument was controlled by Xcalibur software (Thermo Fisher). The nano-electrospray ion source (Thermo Fisher) was used with a spray voltage of 2.2 kV. The ion transfer tube temperature was 275°C. Samples were loaded on a cartridge pre-column PepMap 100 5 * 0.3 mm (Thermo Fisher) in 2% ACN, 98% H2O, 0.1% TFA at 10 µL/min, and then separated either with an easy Spray C18 or a 75 µm * 50 cm 2 µm PepMap 100 column (Thermo Fisher). Separation was done in a linear gradient of ACN, 0.1% FA (buffer B) in H2O, 0.1% FA (buffer A, from 4% to 32% B in 2 hr) at 0.25 µL/min at 45°C. To avoid sample carryover between different samples, both pre-column and column were washed with 3 * 10 min gradients from 2% to 95% B (3 min at 2% B – 3 min from 2% to 95% B – 3 min at 95% B – 1 min from 95% to 2% B). MS analysis was done in DDA mode with 1 MS1 scan, followed by 20 dependent MS2 scans. MS1 parameters were 120,000 resolution, 3e6 AGC target, maximum IT 100 ms, with a scan range of 300–2000 m/z. MS2 parameters were 15,000 resolution; 2e5 AGC target; maximum IT 15 ms; isolation window 1.2 m/z; isolation offset 0 m/z; fixed first mass 110 m/z; (N) CE 30; minimum AGC 8e3; exclude unassigned and 1; 6–8 charges; preferred peptide match; exclude isotopes, and dynamic exclusion was set to 40 s. All mass spectrometry raw data were deposited in MassIVE (https://massive.ucsd.edu/ProteoSAFe/static/massive.jsp) with accession numbers MSV000084270, MSV000084279, and MSV000084375.
A combination of Progenesis QI v2.2 (Waters) and Proteome Discoverer v22.214.171.124 (Thermo Scientific) was used to estimate the relative protein abundance of protein and peptide using a label-free approach. Thermo Raw MS files were imported into Progenesis QI v2.2 (Waters) and the match-between-runs feature was enabled. Subsequently, the matching and alignment time window were performed according to m/z and retention time enable using default settings. Filtering only ions with a charge state of up to +4 was considered. The aligned ion intensity map was carried out using default Peak Peaking settings, and no further imputation analysis was performed. The aligned ion intensity map was exported in .pepXml files and imported into Proteome Discoverer v126.96.36.199 (Thermo Scientific) for further protein and peptide identification and searched against the SwissProt mouse reference database by using an in-house MASCOT server (v2.5.1, Matrix Science Ltd, London, UK). Database searches were performed with the following parameters: Fixed Mod: cysteine carbamidomethylation; Variable Mods: methionine oxidation; Trypsin/P digest enzyme (maximum two missed cleavages); Precursor and fragment mass tolerance was set to 10 ppm and 0.8 Da, respectively. Identified peptides and proteins were filtered using a false discovery rate (FDR) set at 1% and a peptide spectrum match (PSM) set at 1%. Subsequently, the MS/MS and ion abundance search was exported in .mgf files and imported into Progenesis QI v2.2 to perform peptide and protein normalisation and relative quantitation using respectively a label-free analysis Hi-N/3 summarisation method and ion abundance normalisation default method (Ferguson et al., 2005; Korthauer et al., 2019; Supplementary file 3, Table S3, Figure 5—figure supplement 1B). In terms of proteomic depth and data management, for each biological sample two technical replicates were combined by summing both protein ion abundance values. Due to the large proteome scale, each organ were analysed separately to maintain the reliability of the statistical analysis and prevent potential homogenisation of the protein expression differences that occurred during the mouse’s lifespan. Protein group databases are listed in Supplementary file 3, Table S3.
To identify the proteins that exhibited significant abundance changes during ageing, each organ dataset was further normalised against the ‘core proteome’, that is, the proteins identified in all six organs over all four time points. Subsequently, statistical tests based on the rank product test were carried out to quantify the dynamic protein expressions changing during the mouse lifespan (Koziol, 2010; Breitling et al., 2004; Figure 5—figure supplement 2A). FDRs were calculated to correct for multiple testing (Tyanova et al., 2016).
Hierarchical heatmap clustering based on Euclidean distance was performed using Perseus software (1.6.0) (Tyanova et al., 2016). Z-scores were calculated by subtracting the mean of protein abundance values in all samples and dividing by the standard deviation. Protein group tables are listed in Supplementary file 4, Table S4.
GO analysis of differentially expressed proteins was performed using the DAVID online tool (v6.8) to obtain the biological processes (BPs) and pathways from the enriched chromatin proteome organs during ageing. FDR 1% was set as the minimum threshold value. Each dataset was independently analysed, and the M. musculus’ (mouse) genome/proteome was used as background for each organ source. To generate a detailed dataset, which provides information on the location and topology of the protein in the cell, the data was sorted according to the Uniprot (SwissProt) subcellular location library. In particular, the subcellular location section was upgraded with the most recent depository protein libraries relative to chromatin studies (Medvedeva et al., 2015; van Mierlo et al., 2019; Christoforou et al., 2016; Xu et al., 2016).
Proteomics data is deposited in public repository as specified in manuscript.
