Mitochondrial dysfunction contributes to the senescent phenotype of IPF lung fibroblasts

Abstract Increasing evidence highlights that senescence plays an important role in idiopathic pulmonary fibrosis (IPF). This study delineates the specific contribution of mitochondria and the superoxide they form to the senescent phenotype of lung fibroblasts from IPF patients (IPF‐LFs). Primary cultures of IPF‐LFs exhibited an intensified DNA damage response (DDR) and were more senescent than age‐matched fibroblasts from control donors (Ctrl‐LFs). Furthermore, IPF‐LFs exhibited mitochondrial dysfunction, exemplified by increases in mitochondrial superoxide, DNA, stress and activation of mTORC1. The DNA damaging agent etoposide elicited a DDR and augmented senescence in Ctrl‐LFs, which were accompanied by disturbances in mitochondrial homoeostasis including heightened superoxide production. However, etoposide had no effect on IPF‐LFs. Mitochondrial perturbation by rotenone involving sharp increases in superoxide production also evoked a DDR and senescence in Ctrl‐LFs, but not IPF‐LFs. Inhibition of mTORC1, antioxidant treatment and a mitochondrial targeting antioxidant decelerated IPF‐LF senescence and/or attenuated pharmacologically induced Ctrl‐LF senescence. In conclusion, increased superoxide production by dysfunctional mitochondria reinforces lung fibroblast senescence via prolongation of the DDR. As part of an auto‐amplifying loop, mTORC1 is activated, altering mitochondrial homoeostasis and increasing superoxide production. Deeper understanding the mechanisms by which mitochondria contribute to fibroblast senescence in IPF has potentially important therapeutic implications.

altering mitochondrial homoeostasis and increasing superoxide production. Deeper understanding the mechanisms by which mitochondria contribute to fibroblast senescence in IPF has potentially important therapeutic implications.

K E Y W O R D S
cyclin-dependent kinase inhibitors, fibroblasts, idiopathic pulmonary fibrosis, mechanistic target of rapamycin complex 1, mitochondria, peroxisome proliferator-activated receptor gamma coactivator 1-alpha, rapamycin, reactive oxygen species and mitoTEMPO

