Pharmacological clearance of senescent cells reverses HFpEF hallmarks by decreasing inammation, endothelial dysfunction and cardiac brosis

Aging and chronic inammation are associated with the development of heart failure with preserved ejection fraction (HFpEF). However, cellular senescence as a potential mechanistic link between both events and its pathophysiological and therapeutic role were yet unexplored. Here we show that ZSF1-obese rats, a model of cardiometabolic HFpEF, have exacerbated systemic inammation and endothelial damage compared to ZSF1-lean littermates. In addition, ZSF1-obese rats accumulated immune and endothelial senescent cells in the peripheral blood and myocardium. Accordingly, the frequency of circulating senescent leukocytes associated with markers of disease severity in HFpEF patients. Notably, systemic treatment of ZSF1-obese rats with Navitoclax, a BCL-2 family inhibitor, reduced senescent cell burden, decreased circulating B-type natriuretic peptide levels, and attenuated inammation, vascular remodeling and cardiac brosis. Our ndings advance cellular senescence as a key mechanistic pathway leading to HFpEF and provide proof-of-concept evidence that senolytics are a promising treatment for this disease. pairwise as with Whiskers or as with ± SEM. when 0.05.


Introduction
Aging is a major risk factor for the development of heart failure with preserved ejection fraction (HFpEF), which still carries a poor prognosis, and outcome-modifying treatments are a large unmet clinical need 1 .
Age-associated chronic systemic in ammation (i.e. in ammaging) heightened by cardiovascular comorbidities, such as diabetes mellitus, hypertension and dyslipidemia, has been suggested as a key driver of cardiac microvascular dysfunction and matrix remodeling, both contributing to increase myocardial stiffness 2 . Accordingly, pharmacological targeting of in ammation provides clinical bene t to patients with high cardiovascular comorbidity load 3,4 . Nevertheless, upstream mechanisms driving in ammation in HFpEF remain poorly understood.
Age-associated accumulation of senescent cells (SCs) is a signi cant contributor to in ammaging and age-dependent tissue dysfunction [5][6][7] . SCs promote aging by acquiring a senescence-associated secretory phenotype (SASP) 8 which is pro-in ammatory, pro-brotic and may spread the senescence program to neighboring cells 9 . During aging, immune-mediated SCs clearance becomes impaired in part due to agerelated changes in the composition and function of immune cell subsets (i.e. immunosenescence) 10 . On the other hand, immune cells can also acquire a senescent phenotype, further contributing to the accumulation of SCs in older individuals. A link between endothelial senescence and HFpEF was recently demonstrated in a mouse model of accelerated aging 11 . However, the impact of SCs in the pathophysiology and treatment of HFpEF, in relevant pre-clinical models, remains unexplored.
Based on the hypothesis that cellular senescence may collectively contribute to the development of a systemic deleterious environment in HFpEF, anti-senescence pharmacological strategies may emerge as new therapeutic avenues. Senolytic drugs induce systemic clearance of SCs by promoting targeted apoptosis 12,13 . Among these, Nav (ABT-263), a BCL-2/BCL-XL inhibitor, has shown effectiveness in diseases marked by vascular remodeling and brosis, namely in experimental models of myocardial infarction and HF with reduced ejection fraction 12,14,15 .
In the present work, we used obese ZSF1 rats that develop an HFpEF-like phenotype between the 14th and 20th weeks of age 16,17,18 to analyze systemic and cardiac in ammation, the immune system and endothelial dysfunction at disease onset (week 15) and onwards. We then characterized the senescence program of immune, circulating endothelial and cardiac cells. In addition, the association between cell senescence with diagnostic and prognostic markers in HFpEF, such as B-type natriuretic peptide (BNP), was investigated in a patient cohort. Finally, we evaluated the therapeutic effects of the systemic administration of a pharmacological agent to eliminate SCs. The results obtained support that premature aging is a key driver of HFpEF and that senolytics hold promise for its treatment.

