|
Gene expression signatures of human cell and tissue longevity
npj Aging and Mechanisms of Disease volume 2, Article number: 16014 (2016) Cite this article
11k Accesses
31 Citations
7 Altmetric
Abstract
Different cell types within the body exhibit substantial variation in the average time they live, ranging from days to the lifetime of the organism. The underlying mechanisms governing the diverse lifespan of different cell types are not well understood. To examine gene expression strategies that support the lifespan of different cell types within the human body, we obtained publicly available RNA-seq data sets and interrogated transcriptomes of 21 somatic cell types and tissues with reported cellular turnover, a bona fide estimate of lifespan, ranging from 2 days (monocytes) to a lifetime (neurons). Exceptionally long-lived neurons presented a gene expression profile of reduced protein metabolism, consistent with neuronal survival and similar to expression patterns induced by longevity interventions such as dietary restriction. Across different cell lineages, we identified a gene expression signature of human cell and tissue turnover. In particular, turnover showed a negative correlation with the energetically costly cell cycle and factors supporting genome stability, concomitant risk factors for aging-associated pathologies. In addition, the expression of p53 was negatively correlated with cellular turnover, suggesting that low p53 activity supports the longevity of post-mitotic cells with inherently low risk of developing cancer. Our results demonstrate the utility of comparative approaches in unveiling gene expression differences among cell lineages with diverse cell turnover within the same organism, providing insights into mechanisms that could regulate cell longevity.
신체 내 세포 유형에 따라 평균 수명은 며칠에서 유기체의 일생에 이르기까지 상당한 편차를 보입니다. 다양한 세포 유형의 다양한 수명을 지배하는 근본적인 메커니즘은 잘 알려져 있지 않습니다. 인체 내 다양한 세포 유형의 수명을 지원하는 유전자 발현 전략을 조사하기 위해, 우리는 공개적으로 이용 가능한 RNA-seq 데이터 세트를 확보하고 2일(단핵구)에서 일생(뉴런)에 이르는 수명의 진정한 추정치인 세포 회전율이 보고된 21개 체세포 유형 및 조직의 전사체를 조사했습니다.
예외적으로 수명이 긴 뉴런은
단백질 대사가 감소하는 유전자 발현 프로파일을 보였는데,
이는 뉴런의 생존과 일치하며 식이 제한과 같은 장수 개입에 의해 유도된 발현 패턴과 유사했습니다.
다양한 세포 계통에 걸쳐
인간 세포 및 조직 턴오버의 유전자 발현 특징을 확인했습니다.
특히 턴오버는
에너지 비용이 많이 드는 세포 주기 및 노화 관련 병리의 수반되는 위험 요인인
게놈 안정성을 지원하는 인자와 음의 상관관계를 보였습니다.
또한 p53의 발현은 세포 회전율과 음의 상관관계를 보였는데, 이는 낮은 p53 활성도가 본질적으로 암 발병 위험이 낮은 유사 분열 후 세포의 수명을 지원한다는 것을 시사합니다. 이 연구 결과는 동일한 유기체 내에서 다양한 세포 회전율을 가진 세포 계통 간의 유전자 발현 차이를 밝히는 데 비교 접근법의 유용성을 보여주며, 세포 수명을 조절할 수 있는 메커니즘에 대한 통찰력을 제공합니다.