| INTRODUCTION
Idiopathic pulmonary fibrosis (IPF) is a lethal disease of unknown aetiology that largely presents in the elderly. Although the pathogenesis of IPF is unclear, it is considered to be a consequence of dysregulated repair responses emanating from an injured epithelium. 1 Epithelial cell hyperplasia and myofibroblast/collagen accumulation in lung parenchyma lead to tissue structural defects that impair gas exchange. 2 Until recently, effective treatment options for IPF were limited to lung transplantation. However, both pirfenidone and nintedanib have been shown to reduce the rate of decline in lung function and increase progression free survival in IPF, albeit with significant side effects. 3,4 These positive developments have encouraged renewed interest in identification of alternative and complementary drug targets.
Cellular senescence follows a DNA damage response (DDR) and subsequent induction of p53-p21 CIPI and/or p16 INK4a -pRB signalling, 5 both of which lead to cell-cycle arrest. Senescent cells are also resistant to apoptosis, evade immune surveillance and acquire a senescence-associated secretory phenotype (SASP). [6][7][8] There is mounting evidence to support an important contribution of senescence in IPF. For example, mutations in genes encoding components of the telomerase complex that accelerate replicative senescence are linked to IPF and senescent markers are detected in the lung of IPF patients. [9][10][11][12] Furthermore, senescent prone mice are more susceptible to experimental pulmonary fibrosis, whereas the targeting of senescent cells is protective. 6,13 Recent evidence suggests that the senescence of type II alveolar epithelial cells (AECIIs) in IPF impairs re-epithelialization following injury, triggering a cascade of events that lead to fibrosis. 14,15 Senescent lung fibroblasts (LFs) are also highly likely have a role in IPF, exhibiting myofibroblast-like characteristics (ie, increased α-smooth muscle actin expression) and a highly activated secretome. 6,11,12,16 In IPF, a failure to eliminate senescent fibroblasts by apoptosis or immune cell clearance is postulated to impede wound resolution and contribute to disease progression. 11,12 Mitochondria are double membrane-bound organelles responsible for energy production. Mitochondrial integrity is maintained through the coordination of several processes such as biogenesis and mitophagy, which have been collectively referred to as mitochondrial quality control (MQC). 17 Dysfunctional MQC and inflammation are key signatures of ageing and several ageing-related diseases. As part of the mitochondrial energy generating process involving the electron transport chain (ETC), high-energy electrons leak and generate reactive oxygen species (ROS). Mitochondrial DNA (mtDNA) is highly sensitive to oxidative damage because of its proximity to the ETC and ROS production, lack of chromatin structure and a reduced capacity for repair. 18 As such, mtDNA damage and subsequent mutations can result in mitochondrial dysfunction, including a collapse in the mitochondrial membrane potential, 19 which in turn contributes to further ROS production, that ultimately leads to dysmorphic and impaired mitochondria. Increased mtDNA levels and mass occur in a number of ageing organs and tissues, including the lung, possibly as an adaptive mechanism to compensate for diminished mitochondrial output. 20 Sustained increases in ROS as a result of mitochondrial dysfunction are thought to reinforce the DDR, and senescence as a consequence. [21][22][23] Senescent LFs from IPF patients (IPF-LFs) have recently been reported to exhibit features of mitochondrial dysfunction, including disrupted cristae and a diminished capacity for oxidative phosphorylation. 24 However, the consequences of these mitochondrial perturbations on superoxide production have not been evaluated, nor any association between senescence and mitochondrial dysfunction. In this study, we characterize for the first time the relationship between the DDR, senescence and mitochondrial dysfunction in IPF-and Ctrl-LFs involving mitochondrial superoxide. We provide evidence that in IPF-LFs, an auto-amplifying loop exists involving mitochondrial-derived ROS which perpetuates senescence, and that the disruption of this cycle has potential important therapeutic implications. Increased activation of the mechanistic target of rapamycin complex 1 (mTORC1) and subsequent alterations in mitochondrial homoeostasis mediate the increased ROS production that follows a DDR and contributes to senescence persistence in LFs.

| Senescence-associated β-galactosidase detection
For senescence-associated β-galactosidase (SA-β-Gal) staining, subconfluent cell cultures in 12-well plates were fixed and stained using a commercial kit (Cell Signaling Technology) according to the manufacturer's instructions. Cells were imaged using an Olympus IX51 inverted microscope. For quantitation, blue stained cells were visualised by light microscopy, and counted in random fields at 100× magnification. Positive cells were expressed as a percentage of total cells, counted in the same fields by phase-contrast microscopy.

| ELISA and protein assays
Levels of interleukin-6 (IL-6), IL-8, monocyte chemoattractant protein-1 (MCP-1 or CCL2), regulated on activation, normal T cell expressed and secreted protein (RANTES or CCL5) and insulin-like growth factor binding protein-5 (IGFBP5) in LF supernatants were measured by specific sandwich enzyme-linked immunosorbent assays (ELISA) using commercial kits (RnDSystems, MN, USA) as according to the manufacturer's instructions. Protein concentration in cell lysates was measured using the BCA assay kit (Thermo Scientific).
For each sample, RNA (100 ng) was hybridised to the inflammatory gene expression code set before purification and immobilization to nCounter cartridges using the nCounter Prep Station (NanoString Technologies). Cartridges were scanned using the nCounter Digital SCHULIGA ET AL.

| 5849
Analyzer and subsequent data analysed with nSolver software (NanoString Technologies). Raw data of each gene (digital counts) were normalized to housekeeping genes (CLCT1, GUSB, HPRT1 and TUBB) before the ratio of the IPF versus Ctrl samples was computed. Heat maps of differentially regulated genes were generated online using Morpheus software (https://software.broadin stitute.org/morpheus/).