Results
Obese ZSF1 rats show alterations in the immune system, in ammaging and endothelial dysfunction The composition of the immune system in obese ZSF1 (ZSF1-Ob, Ob) and lean ZSF1 rats (ZSF1-Ln, Ln) is relatively unknown. Therefore, peripheral mononuclear blood cells (PBMCs) of ZSF1-Ob and Ln rats were characterized by ow cytometry from week 15 to week 28 following the gating strategy described in Extended Data Fig. 1a. ZSF1-Ob rats showed higher percentage ( Fig. 1a-b) and number (Fig. 1e) of myeloid (CD45R − TCR − CD161a − CD11bc + ) and activated natural killer (NK dim ) cells 19 (TCR − CD161a dim ) as compared to ZSF1-Ln littermates, an effect that increased with aging. In addition, ZSF1-Ob rats showed higher M1/M2 (CD40/CD163) myeloid polarization, indicative of a pro-in ammatory pro le (Extended Data Fig. 1b). These alterations were accompanied by a decrease in the percentage of T (TCR + CD161a − ) and B (CD45R) cells in ZSF1-Ob rats ( Fig. 1c-d) whereas some uctuation in the percentage of NK bright cells, NKT cells and T subpopulations was observed overtime, without a clear trend (Extended Data Fig. 1c-f). This phenotype also extended to main hematopoietic organs such as the spleen, lymph nodes and bone marrow (BM) (Extended Data Fig. 2a-f). Moreover, a decrease in the relative percentage of hematopoietic progenitors (HPC) (Fig. 1f) in BM of ZSF1-Ob rats was accompanied with a higher capacity of these cells to differentiate into the granulocytic-monocytic lineage (Extended Data Fig. 2g-h).
Due to the pro-in ammatory nature of myeloid cells, we further investigated plasma cytokine and chemokine pro les. Our results indicate signi cantly higher levels of pro-in ammatory cytokines and chemokines in the plasma of ZSF1-Ob than ZSF1-Ln rats such as MIP-1A, IL-18 and IP-10 at week 22 and IL-1α, MIP-1A, IL-6, IL-18, MCP-1 and TNF-α at week 26 (Fig. 1g). We then evaluated the impact of this proin ammatory milieu in cardiac endothelial cells, a driver of myocardial dysfunction in HFpEF 2,20 . For this purpose, we exposed human cardiac microvascular endothelial cells (MVECs) to plasma (Fig. 1h) or conditioned medium collected from bone marrow macrophages (CM BM-Mac) (Extended Data Fig. 2i).
Remarkably, the plasma of ZSF1-Ob rats was able to activate MVECs, as demonstrated by an increased percentage of ICAM-I + and VCAM-I + cells (Fig. 1i). This activation effect was less pronounced in response to CM BM-Mac (Extended Data Fig. 2j). In addition, a higher percentage of activated MVECs were positive for the conventional apoptotic marker annexin V, indicating that these cells entered in an apoptotic program (Fig. 1j). Strengthening the translational relevance of these ndings, human HFpEF patient plasma also induced MVECs activation, as demonstrated by higher percentage of P-selectin + MVECS in response to control patient's plasma ( Fig. 1k-l). Of note, circulating endothelial cells (CECs; CD31 + CD45 + ), recognized as markers of endothelial injury 21 , were increased in the blood of ZSF1-Ob than of ZSF1-Ln rats from week 18 onwards ( Fig. 1m-n). At week 28, the number of CECs in the blood of ZSF1-Ob decreased relatively to week 22 (Fig. 1n), which is likely due to the higher number of apoptotic CECs at week 28 than 22 (Fig. 1o).
Altogether, our results suggest that alterations in the genesis and production of main immune cell populations occur earlier in ZSF1-Ob animals and contribute to the establishment of a systemic deleterious pro-in ammatory environment that promote endothelial alterations.