The proteasome: structure, function, and role in the cellCancer Treatment Reviews 29:3–9.https://doi.org/10.1016/S0305-7372(03)00081-1
Changes in 20S proteasome activity during ageing of the LOU ratMolecular Biology Reports 26:89–93.https://doi.org/10.1023/A:1006968208077
Macro histone variants are critical for the differentiation of human pluripotent cellsThe Journal of Biological Chemistry 288:16110–16116.https://doi.org/10.1074/jbc.M113.466144
Age-related alterations of proteasome structure and function in aging epidermisExperimental Gerontology 35:767–777.https://doi.org/10.1016/S0531-5565(00)00136-4
Progeria syndromes and ageing: what is the connectionNature Reviews. Molecular Cell Biology 11:567–578.https://doi.org/10.1038/nrm2944
Impact of ageing on proteasome structure and function in human lymphocytesThe International Journal of Biochemistry & Cell Biology 35:728–739.https://doi.org/10.1016/S1357-2725(02)00356-4
Proteasome function determines cellular homeostasis and the rate of agingAdvances in Experimental Medicine and Biology 694:38–46.https://doi.org/10.1007/978-1-4419-7002-2_4
A draft map of the mouse pluripotent stem cell spatial proteomeNature Communications 7:9992.https://doi.org/10.1038/ncomms9992
UpSetR: an R package for the visualization of intersecting sets and their propertiesBioinformatics (Oxford, England) 33:2938–2940.https://doi.org/10.1093/bioinformatics/btx364
MacroH2A histone variants maintain nuclear organization and heterochromatin architectureJournal of Cell Science 130:1570–1582.https://doi.org/10.1242/jcs.199216
Stress biology and aging mechanisms: toward understanding the deep connection between adaptation to stress and longevityThe Journals of Gerontology. Series A, Biological Sciences and Medical Sciences 69 Suppl 1:S10–S16.https://doi.org/10.1093/gerona/glu055
Recognition and Processing of Ubiquitin-Protein Conjugates by the ProteasomeAnnual Review of Biochemistry 78:477–513.https://doi.org/10.1146/annurev.biochem.78.081507.101607
The STAGA Subunit ADA2b Is an Important Regulator of Human GCN5 CatalysisMolecular and Cellular Biology 29:266–280.https://doi.org/10.1128/MCB.00315-08
HP1BP3 is a novel histone H1 related protein with essential roles in viability and growthNucleic Acids Research 43:2074–2090.https://doi.org/10.1093/nar/gkv089
HP1BP3 expression determines maternal behavior and offspring survivalGenes, Brain, and Behavior 15:678–688.https://doi.org/10.1111/gbb.12312
Readers, writers, and erasers: chromatin as the whiteboard of heart diseaseCirculation Research 116:1245–1253.https://doi.org/10.1161/CIRCRESAHA.116.303630
Redox theory of aging: implications for health and diseaseClinical Science (London, England 131:1669–1688.https://doi.org/10.1042/CS20160897
Regeneration of the spleen and splenic autotransplantationThe British Journal of Surgery 78:270–278.https://doi.org/10.1002/bjs.1800780305
Roles of SIRT3 in heart failure: from bench to bedsideJournal of Zhejiang University. Science. B 17:821–830.https://doi.org/10.1631/jzus.B1600253
Writing, erasing and reading histone lysine methylationsExperimental & Molecular Medicine 49:e324.https://doi.org/10.1038/emm.2017.11
Single Muscle Fiber Proteomics Reveals Distinct Protein Changes in Slow and Fast Fibers during Muscle AtrophyJournal of Proteome Research 17:3333–3347.https://doi.org/10.1021/acs.jproteome.8b00093
Protein degradation by the ubiquitin-proteasome pathway in normal and disease statesJournal of the American Society of Nephrology 17:1807–1819.https://doi.org/10.1681/ASN.2006010083
Developmental potential of Gcn5−/− embryonic stem cells in vivo and in vitroDevelopmental Dynamics 236:1547–1557.https://doi.org/10.1002/dvdy.21160
A Three‐Dimensional Display for Cardiac Activation MappingPacing and Clinical Electrophysiology 14:538–545.https://doi.org/10.1111/j.1540-8159.1991.tb02826.x
The Search for Longevity and Healthy Aging Genes: Insights From Epidemiological Studies and Samples of Long-Lived IndividualsThe Journals of Gerontology. Series A, Biological Sciences and Medical Sciences 67:470–479.https://doi.org/10.1093/gerona/gls089
The role of nuclear architecture in genomic instability and ageingNature Reviews. Molecular Cell Biology 8:692–702.https://doi.org/10.1038/nrm2238
Dynamic protein interactions of the Polycomb Repressive Complex 2 during differentiation of pluripotent cellsMolecular & Cellular Proteomics 15:3450–3460.https://doi.org/10.1074/mcp.M116.062240
Increase of Oxidatively Modified Protein Is Associated With a Decrease of Proteasome Activity and Content in Aging Epidermal CellsThe Journals of Gerontology Series A 55:B220–B227.https://doi.org/10.1093/gerona/55.5.B220
Mechanistic insights into aging, cell-cycle progression, and stress responseFrontiers in Physiology 3:183.https://doi.org/10.3389/fphys.2012.00183
Sirtuins: Molecular traffic lights in the crossroad of oxidative stress, chromatin remodeling, and transcriptionJournal of Biomedicine & Biotechnology 2011:368276.https://doi.org/10.1155/2011/368276
Constitutive heterochromatin formation and transcription in mammalsEpigenetics & Chromatin 8:3.https://doi.org/10.1186/1756-8935-8-3
Protein Posttranslational Modifications: Roles in Aging and Age-Related DiseaseOxidative Medicine and Cellular Longevity 2017:5716409.https://doi.org/10.1155/2017/5716409
Regulation of proteasome activity in health and diseaseBiochimica et Biophysica Acta (BBA) - Molecular Cell Research 1843:13–25.https://doi.org/10.1016/j.bbamcr.2013.08.012
Age‐related behavioral changes from young to old age in male mice of a C57BL/6J strain maintained under a genetic stability programNeuropsychopharmacology Reports 39:100–118.https://doi.org/10.1002/npr2.12052
The function of chromatin modifiers in lineage commitment and cell fate specificationThe FEBS Journal 282:1692–1702.https://doi.org/10.1111/febs.13132
STRING v10: protein–protein interaction networks, integrated over the tree of lifeNucleic Acids Research 43:D447–D452.https://doi.org/10.1093/nar/gku1003
Human Ageing Genomic Resources: Integrated databases and tools for the biology and genetics of ageingNucleic Acids Research 41:D1027–D1033.https://doi.org/10.1093/nar/gks1155
Epigenetic differences between naïve and primed pluripotent stem cellsCellular and Molecular Life Sciences 75:1191–1203.https://doi.org/10.1007/s00018-017-2703-x
Determinants of postnatal spleen tissue regeneration and organogenesisNPJ Regenerative Medicine 3:39.https://doi.org/10.1038/s41536-018-0039-2
The proteasome: overview of structure and functionsProceedings of the Japan Academy, Series B 85:12–36.https://doi.org/10.2183/pjab.85.12
The conserved role of sirtuins in chromatin regulationThe International Journal of Developmental Biology 53:303–322.https://doi.org/10.1387/ijdb.082675av
Recruitment of the Mammalian Histone-modifying EMSY Complex to Target Genes Is Regulated by ZNF131The Journal of Biological Chemistry 291:7313–7324.https://doi.org/10.1074/jbc.M115.701227
Accurate quantification of more than 4000 mouse tissue proteins reveals minimal proteome changes during agingMolecular & Cellular Proteomics 10:M110.https://doi.org/10.1074/mcp.M110.004523
OUP accepted manuscriptNucleic Acids Research 45:D264–D270.
Martin Sebastian DenzelReviewing Editor; Altos Labs, United Kingdom
Matt KaeberleinSenior Editor; University of Washington, United States
In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.
