(252b) Defining Functional Subtypes of Senescence at Single-Cell Resolution

Background. Aging is characterized by the accumulation of damage and physiological dysfunctions that drive age-associated diseases^1,2. According to US census data, over the next 30 years, an estimated 1 in 5 persons will be above the age of 65³. Unfortunately, living longer does not always equate to living healthier. As such, there is a dire need to identify key contributors of age associated pathologies that could reduce disease burden. Recently, cellular senescence has emerged as a key mechanistic driver of age-associated diseases, connecting molecular and cellular changes to physiological dysfunctions ^4,5.

Cellular senescence is a state of cell cycle arrest that is characterized by the secretion of pro-inflammatory factors, termed senescence-associated secretory phenotype (SASP)⁶. In humans, senescent cells (SnC) accumulate with age, and can be induced in response to stresses, such as DNA damage. However, it is unclear when senescence is beneficial vs. unfavorable, and how we can accurately identify and classify senescent cells ⁴. Currently, senescent cells are identified based on protein biomarkers that include p16 and beta-galactosidase (BGAL), but there remains no universal biomarker of senescence ⁶. Recent studies have indicated that based on the type of senescent induction, the resulting senescent cells secrete differential profiles of SASP and respond differently to senolytics—a new class of drugs that eliminate senescent cells ⁷. As such, we hypothesize that functional subtypes of senescent cells exist, and is encoded within the morphological phenotypes of single cells.

Methods. To elucidate the notion of functional subtypes of senescent cells, we will develop a high-throughput morphological reference map that connects distinct morphological properties of single cells with protein biomarkers. Using a panel of primary dermal fibroblasts taken from healthy aging individuals, we induced senescence using four standard approaches, e.g. DNA damage, reactive oxygen species, and replication. Following induction, cells were stained for multiple senescence-associated biomarkers (p21, p16, Lamin B1, HMGB1, and BGAL) and imaged using confocal microscopy. Cells/Nuclei were segmented by a combination of computational and deep learning approaches, and cells were computationally clustered on individual morphological parameters. This analysis was performed on ~45,000 single cells.

Results. From these experiments we have developed approaches to: (1) Identify and classify senescent from non-senescent cells based on a combination of morphological and protein-based senescence biomarkers. (2) Classify functional subtypes of senescence based on a morphological reference map that enables the prediction of protein-based biomarker expression. (3) Predict senescence-burden as a function of induction kinetics and aging conditions. (4) Quantify the responses of senescence subtypes to senolytic treatments.

Conclusion. We have developed a framework to classify functional subtypes of senescence with potential applications in precision aging in health and disease.

[1] Phillip, et al., Annual Review BME (2015) [2] Otín, et al., Cell (2013) [3] Ortman, et al., US Census (2014) [4] Di Micco, et al., Nature Rev (2020) [5] Baker, et. al., JCB (2018 (2014) [6] Lozano, et al. Nature Rev (2019). [7] Campisi, et al., Annual Rev Path. (2014)