Type 2 Diabetes, Independent of Obesity and Age, Is Characterized by Senescent and Dysfunctional Mature Human Adipose Cells

Aging is a major risk factor for many chronic diseases, including diabetes, and people with diabetes are also more likely to develop other age-related diseases, such as cardiovascular disease, cognitive disorders, and cancer (1). Cell senescence with proliferative cell arrest has emerged as an important contributor to aging and age-related diseases (1–4). It is a consequence of several factors, including telomere shortening, DNA damage, inflammation, and increased reactive oxygen species production, and by senescent cell–secreted factors, such as senescence-associated secretory phenotype (SASPs), that also exert negative effects on neighboring nonsenescent cells and contribute to the age-related disorders (3,5,6). Genetic deletion of senescent cells in animal models or use of senolytic agents has shown improvements of several different aging-associated diseases, including atherosclerosis and diabetes (7,8). Studies are currently ongoing with senolytic agents in humans, but firm clinical results are not yet available.

Obesity in adult humans is usually associated with expanded and dysfunctional subcutaneous adipose cells and is the major cause of the ongoing epidemic of insulin resistance and type 2 diabetes (T2D). We have recently shown that senescence is increased in human subcutaneous adipose progenitor cells in obesity and T2D, which is associated with reduced adipogenic differentiation of the cells and induced by secreted SASPs from senescent progenitor cell culture medium (9).

Dysregulated adipose tissue is a well-recognized cause of systemic insulin resistance in obesity (10,11). Extensive studies, mainly in young obese mouse models, have attributed this to increased inflammation and tissue infiltration of macrophages and other inflammatory cells, which lead to dysfunctional adipose cells with increased lipolysis, altered adipokine secretion, and reduced glucose metabolism and insulin sensitivity. However, recent studies have questioned the importance of inflammatory cells in inducing insulin resistance in human adipose tissue in obesity (12,13). Here, we here asked whether mature, differentiated, and lipid-filled adipose cells (henceforth referred to as mature adipose cells), which do not undergo proliferation, from middle-aged adults who are lean, obese, and obese with T2D become senescent, and if so, what the cellular consequences are.

Hyperinsulinemia has recently been shown to activate the early cell cycle regulated by cyclin D1 and induce endoreplication and subsequent senescence in mature human adipose cells in obesity (14). However, it is not clear whether this is the only mechanism or whether mature adipose cells are also susceptible to SASPs from ambient cells promoting their cell senescence. Here, we found adipose cell senescence to be considerably higher in cells from individuals with T2D than from matched obese individuals without diabetes. We also show that fully differentiated human adipose cells are responsive to prosenescent agents, including SASPs, inducing cyclin D1, senescence, and dysfunctional cells, and that this can be counteracted by senolytic agents. We conclude that senescence is increased in T2D mature adipose cells, also when matched for donor age and BMI, and is a likely contributor to dysfunctional mature adipose cells, tissue inflammation, and insulin resistance.

Gene expression and protein levels were studied in isolated mature adipose cells and intact human subcutaneous adipose tissue (SAT) (needle biopsies) from 42 subject after an overnight fast, including 10 healthy lean control subjects, 9 obese control subjects of similar age (Ob. 1 group), and 12 obese subjects (Ob. 2 group) and 11 obese subjects with T2D treated only with oral medication, mainly metformin, and of similar age and BMI. Table 1 lists additional details. An additional 12 subjects with T2D were included for the plasma insulin assays (Supplementary Table 1). HOMA for insulin resistance (HOMA-IR) was calculated using the equation [fasting glucose (mmol/L) × fasting insulin (mU/L)]/ 22.5. We also included for experiments with conditioned medium (CM), immunofluorescence, doxorubicin, dasatinib, and quercetin treatment an additional 13 subjects aged of 45.8 ± 10.7 years, BMI 27.2 ± 3.1 kg/m², and adipose cell size 98.4 ± 9.2 μm.

The ethics committee of the University of Gothenburg approved the study design (approval nos. S655-03 and T492-17). The study was performed in agreement with the Declaration of Helsinki. Written informed consent was received from study subjects before inclusion.