[Editors' note: this paper was reviewed by Review Commons.]https://doi.org/10.7554/eLife.73524.sa1
1. The supplementary data was missing, so it is difficult to look at the quality of the data itself.
We apologize for this oversight. We shared with the reviewers the mass spectrometry RAW data deposited on MassIVE:
LINK-1- mESC: ftp://massive.ucsd.edu/MSV000084270/
LINK-2- Tissue Brain Total Lysate: ftp://massive.ucsd.edu/MSV000084279/
LINK-3- Tissue six organs Chromatin Lysate: ftp://massive.ucsd.edu/MSV000084375/
We also attach the large dataset from the ProteinGroup file of our mass spectrometry analysis to the manuscript.
In the supplementary material we updated all the ProteinGroup.file (Proteome Tables) in excel format. They correspond to:
1. Proteome of mouse embryonic stem cell obtained from whole cell lysate and chromatin enriched lysate.
File name: Supplementary material, Table S1
2. Brain proteome whole cell lysate during adult mouse lifespan
File name: Supplementary material, Table S2
3. proteomics of six mouse organs over time
File name: Supplementary material, Table S3
4. Uniquely regulated proteins organ-specific during ageing.
File name: Supplementary material, Table S4
2. Table 2: please remove the sum of all proteins which is not helpful. The combined number (currently in parentheses) is all we need. Please add the number of unique proteins for each organ.
We adjusted Table 2 as suggested.
3. It is interesting that the spleen has less changes over age than other organs. the authors may want to discuss this considering the differences in longevity of cell types.
We agree. We included a more detailed presentation of the spleen results in the manuscript and made this observation more prominent in the abstract, results and Discussion sections.
4. Generally, there is quite a bit of information missing in many figure legends or the figures themselves. Some examples are below. Please check carefully.
Thanks for this important comment. We have revised the legends and figures accordingly.
5. Please provide details about statistical analysis of proteomics data. Was ANOVA used? Was data FDR controlled?
We revised the description of data analysis (material and method section), line 843
The proteome dataset was obtained from organ samples harvested during adult mice lifespans. We focused on proteins that changed in one direction during the ageing of the organ, or linearly throughout the lifespan of the animal. A statistical strategy was applied to detect the most significant chromatin proteome changes associated with ageing. We used rank product tests to identify proteins that exhibited significant abundance changes during ageing (supplementary material, Figure 5, Panel D). We called these “differentially regulated proteins”. Depending on the experiment’s intrinsic data structures and variations, rank product tests was applied (DOI: https://doi.org/10.1074/mcp.RA119.001777). Significance thresholds were then calculated via a permutation-based false discovery rate using adjusted q-values calculated to correct for multiple testing (-log10 qValue < 0.1 cutoff)
6. I am not aware how Progenesis works. Is there some form of imputation when there are missing values?
We revised the description of data analysis (material and method section):
A combination of Progenesis QI v2.2 (Waters) and Proteome Discoverer v188.8.131.52 was used to estimate the relative protein abundance of protein and peptide using in label-free approach. For a better estimation and comparison of protein expression across different biological conditions, the Progenesis QI v2.2 integrated the MS files according to m/z and retention time features (Peak Peaking Strategy). The outcome increased the reliability and precision of peptide abundance measurements. After a Peak Peaking strategy was applied, no further imputation analysis was performed. The aligned ion intensity map was exported in.pepXml files and imported into Proteome Discoverer v184.108.40.206 (Thermo Scientific) for further protein and peptide identification.
By, Progenesis QI peptide ion abundance was normalised using the default method (Ferguson et al., 2005: "Housekeeping proteins: a preliminary study illustrating some limitations as useful references in protein expression studies", DOI: 10.1002/pmic.200400941). Progenesis QI for proteomics uses ratiometric data in log space, along with a median and mean absolute deviation outlier filtering approach, to calculate the scalar factor. This is a more robust approach, which is less influenced by noise in the data and any biases owing to abundant species, as the absolute values of abundance are disregarded.
7. Please explain better the heatmap/clustering. I assume the data represents z-scores? What are the colours to the right of the heatmaps?
We revised the heatmaps and clustering in Figure 5 and revised the description of data analysis line 848.
Hierarchical heatmap clustering based on Euclidean distance was performed using Perseus software (1.6.0) (100). Z-scores were calculated by subtracting the mean of protein abundance values in all samples and dividing by the deviation.
The colours on the right of side of the heatmaps indicate blue chromatin associated protein, green nuclear protein, and yellow protein associated with other cellular components. Description present in the figure legend.
8. Figure 5c, please provide the individual months in the bar graphs. To what has the data been normalised?
Figure 5c was adjusted following reviewer’s suggestions.
1. The link to the "mass spectrometry raw data" just has "XXX" as the location of the data. The processed data table is also not shared; what are the peptide and protein quantifications for every gene at every timepoint? The value of this study is largely because it will serve as a useful reference for future studies, but the data are obscured; figure 5C is the only place we actually see the fold changes across age for any named proteins. Or did I miss this somewhere? I looked through twice as I was surprised such a critical piece of information was missing.
We apologize for this omission.
We share the mass spectrometry RAW data deposited on MassIVE:
LINK-1- mESC: ftp://massive.ucsd.edu/MSV000084270/
LINK-2- Tissue Brain Total Lysate: ftp://massive.ucsd.edu/MSV000084279/
LINK-3- Tissue six organs Chromatin Lysate: ftp://massive.ucsd.edu/MSV000084375/
We also attach the large dataset from the ProteinGroup file of our mass spectrometry analysis to the manuscript.
In the supplementary material we updated all the ProteinGroup.file (Proteome Tables) in excel format. They correspond to:
1. Proteome of mouse embryonic stem cell obtained from whole cell lysate and chromatin enriched lysate.
File name: Supplementary material, Table S1
2. Brain proteome whole cell lysate during adult mouse lifespan
File name: Supplementary material, Table S2
3. Chromatin proteomics of six mouse organs over time
File name: Supplementary material, Table S3
4. Uniquely regulated proteins Organ-specific during ageing.
File name: Supplementary material, Table S4
2. In the introduction, the authors state that "one adult mouse month equivalent to approximately three human years".
At note (2) the reviewer points out the lack of reference in the following sentence from the Introduction paragraph: “Mice have a relatively short lifespan, with one adult mouse month equivalent to approximately three human years”. We agree with the reviewer and we have added the relevant reference.
The two articles from Pallav Sengupta and co-authors attempt to compare rodent and human age at different phases of their life. The precise correlation between the age of laboratory rodents and humans is still a subject of debate. Both articles try to fill this gap between approximation and accuracy, and they suggest a precise relation between murine and human age using multiple assessments central to a comparison of phenotypes and physiology features.
3. Figure 2C – can you also add a histogram for "intersect of all six tissues"? A Venn diagram would be overly complex here but you can use the "UpSetR" R package to make a comprehensible histogram showing overlap. Ok, I wrote this comment before reading the rest of the paper – I see you already did that. I think Figure 4A can already go directly here in Figure 2C? We don't really need to go through the analysis of Figure 3 to get here.