| PCR analysis
Real time polymerase chain reaction (PCR) was conducted to quantify mRNA and mitochondrial DNA. RNA and DNA were purified from cells using RNeasy and QIAamp DNA mini spin columns (Qiagen), respectively. RNA was reverse transcribed into cDNA using the iScript Advanced cDNA kit (Bio-Rad). DNA was amplified by PCR using the iTaq Universal SYBR Green Supermix (Bio-Rad) in an ABI Prism 7500HT sequence detection system (Applied Biosystems) with the relevant PCR primers (Table S3)

| Immunoblotting
Cell lysates (4 μg protein) were subjected to SDS polyacrylamide gel electrophoresis (SDS-PAGE) using 4%-15% Mini-Protean TGX 15 well stain free gels (Bio-Rad). Fluorescent detection of protein after electrophoresis was imaged using a Bio-Rad Gel Doc imaging system. Gels were then electroblotted as described previously 27

| IPF-LFs exhibit senescence-like characteristics
A composite set of parameters were measured to ascertain the extent of IPF-LF senescence. The expression of the cell-cycle arrest proteins p16 and p21 and the percentage of cells positive for SA-β-Gal were higher in cultures of IPF-LFs, compared to the Ctrl-LFs ( Figure 1A and B). Furthermore, a higher percentage of nuclei in the IPF-LFs exhibited foci of phospho-p53 (pp53), a DDR marker (Figure 1C). Like p53, phosphorylated histone H2A.X (γH2A.X) rapidly forms complexes at DNA double-stranded breaks and is a sensitive marker of the DDR. The formation of γH2A.X nuclear foci was also increased in IPF-LFs, when compared to Ctrl-LFs ( Figure 1D). An important pathological feature of senescent fibroblasts is the SASP and while the SASP of IPF-LFs has previously been described, its characterization was limited to the gene expression of a few inflammatory genes. 24 In this study, the levels of IL-6, CCL2, CCL5 and IGFBP5 in IPF-LF conditioned medium were shown to be higher than Ctrl-LF conditioned medium ( Figure 1E). Using nanostring technology, 43 inflammatory genes were differentially regulated between IPF-and Ctrl-LFs (P < 0.05), with 39 increased in IPF ( Figure 1F).
Collectively, these data confirm that LFs of IPF patients are more senescent-like than Ctrl-LFs, corresponding with an intensified DDR and SASP.

| IPF-LFs display mitochondrial dysfunction
Evidence is accumulating that mitochondrial-derived superoxide and its ROS by-products elicit a DDR to reinforce senescence. 22,23 Adding further evidence, the levels of mitochondrial superoxide and intracellular ROS in this study were significantly higher in IPF-LFs than Ctrl-LFs (Figure 2A  Concomitantly, rotenone induced long-term increases in mitochondrial superoxide production and stress ( Figure 3I and J). Like etoposide, rotenone had little effect on senescence markers in IPF-LFs ( Figure 3).

| NAC suppresses etoposide-induced senescence
The role of ROS in LF senescence was investigated using N-acetyl

LF senescence
Our data suggest that alterations in mitochondrial homoeostasis involving increased superoxide production have a causative role in LF senescence. To investigate further this possibility, we used rapamycin, the pharmacological inhibitor of mTORC1, which has a pivotal role in regulating mitochondrial biogenesis and activity, including PGC-1α expression. Pre-treatment of Ctrl-LFs with rapamycin (100 nM) attenuated the effects of etoposide on senescent markers, PGC-1α gene expression and mitochondrial stress, mass and DNA levels ( Figure 5).

