CD36 is induced in multiple senescence contexts
In previous work, we showed that targeted chemical inhibition of the epidermal growth factor receptor (EGFR) in primary human bronchial epithelial (HBE) cells is sufficient to trigger a comprehensive senescent phenotype within 3 days
9. Taking advantage of the efficiency of this method, we developed an unbiased gene expression profiling approach to compare senescent HBE cells with their proliferating counterparts in order to identify novel signaling molecules that function to regulate senescence initiation and the SASP. HBE cells, treated with either erlotinib or DMSO, were incubated with the fluorescent substrate C12FDG for the senescence‐associated beta‐galactosidase (SA‐βGal), and senescent cells were purified using flow cytometry as previously described
19. The biological triplicated samples of senescent and proliferating cells were then screened by transcriptional profiling to identify genes differentially expressed during the early phase of senescence establishment (Fig
1A). This method revealed 331 genes with at least 1.5‐fold transcriptional induction or suppression at 18 h after erlotinib‐mediated senescence initiation (in press). Of these, 10 candidates were chosen for further investigation based on their plasma membrane localization (GeneCards confidence > 4), and time‐dependent mRNA upregulation was validated by qPCR (Fig
EV1A). As an indication that our method can detect bona fide senescence regulators, this subset of membrane proteins included the interleukin‐1 receptor (IL‐1R) and Notch3, both of which have been previously shown to modulate senescence but not SASP initiation per se
20. However, among the candidates, the scavenger receptor CD36 stood out as the one with the strongest and most rapid induction (Fig
EV1A). In fact, CD36 mRNA expression was upregulated approximately ten‐ and thirty‐fold at the 6‐ and 48‐h time points, respectively (Fig
1B). Normally at 6 h, most features of the senescent phenotype are not yet evident and signal transduction is still in its initiation phase. Thus, we hypothesized that CD36 might be functional in the senescence programming process.
To determine whether CD36 is also induced in other senescence models, we analyzed CD36 expression in replicative and oncogene‐induced senescent fibroblasts using the publicly available Gene Expression Omnibus (GEO) database. In agreement with our chemical‐induced senescence results, CD36 was found to also be significantly upregulated in both of these microarray datasets (Fig
EV1B).
To further validate our microarray results, we replicated and confirmed the models of erlotinib‐induced, oncogene‐induced, and replicative senescence using canonical senescence markers (Fig
EV1C–H). We then measured CD36 expression in those senescent contexts. Consistently, a strong induction of CD36 mRNA and protein was observed in senescent cells induced by all three stimuli (Fig
1C–E). It is known that senescent cells accumulate within aged tissues
22, and this phenomenon is conserved across different species and proposed to be functionally responsible for the development of major aging‐related phenotypes
23. To assess whether CD36 expression is correlated with senescent cell accumulation in aging organs, we measured CD36 expression in lung, liver, and muscle tissue of both young (1 month) and old (30 months) mice. Importantly, all tissue types tested contained elevated CD36 expression. Moreover, in lung tissue, which is the origin of HBE cells, CD36 had a ~100‐fold induction in aged mice, indicating a strong correlation between CD36 induction and bronchial cell senescence and aging (Fig
EV1I). Taken together, these results demonstrate that CD36 expression is consistently upregulated in a wide range of cell and tissue types by various senescent stimuli including replicative stress, oncogene activation, EGFR inhibition, and the natural aging process.
Short‐term CD36 expression initiates production of secretory molecules
Based on its strong induction during the early stages of senescence initiation, we hypothesized that CD36 might have a functional impact on senescence programming. To investigate its possible sufficiency in inducing cellular senescence, we ectopically expressed CD36 in HBE cells using a Tet‐on inducible expression system. Since there are no published reports describing a role for CD36 in cellular senescence, we first examined established features of the senescent phenotype including cell proliferation, CDK inhibitor expression (p16 and p21), SA‐βGal activity, and the SASP. First, we performed SA‐βGal staining (Fig
2A), 5‐ethynyl‐2A‐deoxyuridine (EdU) incorporation (Fig
2B), and p16 and p21 Western blot assays (Fig
2C) using HBE cells engineered to overexpress CD36 for 7 days or control cells. However, these proliferation‐related assays failed to reveal significant differences between CD36‐expressing and control HBE cells at the 7‐day time point, indicating that forced CD36 expression is insufficient to drive cell cycle arrest in the short term.