(5) Table 2 and Figure 2C are redundant; just color Figure 2C the same way with the blue/green/yellow and then the ratios will be clear without the need for the table. Table 3 is similarly redundant with Figure 4A.
Reviewer #2 suggested some editing of Figure 2 and Table 2 to avoid redundancy between tables and figures in the manuscript. We thank Reviewer #2 for the advice, but we are also aware that those changes may affect the meaning of Figure 2. We therefore maintain Figure 2 and Table 2 in order to maintain consistency in the manuscript. We rephrased the manuscript text in order to avoid any redundancy or misleading information between Figure 2C and Table2. In the paragraph “Quantitative chromatin proteomics of ageing mouse organs”, Figure 2 showed the overall number of proteins identified across six organs over time, while Table 2 highlighted a more in-depth meaning to Figure 2. Table 2 showed the proteins detected in six mouse organs across time and displayed the number of annotated proteins associated with their subcellular location. Table 2 also showed the number of unique proteins in each organ for each subcellular location, suggesting a difference at the protein level appearing across all organs.
4. Why does Figure 3B only show one square per condition? Is this the average of the three replicates per time? Are these biological replicates in Figure 3A, or technical? (I assume biological, but I don't see it unambiguously stated.)
We adjusted Figure 3 Panel B by introducing the biological replicates per time point in the Pearson correlation coefficients cluster, line: 238. Previously the average of each biological replicate across each time point was included in the analysis. Reviewer #2 pointed out an interesting observation and we adjusted the figure by introducing all the biological replicates for all samples across all time points. We demonstrated reproducibility of each biological replicate and confirmed the robustness of our biochemical and proteomics methodology. We observed that this round of analysis did not differ from the previous analysis. The Pearson correlation coefficients Cluster (figure 3B) showed the consistency and reproducibility of analysis of three biological replicates at all time points and revealed distinct organ proteome profiles. We thank Reviewer #2 for the suggestions which improves the quality of Figure 3B.
Notes (6), (7), (8) and (13) from Reviewer #2 focused on the GO term analysis. Notes (6), (7) and (8) corresponded to Figure 4C and note (13) related to Figure 6. We thank Reviewer #2 for the constructive feedback providing useful insights that contributes to the GO term analysis in the manuscript.
6. The authors mention that "the core chromatin proteome contained proteins associated with the major transcriptional/epigenetic chromatin complexes such as…" which I don't doubt, but the way it is presented it could also just be cherry-picking. Please run the protein composition through DAVID to show whether those categories of proteins are enriched. Make sure to also upload a specific "Background" list to DAVID that includes only the proteins measured, as by default it will compare your uploaded list against all genes, which is not accurate for proteomics since the 863 measured proteins are going to be a biased sample. Please do the same thing for every subsequent list of genes that is listed, e.g. the "they included the histone binding protein of the methyltransferase category…" for brain in the following paragraph, spleen in the following paragraph, etc. For each, please include a supplemental table with the genes uploaded for each DAVID analysis as well as the different background list for each analysis. OK, I'm reading sequentially while I write this review, and I see such a GO analysis was done later on the page, but the analysis should be included directly with the first time the list of proteins are mentioned (e.g. lines 280-288 should go directly after line 252).
We re-edited the manuscript in order to avoid any confusion about the GO term analysis performance, line: 835. In the “Material and Method” section, <
7. The GO analysis results don't always agree with the earlier text. For instance Figure 4B agrees with the text (i.e. that translation genes are enriched in the common 863), but figure 4C with the brain proteins does not list any methyltransferase or histone deacetylase category-or indeed anything related to translation. Thus it seems from the GO analysis that the genes selected in lines 259-260 were cherry picked rather than representatives of a statistically significant cluster.
In order to investigate a comprehensive functional analysis of the proteomics datasets across all the organs during ageing we applied an UpSetR plot cluster analysis. We first observed a shared proteome, with 863 proteins present across all organs over time (Figure 4, Panel A, Table 3). We labeled the 863 shared proteins as the “Core” proteome. As expected from the overlapping of six proteome organs across time, functional analysis of the “Core” proteome anticipated the presence of categories associated with aging response, such as “response to stress”, “system development”, “cell cycle”, and “regulation of cell death”. Since our strategy was performed in the chromatin and nuclear environment the resulting GO term categories matched with chromatin cellular environments, such as “nucleosome assembly”, “chromatin remodeling”, and “covalent chromatin modification”.
From UpSetR plot cluster analysis (Figure 4A) we observed that relatively large numbers of unique proteins in each organ likely reflected the inherent features of the individual organs and the diversity of cell types and physiology.
As we observed in previous analysis the largest proteomes were observed in mouse brains (3110) and spleens (3125). Consequently, the two organs exhibited a large number of unique proteins suggesting a distinctive chromatin proteome profile. As expected, due to the chromatin enrichment, we listed the unique epigenetic enzymes relative to brain and spleen. We detected protein associated with transcriptional gene regulation, included histone binding protein of the methyltransferase enzymes and several epigenetic complexes.
8. * "We observed distinctive organ-specific categories, "age-classes and age-development" and categories that reflected their organ source". OK, certainly proteins like "nervous system development" are clearly brain related and "liver development" clearly liver related, so it's nice that those categories are enriched but this isn't really very interesting… it just means the samples are not mislabeled. Regarding the more interesting distinctive "age-classes" and "agedevelopment" that I do not see. What are the "age-classes"? And "age-development" seems vague. True developmental genes will not be especially important at any timepoint the authors' selected, as the first age selected (3 months) is already full adulthood and beyond puberty.
We rephrased this sentence, line 314:
We observed distinctive organ-specific categories (“age-classes and age-development”) and categories that reflected their organ source (Figure 4, Panel C). For instance, unique Go term categories were associated with each organ:
“Nervous system development” and “chemical synaptic transmission” related to the brain;
“Cardiac myofibril assembly” and “adult heart development” were attributed to the heart;
“Steroid metabolic process” and “liver development” were distinctive to the liver;
“Transport” and “sodium ion transport” categories related to the kidney, and;
“Angiogenesis” and “respiratory gaseous exchange” were present in the lung.
These results confirmed that many proteins found in individual organs are likely to confer organspecific functions (Figure 4, Panel C). Our strategy showed that a robust enrichment of the chromatin-associated protein in mouse organs was obtained by using GO term analysis.
We reported a significant enrichment of annotated categories listed as age-related, including a large class of molecular features associated with the “core” chromatin environment present in all organs. Our analyses ensured the success of our strategy and confirmed the hypothesis that mammalian organs have ageing-dependent signatures and unique chromatin-associated proteins.