Similar content being viewed by others
Aging is associated with a systemic length-associated transcriptome imbalance
Article Open access09 December 2022
A single-cell transcriptomic atlas characterizes ageing tissues in the mouse
Article 15 July 2020
Temporal changes in the gene expression heterogeneity during brain development and aging
Article Open access05 March 2020
Introduction
Nature can achieve exceptional organismal longevity, >100 years in the case of humans. However, there is substantial variation in ‘cellular lifespan’, which can be conceptualized as the turnover of individual cell lineages within an individual organism.1 Turnover is defined as a balance between cell proliferation and death that contributes to cell and tissue homeostasis.2 For example, the integrity of the heart and brain is largely maintained by cells with low turnover/long lifespan, while other organs and tissues, such as the outer layers of the skin and blood cells, rely on high cell turnover/short lifespan.3–5 Variation in cellular lifespan is also evident across lineages derived from the same germ layers formed during embryogenesis. For example, the ectoderm gives rise to both long-lived neurons4,6,7 and short-lived epidermal skin cells.8 Similarly, the mesoderm gives rise to long-lived skeletal muscle4 and heart muscle9 and short-lived monocytes,10,11 while the endoderm is the origin of long-lived thyrocytes (cells of the thyroid gland)12 and short-lived urinary bladder cells.13
How such diverse cell lineage lifespans are supported within a single organism is not clear, but it appears that differentiation shapes lineages through epigenetic changes to establish biological strategies that give rise to lifespans that support the best fitness for cells in their respective niche. As fitness is subject to trade-offs, different cell types will adjust their gene regulatory networks according to their lifespan. We are interested in gene expression signatures that support diverse biological strategies to achieve longevity. Prior work on species longevity can help inform strategies for tackling this research question. Species longevity is a product of evolution and is largely shaped by genetic and environmental factors.14 Comparative transcriptome studies of long-lived and short-lived mammals, and analyses that examined the longevity trait across a large group of mammals (tissue-by-tissue surveys, focusing on brain, liver and kidney), have revealed candidate longevity-associated processes.15,16 They provide gene expression signatures of longevity across mammals and may inform on interventions that mimic these changes, thereby potentially extending lifespan. It then follows that, in principle, comparative analyses of different cell types and tissues of a single organism may similarly reveal lifespan-promoting genes and pathways. Such analyses across cell types would be conceptually similar, yet orthogonal, to the analysis across species. Publicly available transcriptome data sets (for example, RNA-seq) generated by consortia, such as the Human Protein Atlas (HPA),17 Encyclopedia of DNA Elements (ENCODE),18 Functional Annotation Of Mammalian genome (FANTOM)19 and the Genotype-Tissue Expression (GTEx) project,20 are now available. They offer an opportunity to understand how gene expression programs are related to cellular turnover, as a proxy for cellular lifespan. Here we examined transcriptomes of 21 somatic cells and tissues to assess the utility of comparative gene expression methods for the identification of longevity-associated gene signatures.
인간의 경우 100세 이상의 놀라운 유기체 수명을 자랑합니다. 그러나 '세포 수명'에는 상당한 차이가 있으며, 이는 개별 유기체 내에서 개별 세포 계통의 턴오버로 개념화할 수 있습니다.1
턴오버는
세포 및 조직 항상성에 기여하는
세포 증식과 사멸 사이의 균형으로 정의됩니다.2
between cell proliferation and death
예를 들어, 심장과 뇌의 완전성은 주로 낮은 회전율/긴 수명을 가진 세포에 의해 유지되는 반면, 피부의 외층과 혈액 세포와 같은 다른 기관과 조직은 높은 세포 회전율/짧은 수명에 의존합니다.3-5 세포 수명의 차이는 배아 발생 중에 형성된 동일한 생식층에서 유래한 계통에서도 분명하게 나타납니다.