Hearts from ZSF1-Ob rats show in ltration of immune cells, endothelial dysfunction and brosis
The non-cardiomyocytic cellular compartment of ZSF1-Ob and ZSF1-Ln ventricles was characterized by ow cytometry at weeks 22 and 28, according to the gating strategy presented in the Extended Data Fig. 3a. The level of in ltrating immune cells (CD45 + ) in the heart of ZSF1-Ob rats was higher than ZSF1-Ln at week 22 (Fig. 2a) and was mainly composed of myeloid cells (CD45 + CD11bc + ) (Fig. 2b). These results were supported by immuno uorescence in myocardial sections, which showed the presence of signi cantly higher percentage of macrophages (CD68 + ) in the heart of ZSF1-Ob rats than of ZSF1-Ln animals ( Fig. 2c-d). Alongside immune in ltration, endothelial cell activation (ICAM-1 + and VCAM-1 + cells) was augmented in the hearts of ZSF1-Ob rats ( Fig. 2e-f). Of note, a reduction in the frequency of cardiac endothelial cells (CD45 − CD31 + ) was observed in ZSF1-Ob rats from week 22 to 28 (Fig. 2g), resulting in poor microvascular density compared to ZSF1-Ln rats ( Fig. 2h-i). Together with these alterations, broblasts (CD140a + ) frequency was increased in ZSF1-Ob hearts (Fig. 2j), as well as the expression of brosis-associated genes in whole heart tissue compared to ZSF1-Ln (Fig. 2k).
Taken together, these ndings support that immune in ltration and endothelial activation precede the development of vascular rarefaction and brosis in ZSF1-Ob rats.
CECs present in peripheral blood of ZSF1-Ob (Fig. 3h), as well as cardiac cells isolated from the heart of ZSF1-Ob rats at week 28 ( Fig. 3i), showed higher cellular senescence than counterparts collected from ZSF1-Ln rats, as con rmed by SA-β-gal staining. In the heart, SA-β-gal + cells were more abundant in immune and endothelial compartments (Fig. 3j) at 20 and 28 weeks. Importantly, cardiac cells from ZSF1-Ob rats expressed higher levels of senescence markers p21 and p53 at week 28 than ZSF1-Ln rats ( Fig. 3k), as well as increased levels of SASP-associated molecules such as MCP1, IL-6 and TGF-β (Fig. 3l). Of note, no differences were found in SA-β-gal levels in cells collected from BM, spleen and subcutaneous fat (Extended Data Fig. 3b-d).
Overall, our results demonstrate the early onset of a cellular senescence program in peripheral blood of ZSF1-Ob animals (at week 15 for immune cells and week 20 for CECs), that extended speci cally to the heart (immune and endothelial compartments from week 20 onwards).

Circulating senescent cells associate with cardiac overload surrogates and key drivers of human HFpEF
To assess the translation potential of our ndings, we investigated the senescence program in PBMCs collected from HFpEF patients and comorbidity-matched control patients (Extended Data Table 1). Interestingly, no signi cant association was found between the burden of circulating senescent (SA-β-gal) cells and age. However, despite similar percentage of SCs in both groups (Fig. 4a), higher HFpEF severity (NYHA III) was associated with increased SCs fraction (Fig. 4b). Accordingly, the number of SCs correlated with BNP levels (Fig. 4c) and pulmonary artery systolic pressure (PASP) (Fig. 4d), key analytical and clinical surrogates of cardiac overload, respectively. Of note, clinically relevant biomarkers of in ammation (high-sensitivity C-reactive protein, hsCRP, and monocytosis), glycemia/diabetes control (glycated hemoglobin, HbA1c) and body mass index (BMI) could be found among explanatory variables of SCs variation (Fig. 4e). Altogether, our results suggest that SCs burden is correlated with disease severity and is highly modulated by major risk factors for developing HFpEF.

Nav-treated ZSF1-Ob rats display lower levels of circulating BNP, in ammation and immunosenescence
Previous studies have shown that Nav, a potent small-molecule inhibitor of BCL family, improved the recovery after myocardial infarction 14 , as well as in ammation and ejection fraction in a model of HF with reduced EF 12 , through systemic elimination of SCs. Therefore, we investigated whether clearance of senescence cells by Nav (i.e. by inhibition of the anti-apoptotic pathway BCL-XL upregulated in ZSF1-Ob rats at week 18) ( Fig. 3e-f) could alleviate the disease progression in ZSF1-Ob rats. Upon the establishment and characterization of the HFpEF phenotype, 18-week-old ZSF1-Ob rats were treated with Nav (50 mg/Kg/day) or vehicle (DMSO + PEG-300) in a 2-cycle administration regimen, with 2 weeks of interval (Fig. 5a). Remarkably, treatment of ZSF1-Ob rats with Nav decreased systemic in ammation ( Fig. 5b) at week 26, reducing the levels of several in ammatory cytokines, and BNP levels at weeks 22 and 26, a marker of cardiac overload and key prognosticator in HF patients 24,25 (Fig. 5c), relative to nontreated ZSF1-Ob rats. Although the Nav treatment in ZSF1-Ob rats had no statistical effect in the body weight and exercise intolerance (Extended Data Table 2), high metabolic risk features such as perirenal fat, hyperglycemia, total cholesterol and circulating urea were alleviated with Nav treatment (Extended Data Table 2). Importantly, at systemic level, Nav treatment decreased the senescence program in PBMCs of ZSF1-Ob rats (week 28), as demonstrated by a decrease in SA-β-gal + cells (Fig. 5d), in the expression of senescence-related gene p21 and of the SASP molecules TGF-β and iNOS (Fig. 5e), and a tendency to decrease BCL-XL/BAX ratio, suggesting increased apoptosis of senescent PBMCs after Nav treatment ( Fig. 5f-g).
Treatment of ZSF1-Ob rats with Nav decreased the percentage of pro-in ammatory cellular orchestrators, particularly myeloid cells ( Fig. 5h-i), and increased the percentage of T cells in PBMCs (Fig. 5j), while no differences were detected at the level of B cells (Fig. 5k). In the BM, Nav treatment restored the proportion of myeloid cells, whereas other immune subsets and hematopoietic progenitors were not affected (Extended Data Fig. 4a-d).
Taken together, our results show that Nav promoted clearance of SCs, decreased systemic in ammation and BNP levels, and partially rescued the composition of the immune system.