Abdominal SAT biopsies were obtained as previously described (9). Briefly, SAT biopsies were obtained under local anesthesia. Snap frozen whole tissue and isolated mature adipose cells were used for RNA and lysates. The biopsies were washed with PBS, cut into small pieces, and digested with 0.28 mg/mL collagenase A (Sigma-Aldrich) in M199 with Hanks’ balanced salt solution (Thermo Fisher Scientific) supplemented with 4% (w/v) BSA (Hyclone; Thermo Fisher Scientific), pH 7.4, for 50 min at 37°C in a shaking water bath. The digest was filtered through nylon mesh with a pore size of 250 μm. The floating adipocytes were washed in DMEM/F12 (Lonza) supplemented with 10% (v/v) FBS, 2 mmol/L l-glutamine, and 1% (v/v) penicillin/streptomycin (Thermo Fisher Scientific). The adipocytes were removed and used for adipose cell size measurements, RNA, and lysates, while the remaining media (saved filtrate) containing the progenitor cells were pooled and centrifuged for 15 min at 1,500g at 20°C. The progenitor cells were seeded in a 175-cm² cell culture flask with DMEM/F12 supplemented with 10% (v/v) FBS, 2 mmol/L l-glutamine, and 1% (v/v) penicillin/streptomycin. Cells were tested repeatedly for mycoplasma infection with MycoAlert Mycoplasma Detection Kit (Lonza). We verified that mature adipose cells were not contaminated by tissue debris by microscope analysis (Supplementary Fig. 1A), and cell size was measured on the isolated and floating cells using a calibrated scale. This method has been used for several years and independently validated against sectioned tissue slices and in another large study where we compared the sizing of isolated adipocytes (15,16). We also examined mRNA markers in the mature adipose cells, which would indicate potential cofractionation of cells that secrete factors that can promote senescence in the mature adipose cells (i.e., primarily inflammatory macrophages [PTPRC/CD45] and adipose progenitor cells [WNT10B]), but we also included a marker of endothelial cells (ANGPT2). As shown in Supplementary Fig. 1B–D, all these markers were at or below the level of detection.

The isolated progenitor cells were expanded in culture using DMEM/F12 cell culture medium with 10% (v/v) FBS. After 3 days of confluence, cells at passages 3–4 were induced to differentiate as previously described (9).

For studies with CM, medium from cells undergoing poor (estimated area with lipids 15 ± 8.7%, n = 3) or good (67 ± 5.8%, n = 3) differentiation were collected between differentiation days 10 and 13. Progenitor cells from three independent subjects were differentiated (9) for 14 days in adipogenic medium (AM) and thereafter incubated with CM and AM at a 1:1 ratio for 72 h. For stimulation with doxorubicin and dasatinib/quercetin, progenitor cells were seeded in 12-well plates and differentiated for 14 days (9) and thereafter incubated with 0.4 μmol/L doxorubicin, 50 nmol/L dasatinib, and 20 μmol/L quercetin (all Sigma-Aldrich) for 48 h. For glucose stimulation, progenitor cells were seeded in 12-well plates and differentiated for 10 days before AM with the final glucose concentrations 5 mmol/L, 12.5 mmol/L, and 22 mmol/L (DMEM) were added. Cells were allowed to differentiate 4 more days before treatment with 0.4 μmol/L doxorubicin for 48 h.

Protein lysates were extracted in ice cold lysis buffer (25 mmol/L Tris-HCl, pH 7.4, 0.5 mmol/L EGTA, 25 mmol/L NaCl, 1% [v/v] IGEPAL CA-630, 1 mmol/L Na₃VO₄, and protease inhibitor cocktail (Sigma-Aldrich). Protein content was determined using Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). Proteins were separated on 4–12% NuPAGE gels (Thermo Fisher Scientific) and transferred to nitrocellulose membranes (Bio-Rad). The membranes were probed with the antibodies shown in Supplementary Table 2, and immunoreactive proteins were visualized by enhanced chemiluminescence (Clarity or Clarity Max Western ECL; Bio-Rad) using the ChemiDoc Imaging System (Bio-Rad) and Image Lab version 6 software (Bio-Rad) for quantitation.