예를 들어, 외배엽은 수명이 긴 신경세포4,6,7와 수명이 짧은 표피 피부 세포를 모두 생성합니다.8 마찬가지로 중배엽은 수명이 긴 골격근4 및 심장 근육9과 수명이 짧은 단핵구를 생성하며,10,11 내배엽은 수명이 긴 갑상선세포12와 수명이 짧은 방광 세포의 기원입니다.13
단일 유기체 내에서 이처럼 다양한 세포 계통의 수명이 어떻게 유지되는지는 명확하지 않지만, 분화가 후성유전학적 변화를 통해 계통을 형성하여 각각의 틈새에서 세포에 가장 적합한 수명을 제공하는 생물학적 전략을 수립하는 것으로 보입니다. 체력은 트레이드 오프의 영향을 받기 때문에 세포 유형에 따라 수명에 따라 유전자 조절 네트워크를 조정합니다. 우리는 장수를 달성하기 위한 다양한 생물학적 전략을 지원하는 유전자 발현 시그니처에 관심이 있습니다. 종의 수명에 대한 선행 연구는 이 연구 문제를 해결하기 위한 전략을 수립하는 데 도움이 될 수 있습니다. 종의 수명은 진화의 산물이며 주로 유전적, 환경적 요인에 의해 형성됩니다.14 장수와 단수 포유류의 비교 전사체 연구와 대규모 포유류 그룹(뇌, 간, 신장을 중심으로 한 조직별 조사)에서 장수 특성을 조사한 분석은 후보 장수 관련 과정을 밝혀냈습니다.15,16 이러한 연구는 포유류 전반에서 장수의 유전자 발현 신호를 제공하고 이러한 변화를 모방하여 수명을 연장하는 중재에 대한 정보를 제공할 수 있습니다. 그런 다음 원칙적으로 단일 유기체의 다양한 세포 유형과 조직을 비교 분석하면 수명을 촉진하는 유전자와 경로를 유사하게 밝혀낼 수 있습니다. 세포 유형에 따른 이러한 분석은 종에 따른 분석과 개념적으로 유사하면서도 직교할 수 있습니다. 인간 단백질 아틀라스(HPA),17 DNA 요소 백과사전(ENCODE),18 포유류 게놈의 기능적 주석(FANTOM)19 및 유전자형-조직 발현(GTEx)프로젝트20와 같은 컨소시엄에서 생성한 공개적으로 사용 가능한 전사체 데이터 세트(예: RNA-seq)가 현재 제공되고 있습니다. 이러한 프로젝트는 유전자 발현 프로그램이 세포 수명의 대리인으로서 세포의 턴오버와 어떻게 관련되어 있는지 이해할 수 있는 기회를 제공합니다. 여기에서는 21개의 체세포와 조직의 전사체를 조사하여 장수 관련 유전자 서명을 식별하기 위한 비교 유전자 발현 방법의 유용성을 평가했습니다.
Results
We interrogated publicly available transcriptomes (paired-end RNA-seq reads) of 21 human cell types and tissues, comprising 153 individual samples, with a mean age of 56 years (Table 1; details in Supplementary Table S1). Their turnover rates (an estimate of cell lifespan4) varied from 2 (monocytes) to 32,850 (neurons) days, with all three germ layers giving rise to both short-lived and long-lived cell lineages. Biological replicates showed Pearson's correlation coefficients above 0.90, indicating reproducibility of the gene expression data (Supplementary Table S2; Supplementary Figure S1). Gene expression patterns were first analyzed by Principal Component Analysis (PCA) (Figure 1a), which revealed that the first three Principal Components (PCs) accounted for ~60% of gene expression variation. The cells and tissues formed several clusters, suggesting related biological functions for these clusters. For example, the gastrointestinal tissues, esophagus, rectum and colon all grouped together, and hematopoietic tissues (bone marrow and spleen) and monocytes also clustered. Because transcriptomes of functionally related cell types often exhibit substantial hierarchical structure,21,22 a neighbor-joining gene expression tree was generated based on mean gene expression levels (Figure 1b). Similar to the PCA results, bone marrow and spleen clustered with monocytes, while skeletal muscle and heart muscle grouped together and were distinct from smooth muscle. Although the PCA and gene expression tree correlated with the broad functional features of the cells and tissues, the clustering did not strictly follow germ layer origin. In agreement, recent data demonstrate that the regulatory DNA landscape (epigenome), but not gene expression (transcriptome), correlate with cell lineage relationships, including germ layer origin.23–25 Thus, for any given cell type, e.g., a neuron, epigenetic marks reflect both the prior (e.g., state in the germ layer and derived cell lineages) and present regulatory landscapes.24
Table 1 Summary of human cells and tissues used in the study
Figure 1
Clustering of gene expression from diverse human cells and tissues. (a) Representations of gene expression in Principal Component (PC) space. Values in parenthesis indicate the percentage of variance explained by each of the PCs. Ostensible germ layer origin is indicated (red: mesoderm; blue: ectoderm, green: endoderm). (b) Gene expression tree. Branches are colored according to germ layer origin (as in a). Estimated cell and tissue turnover (‘cellular lifespan’; in days) is shown next to cell and tissue names. The tree was generated by neighbor-joining (NJ) method (bootstrap=1,000,000). Bootstrap values are indicated by circles: white ⩾0.9; yellow ⩽0.9.