Treatment of ZSF1-Ob rats with Nav decreased heart cell senescence, brosis and endothelial dysfunction
A recent study has shown that Nav is able to improve vascular and myocardial function in aged hearts through the elimination of SCs 1 . However, the therapeutic potential of Nav remains to be determined in the context of HFpEF. Echocardiographic analyses at week 28 con rmed that the Nav treatment in ZSF1-Ob rats did not affect cardiac function (Table 1). ZSF1-Ob rats treated with Nav showed lower levels of SA-β-gal in cardiac cells (Fig. 6a), particularly in immune and endothelial cells, as compared to nontreated ZSF1-Ob rats (Fig. 6b), decreased expression of p21 and BNP (Fig. 6c), reduced SASP markers, namely TGF-β, TNF-α and iNOS (Fig. 6c), and a non-statistical decrease in the percentage of cells with DNA damage (Extended Data Fig. 4e-f).
Next, the impact of Nav treatment in myocardial brosis and immune cell in ltration, cardiomyocyte physiology and cardiac endothelial cell number and activation was investigated at week 28. The Nav treatment reduced the frequency of myocardial macrophages (Fig. 6d-e and Extended Data Fig. 4g) and broblasts (Fig. 6f), and a decrease in brosis as con rmed by picrosirius red staining (Fig. 6g-h and Extended Data Fig. 4h) and brosis-associated genes, namely of collagen type 3 and alpha smooth muscle actin (α-SMA) (Fig. 6i). In addition, Nav treatment of ZSF1-Ob rats had no impact in cardiomyocyte hypertrophy (assessed by cross-sectional area), in agreement with the LV mass estimation by echocardiography (Extended Data Figs. 4i-j). The proportion of cardiac endothelial cells was not signi cantly increased (Fig. 6j) but their activation pro le was improved (Fig. 6k), as was the myocardial microvasculature density, assessed by Isolectin B4 (Fig. 6i-m and Extended Data Fig. 4k).
Altogether, our results demonstrate that ZSF1-Ob rats treated with Nav showed decreased cardiac overload, cardiac cell senescence, immunosenescence, brosis and endothelial activation.
Treatment of ZSF1-Ob rats with Nav decreased circulating endothelial cells, vascular cell senescence and hypertrophy We next investigated whether vascular remodelling was extended to the peripheral system, besides the myocardium. Nav treatment abrogated the increase in frequency of CECs in the peripheral blood of ZSF1-Ob rats, when compared to ZSF1-Ln animals (Fig. 7a). Nav treatment also signi cantly decreased the frequency of SA-β-gal + (Fig. 7b) and Annexin V + cells (Fig. 7c), and reduced p21 levels in the aorta of obese rats (Fig. 7d-e), as con rmed by RNA-ISH analysis. In addition, ZSF1-Ob rats treated with Nav showed decreased cross-sectional area and a trend of lower media wall thickness and larger lumen area, which indicates that the treatment was able to attenuate hypertension-induced vascular remodeling In summary, our results demonstrated that Nav treatment of ZSF1-Ob rats restored vascular compartment by decreasing the frequency of SCs in vascular endothelium and ameliorating aortic remodelling.