9. * In Figure 5B, the authors find a number of genes that change with age and that are mostly tissue-specific. That's fine, but due to the relatively small sample size and unique tissue signatures, it is hard to be confident in really any of these findings. The authors findings would benefit tremendously with a meta-analysis. For instance, David Glass's lab has a 2019 transcriptional study on liver and kidney gene expression across aging in a conceptually similar study; the hits should be compared. (They also checked hippocampus and skeletal muscle, which may be a worse comparison with brain and heart-but possibly worth checking anyway.) PMID 31533046. This is mRNA and in rats, so much may be different, but when there are overlapping results, that would raise confidence. There are other conceptually-similar studies to compare to as well; off the top of my head there is a 2011 Matthias Mann paper (PMC3033683) that looks at proteome in heart, kidney, and brain. It does not have as many timepoints as yours or the Glass paper, but in other ways is similar.
Glass and Co-authors (PMID 31533046) produced a multi-time point age-related gene expression signature (AGES) from liver, kidney, skeletal muscle, and hippocampus from rats, comparing 6, 9, 12, 18, 21, 24, and 27month old animals. Glass and Co-authors focus on genes that changed in one direction throughout the lifespan of the animal, either early in life (early logistic changes), at middle-age (mid-logistic), late in life (late logistic), or linearly throughout the lifespan of the animal.
The overlap between our proteomic strategy and the transcriptome analysis from Glass and Coauthors may reveal a universal aging signature across mammalians, more specifically across murine. However, there are substantial differences between our strategy and Glass and Coauthors’ approach which may need to be discussed before any meta-analysis is carried out.
Glass and Co-authors used a different murine ageing model in the rat. Rats were chosen because previously studies showed that rats are an excellent model for sarcopenia, which is the age-related loss of skeletal muscle. However, no mention of rat gender was described in the article. Our ageing model was male mouse C57BL/6J. There is extensive literature documenting the moderating effects of healthy aging and gender on cognition in humans (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3181676/) and aging-related gender differences at a bio-psycho-social level (https://www.karger.com/Article/Pdf/323154). It may be risky to compare data through meta-analysis if the animal models have different genders.
Another critical point is the lack of overlapping between the murine lifespans. Glass and Coauthors harvested organs from 6, 9, 12, 18, 21, 24, and 27 month old animals, which corresponds to middle‐aged, mature adult and old adult mice. We expected a drastic change near where mortality occurs. Our strategy does not completely overlap with the Glass and Coauthors’ rodent lifespan analysis, as only 3 out of the 7 time points are related between the two studies. To exclude any age-related changes due to the alteration of social behavior, including physical characteristics such as motor function and locomotor activity affecting the in vivo dynamics of chromatin changes during ageing, we exclude from our analysis late-stage adult mice in the 18‐24 month old age bracket. Our strategy monitors changes in the chromatinenriched proteomes of mouse organs over a time course that mimics adult development and ageing, from early adult stage upper to middle-aged adult, excluding the stages of lifespan where mortality occurs.
Lastly, Glass and Co-authors’ transcriptome analysis was measured using RNA sequencing technology. The RNA libraries were prepared using the Illumina TruSeq. For this reason, we do not know how large the RNA libraries are, how many gene were included in the custom library, or how many gene readings were performed. The custom gene library may lack of chromatin or nuclear proteins, thus limitation of the meta-analysis may occur during the comparison with our enriched chromatin study.
We found the suggestion from Reviewer #2 extremely interesting. The manuscript certainly contains a good meta-analysis. In the paragraph beginning “Distinct organ ageing profiles are defined by unique protein expression patterns”, line 391, we showed a consistent overlapping between our analysis and a series of human protein biomarker candidates for ageing obtained from the “Human Ageing Genomic Resources” and “GenAge machine learning databank” (supplementary material Figure 5, Panel F) (50) (51). In the brain protein IREB2 is associated with Alzheimer’s disease, whereas the brain protein MAOB is associated with both Alzheimer’s and Parkinson’s diseases. The heart proteins ADD3, PTGIS and COL1A2 are candidates for hypertension and myocardial infarction. The liver proteins INSR, PTPN1 and ENPP1 are associated with diabetes mellitus type 2 and obesity. Lung protein CYP2E1 is related to lung adenocarcinoma and MMP9 is associated with lung neoplasms. In the spleen, the KLK1 protein is a biomarker candidate for hypertension (supplementary material Figure 5, Panel F). In summary, we observed how differentially regulated proteins changed in mice during ageing just like those found in humans.
We inserted the Reviewer #2 suggestion in the discussion, line: 663
10. * Generally, there are a lot of examples picked out for which it is impossible to determine the significance. The authors mention (line 342) that HDAC2 is downregulated in liver and that GATA2AD and TRIM28 are upregulated in kidney. Certainly this could be accurate and mechanistically meaningful, but there are a lot of genes here, we will see a lot of patterns coincidentally. No significance is shown on any of these plots. Based on the fold changes and error bars it looks like it's probably significant, but there's also a ton of multiple testing going on here, which is of course a common issue in exploratory omics analysis. Same concern for the "Reader" enzymes being upregulated in brain and others downregulated in liver. Is this noise? When you do have proteins that are significant in more than one tissue, what's the probability they go in tandem versus opposite? Is it 50/50?
Temporal analysis of the overall dataset, to quantify the protein expressions profile changes during ageing, was performed based on a rank product test. Adjusted q-values were calculated to correct for multiple testing (-log10 qValue < 0.1 cutoff). We used an original approach based on a PolySTest analysis (DOI: https://doi.org/10.1074/mcp.RA119.001777) from Veit Schwämmle and Co-authors were they developed a robust statistical toolbox to improve coverage and depth of large-scale proteomics analysis.
Regarding Figure 5, we observed protein belong to the same chromatin remodeling complex changed expression. For instance, "Reader" enzymes being upregulated in brain and others downregulated in liver. Recently, studies shown the alteration of chromatin complex assembly is largely dynamic. Depending to the cell fate state content the architecture of the chromatin remodelling complex is largely affected. This may explained why "Reader" enzymes being upregulated in brain (Glyr1, Baz1b) and others downregulated in liver (Dpf2, Brd2, Chd2). Also, another reason may the content of the tissue morphology and the diverse “reader” behavior to regulate the gene expression in different cellular contexts.
This result is consistent with the highly dynamic nature of chromatin remodelling complexes. Some protein subunits are only present in a complex at distinct time-points to provide a unique function or feature.