Differential gene expression of cell and tissue groups
We selected several lineage groups and individual cells and tissues and compared each of them against all other cells and tissues in order to identify associated specific expression patterns (Supplementary Tables S3 and S4; Supplementary Results and Discussion). The results for heart (muscle), thyroid gland, hematopoietic tissues and monocytes are presented in Supplementary Figure S2. In heart and skeletal muscle, 455 out of 12,044 genes were differentially expressed (phylogenetic analysis of variance (ANOVA) P value ⩽0.01) compared with other cells and tissues (Figure 2a). Approximately 44% of these genes were associated with the tricarboxylic acid (TCA) cycle and respiration, in agreement with the metabolic organization and energy sources of these tissues.26
Figure 2
Genes differentially expressed between selected cell lineages and tissues. (a) Heart and skeletal muscle. Left panel, heat map of genes differentially expressed in heart and skeletal muscle relative to the other cells and tissues (standardized expression level; red: high expression; blue: low expression). Right panel, significantly enriched (FDR P⩽0.05) pathways. (b) Neuron. Left panel, heat map of genes differentially expressed in neurons, colored and annotated as in a. Right panel, significantly enriched (FDR P⩽0.05) pathways ‘ribosome’ and ‘proteasome’. FDR P denotes false discovery rate-adjusted P value. Lower panel, box plot showing p53 (TP53) expression. Error bars indicate s.e.m.
Neurons, which are critical for cognitive and motor functions, have cell lifespans that likely exceed the lifespan of the organism.7,27 Comparing neurons to shorter-lived cells and tissues is conceptually similar to comparing gene expression of long-lived mammals to related short-lived species, e.g., examining African mole rats against other rodents.15 Accordingly, neurons should possess a gene expression signature associated with low turnover/long lifespan, in addition to the patterns indicative of neuronal function. Out of 12,044 genes 1,438 were differentially expressed in neurons (P⩽0.01; Figure 2b; Supplementary Table S3) and gene set enrichment analysis showed enrichment for functions associated with lysosomes, proteasomes, ribosomal proteins and apoptosis (Supplementary Table S3). Neurons presented with reduced expression of 27 ribosomal proteins and multiple 20S proteasome subunit genes (Figure 2b), consistent with distinct protein metabolism required to fine-tune self-renewal and synaptic plasticity.28 This group of genes was not correlated with cell and tissue turnover (see section below), suggesting that this expression pattern is unique to long-lived neurons. Reduced protein metabolism, which may be induced by dietary restriction and other interventions, is known to associate with extended lifespan in a number of model organisms.29,30 Furthermore, expression of the tumor suppressor p53 (TP53) was significantly reduced (P⩽0.001) in neurons, where it was expressed at a level 5–30 times lower than that in the other cells and tissues (Figure 2b). Reduced p53 expression is associated with a concomitant reduction in cell cycle-related proteins in neurons following their terminal differentiation from neuroblasts.31
Gene expression patterns of cell and tissue turnover
We identified genes whose expression correlated with cell and tissue turnover. Available turnover times for a number of tissues and cell types (in days)3 were supplemented with estimates from the literature and used as a bona fide measure of lifespan (‘lifespan trait’). We applied generalized least squares regression,32 tested different evolutionary models and selected the best fit model by maximum likelihood (see Extended Experimental Procedures). Two hundred eight out of 12,044 genes showed significant correlation with turnover at a false discovery rate (Q-value) of 0.05, with 75% (155 genes) in negative correlation and 25% (53 genes) in positive correlation (Supplementary Table S5). Notable genes with a positive correlation included the complex SNRPN-SNURF locus, which gives rise to a number of proteins and short non-coding RNAs (Supplementary Figure S3; Supplementary Results and Discussion). We visualized the protein–protein interaction network represented by these 208 genes, revealing significant enrichment (FDR P⩽0.05) for genes involved in cell cycle, immune signaling (NF-κB) and p53 signaling (Figure 3a,b and Supplementary Tables S6 and S7). In our data set, hematopoietic tissues (bone marrow and spleen) and monocytes constituted the samples with the shortest turnover. Removal of these data points in the regression analysis retained the ‘turnover signature’, with the overlapping gene set comprising critical cell cycle and apoptosis associated genes, such as CHEK1, CHEK2, MKI67, FOXM1, TP53 and BCL10, while a correlation with immune signaling-associated genes was lost (Table 2 and Supplementary Table S8).