Discussion
Limited success has been achieved so far in improving HFpEF outcomes 26,27,28 . One of the main di culties in tackling this problem stems from an incomplete understanding of upstream mechanisms leading to early disease manifestations, including systemic in ammation and endothelial dysfunction, important triggers of structural and functional cardiac alterations. Here, we show that immune dysregulation and accumulation of circulating and myocardial SCs may contribute to the establishment of systemic in ammation and endothelial dysfunction in a pre-clinical animal model of HFpEF. We further show that higher levels of SCs in patients associate with major risk factors and prognosticators in HFpEF, specially with cardiac overload surrogates. Moreover, circulating SCs burden positively associates with hsCRP and HbA1c, important biomarkers of in ammation and glycemic control, whereas a negative correlation was showed for BMI. Curiously, mildly elevated BMI levels are known to be protective regarding mortality in HFpEF (obesity paradox) by mechanisms yet to be clari ed 29 , but among which senescence can potentially play a role.
Importantly, the treatment of ZSF-1-Ob rats with the senolytic compound Nav decreased the number of SCs, systemic in ammation, endothelial dysfunction and circulating BNP levels, as well as rejuvenated the immune system, decreased cardiac brosis and vascular remodeling. These ndings support cellular senescence as a likely driver of HFpEF and indicate senolytics as a potential pharmacological alternative for this prevalent syndrome.
Age-associated immune dysfunction and accumulation of SCs in blood and heart is observed at the onset of HFpEF disease. ZSF1-Ob animals display an increase in the frequency and absolute numbers of circulating myeloid cells, displaying a pro-in ammatory and senescent phenotype, and of mature (NK dim ) cells. Similar alterations have been described in HFpEF patients, namely the increase in myocardial tissue macrophage density and number of circulating pro-in ammatory monocytes 30,31 . Herein, alterations in immune subsets of ZSF1 obese rats were accompanied with the emergence of SCs in the immunemyocardium-endothelium axis, while other organs, like fat and spleen, failed to show features of cellular senescence. Although endothelial cell senescence has been previously described in an aging mouse model of HFpEF 11  Despite the known role of in ammation on HFpEF pathophysiology and promising data from trials 3,4 , anti-in ammatory therapies have not yet been included in the disease management routine care. Herein, we show that the systemic treatment of a preclinical model of HFpEF with a senolytic drug 37,38 signi cantly decreased age-related in ammation, endothelial dysfunction, BNP levels and cardiac brosis. To the best of our knowledge, this is the rst study to use an anti-senescence therapy to treat HFpEF. Nav was selected because it targets the BCL family of proteins, which were found to be dysregulated in ZSF1 obese animals, and has previously shown e cacy in the elimination of SCs from the main systems affected by the senescence program in the ZSF1 obese, i.e. the endothelium 39 , the hematopoietic system 34,40 and the heart 14 . The selective depletion of SCs stimulated immune recovery towards a physiological and less in ammatory phenotype. A similar outcome has been previously reported following Nav administration to sub-lethally irradiated or normally aged mice, which effectively cleared senescent hematopoietic bone marrow stem cells 40 . Conversely, the bone marrow of ZSF1 obese showed no signs of SCs accumulation, despite having an aged phenotype associated with a myeloid skewing in differentiation.
In summary, our study provides evidence for the etiology of HFpEF indicating that immunosenescence and accumulation of SCs in the immune-myocardium-endothelium axis are major generators of in ammaging and endothelial dysfunction, and offers a proof-of-principle that systemic clearance of SCs attenuates HFpEF phenotype.

Methods
Detailed protocols are also available in Supplementary Information.

Organ collection and blood processing
During sacri ce, organs were weighed, washed in PBS and immediately frozen for subsequent analysis. Blood was collected from subclavian artery in heparinized tubes and centrifuged at 1200 g to isolate plasma and buffy coat. Plasma was then centrifuged for 30 min at 4°C. To isolate peripheral mononuclear cells (PBMCs), buffy coat was overlaid in lymproprep (Axis-Shield) accordingly to manufacture instructions. Single cell suspensions from the heart were obtained by tissue dissociation with 600 U/mL of collagenase II and 60 U/mL of DNase I using the GentleMACs dissociator (Miltenyi Biotec) as reported elsewhere 43 . Red blood cells (RBCs) were removed from cell suspensions using RBC lysis buffer (Supplementary Table 4).