We address the Reviewer #2 suggestion in the manuscript line: 380. We also suggested relevant literature about the alteration of chromatin remodelling assembly
Dynamic Protein Interactions of the Polycomb Repressive Complex 2 during Differentiation of Pluripotent Cells: doi: 10.1074/mcp.M116.062240
Dynamic Competition of Polycomb and Trithorax in Transcriptional Programming doi: https://doi.org/10.1146/annurev-biochem-120219-103641
A Family of Vertebrate-Specific Polycombs Encoded by the LCOR/LCORL Genes Balance PRC2 Subtype Activities doi: https://doi.org/10.1016/j.molcel.2018.03.005
11. ** The shape of the volcano plot in Supplemental Figure 5D – an almost perfect U – indicates a rather aggressive normalization procedure that almost ties fold change to p-value. This is maybe unavoidable, but it's not ideal. The log2 fold changes are also very large – log2FCs of -5 and +7? So some proteins are upregulated 128-fold in the spleen across age? In general the log2FCs here are far larger than I've seen in any study. For instance, the Mann paper above (PMC3033683) finds only 1% of proteins that change by more than 2-fold between 5 and 26 months. My own studies (although not on aging) also find far smaller fold changes. I don't think I've ever seen a fold change of larger than 23 unless the starting value was so low as to be essentially noise. That would be also my concern here; are the FCs that are like 28 changing from say, 0.01 to 1.0? The aggressive normalization would do this kind of distortion. In my experience normally the very large FCs also have relatively low p-values for that exact reason.
While mouse strains should provide the lowest variance out of all appropriate model animals, often the variance between them is not as low as expected. Also, to avoid any strong bias towards both mouse organ and ageing, each organ has been normalised independently (supplementary figure S4, boxplot analysis)
Due to high complexity of the in vivo model we applied a powerful statistical tool using an original approach based on a PolySTest analysis (DOI: https://doi.org/10.1074/mcp.RA119.001777). PolySTest is a robust statistical toolbox to improve coverage and depth of large-scale proteomics analysis based on large numbers of different variables.
Temporal analysis of the overall dataset able to quantify the protein expressions profile changes during ageing was performed based on a rank product test. Adjusted q-values were calculated to correct for multiple testing (-log10 qValue < 0.1 cutoff).
The proteome dataset was obtained from organ samples harvested during adult mice lifespans. We focused on proteins that changed in one direction during the ageing of the organ, or linearly throughout the lifespan of the animal. We applied a novel statistical strategy to detect the most significant chromatin proteome changes associated with ageing. We used rank product tests to identify proteins that exhibited significant abundance changes during ageing (supplementary material, Figure 5, Panel D). We called these “differentially regulated proteins”.
We attached on supplementary material Figure S4 the brain chromatin enrichment undergo PolySTest analysis. We observed how the PolySTest analysis help our data interpretation and provide a robust strategy to identified significant and differentially regulated proteins over time.
12. * Are the tissues normalized together, or each separately? It does not seem like the values are ever directly compared, in which case normalizing separately would probably be better due to the relative paucity of common proteins across all six tissues, but I could see an argument made for either case. In the volcano plot for liver in Supplemental Figure 5D, why is the volcano not centered on FC=0? It is shifted dramatically to the right; this looks like a mistake in the drawing as the significance for minus log2FC starts at like FC=21 (red), but for positive log2FC it is only significant at like FC=24 which doesn't make sense for the U shape.
The proteome dataset was obtained from organ samples harvested during adult mice lifespans. We focused on proteins that changed in one direction during the ageing of the organ, or linearly throughout the lifespan of the animal. A statistical strategy was applied to detect the most significant chromatin proteome changes associated with ageing.
First, to avoid any strong bias towards both mouse organ and ageing, each organ has been normalised separately (supplementary figure S4, boxplot analysis)
Subsequently, we used rank product tests to identify proteins that exhibited significant abundance changes during ageing (supplementary material, Figure 5, Panel D). We called these “differentially regulated proteins”.
Depending on the experiment’s intrinsic data structures and variations, rank product tests indicated where performance was required to improve in a complementary manner (DOI: https://doi.org/10.1074/mcp.RA119.001777). An example of this is where we showed a comparison of protein expression profiles in a mouse brain aged from 3 to 15 months using different statistical tests. By maintaining the same threshold cut-off across all tests, a different “V” Volcano shape was observed. We hypothesized the alteration of the canonical “V” shape due to:
a)Large differences in the protein expression profile from young adult mice (3 months) in comparison to “upper middle-aged” mice (15 months), possibly due to the label-free strategy;
b)The intrinsic biological differences present in the organ; for example, the large and different morphology and cell type compositions.
To avoid any strong bias towards both mouse organ and ageing, the rank product test method was selected. In common with Rainer Breitling and Co-author (DOI: 10.1016/j.febslet.2004.07.055), we observed that the rank product test performed well for our proteome strategy and it identified significant biological feature changes during time. Subsequently, we performed the same statistical analysis for the rest of the organs. We recommend the use of the PolySTest tool for future proteome analysis with a large number of variables. PolySTest tool is a free statistical analysis platform available on:
13. * Near line 381, the authors mention that" the highest unique changes were observed in brain, heart and kidney organs relative to.… "development/ageing"", but not for lung, liver, and spleen (I thought the authors said they removed spleen from future analyses in line 330 so I am surprised to see it in Figure 6, but I digress). What is this annotation of "gene expression" "ageing/development" etc from? Is it the authors or from GO? I have not seen it clustered like that in DAVID, but maybe it is a new feature. In any case, the categorization seems somewhat strange to me. For instance, "cellular respiration" is listed as "ageing / development" – but cellular respiration is involved in basically everything from "cellular metabolic process" to "gene expression" and more. This is even more confusing when "ATP metabolic process" is listed as "cellular metabolic process" but "cellular respiration" is listed as "ageing/development". Why would "ATP metabolic process" be separate from "cellular respiration"? I know these are not exactly the same, but they are so similar I would certainly cluster them together in terms of basic ontological function. Why is "heart contraction" listed as an "ageing/development" function? Some processes in "ageing/development" make sense (e.g. "double-strand break repair") but the four broad ontologies seem largely arbitrary.
We rephrased the sentence, Line 415:
“We reported a high proportion of uniquely annotated categories relative to each organ. For instance, the highest unique changes observed in the brain were relative to “gene expression”, “ageing/development”, and “structure organization and biogenesis”; in the heart and kidney significant changes were observed relative to “structure organization and biogenesis”; the liver showed changes across the “gene expression” and “structure organization and biogenesis” categories; the lung showed the highest unique changes in the “ageing/development” category, and; relative changes in the spleen were detected at the “gene expression” level”
Reviewer #2 pointed out that the four broad ontology groups in Figure 6 are largely arbitrary. In the paragraph “Functional analysis of ageing organ chromatin-enriched proteomes”, we applied an extensive Gene Ontology (GO) analysis to characterize the overall common and unique pathways differentially regulated during ageing across all the organs.