Total RNA from cultured human progenitor cells was isolated with E.Z.N.A. Total RNA Kit (Omega Bio-Tek). Frozen intact adipose tissue and isolated mature adipose cells were first suspended in QIAzol Lysis Reagent (QIAGEN) before homogenization with Mixer Miller (MM400; Retsch GmbH). RNA was separated by the addition of chloroform. After separation, the water phase containing RNA was used for isolation with the E.Z.N.A. Total RNA Kit. Quantification of RNA was performed with a NanoDrop 1000 Spectrophotometer. cDNA synthesis was performed with the High-Capacity cDNA Kit (Thermo Fisher Scientific) according to the manufacturer’s recommendations. Gene expressions were analyzed with the QuantStudio 6 sequence detection system. Gene-specific primers and probes were either designed using Primer Express software or purchased as Assays-on-Demand. Relative expression was calculated using the ΔΔCt method with normalization to endogenous control, either 18S or RPLPO (reagents and instruments from Thermo Fisher Scientific). Assays-on-Demand, probes and primers, and endogenous controls are listed in Supplementary Table 3.

Human adipose tissue stromal cells were grown on four-well glass detachable chamber slides (Sarstedt). Differentiation was initiated as above and at differentiation day 8, incubated with 0.4 μmol/L doxorubicin, 50 nmol/L dasatinib, and 20 μmol/L quercetin for 24 h, then fixed with 4% formaldehyde for 15 min and permeabilized in 0.5% Triton X-100 for 5 min. Cells were blocked with 20% FBS for 30 min followed by incubation with γH2AX, p21, and perilipin-1 antibodies (Supplementary Table 2) for 1 h. Cells were then washed and incubated with secondary antibodies conjugated with Alexa Fluor 594 or Alexa Fluor 488 (Supplementary Table 2) for 1 h, followed by nuclear counterstaining with DAPI and mounting with ProLong Diamond Antifade Mountant (Thermo Fisher Scientific). Images were collected with a ZEISS Axio Observer microscope camera.

The experimental data are presented as means ± SEM of at least three independent experiments as indicated in the figure legends. Data analysis was performed with GraphPad Prism 9 (GraphPad Software). Normal data distribution was determined using Shapiro-Wilk test. Unless otherwise stated, single comparisons with basal samples were performed using two-tailed Student t test and among more than two groups by one-way ANOVA adjusted for multiplicity. For data that were not normally distributed, the Kruskal-Wallis test with Dunn’s multiple comparison test was used as indicated in the figure legends. Correlations between pairs of columns were made using Spearman rank correlation. The association analyses were adjusted for cell size or age by using multiple linear regression models. Differences were considered statistically significant at P < 0.05.

All data generated or analyzed during this study are included in the published article (and Supplementary Material) and available from the corresponding author upon request.

To study whether human postmitotic SAT mature adipose cells become senescent and whether this is different in obesity and T2D, we examined cells from 42 donors (Table 1). Of these, 10 were lean and 9 obese (Ob. 1), with different BMIs (24.5 vs. 32.6 kg/m²) but similar ages (41.2 vs. 36.8 years). Since senescence is age related, we also included 12 obese (Ob. 2) subjects with similar BMIs (30.4 vs. 32.5 kg/m²) and ages (59.7 vs. 62.3 years) and 11 obese subjects with T2D.

First, we examined the degree of whole-body insulin sensitivity, measured as HOMA-IR, in the lean and obese groups (HOMA-IR is not a good marker for T2D) and its relation to adipose cell size and markers of cell senescence. As expected, the degree of HOMA-IR correlated significantly with adipose cell size (Fig. 1A). We then examined whether the mature adipose cells also became senescent and whether this is associated with insulin resistance. Interestingly, we found the senescence markers β-galactosidase (β-gal) (GLB1) and serpin family E member 1 (SERPINE1) in the SAT mature adipose cells to be increased, and this was strongly related to degree of insulin resistance (Fig. 1B and C).

To further evaluate senescence in mature adipose cells and its relation to clinical phenotype, we analyzed the senescence markers tumor protein p53 (TP53), GLB1, SERPINE1, and cyclin dependent kinase inhibitor 2A (CDKN2A) in isolated mature adipose cells and in intact adipose tissue biopsies from the same subjects. All these senescence markers were highly expressed in mature adipose cells and correlated closely with their expression in intact adipose tissue (Fig. 1D–G).

Adipose cell size is positively related to progenitor cell senescence in intact adipose tissue (9), and a similar relation between senescence markers and mature adipose cell size was found, further supporting the association between cell senescence and hypertrophic obesity in adult humans (Fig. 1H–K). Taken together, these data show that mature adipose cells, like progenitor cells, become senescent, and this is strongly related to adipose cell size and degree of insulin resistance.