Senolytic treatment
At the 18th week, ZSF1 Ob rats were randomly allocated to receive 50 mg/Kg/day Nav (Selleckchem; n = 8) or vehicle (10% DMSO + 90 % PEG-300; n = 8) by oral gavage. ZSF1 Lean rats were not exposed to any treatment during the study. Administration followed a 2 cycle-regime of 7 days daily administration, with two weeks of interval. Weight gain was recorded every week along the study and every day prior to administration of Nav (or vehicle). At the 28th week, animals underwent cardiorespiratory capacity evaluation and echocardiography, prior to sacri ce.

Flow Cytometry
Single-cell suspensions were incubated with anti-rat CD32 in FACS medium during 15 min and subsequently incubated with mix of conjugated antibodies during 20 min, 4 ºC. When apoptosis was evaluated, a subsequent step was performed by incubating cell suspension with annexin-V during 15 min (ThermoFisher). When SA-β-gal was assessed, cells were resuspended in staining medium and the assay was done using Fluorescein di-β-D-galactopyranoside (ThermoFisher), accordingly with manufacturer's instructions. Nonviable cells were detected with 0.5 % propidium iodide (Sigma-Aldrich). For quantitative ow cytometry, precision count beads (Biolegend) were added to cellular suspension (1:4 ratio). Samples were acquired in the cytometer FACS Canto II (BD Biosciences) or in Accuri C6 (BD Biosciences). Data analysis was performed in FlowJo software.

Plasma proteomics
For cytokine and chemokine measurements, plasma-heparin samples from ZSF1 rats with 22 and 26 weeks were analysed at Eve Technologies for the relative quanti cation of a large array of SASP factors using Rat Cytokine Array/Chemokine Array 27 Plex, RD27.
Immunoblotting Samples were homogenized in RIPA buffer and protein concentration was quanti ed using the Bradford assay (Bio-Rad). Proteins (25-60 µg) were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) with a 10-15 % polyacrylamide gel and then electroblotted onto nitrocellulose membranes (Bio-Rad). Blots were blocked in 5 % bovine serum albumin (BSA) in TBS, and then incubated with primary antibodies to BCL-2 (Cell Signaling, #2876), BCL-XL (Cell Signaling, #2764), BAX (Cell Signaling, #2772) and GAPDH (Abcam, #ab8245) in TBS-T (0.1 % Tween-20). Immunoblots were washed and incubated with secondary antibodies coupled to 700 nm or 800 nm infra-red dye in 0.5 % non-fat dry milk in TBS-T. Membrane was washed and imaged with Odyssey Infrared Imaging System (LICOR Biosciences). GAPDH was used as internal control and the control group was set as reference.
To decrease auto uorescence, Vector TrueView™ Auto uorescence Quenching Kit (Vector Laboratories) was used.
For isolectin, pH2AX, CD68 immunolabelling, 10 µm or 30 µm cryo-sections from the heart of ZSF1 rats were used. Sections were cut using cryomicrotome (Microtom HM 550, Thermo Scienti c), processed and stained as described elsewhere 3 . Nuclei was stained with DAPI (10 µg/mL) and mounted in Fluoroshield mounting media. Images were obtained in a high-content uorescence microscope (IN Cell 2200) or LSM 710 confocal microscope. Quanti cations were performed using ImageJ software.

Immunohistochemistry
Descending aortas from rats were collected, formalin-xed para n-embedded. Sections of 4 µm were used for haematoxylin eosin (H&E), picrosirius red (SR) and Verhoeff-Van Gieson (VVG) staining, according to manufacturer speci cations. H&E allowed aortic measurements by de ning a region-ofinterest. For collagen and elastin quanti cation, SR or VVG staining's were used, respectively, and a threshold applied to identify collagen or elastin bers. Elastin fragmentation was de ned as the number of visible breaks of continuous elastin bres. Left ventricle from ZSF1 rats was also collected for detection of collagen content (SR staining). Images were acquired under Zeiss Axio Scan.Z1 slide scanner and quanti cations performed with QuPath software.

RNA In Situ Hybridization (RNA-ISH)
RNA-ISH in aortic rings was performed with RNAScope® Multiplex Fluorescent V2 Assay (Advanced Cell Diagnostics), with minor modi cations to manufacturer protocol. Images were acquired using confocal microscope (LSM 710). Between 3 to 5 aortic rings were analysed per animal. Percentage of p21 + cells was assessed based on the number of cells with ≥ 1 dot/cell. Quanti cations were performed in QuPath software.