From Figure 5B we first extrapolated the “unique differentially regulated proteins” in each organ. Subsequently, we applied Gene Ontology (GO) analysis across all the “unique differentially regulated proteins” to list the overall common trends and processes being differentially regulated during ageing across all the organs. We sorted the annotated categories by their relative GO term category and separated them by their main family source.
Several pathways were observed to change the chromatin conformation induced by stress responses affected by ageing. We also detected categories associated with modification of the cell structure environment due to organ development. By listing each single category, we described a comprehensive biological profile and listed the main common pathways affected by age-responses present in all organs and related to differentially expressed proteins (Figure 6).
In line with the previous GO analysis (Figure 4B and 4C) we observed the most significant categories changed in chromatin conformation and cell development induced by stress responses affected by ageing. Therefore, we largely sorted four constitutive parts and labeled them as “Gene Expression”, “Ageing / Development”, “Cellular metabolic process”, and “Structure organization and biogenesis” to create a comprehensive cell biology profile during ageing for all organs. Subsequently, the differentially regulated proteins were sorted by their organ source and subjected to further GO analysis in order to distinguish unique organ pathway responses as distinct to those associated with ageing (Figure 6).
Overall, using GO term analysis we dissected the biology landscape aging-related through adult mouse lifespan. By breaking down common and unique regulated categories we aimed to address new biological avenues not yet assigned to be known as ageing-related pathways and improve gene-annotation enrichment analyses related to chromatin cell environment
15. ** The western blots in Figure 7 are a reasonable approach towards such validation, but without the raw data, I don't know what is exactly expected for each of these genes (although the green and red colors at top give me a clue, they are not precise). Is it expected that RNF20 goes down in the aging brain? This gene is not among those shown in Figure 5C. The upregulation of proteasome genes in kidney is surprising; proteasome normally declines with age. Why were these proteins selected? Did you select other proteins that you did not show because they did not validate? Are these measurements in the exact same tissues as the proteomics? That would be an OK validation but essentially just show that the digestion and mass spectrometry worked, rather than independently showing that these proteins change with age – rather like doing a qPCR to validate an mRNA change in the same samples that you ran RNA-seq. This is nice enough to do and appreciated, but it validates the technology and data measurement rather than the actual biology. Also how many gels am I looking at here? I see bactin and RING1b both at the same location in the brain, so where is the loading control for RING1b? I assume this was correctly done as the westerns look well done, but the authors should show the full gel and/or exact loading control. Similar concerns with all of the other tissues, e.g. SIRT3 and SIRT5 are the same molecular weight, so what was the loading control for those 8 lanes?
Following Reviewer #2’s Note (15), we have taken the comments on board to improve and clarify the manuscript. Figure 7 attempted to confirm independently, by a semi-quantitative approach (western blotting), the observations noted in our wider dataset, specifically the unique co-expression network generated. We looked in more detail at the most significant organspecific up-regulated and down-regulated biological processes during mouse ageing.
To avoid cherry-picking in the validation, we divided our strategy into two parts:
Part A: We used a STRING database to investigate the protein/protein interaction (PPI) of chromatin associated proteins belonging to the most significant regulated organ-specific pathways (up-regulated and down-regulated) changes during mouse ageing;
Part B: Integrating protein/protein interaction data with the quantitative protein expression profile.
By integrating protein/protein interactions and protein expression we derived co-interaction and co-expression networks to improve our understanding of the biological mechanisms involved in ageing. We attempted to confirm independently, by immunoblotting, the observations noted in our wider dataset, specifically the co-expression network (Figure 7, Panel B). We then attempted to validate the most significant organ-specific biological processes that are regulated during mouse ageing, particularly at the chromatin level, as we hypothesized that changes in protein expression affects each organ in a distinct manner.
We validated the protein up-regulation changes in the brain, heart, and kidney and conversely the down-regulation changes in the liver and kidney. As we explained in the manuscript, the spleen was not validated as only a few differentially expressed proteins were observed there, and they did not allow for useful GO and protein network analysis. In Figure 7B the biochemical validation of four proteins module responses to ageing were analyzed. Chromatin lysates (from the brain, heart, liver, and kidney) were immunoblotted with the indicated antibodies. Loading controls were used for the proper interpretation of the western blot results. Β-actin and Histone H3 confirmed that protein loading was the same across the samples during ageing.
To compare the trend change (up-regulated and down-regulated) in all four-protein modules further control validations were applied. We probed proteins which exhibited either smaller expression changes or opposite trends. For example, in the brain module, the RNF20 antibody was probe for antagonists in the up-regulation trend. RIG1B showed only minor protein expression changes. The same loading strategy was used with the rest of the organ modules. A useful tool for the analysis of all four protein modules was the integration of quantitative data: a bar above the blots corresponding to the quantitative protein expression levels determined in our proteomics experiments over time (green for up-regulated and red for down-regulated).
1. Regarding the preparation methods, I think it would be more accurate to label the proteome samples as "nuclear proteome" instead of "chromatin-enriched proteome".
We agree that the discrimination between nuclear and chromatin compartments is relatively small. This is in part due to the lack of protein databank cell localization data for the chromatin compartment. We demonstrate that our protein isolation protocol successfully enriches for chromatin. In Figure 1A, we demonstrated that the polycomb protein SUZ12 was detected in the chromatin enriched fraction but not in the nuclear fraction. In chromatin fractions isolated from mESC cells we demonstrated that many known chromatin-associated protein complexes (such as Polycomb Repressive Complex 2 (PRC2), Nucleosome Remodeling and Deacetylase (NuRD), BRAF-HDAC complex (BHC), and mixed-lineage leukemia (MLL) complex) were detected by mass spectrometry (Figure 1B). In the mouse brain we showed, using this protocol, that the chromatin marker histone H3 was present in the chromatin enriched fractions while the cytosolic marker GAPDH was reduced (Supplementary material, S2). We determined that these fractions, while they contained many nuclear components, are enriched chromatin-associated proteins in comparison to total brain lysate. The dynamic range plot was used to visualize the mESC chromatin fraction which was highly enriched for chromatin-associated proteins as compared to the whole cell lysate (Figure 1, panels B and C).
In conclusion we demonstrated that the chromatin enrichment protocol used for mESCs is applicable to mouse organs as shown in Supplementary Material Figure 2.
Sanders et al. 1978 (Sanders 1978)- original salt extraction protocol
Henikoff et al. 2009 (Henikoff et al. 2009)- updates the Sanders protocol and describes isolating different types of chromatin (and the associated proteins, but the focus is on the chromatin itself).
Teves & Henikoff 2012 (Teves and Henikoff 2012)- details the protocol used in their 2009 paper.