Figure 3
Overview of genes correlating with cellular turnover. (a) Heat map analysis. Upper panel, cellular turnover of the cells and tissues (in ascending order). Lower panel, heat map displaying gene expression of genes correlating with turnover (standardized expression levels; red: high expression; blue: low expression). (b) Network analysis. Protein–protein interaction network of genes correlating with turnover. The interaction network was created by interrogating the STRING database (evidence view). Lines (strings) indicate protein interactions. Proteins without interacting partners were omitted. Selected pathways are indicated by colored rings. (c) Expression of the cell proliferation marker Ki-67 (MKI67) correlates negatively with turnover. Error bars indicate standard error of the mean. FDR P denotes false discovery rate P value. (d) Genes associated with RAD51-mediated genome stability control in the replication phase of the cell cycle. Names highlighted in blue showed negative correlation with turnover. CDK: cyclin-dependent kinase; CCNA2: cyclin A2; CCNB2: cyclin B2; CHEK1: checkpoint kinase 1, CHK1; CHEK2: checkpoint kinase 2, CHK2; CLSPN: claspin; BRCA2: breast cancer 2, early onset; RAD51: RAD51 recombinase; RAD51AP1: RAD51 associated protein 1; MMS22L: MMS22-like, DNA repair protein; TONSL: tonsoku-like, DNA repair protein.
Table 2 Overlap of genes correlating with turnover before and after removal of immune system-associated cells and tissues (monocytes, bone marrow and spleen)
Negative correlation between cell cycle and associated genome integrity pathways and cell and tissue turnover
Gene ontology, KEGG pathway analysis and manual interrogation of genes correlating with turnover revealed that 30–40% of these (Supplementary Tables S5 and S8) have roles in the cell cycle, a highly complex multi-step process. They spanned all phases of the cell cycle, with the majority of the genes showing decreased expression associated with increased turnover. As expected, the classic DNA replication marker, Ki-67 (MKI67), which is only expressed during the cell cycle,33 was present at high levels in rapidly proliferating tissues such as bone marrow, rectum and colon. It was expressed at very low levels in monocytes (the majority of which do not proliferate) and in post-mitotic skeletal muscle and neurons (most of which are incapable of cell division; Figure 3c). Ki-67 was expressed at a moderate level in heart muscle, which can proliferate at a limited rate,9 and where Ki-67 may promote postnatal cardiac remodeling.34
A negative correlation between the cell cycle machinery and cell and tissue turnover is not surprising, as proliferative homeostasis, a balance between cell growth and death, is essential for normal turnover (as occurs in epidermal cells, for example). In contrast, terminally differentiated cells comprising heart muscle, skeletal muscle and neurons, have permanently or largely exited the cell cycle. They are, therefore, expected to express low levels of genes related to cell cycle checkpoints and the maintenance of replication fidelity. We noted that a number of genes essential for genome stability during the premitotic phase (G2), a process tightly linked to tumor development,35 were negatively correlated with turnover (Figure 3d). This included regulators of cyclin-dependent kinases, such as cyclin A (CCNA2) and B (CCN2B), checkpoint kinase 1 (CHEK1) and its regulator claspin (CLSPN), and CHEK2, which together with breast cancer 2 early onset (BRCA2), RAD51 recombinase (RAD51), its enhancer RAD51AP1 (RAD51AP1), and MMS22-like DNA repair protein (MMS22L) promote genome stability.35 These data support recent work showing that RAD51 and BRCA2 are major facilitators of genome integrity in proliferating cells.35–37 Genes in the p53 pathway can halt progression of the cell cycle (induce senescence), or inhibit apoptosis in proliferating cells, and blocking apoptosis is crucial for the survival of differentiated post-mitotic cells.38 In agreement with a reduction in the involvement of cell cycle in longer-lived cells and tissues, multiple p53 pathway associated genes, including p53 itself (TP53), were negatively correlated with turnover (Supplementary Figure S4; Supplementary Tables S5 and S8). They encompassed several checkpoint kinases (CHEK1 and CHEK2), BCL2-associated X protein (BAX), which activates apoptosis, and ASC/TMS1 (PYCARD), which when downregulated inhibits BAX translocation to mitochondria.