Since CD36 is known to activate an inflammatory phenotype in certain immune system cell types
15, we next considered the possibility that it might have a similar function during senescence reprogramming. SASP has been shown to be mediated through the activity of the NF‐κB transcription factor complex, and a conventional indicator of NF‐κB activation is the phosphorylation status of its functional subunit p65; we therefore measured p65 phosphorylation in CD36‐overexpressing and control cells by Western blotting
25. Interestingly, these assays revealed increased phosphorylation of p65 along with activation of its upstream tyrosine kinase c‐Src and MAP kinase p38 (Fig
2D). Luciferase reporter assays further verified the activation of NF‐κB signaling in CD36‐expressing HBE cells (Fig
2E). Since the conventional role of NF‐κB is regulating cytokine production and secretion, the finding of CD36 driving NF‐κB activation suggests that CD36 might promote the SASP through stimulating the Src–p38–NF‐κB axis. To comprehensively explore a relationship between CD36 and the SASP, we performed quantitative PCR (qPCR)‐based profiling analysis of 78 molecules previously reported as components of the SASP
26. Strikingly, across HBE cells from three independent human donors, 22 of these molecules showed a statistically significant (> 1.5‐fold) upregulation upon ectopic CD36 expression (Fig
2F). Among these, many well‐recognized conventional SASP components were found to be consistently produced upon CD36 expression, including interleukin 6 (IL‐6) and interleukin 8 (IL‐8)
11. These results demonstrate that short‐term expression of CD36 in HBE cells, while unable to trigger a cell cycle exit, can promote NF‐κB signaling and production of a large set of SASP components.
Long‐term CD36 expression promotes a comprehensive senescent phenotype
Some senescent stimuli, such as replicative exhaustion, require a period of weeks to yield cell cycle arrest
29. To further assess a possible effect of CD36 on the cell cycle, HBE cells expressing CD36 were maintained for a period of 14 days. Interestingly, prolonged CD36 expression led to a striking phenotype of cell cycle exit that was associated with increased levels of cyclin‐dependent kinase inhibitors and SA‐βGal activity (Fig
3A–C). Further signaling pathway analysis revealed a sustained activity of the Src–p38–NF‐κB axis, indicating that long‐term activation of CD36‐dependent pro‐inflammatory signaling might be responsible for proliferative arrest (Fig
3D–E). Overall, these results suggest that long‐term CD36‐dependent SASP signaling activation can trigger a comprehensive senescence phenotype in primary human epithelial cells.
Based on the fact that CD36 is broadly induced during epithelial cell and fibroblast senescence and is sufficient to promote cellular senescence in HBE cells, we next asked whether the effects of CD36 on NF‐κB activation and cell cycle arrest are conserved across cell types. To investigate this, we ectopically expressed CD36 for 7 days in IMR90 human diploid fibroblasts. This short‐term CD36 overexpression in IMR90 cells did not produce significant NF‐κB activation (Fig
4A), cell cycle arrest (Fig
4C), or SA‐βGal activity (Fig
4E). However, as in HBE cells, upon extended (17 days) CD36 expression, we observed a significantly decreased proliferative capability (Fig
4D) associated with increased SA‐βGal staining (Fig
4F) and increased levels of p16, p21, and activated forms of NF‐κB signaling pathway components (Fig
4B). Together, these findings suggest that the CD36–NF‐κB–SASP signaling cascade exists in both human epithelial cells and diploid fibroblasts and prolonged exposure to this signaling is capable of inducing stable cell cycle arrest and appearance of a senescent phenotype.
The SASP comprises a combination of secreted molecules that collectively function to reinforce the overall senescent phenotype
27. To investigate the biological function of the CD36‐dependent SASP, we applied conditioned media from long‐term CD36‐expressing or control IMR90 cells to naïve fibroblasts. Following 10 days of culture, a strong senescent phenotype was observed only in the group treated with supernatant from CD36‐expressing cells (Fig
EV2A–C), suggesting that upon CD36‐mediated NF‐κB activation, fibroblasts secrete a cohort of cytokines which act to accelerate the onset of senescence. To test whether NF‐κB and its downstream SASP components are the major factors driving cell cycle arrest, we administered the NF‐κB inhibitor Bay 11‐7082 to block CD36‐dependent NF‐κB‐mediated cytokine production and secretion and then measured fibroblast senescence. Importantly, NF‐κB inhibition resulted in diminished SA‐βGal staining, reduced p16 and p21 expression, and partially restored proliferative capacity in long‐term CD36‐expressing IMR90 cells (Fig
EV2D–F). Based on these results, we conclude that long‐term activation of the CD36–Src–p38–NF‐κB signaling axis is sufficient to drive a comprehensive senescent phenotype.
