Senescent cells and the incidence of age-related diseases

Abstract

Age-related diseases such as cancer, cardiovascular disease, kidney failure, and osteoarthritis have universal features: Their incidence rises exponentially with age with a slope of 6-8% per year and decreases at very old ages. There is no conceptual model which explains these features in so many diverse diseases in terms of a single shared biological factor. Here, we develop such a model, and test it using a nationwide medical record dataset on the incidence of nearly 1000 diseases over 50 million life-years, which we provide as a resource. The model explains incidence using the accumulation of senescent cells, damaged cells that cause inflammation and reduce regeneration, whose level rise stochastically with age. The exponential rise and late drop in incidence are captured by two parameters for each disease: the susceptible fraction of the population and the threshold concentration of senescent cells that causes disease onset. We propose a physiological mechanism for the threshold concentration for several disease classes, including an etiology for diseases of unknown origin such as idiopathic pulmonary fibrosis and osteoarthritis. The model can be used to design optimal treatments that remove senescent cells, suggeting that treatment starting at old age can sharply reduce the incidence of all age-related diseases, and thus increase the healthspan.

Keywords: age-related disease; aging; cancer; cellular senescence; diabetes; electronic medical records; fibrosis; idiopathic pulmonary fibrosis; incidence rate; mathematical model; osteoarthritis.

FIGURE 1

Diseases caused by threshold‐crossing of a parameter affected by senescent cells are predicted to have an exponential incidence curve with a decline at old ages. (a) Incidence curves for several age‐related diseases, from (Public Health Agency of Canada, 2011; National Cancer Institute et al., 2018; Navaratnam et al., 2011; Oliveria et al., 1995). (b) We assume that disease onset occurs when a physiological parameter ϕ exceeds a threshold, ϕ _c. (c) ϕ is a rising function of senescent‐cell level, $X$, so that ϕ _c is crossed when $X$ exceed a disease threshold $X_c$. (d) The senescent‐cell levels of three susceptible individuals simulated by the SR model. The disease arises as a first‐passage‐time process when $X$ crosses $X_c$. (e) In the three‐parameter model, the threshold $X_c$ for each person in the susceptible fraction of the population is drawn from a Gaussian distribution with mean $\overline{X_c}$ and standard deviation $\sigma$. (f) Effect of the model parameters on the incidence curve. The parameters are $X_c=14,s=0.05$ (black), $X_c=14,s=0.15$ (dashed red), $X_c=16,s=0.05$ (dashed black), $\overline{\mathit{Xc}}=14,s=0.05,\sigma =3$ (dashed green)

FIGURE 2

The model describes the incidence curves of a wide range of age‐related diseases. (a) The two‐parameter (2p) and three‐parameter (3p) models fit the incidence curves of many age‐related diseases. Data from Clalit ICD9 codes for females (similar results for males in Figure S2). (b) Examples where the three‐parameter model provides an excellent fit, but not the two‐parameter model. (c) The model does not describe well the incidence of osteoporosis in women (left panel). It cannot capture the incidence curve of Alzheimer disease and dementia using the maximal value of $X_c=X_{\mathit{death}}=17$; the fit is improved with $X_c\approx 20$ for dementia and $X_c\approx 23$ for Alzheimer's disease (black line). (d) Coefficient of determination R ² for fits of the two‐parameter (2p) and three‐parameter (3p) models to incidence of ICD9 codes as a function of mean slope of incidence between ages 30 and 80. (e) Percent of ICD9 codes with R ² > 0.9 as a function of slope. Inset: number of ICD9 codes as a function of slope. Error bars are 95% CI

FIGURE 3

Threshold‐crossing of the ratio of progenitor removal to proliferation can explain the incidence of idiopathic pulmonary fibrosis and osteoarthritis. (a) General scheme of “frontline” tissues, in which stem or progenitor cells, S, are as exposed to damage as their differentiated progeny, D. (b) The lung alveoli progenitor AT2 cells lie within the same layer as the differentiated AT1 cells. (c) In joints, cartilage‐derived stem/progenitor cells (CSPC) are at the superficial zone and face the same amount of damage as the differentiated chondrocytes (CH). (d) Homeostasis is maintained by signals secreted from the cells that act on the proliferation and differentiation rates. (e) When the physiological parameter ϕ = r ₁/p, the ratio of progenitor removal and proliferation rates, exceeds ϕ _c = 1, the number of cells in the tissue, S + D, crashes. (f) Senescent cells slow progenitor proliferation due to SASP from both local and systemic senescent cells (SnC). Senescent cells can also disrupt the extracellular matrix and increase removal rate r ₁. (g) When senescent cells cross a threshold $X_c$, tissue collapse is predicted to occur. (h) Simulated tissue dynamics show that when senescent cells cross a threshold, the number of differentiated cells collapse, triggering the onset of the disease. (i) The model fits the incidence curves of IPF (Navaratnam et al., 2011) and OA (knee and hip) well. (j) Incidence of knee OA stratified by BMI, see Figure S4 for hip OA. (k) Effect of BMI on best‐fit parameters for knee OA incidence, with s in percent. OA data from (Reyes et al., 2016)

FIGURE 4

Cancer incidence can be explained by threshold‐crossing of the ratio of cancer growth rate to removal rate. (a) Cancer cells C proliferate at rate p, and are removed at rate r. With age, rising senescent‐cell (SnC) levels cause immune saturation by taking up some of the removal capacity of NK cells and macrophages (Karin et al., 2019). Inflammation driven by senescent cells increases proliferation p for some cancer types. (b) Both effects, raising p and lowering removal r, cause the parameter ϕ to increase, ϕ(X) = p(X)/r(X). Thus, there exists a threshold $X_c$ where ϕ exceeds the critical value of 1 and cancer cells proliferate more than they are removed, reaching a clinically detectable disease. (c) The models fit various types of cancer very well. (d) Example of cancer types in which the three‐parameter model provides an excellent fit but the two‐parameter model does not. (e) Example of cancer types not described by the models. In the case of Hodgkin Lymphoma, the model describes well the incidence curve above age 50 (black line)

FIGURE 5

Infrequent treatment that removes senescent cells starting at old age can reduce disease incidence in the model. (a) Treatment with senolytics shifts the incidence rate in the model by 25 years. We assumed a conservative case in which only 25% of the senescent cells are drug‐sensitive (Karin et al., 2019). In this example, the treatment is given every 30 days, and starts at the age of 60 years. We used typical disease parameters ($X_c=14,s=0.1)$ to calculate the incidence curves. (b) Shift of the incidence curve to younger ages (years) as function of the time interval between treatments and the effectiveness of the treatment defined as the percentage of drug‐sensitive senescent cells that it removes. (c) The incidence curves for different choices of the age in which the senolytic treatment starts. (d) The shift of the incidence curve is larger the later the treatment starts. Panels c and d use the same treatment and disease parameters as panel a.