To further investigate the connection between gene expression and turnover, we turned to the library of integrated network-based cellular signatures (LINCS) to identify perturbations (gene overexpression and knockdown) that produce a similar overall gene expression profile to our regression analysis. We found that the most similar profiles came from genes associated with cell cycle control and associated DNA repair (Supplementary Table S11). This included overexpression of the major cell cycle inhibitors p18 (also known as INK4C; CDKN2C), p21 (CDKN1A) and p27 (CDKN1B), in agreement with the negative correlation between downstream genes and turnover in our data set.
No correlation between oxidative phosphorylation gene expression and cell and tissue turnover
It is clear that mitochondrial dysfunction is a hallmark of aging. Properly functioning mitochondria are essential for energy production and cell survival, and hence, are crucial for longevity and resistance to age-related disease.39 We calculated pair-wise correlations of the expression of 97 nuclear-encoded genes associated with oxidative phosphorylation (OXPHOS)40 across the data set, revealing a strong correlation across all samples with the exception of heart muscle and skeletal muscle (Figure 4a). As shown in Figure 4b, expression of individual OXPHOS genes varied by tissue, in agreement with a previous microarray study,40 but there was no overall correlation with cell and tissue turnover.
Figure 4
Expression of nuclear-encoded oxidative phosphorylation (OXPHOS) genes in 21 human somatic cells and tissues. (a) Pairwise correlation matrix of 97 OXPHOS genes. Darker blue shading indicating stronger positive correlation. Blue corresponds to a correlation of ~1 and white to ~0. (b) Upper panel, cellular turnover of the cells and tissues (in ascending order). Lower panel, heat map displaying gene expression of individual OXPHOS genes. Scaled log2 transformed normalized counts (Z-score) are plotted in blue–red color, with red indicating high expression and blue indicating low expression.
Discussion
Several interrelated evolutionary and mechanistic theories have been proposed that provide insights into the evolution of lifespan and suggest the involvement of a large armamentarium of genes.41,42 In this study, we employed gene expression data from 21 somatic cell types and tissues and sought to identify genes and pathways associated with cell and tissue turnover, to our knowledge the current best estimate of cellular lifespan,1 in one of the longest-lived mammals, human. We first tested for differential expression in selected groups of cells and tissues, revealing expression patterns that fit well with expected biological functions, including the TCA cycle and respiration in cardiac and skeletal muscle; immune function genes in bone marrow, monocyte and spleen, and reduced protein metabolism in neurons.
The major insights of our study centered on the relationship between gene expression and cell and tissue turnover. Multiple genes taking part in the energetically expensive cell cycle and associated repair (genome stability) were negatively correlated with turnover. By interrogating the Broad Institute’s Library of Integrated Cellular Signatures resource, we found that very similar gene expression patterns can be achieved by targeted overexpression and knockdown of single genes, suggesting that gene expression patterns associated with cell and tissue turnover, and by proxy cell lifespan, may in principle be achieved by genetic, pharmacological and perhaps dietary interventions. The observation that turnover negatively correlates with cell cycle genes may seem tautological. Indeed, one would expect tissue or cell populations with high turnover to exhibit more cells in the cell cycle and, consequently, high expression of cell cycle and associated genes. However, it is recognized that the rate of aging and longevity is indeed fine tuned by the balance of cell division and death (i.e., cell turnover).2 Interestingly, a recent study examined the transcriptional response to long-term calorie restriction in humans, revealing that calorie restriction shifts cellular metabolism of skeletal muscle from proliferation to maintenance and repair.43 The concomitant reduction in cell cycle gene expression presumably results in a ‘younger’ transcriptional signature that contributes to the lifespan-extending properties of calorie restriction.