To gain further insight into the impact of obesity and T2D on senescence in mature cells, we examined the age-matched lean versus Ob. 1 and age- and BMI-matched Ob. 2 versus T2D groups separately. All senescence markers, including the novel senescence marker regulator zinc finger matrin-type 3 (ZMAT3) (17), were increased in the Ob. 1 versus lean group, but only GLB1 and SERPINE1 were significantly different (Fig. 2A–E). In contrast, T2D versus Ob. 2 had a significant increase in all senescence markers except CDKN2A (Fig. 2G–K). Furthermore, the proinflammatory marker tumor necrosis factor-α (TNFA) was increased in the T2D but not in the Ob. 1 group (Fig. 2F and L). Together, these results show that senescence is increased in mature adipose cells in obesity and considerably more pronounced in T2D independent of BMI or age differences. Thus, expanded adipose cell size in human adult adipose tissue is a marker not only of obesity but also of increased senescence in mature adipose cells.

We then examined senescent and adipose cell functional protein markers focusing on lean versus T2D cells, recognizing that these groups differed in both age and BMI, but their respective importance was addressed in the previous data. In agreement with the increased transcriptional activation of senescence, we also found the senescent protein levels of β-gal, p16^INK4A (encoded by CDKN2A), and plasminogen activator inhibitor 1 (PAI-1) (encoded by SERPINE1), and gene expression, to be increased in mature adipose cells from obese subjects with T2D compared with cells from lean subjects without diabetes (Fig. 3A–C, J, and K). Phosphorylation of serine/threonine kinase 1 (AKT) on Ser473 as a marker of insulin sensitivity and action was markedly reduced in the T2D mature adipose cells (Fig. 3D and L) as were peroxisome proliferator–activated receptor-γ (PPAR-γ) and the marker of adipogenic terminal differentiation GLUT4 (encoded by solute carrier family 2 member 4 [SLC2A4]) (Fig. 3E, F, L, and M). Taken together, our findings show that mature adipose cells also become senescent and that this is associated with local cell- and whole-body insulin resistance, with reduced PPAR-γ and GLUT4 as markers of dysfunctional adipose cells. Mechanistically, this could be due to the associated hyperinsulinemia in obesity, recently shown to promote endoreplication and senescence in the nonreplicating mature adipose cells, and/or to increased adipose tissue inflammation caused by increased macrophage infiltration (18) and/or increased progenitor cell senescence secreting SASPs targeting the mature adipose cells. Consistent with increased inflammation, we found increased phosphorylation of mitogen-activated protein kinase 8 (also known as JNK1) on Thr183 and Tyr185 (pJNK) in T2D mature adipose cells (Fig. 3G and M). However, we did not find any evidence of increased macrophage binding to the isolated mature adipose cells since adhesion G protein-coupled receptor E-1 (ADGRE1) (also known as EMR1) was similar in all groups (Supplementary Fig. 2A); nitric oxide synthase 2 (NOS2) (also known as INOS) and mannose receptor C-type 1 (MRC1) (also known as CD206) expression as markers of macrophage M1 and M2 phenotypes were similar (Supplementary Fig. 2B and C). Thus, the increased inflammation in the mature cells may well be secondary to increased SASPs released by the adipose tissue progenitor cells.

Normal adipogenic development requires downregulation of the tumor suppressor and prosenescence factor p53; this is impaired in human senescent progenitor cells (9) and has been shown in experimental animal models to prevent normal adipose tissue development and induce insulin resistance (19). The p53 protein and gene expression were increased in T2D mature adipose cells (Fig. 3H and N), and activation of MDM2, which degrades p53, measured as S166MDM2 phosphorylation, was reduced (Fig. 3I and N).

Hyperinsulinemia in obesity activates the cell cycle in mature postmitotic human adipose cells through the early cell cycle G1/S regulator cyclin D1 (encoded by CCND1), and this leads to endoreplication and increased de novo DNA synthesis, but not mitosis, and cells undergo senescence (14,20). Therefore, we examined CCND1 expression in our cohorts. First, we compared the expression of CCND1 in mature adipose cells and intact biopsies and found a significant correlation and a marked increase (74%) in isolated mature adipose cells from the T2D group (Fig. 4A). There was also a close correlation between CCND1 expression and adipose cell size (Fig. 4B). CCND1 was increased in obese versus lean mature cells and in T2D cells from equally obese subjects of similar age (Fig. 4C and D). We analyzed adipose cell cyclin D1 protein levels in the lean and T2D groups and found a >95-fold increase in the T2D group (Fig. 4E). In fact, and consistent with our current results, there was a gradual increase in cyclin D1 as well as the other senescence markers β-gal and p16 among the lean, obese, and obese T2D groups (Fig. 4S).