Next, we explored the involvement of individual SASP components in CD36‐driven cell cycle arrest. Both paracrine signaling and autocrine signaling are known to contribute to the senescent process, and canonical SASP cytokines such as IL‐6 and IL‐8 have been shown to promote fibroblast proliferative arrest
21. IL‐6 and IL‐8 are among the secreted factors upregulated in HBE cells in response to ectopic CD36 expression (Fig
2F). To test whether these cytokines are capable of driving epithelial cell senescence, we treated HBE cells with recombinant IL‐6 or IL‐8 for 9 days, a procedure that resulted in increased SA‐βGal activity (Fig
EV3A), reduced proliferative potential (Fig
EV3B), and mild but consistent upregulation of p16 and p21 (Fig
EV3C). Consistent with previous reports, IL‐6 administration produced a strong senescent phenotype in IMR90 fibroblasts (Fig
EV3D), indicating cell type‐specific differences in the ability of individual SASP components to induce senescence. These results suggest that at early time points, CD36 functions to drive NF‐κB‐mediated secretion of canonical SASP components, which in turn act in a feed‐forward manner to promote stable cell cycle arrest and establish the senescent state.
Ligands are required for CD36‐dependent SASP production and senescence establishment
Like many other cell surface receptors, CD36's signaling activity requires an interaction with its cognate ligand. Extracellular oxidized low‐density lipoprotein (oxLDL) and amyloid‐beta 1–42 (Aβ) are known to bind CD36 to drive Src–p38‐dependent cytokine and chemokine expression in immune cell types
16. To identify ligands involved in stimulating CD36‐dependent NF‐κB signaling in HBE cells and fibroblasts, we ectopically expressed CD36 in HBE cells and supplemented them with purified oxLDL or Aβ for 24 h, using phospho‐p65 as a NF‐κB activation readout. As shown in Fig
EV4A, the addition of oxLDL had no measurable effect on NF‐κB activation in either the presence or absence of CD36 overexpression, indicating that this ligand is unable to activate inflammatory signaling in this particular cell context, perhaps due to the absence of certain co‐receptors or intracellular adaptor proteins. In contrast, supplementation with recombinant Aβ produced strong NF‐κB activity when combined with CD36 expression. Importantly, Aβ addition alone stimulated lesser but detectable levels of p65 phosphorylation, likely due to low levels of CD36 present under basal conditions (Fig
EV4A). Aβ supplementation was also sufficient to induce NF‐κB activation in CD36‐expressing fibroblasts (Fig
EV4B), suggesting that Aβ is the major ligand driving the CD36–NF‐κB‐dependent SASP in cultured human cells.
If CD36 is involved in the generation of a SASP and Aβ is necessary for CD36‐dependent signaling, then this ligand should be detectable under standard cell culture conditions. To examine this, we measured Aβ levels in HBE cell culture medium by ELISA and found that Aβ is indeed present at a concentration of approximately 75 pg/ml (Fig
EV4C). In parallel assays, IMR90 fibroblast cultures, which do not activate NF‐κB signaling in response to short‐term CD36 expression (Fig
4A), do not contain measurable levels of extracellular Aβ at the 7‐day time point (Fig
EV4C). Aβ production is regulated at several levels including by transcription of amyloid precursor protein (APP) mRNA as well as post‐translational cleavage by multiple proteolytic secretase proteins including beta‐secretase 1 (BACE1)
30. To investigate Aβ production, we directly measured the expression levels of APP and BACE1 in CD36‐expressing HBE cells and IMR90 cells. Interestingly, we noted an induction of APP and BACE1 at the 7‐day time point in both cell types, suggesting the presence of a positive feedback loop from CD36 signaling to generate its ligand Aβ (Fig
EV4D). Based on these results, we conclude that ligand‐dependent receptor activation is necessary for CD36's SASP‐ and senescence‐promoting activity.