Expression of p53 (TP53), often referred to as the ‘guardian of the genome’, was negatively correlated with cell and tissue turnover; with particularly low levels of expression in long-lived neurons. Evidence is emerging that p53 has an evolutionarily ancient lifespan-regulating function, in addition to its role as tumor suppressor. It is appreciated that p53 promotes organismal longevity by preventing survival of abnormal cells; however, several investigators have speculated that p53 protects against cancer in proliferating cells at the cost of accelerated aging.3,44,45 Thus, in non-dividing cells p53 may, in effect, reduce lifespan. It was found that suppression of TP53 orthologs in animal models such as the mouse and fruit fly can extend organismal lifespan (reviewed in ref. 46). Dominant-negative Drosophila p53 (Dmp53) significantly extends organismal lifespan when expressed in adult neurons, but not other tissues (such as muscle) via insulin/insulin-like signaling (IIS) and TOR (target of rapamycin) pathways.47,48 Several p53 (TP53) retrogenes have been recently reported in the elephant genome,49 however, while elephant lymphocytes and fibroblasts show an increased response to DNA damage compared with human cells,49 it is currently not known how many of these retrogenes are actively translated and exhibit p53 function, nor whether the expression of p53 is appropriately activated and/or elevated in all elephant cells. Thus, the elephant’s resistance to cancer may stem from other mechanisms and it would be of interest to examine the expression of TP53 in elephant neurons and other cell types. Interestingly, the cancer-resistant long-lived blind mole rat, Spalax, has evolved an enhanced necrotic and reduced apoptotic defense (via a dominant negative form of p53) against cancer, possibly to adapt to an oxygen-poor underground environment, which would normally result in extensive p53-mediated cell death.50,51 Whether Spalax p53 contributes to the longevity of this cancer-resistant rodent is not known, but is an exciting possibility under the hypothesis that reduced p53 activity exerts beneficial effects on cellular lifespan if tumor formation is avoided. Taken together, we speculate that very low levels of endogenous p53 contribute to the exceptional lifespan of cells and tissues with low turnover, such as neuron, heart muscle and skeletal muscle, and perhaps also organismal longevity.
It is now appreciated that expression of OXPHOS genes decreases with age in diverse cells and tissues in species ranging from nematode to human.39 However, the link between mitochondrial homeostasis and lifespan is currently enigmatic.39 Interestingly, lifespans of different strains of the single-celled yeast is associated with upregulation of OXPHOS genes.52 In our data set of 21 human cells and tissues, expression of individual OXPHOS genes, and more generally metabolic genes, did not correlate with cell turnover. Thus, we propose that while cells and tissues share a gene expression signature manifested as reduced mitochondrial function with age, the overall integrity of mitochondria in long-lived human cell types is achieved by distinct gene expression strategies.