Herrmann et al. 2017 (Herrmann et al. 2017)- details a protocol focused on the isolation of chromatin-associated proteins from total chromatin (as opposed to isolating different types of chromatin).
2. Line 183: 15 months of age is not an old mouse and the time points do not actually probe the lifespan of a mouse. However, the selection of the time points is rational, and also justified in the changes that can be observed. It might be good to point out that the time point of 15 months is in mid-life before mortality begins to diminish a population.
We thank the reviewer for this helpful comment. We agree that 15 months represents the upper middle age rather than old age in mice. As the reviewer mentioned, after this time point mortality begins to diminish in the population and it is challenging to harvest enough mice for analysis after this time point. However, we do observe distinct ageing profiles across organs for the time points examined in our study which suggests that this time point is sufficient to observe age related changes. We revised the text in the section “Quantitative chromatin proteomics of ageing mouse organs” to make this more clear.
3. The nuclear enrichment ranges from 30% to 65% across tissues. It would be important to discuss how this difference might affect comparisons.
We thank the reviewer for sharing this observation. Differences in nuclear and chromatin enrichment across tissues are an important consideration. We added a sentence in the section “Quantitative chromatin proteomics of ageing mouse organs”, line 208 to discuss how the difference in enrichment ranges might affect the organ comparison.
4. In Lines 238 and below: for a better understanding of the experiment […]
We thank the reviewer for their helpful suggestion. We changed the section title to “Quantitative proteomics define biological changes in the ageing process” for a better overall meaning of the paragraph.
5. Table 3 it is not clear
We revised the text to convey the concept: line 253
To define ageing signatures across mouse lifespan in each organ, we performed a comprehensive functional analysis of the proteomics datasets using an UpSetR plot to investigate ageing markers.
To gather more information from the UpSetR plot, we retrieve the dataset to estimate the amount of protein shared or not in all mouse organs over time through to their cell compartments. In line with our results Spleen and Brain showed the highest number of unique chromatin associated protein. We also estimated a large number of proteins found in individual organs are likely to confer organ-specific functions.
6. With the collection of nuclear extracts from the various organs, the study results might be confounded by tissue heterogeneity and temporal changes in organ composition. At baseline, what are potential contributions from the various cell type in the organs? Considering the many cell types contributing, for example, to brain extracts, how should the results be interpreted for neurons, glia, etc.? What is the contribution of blood and blood vessels and immune cells? Finally, their relative contribution is likely to change with age, as for example inflammation increases with time. It would be great if the authors could discuss this in the manuscript.
We agree that our results observed in bulk tissues might not be reflective of the alteration of the chromatin landscape at the singular cell level. We address this in line 696.
We are aware that the nuclear and chromatin proteome analysis from the various organs may not be reflective of the overall cellular heterogeneity compositions of the tissues. Chromatin enrichment was, however, observed among each tissue demonstrating the high reproducibly and reliability of the experimental approach, with a coefficient variation estimated at less than 10% across all the samples.https://doi.org/10.7554/eLife.73524.sa2
- Ole N Jensen
- Ole N Jensen
- Giorgio Oliviero
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
We thank professor Bente Finsen, Institute of Molecular Medicine, University of Southern Denmark, Odense, Denmark, for providing C57Bl6 mice. We also acknowledge Dr. Benjamin P Garfinkel (Harvard TH Chan School of Public Health), for the kind gift of HP1BP3 antibody for the immunoblotting experiment. We also thank Aliaksandra Radsizheuskaya (The Memorial Sloan Kettering Institute, New York) for providing mouse embryonic cell pellets for the initial chromatin-proteome experiment. SK and AR-W were supported by a grant from the Independent Research Fund Denmark—Natural Sciences (to ONJ). GO was supported by a grant to the Center for Epigenetics from the Danish National Research Foundation (DNRF #82). Proteomics and mass spectrometry research at SDU are supported by generous grants to the VILLUM Center for Bioanalytical Sciences (VILLUM Foundation grant no. 7292 to ONJ) and PRO-MS: Danish National Mass Spectrometry Platform for Functional Proteomics (grant no. 5072-00007B to ONJ).
Male C57BL/6J mice were obtained from a study approved by the Danish Animal Ethics Inspectorate (J.nr. 2011/561-1950).
- Matt Kaeberlein, University of Washington, United States
- Martin Sebastian Denzel, Altos Labs, United Kingdom
© 2022, Oliviero et al.
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.
Cylicins are testis-specific proteins, which are exclusively expressed during spermiogenesis. In mice and humans, two Cylicins, the gonosomal X-linked Cylicin 1 (Cylc1/CYLC1) and the autosomal Cylicin 2 (Cylc2/CYLC2) genes, have been identified. Cylicins are cytoskeletal proteins with an overall positive charge due to lysine-rich repeats. While Cylicins have been localized in the acrosomal region of round spermatids, they resemble a major component of the calyx within the perinuclear theca at the posterior part of mature sperm nuclei. However, the role of Cylicins during spermiogenesis has not yet been investigated. Here, we applied CRISPR/Cas9-mediated gene editing in zygotes to establish Cylc1- and Cylc2-deficient mouse lines as a model to study the function of these proteins. Cylc1 deficiency resulted in male subfertility, whereas Cylc2-/-, Cylc1-/yCylc2+/-, and Cylc1-/yCylc2-/- males were infertile. Phenotypical characterization revealed that loss of Cylicins prevents proper calyx assembly during spermiogenesis. This results in decreased epididymal sperm counts, impaired shedding of excess cytoplasm, and severe structural malformations, ultimately resulting in impaired sperm motility. Furthermore, exome sequencing identified an infertile man with a hemizygous variant in CYLC1 and a heterozygous variant in CYLC2, displaying morphological abnormalities of the sperm including the absence of the acrosome. Thus, our study highlights the relevance and importance of Cylicins for spermiogenic remodeling and male fertility in human and mouse, and provides the basis for further studies on unraveling the complex molecular interactions between perinuclear theca proteins required during spermiogenesis.
Previously we showed that 2D template matching (2DTM) can be used to localize macromolecular complexes in images recorded by cryogenic electron microscopy (cryo-EM) with high precision, even in the presence of noise and cellular background (Lucas et al., 2021; Lucas et al., 2022). Here, we show that once localized, these particles may be averaged together to generate high-resolution 3D reconstructions. However, regions included in the template may suffer from template bias, leading to inflated resolution estimates and making the interpretation of high-resolution features unreliable. We evaluate conditions that minimize template bias while retaining the benefits of high-precision localization, and we show that molecular features not present in the template can be reconstructed at high resolution from targets found by 2DTM, extending prior work at low-resolution. Moreover, we present a quantitative metric for template bias to aid the interpretation of 3D reconstructions calculated with particles localized using high-resolution templates and fine angular sampling.