| Rapamycin and mitoTEMPO decelerate IPF-LF senescence
We next investigated the effects of long-term treatment with rapa-

| 5855
Promisingly, the mitochondrial-selective antioxidant, mitoTEMPO, had a comparable effect on IPF-LF senescence. Overall, this study provides substantial evidence that senescence and mitochondrial dysfunction are highly interrelated processes in IPF-LFs and that targeting either mitochondrial-derived superoxide or the DDR-driven processes that contribute to its increased production may limit the damaging impact of senescence in IPF.
The IPF-LFs in this study displayed many of the senescence characteristics reported previously, including increased p16 and p21 expression and SA-β-Gal activity. 12, 24 In addition, for the first time to our knowledge, IPF-LFs were associated with increased nuclear acti- fibroproliferative phase of the disease. Senescence can also spread from senescent to naïve fibroblasts in the ageing fibrotic lung as a consequence of the "bystander effect". 28 Regardless of the reason(s) for increased IPF-LF senescence, these cells are likely to have important roles in IPF because of their resistance to apoptosis, myofibroblast-like features and development of a SASP.
The SASP is an important pathological feature of senescence. 6,29 In this study, IPF-LFs exhibited increased production of cytokines typically associated with the SASP, as well as a markedly pro-inflammatory gene expression profile. Profiling revealed the expression of several transcription factors was higher in IPF, including NF-κB and C/EBPβ, important mediators of the secretory component of senescent cells. 30 In senescence, NF-κB signalling follows the DDR via activation of p38 MAPK and its downstream mediator, MAPKAPK2, both of which were up-regulated in IPF-LFs. 30 Furthermore, the expression of two key modulators of the innate immune system that regulate NF-κB activation, MyD88 and toll-interacting protein (Tollip), was also higher in IPF-LFs. Evidence is accumulating that chronic low-level activation of the innate immune system has an important role in IPF, with several Tollip gene polymorphisms linked with the disease. 31 A negative regulator of acute inflammation, Tollip also augments chronic low-grade inflammation in a process that involves its translocation to mitochondria. 32 Many of the inflammatory gene changes reported in this study align with our current understanding of IPF pathology and/or processes of cellular senescence.
Mitochondrial dysfunction is a feature of both senescence and ageing. 19 Mitochondrial DNA, membranous lipids and proteins are highly susceptible to oxidative damage, causing respiratory uncoupling and increased formation of superoxide. 33 The latter in turn contributes to further mitochondrial damage, as part of a cycle that ultimately leads to dysmorphic and impaired mitochondria. In this study, IPF-LF mitochondrial dysfunction was shown by increases in MitoSOX and NAO fluorescence, which were not evident in Ctrl-LFs. and airway smooth muscle (ASM) cells in severe asthma. 23,36,37 In this study, increased levels of mtDNA and MitoTracker Green staining suggest IPF-LFs also exhibit a net increase in mitochondrial mass as compared to Ctrl-LFs, possibly a consequence of increased biogenesis. In support of this hypothesis, the IPF-LFs showed an increase in mTORC1 activity and PGC-1α expression. PGC-1α is a key regulator of mitochondrial biogenesis and is directly up-regulated by mTORC1 at a transcriptional level. Impaired recycling of mitochondria (mitophagy) may also explain our findings of increased mitochondrial mass in IPF-LF cultures. While not investigated in this study, mitophagy has been reported to be suppressed in IPF-LFs. 38 Contrary to our observations, Rojas and colleagues reported a decrease in the mitochondrial mass of fibroblasts isolated from IPF lung. 24 Variations in fibroblasts isolation and culture conditions between these two studies may contribute to these divergent findings. In particular, our fibroblast culture protocol used a high glucose containing medium favouring increased mitochondrial biogenesis. 5′ AMP-activated protein kinase (AMPK) is activated when energy stores are low to regulate mitochondrial biogenesis by inhibiting mTORC1, 39 which may have been relevant in the Rojas study.