Our study has several limitations. Future studies should attempt to more accurately determine the turnover of cells and cell populations (tissues) in the body, and sequence the transcriptomes of additional long-lived post-mitotic cells, such as osteocytes.53 Single-cell RNA sequencing is rapidly evolving54 and would greatly advance the study of cell turnover, especially as it would avoid the analysis of organs that are composed of heterogeneous cell types. Multiple cells and tissues from the same individual are also becoming available.55 Finally, the contribution of epigenetics56,57 and long-lived proteins58,59 to cell and tissue turnover is largely unknown but likely important. Owing to a lack of a matching set of samples and limited number of biological replicates, we did not interrogate other species, such as the mouse, for turnover-associated genes. It is currently unknown whether individual genes identified through our procedure would overlap among species,60 but is an important question for future studies. Although some common features may be observed, human cell types may also harbor signatures quite distinct from other animals, including other primates, since humans are one of the most exceptionally long-lived species.61
Overall, our analysis, employing cellular turnover, as a proxy of lifespan, is a first step to a molecular understanding of cell and tissue longevity. We reveal a gene signature of exceptionally long-lived post-mitotic neurons, and genes and pathways that correlate with turnover across 21 somatic cells and tissues. The data suggest that human cell lineages utilize both common and lineage-specific strategies to alter their lifespan. This new perspective should provide further impetus to the study of the lifespan trait (longevity) and the aging process.
Materials and methods
See Supplementary Information for detailed methods.
Biological samples
Our analysis was restricted to 21 adult somatic cells and tissues with more than three biological replicates and bona fide lifespan estimates (cell turnover in days) derived from a recent comprehensive review3 and the additional data collected through primary literature searches (see Supplementary Experimental Procedures).
Transcriptome data relationship inference
The relationship of publicly available transcriptomes (RNA-seq data) from the 21 cells and tissues was investigated by principal component analysis and a gene expression tree (see Supplementary Experimental Procedures).
Identification of genes differentially expressed between cell and tissue groups
Differentially expressed genes in a particular group (e.g., heart and skeletal muscle) were identified by ANOVA, taking into account the hierarchical relationship between samples in our data set (tissue autocorrelation by ‘phylogenetic ANOVA’). See Supplementary Experimental Procedures for further details.
Identification of genes correlating with cell and tissue turnover
To identify genes correlating with cell turnover (a bona fide lifespan estimate) we employed generalized least squares regression (Supplementary Experimental Procedures).
References
Spalding, K. L. et al. Dynamics of fat cell turnover in humans. Nature 453, 783–787 (2008).
Pellettieri, J. & Sanchez Alvarado, A. Cell turnover and adult tissue homeostasis: from humans to planarians. Annu. Rev. Genet. 41, 83–105 (2007).
Richardson, R. B., Allan, D. S. & Le, Y. Greater organ involution in highly proliferative tissues associated with the early onset and acceleration of ageing in humans. Exp. Gerontol. 55, 80–91 (2014).
Spalding, K. L., Bhardwaj, R. D., Buchholz, B. A., Druid, H. & Frisen, J. Retrospective birth dating of cells in humans. Cell 122, 133–143 (2005).
Bergmann, O. et al. Dynamics of cell generation and turnover in the human heart. Cell 161, 1566–1575 (2015).
Huttner, H. B. et al. The age and genomic integrity of neurons after cortical stroke in humans. Nat. Neurosci. 17, 801–803 (2014).
Magrassi, L., Leto, K. & Rossi, F. Lifespan of neurons is uncoupled from organismal lifespan. Proc. Natl Acad. Sci. USA 110, 4374–4379 (2013).
Halprin, K. M. Epidermal ‘turnover time’--a re-examination. Br. J. Dermatol. 86, 14–19 (1972).
Bergmann, O. et al. Evidence for cardiomyocyte renewal in humans. Science 324, 98–102 (2009).
Fahy, R. J., Doseff, A. I. & Wewers, M. D. Spontaneous human monocyte apoptosis utilizes a caspase-3-dependent pathway that is blocked by endotoxin and is independent of caspase-1. J. Immunol. 163, 1755–1762 (1999).
Mangan, D. F. & Wahl, S. M. Differential regulation of human monocyte programmed cell death (apoptosis) by chemotactic factors and pro-inflammatory cytokines. J. Immunol. 147, 3408–3412 (1991).
Coclet, J., Foureau, F., Ketelbant, P., Galand, P. & Dumont, J. E. Cell population kinetics in dog and human adult thyroid. Clin. Endocrinol. (Oxf) 31, 655–665 (1989).
Leblond, C. P. & Walker, B. E. Renewal of cell populations. Physiol. Rev. 36, 255–276 (1956).