Endothelial Activation Assay
Human cardiac microvascular endothelial cells (MVECs; #7130, Lonza) were seeded at density of 5000 cells/cm 2 in EGM-2MV microvascular endothelial cell growth medium BulletKit (EGM; Lonza). When cells reached 75-80 % con uency, medium was changed to the following conditions: 1) EBM supplemented with 2 % plasma of rats without lipidic phase; 2) EBM supplemented with 10 % of CM BM-MAC of rats; 3) EBM supplemented with 10 % of plasma of HFpEF or control patients. After 24 h, cells were detached with trypsin/EDTA and MVECs activation was assessed by ow cytometry.

Statistical Analysis
Statistical testing was performed using GraphPad® Prism 8.0 Software. Kolmogorov-Smirnov normality test was used to evaluate normal distribution of data. Normally distributed data was tested with independent sample Student's t test and one-way ANOVA (Bonferroni´s post hoc test) for two or three groups, respectively. Outliers were excluded by ROUT analysis. Non-normally distributed data was tested with Mann-Whitney U test and Kruskal-Wallis one-way ANOVA for two or three groups, respectively. Data with two independent variables were compared with two-way ANOVA followed by pairwise comparison using Bonferroni's post-hoc test. Results are presented as Box Plot with min/max Whiskers or as column bars with Mean ± SEM. Differences between groups were considered signi cant when p < 0.05.
The association between circulating SCs and markers of cardiac overload (BNP and PASP) was assessed using a χ2 test for linear trend. To optimize the statistical analysis model, variables with skewed distribution, namely BNP, PASP and hsCRP, were transformed to their natural logarithm. A linear regression model was used to assess the association of SC percentage (dependent variable) with biomarkers of in ammation (hsCRP and monocytes), glycemic control (HbA1c) and BMI (independent

Competing interests
Andreia Silva is currently an employee of AstraZeneca plc.

Figure 6
Treatment of ZSF1-Ob rats with Nav decreased heart cell senescence, brosis and endothelial dysfunction. a. Percentage of SA-β-gal+ cells in the heart and b. in different non-cardiomyocyte populations from Ln, Ob and Ob (Nav) rats with 28 weeks (n ≥ 7/group). c. Transcriptomic pro le of the heart for senescence markers, SASP and BNP at 28 weeks (n ≥ 6/group), presented as a fold change relative to Ob rats. d. Representative images and e. respective quanti cation of macrophages (CD68; red) in Ln, Ob and Ob (Nav) hearts at 28 weeks (n=8/group). Scale bars 25 µm. f. Percentage of broblasts (CD45-CD31-PDGFRα+) in the heart at 28 weeks (n ≥ 6/group). g. Representative images and h.
respective quanti cation of picrosirius red at 28 weeks (n=8/group). Scale bars 50 μm. i. Expression of broblast activation-associated genes in the myocardium at 28 weeks (n=5/group) presented as ratio to the levels of Ob rats. j. Percentage of endothelial cells (CD31+CD45-) of total cardiac cells isolated at 28 weeks (n=7/group). k. Percentage of cardiac endothelial cells expressing ICAM-1 and VCAM-1 (n ≥ 6/group). l. Representative images and m. respective quanti cation of Isolectin B4 (red) staining in the heart at 28 weeks. Scale bar 50 μm. Values are presented as mean ± SEM or as box and whiskers with min to max values. Treatment of ZSF1-Ob rats with Nav decreased circulating endothelial cells, vascular cell senescence and hypertrophy. a. Proportion of circulating endothelial cells (CECs; CD31+CD45-) at 22 weeks (n > 7/group).
b. Percentage of SA-β-gal+ cells in CECs at 28 weeks (n > 7/group). c. Apoptotic cells (Annexin V+) in CECs at 28 weeks (n>7/group). d. Representative images and e. respective quanti cation of IHC RNA-ISH staining for p21 (red) at 28 weeks in aortic rings (n=6/group). Scale bar 20 µm. f. Representative images of aortic wall sections stained with haematoxylin eosin (H&E) and Verhoeff's Van Gieson (VVG). Scale bar 50 µm. g-i. Quanti cation of morphological parameters of the aortic wall by H&E. Region-of-interest was de ned in the media aortic wall (continuous red line) of the lumen (n>7/group). j. Quanti cation of elastin fragmentation in the aortic wall (yellow arrows) (n>7/group