In this study, we provide evidence of a causal link between LF senescence and mitochondrial dysfunction. mTORC1 is a key mediator of cellular senescence, which links the DDR with alterations in mitochondrial homoeostasis in an auto-amplifying loop that perpetuates senescence. 40 As part of this loop, ATM via AKT activates mTORC1, which regulates mitochondrial biogenesis through PGC-1α/ β expression and by inhibiting eukaryotic translation initiation factor 4E (eIF4E)-binding proteins (4E-BPs). 41 Furthermore, mTORC1 suppresses mitophagy and the expression of the antioxidant enzyme, mitochondrial superoxide dismutase (MnSOD). 38 Inhibition of the mTORC1 pathway is reported to attenuate γ-irradiation-induced senescence in fibroblasts by suppressing mitochondrial biogenesis and subsequent superoxide production. 23 In this study, rapamycin attenuated etoposide-induced Ctrl-LF senescence and decelerated IPF-LF senescence in long-term culture. Furthermore, the knockdown of PGC-1α in IPF-LFs resulted in a marked reduction in expression of senescence markers. Overall, these observations provide additional evidence of the importance of mitochondria and mTORC1 signalling in the maintenance of LF senescence. Increased mTORC1 activation within the fibrotic foci of IPF patients suggests this pathway has a role in pulmonary fibrosis. 42 Recently the anti-fibrotic potential of the pan-PI3 kinase/mTOR inhibitor, GSK2126458 was evaluated in ex vivo models of IPF with promising results. 43 However, rapamycin has also been associated with a more rapid disease progression in a cohort of IPF patients. 44 As mTORC1 regulates a range of processes outside of mitochondrial homoeostasis, with many potential biological effects, alternative targets may be required to limit the deleterious effects of dysfunctional mitochondria on senescence in the treatment for IPF. In this regard, mitoTEMPO and other mitochondrial-selective antioxidants such as MitoQ show potential as therapies for lung fibrosis. Here we observed that mito-TEMPO decelerates IPF-LF senescence in vitro, and in pre-clinical studies for asthma and kidney disease, mitoTEMPO exhibits antifibrotic activity. 1,45,46 In this study, the expression of PGC-1α was increased in IPF-LFs  48 These splice variants are regulated differently to PGC-1α1 and evoke distinct biological programs. 49 An important function ascribed to PGC-1α, asides from its role in mitochondrial biogenesis, is the induction of antioxidant expression. 50 The antioxidant role of PGC-1α is dependent on the energy status of the cell and involves activation of its upstream regulator AMPK under conditions of energetic stress. 50  In this study, we provide evidence that an auto-feedback loop exists between dysfunctional mitochondrial and senescence in LFs, involving increases in mitochondrial superoxide and mTORC1 activation. However, there are inherent difficulties in confirming the existence of such a loop, particularly as senescence and mitochondrial dysfunction have a range of overlapping pathological features and effects. Another contributing factor is the non-selectivity of the tools that can be utilized with primary cultures of LFs to delineate the role of mitochondria in the acquisition and reinforcement of senescence.
We have made mention of the limitations of the approaches employed to target mitochondrial function and biogenesis, including the blunt effects of rapamycin and PGC-1α knockdown. To date, one of the more convincing studies to show the important contribution of mitochondria to senescence was by Passos and colleagues, who used recombinant genetics to remove mitochondria from embryonic fibroblasts (ie, MRC-5s). 23 However, to conduct such mechanistic studies is less feasible with primary cultures of LFs from aged donors/patients.
What our studies do confirm is that mitochondrial dysfunction involving increased superoxide occurs in parallel with senescence in IPF-LFs.
These processes are highly intertwined and that a range of mitochondrial targeting approaches attenuates/decelerates senescence.
In summary, LFs isolated from IPF patients exhibit senescencelike characteristics in association with an increased DDR, SASP and changes in mitochondrial homoeostasis. Our data also support roles of increased mitochondrial stress and superoxide production in the induction and maintenance of the senescent phenotype in LFs. The involvement of dysfunctional mitochondria in LF senescence may provide alternate targeting opportunities in the treatment of IPF.

CONFLI CTS OF INTEREST
The authors confirm that there are no conflicts of interest.