CD36‐dependent secretory molecule transcription and senescence establishment require ligand–receptor interactions
The results presented above indicate that expression of CD36 and its ligands is sufficient to trigger a NF‐κB‐mediated SASP, and prolonged signaling through this pathway can promote a comprehensive senescence phenotype. This prompted us to ask whether CD36 is required for SASP establishment and cell cycle arrest during conventional senescence reprogramming. To address this, we first blocked the induction of CD36 using small‐hairpin RNAs and then induced HBE cell senescence by administering the chemical inhibitor erlotinib (Fig
5A). Although CD36 knockdown did not rescue the cell cycle arrest or SA‐βGal activity phenotypes caused by EGFR inhibition, we observed a major impairment in Src–p38–NF‐κB activation after senescence induction when CD36 upregulation was suppressed (Fig
5B and C). Consistent with this, a large cohort of SASP components was not induced during senescence after CD36 knockdown (Fig
5D), and this subset largely coincides with the group of pro‐inflammatory molecules induced by ectopic CD36 expression (Fig
2F). To determine whether CD36 is also necessary for SASP generation in other senescent contexts, we performed similar experiments using fibroblasts triggered to senescence by HRAS activation and also observed decreased levels of active NF‐κB (Fig
5E) upon CD36 knockdown. These results demonstrate a functional requirement of CD36 for NF‐κB‐mediated SASP establishment in response to multiple senescent stimuli.
In this study, we discovered the CD36 receptor and its ligand Aβ as novel regulators of cellular senescence via their cooperative ability to initiate a SASP. Although the NF‐κB and CEBPβ transcription factors are previously known to be major mediators of the SASP and their immediate upstream signaling cascades are well characterized, the identity of the specific molecular trigger responsible for activating these pathways during the onset of senescence has remained elusive
26. It is important to note that CD36 is previously known for its ability to stimulate cytokine and chemokine production in immune cell types such as monocytes
16. Here, we propose that the ability of senescent cells to adopt an immune‐like secretory phenotype largely stems from their capacity to upregulate the expression of CD36 in response to various senescent stimuli. Further research will be required to determine the precise molecular mechanisms responsible for CD36 transcriptional upregulation during senescence induction.
Based on results obtained using epithelial and fibroblast model systems, we conclude that ligand‐mediated CD36 activation is necessary for expression of the full‐senescence secretome. In particular, interaction with Aβ appears to be a critical event in this process. In both HBE and IMR90 cells, APP and BACE are upregulated during senescence resulting in sufficient levels of Aβ to activate CD36 and generate a SASP (Fig
EV4C). Recent
in vivo analysis has identified Aβ as mediating phenotypic changes in intestinal epithelial cells, and it will be interesting to determine whether a similar relationship holds for fibroblast cells and the molecular mechanisms responsible for such lineage‐specific distinctions
32. Because Aβ is most well characterized as the primary component of the amyloid plagues found in the brains of Alzheimer's patients, an intriguing area for future investigation concerns the possible relationship between Aβ accumulation in neurodegenerative diseases and processes related to cellular senescence and inflammation
33.
Several recent reports have implicated the cyclic GMP‐AMP synthase (cGAS)–STING cytosolic DNA sensing pathway in SASP induction
35. In brief, these studies show that multiple types of senescent cells engage the cGAS–STING pathway through recognizing cytosolic DNA fragments, resulting in IRF3‐mediated IFN‐β upregulation and the subsequent production of many known SASP components. Like the CD36 pathway reported here, cytosolic DNA sensing allows senescent cells to co‐opt signaling cascades normally utilized by the immune system in order to generate an inflammatory phenotype. However, two major differences between these distinct SASP regulatory pathways should be noted. First, CD36‐mediated cytokine and chemokine production proceeds directly through NF‐κB and does not involve an IFN‐β intermediate. Thus, CD36 is likely to be a more direct and rapid pathway for SASP activation in response to senescent stimuli. Indeed, nearly ten‐fold CD36 upregulation was observed here only 6 h after senescence induction (Fig
1B). Second, the CD36‐mediated secretome does not require the presence of cytosolic DNA but instead depends on an extracellular CD36‐activating ligand. Therefore, the CD36–NF‐κB–SASP pathway might be predicted to be utilized by a broader range of senescent cell types, perhaps including those known to be involved in embryonic development
4.
In summary, here we report a critical role for the scavenger receptor CD36 in the early initiation of the SASP in response to multiple senescent stimuli. Through this study, we have also demonstrated the dynamic nature of the senescent program, from the initiation of the process to the establishment of the senescent state dependent on its enforcement by multiple secreted components. Although the two distinct features of cellular senescence, cell cycle exit and the SASP, are mediated by different signaling mechanisms, they are closely linked by the pathway identified in our study. Given the expanding appreciation of the relationship between cellular senescence and the aging process, future studies should aim to address the role of ligand‐dependent CD36 activation as it relates to the SASP in various age‐related diseases including neurodegeneration and cancer.