To address potential mechanisms, we examined CCND1 and its relation to insulin levels in our nondiabetic groups and other senescence markers in the whole cohort. CCND1 expression in lean and obese subjects correlated with serum insulin levels of the donors, albeit of borderline significance (Fig. 4F), while it was very closely correlated with the cell senescence markers, including ZMAT3 (Fig. 4G–J). The correlation between insulin levels and the individual cell senescence markers in the nondiabetic groups was quite variable (Fig. 4K–O). Thus, CCND1 seems a considerably better marker of senescence in mature adipose cells than the ambient insulin levels. Furthermore, because of lack of fasting insulin values in our T2D cohort, we measured insulin in 12 new subjects with T2D, and they had similar insulin levels as the obese subjects with normoglycemia (Supplementary Table 1). However, T2D had an approximately twofold increase in cell CCND1 expression (Fig. 4D). Thus, it is unlikely that the increased senescence in mature adipose cells from T2D can be solely explained by elevated insulin levels. In support of this, we found a close correlation between CCND1 expression and ambient blood glucose levels (Fig. 4P), suggesting the importance of other diabetes-related factors. Glucose may be a marker rather than a cause since increased glucose levels per se in the cell culture medium did not increase the senescence markers p21 and cyclin D1 in differentiated progenitor cells, supporting the importance of other diabetes-related factors (Supplementary Fig. 3). We examined the possibility that other endogenous mitogenic factors, such as IGF1 and IGF2, were increased in adipose cells. We found a close correlation between IGF1, but not IGF2, and CCND1 in our obese subjects (including T2D) (Fig. 4Q and R).

We then validated that also fully differentiated progenitor cells can become senescent by incubating them with doxorubicin for 48 h. Senescence was induced by doxorubicin in these cells with increased levels of p53, p21, MDM2, cyclin D1, ZMAT3, and nuclear γH2AX, supporting DNA damage (Figs. 5A and 6A). Similarly, PPAR-γ, adiponectin, and fatty acid binding protein 4 (FABP4) were reduced, indicating dedifferentiation of the cells after only 48 h of induced senescence (Fig. 5A). Thus, fully differentiated human progenitor cells can become senescent and dysfunctional similar to what we see in obese T2D cells.

We have previously shown that CM from cells undergoing poor differentiation decreased adipogenesis of progenitor cells and reduced adipocyte markers (9). Now, we show that CM from cells with poor differentiation increased cyclin D1 in fully differentiated progenitor cells (Fig. 5B). Taken together, these data support that also differentiated progenitor cells can become senescent and dysfunctional, as we see in T2D cells, and they are responsive to secreted factors by ambient senescent cells increasing cyclin D1.

To study whether senescence in human mature adipose cells in obesity/T2D can be treated by senolytic agents, we examined whether the fully differentiated human progenitor cells are targeted by the senolytic agents dasatinib and quercetin (8). Doxorubicin-induced senescence and inflammatory markers were reduced by dasatinib/quercetin, including SERPINE1, C-C motif chemokine ligand 2 (CCL2), interleukin-6 (IL6), and IL8 (Fig. 5A and D–G), and dasatinib/quercetin counteracted the effects of doxorubicin and increased differentiation markers PPAR-γ and adiponectin, as well as CCAAT enhancer binding protein α (C/EBPα) expression (Fig. 5A and C). Cleaved caspase-3, as a marker of apoptosis, was slightly increased by dasatinib/quercetin (Fig. 5A). To validate that doxorubicin induces senescence in differentiated progenitor cells, we performed immunofluorescence staining with both p21 and perilipin-1 (Fig. 6B). Doxorubicin also induced DNA damage, as measured as γH2AX, and this was reduced by dasatinib/quercetin (Fig. 6A).

Taken together, these data show that human mature adipose cells, like progenitor cells, become senescent; this is increased in hypertrophic obesity, further pronounced in T2D, and associated with reduced PPAR-γ and other markers of impaired terminal differentiation of the adipose cells, increased inflammation, and local and whole-body insulin resistance. A likely scenario is that the senescent progenitor cells in obesity and T2D secrete SASPs, which also target mature adipose cells, and that cyclin D1, like ZMAT3, is a prominent marker of senescence and reduced by senolytic agents.

Aging is a common cause of increased cell senescence, and the recent large Tabula Muris study (21) showed that mice gradually developed cell senescence in most tissues over the 21 month study period, including adipose tissue, and that this is associated with reduced cell growth, tissue size and function of most tissues, increased inflammation, and aggregated DNA mutations (21). Here, we examined human SAT cells, both differentiated progenitor cells, intact tissue specimens, and isolated mature adipose cells from the same subjects who were middle-aged lean, obese without diabetes, or obese with T2D. Both transcriptional activation of senescence markers and their cell proteins were most markedly increased in cells from obese individuals with T2D but also were increased in obese compared with lean individuals of similar age.

We have previously shown that adipose progenitor cells from subjects with obesity and T2D are characterized by increased senescence and a reduced ability to undergo differentiation (9). However, when we started this work, we did not expect fully mature and lipid-filled adipose cells to exhibit senescence since they are not mitotic. During the preparation of this article, a study was published showing that mature adipose cells from obese individuals become senescent and that this can be caused by hyperinsulinemia promoting cyclin D1–induced G1/S cell cycle activation, but not mitosis, leading to endoreplication and induction of senescence (14). However, what happens in T2D was not studied. Here, we confirm that mature adipose cells from obese individuals indeed show increased senescence but that cells from similarly obese individuals with T2D exhibit considerably higher senescence together with several markers of dysfunctional adipose cells. We also show that cyclin D1 is an excellent marker of cell senescence and that the senolytic agent dasatinib/quercetin reduces cyclin D1.

Cyclin D1 regulates the early cell cycle but has many cell connections, including with different transcription factors, chromatin-modifying enzymes, and cytosolic proteins (22), and is also one of the most common cell cycle regulators increased in cancer cells. As shown recently, insulin, like other growth factors, increases cyclin D1 (13), and we show here that induced senescence, which inhibits cell proliferation, also leads to increased cell cyclin D1. The mechanism for this is unclear and needs to be studied further.

The reason for the enhanced senescence in T2D cells is unclear at present, but it is unlikely that it is only related to the degree of hyperinsulinemia. The subjects with T2D had a mild T2D, and we measured similar fasting insulin levels in individuals with obesity and T2D. Interestingly, we found a close correlation between another mitogenic factor, adipose cell IGF1, and CCND1 in both obesity and T2D. IGF-I is an important growth factor for adipose tissue, and mouse data have shown that deleting IGF-I specifically in the adipose tissue reduces its growth and mass (23). We also found a close correlation between cell senescence and ambient glucose levels, and elevated glucose has been shown to increase senescence in human fibroblasts in culture (24). However, when we examined the effect of glucose levels in the human differentiated progenitor cells, we did not see any direct effect of hyperglycemia on induction of cell senescence. Thus, we suggest that the correlation with ambient glucose levels may reflect other diabetes-related mediators.

Taken together, it is clear that both adipose tissue progenitor cells and mature adipose cells undergo senescence and that this is increased in obesity and further enhanced in T2D. This also leads to partial dedifferentiation and dysfunction of the cells with reduced PPAR-γ, GLUT4, adiponectin, and FABP4, all of which are associated with insulin resistance. The positive effects of dasatinib/quercetin in reducing senescence and cyclin D1 and increasing PPAR-γ in adipose cells raise the possibility that dasatinib/quercetin may be beneficial in treating T2D, and such studies are in progress in our laboratory.

Funding. This study was supported by the Knut and Alice Wallenberg Foundation (2020.0118 to U.S.), Swedish Research Council (2018-02587 to U.S.), Novo Nordisk Foundation (NN20OC0063541 to U.S.), Swedish Diabetes Association (DIA2020-506 to U.S.), Heart and Lung Foundation (20180492 to U.S.), and Swedish ALF (GBG-718601 to U.S.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. B.G. and A.N. designed the studies, conducted the experiments, acquired and analyzed data, and wrote the manuscript. R.S. analyzed data. F.B. and U.S. designed the studies, analyzed data, and wrote the manuscript. U.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.