Abstract

   Heart failure with preserved ejection fraction (HFpEF) is increasingly
   common but its pathogenesis is poorly understood. The ability to assess
   genetic and pharmacologic interventions is hampered by the lack of
   robust preclinical mouse models of HFpEF. We developed a novel
   “two-hit” model, which combines obesity and insulin resistance with
   chronic pressure overload to recapitulate clinical features of HFpEF.
   C57Bl6/NJ mice fed a high-fat diet (HFD) for > 10 weeks were
   administered an AAV8-driven vector resulting in constitutive
   overexpression of mouse Renin1d. HFD-Renin (aka “HFpEF”) mice
   demonstrated obesity and insulin resistance, moderate left ventricular
   hypertrophy, preserved systolic function, and diastolic dysfunction
   indicated by echocardiographic measurements; increased left atrial
   mass; elevated natriuretic peptides; and exercise intolerance.
   Transcriptomic and metabolomic profiling of HFD-Renin myocardium
   demonstrated upregulation of pro-fibrotic pathways and downregulation
   of metabolic pathways, in particular branched chain amino acid
   catabolism, similar to human HFpEF. Treatment with empagliflozin, an
   effective but incompletely understood HFpEF therapy, improved multiple
   endpoints. The HFD-Renin mouse model recapitulates key features of
   human HFpEF and will enable studies dissecting the contribution of
   individual pathogenic drivers to this complex syndrome. Additional
   preclinical HFpEF models allow for orthogonal studies to increase
   validity in assessment of interventions.

Supplementary Information

   The online version contains supplementary material available at
   10.1038/s41598-024-84515-9.

   Keywords: Cardiovascular disease, Heart failure with preserved ejection
   fraction, Mouse models, Diastolic dysfunction

   Subject terms: Heart failure, Obesity, Hypertension, Molecular
   medicine, Experimental models of disease

Introduction

   Heart failure with preserved ejection fraction (HFpEF) is a complex
   disease, and despite many advances in research, its underlying
   mechanisms are not yet fully understood^[59]1. This is due, in part, to
   the lack of pre-clinical small animal models that faithfully
   recapitulate the human condition. There is no consensus on the required
   elements to define rodent HFpEF^[60]2–[61]5. Consequently, current
   rodent models have notable downsides, including significant lead time
   hindering rapid experimentation^[62]6, need for surgical
   expertise^[63]7,[64]8, predominant hypertensive phenotype without
   metabolic dysregulation^[65]9, or inability to use genetic
   manipulation, as in the ZSF-1 rat or leptin signaling mutant
   models^[66]10–[67]13. Additionally, the best characterized mouse model
   to date using L-NAME and high-fat diet (HFD)^[68]14, induces
   endothelial dysfunction using an artificial, pharmacological inhibitor
   of vascular NO signaling and some, including our lab, have shown that
   this model is not consistently reproducible^[69]15,[70]16.

   We sought to devise a complementary “two-hit” model of mouse HFpEF that
   combines the most common human drivers of the metabolic phenogroup of
   human HFpEF, diet-induced obesity (DIO), and chronic hypertension.
   Among HFpEF patients, the phenogroups with obesity/diabetes, abnormal
   systemic metabolism, liver and renal dysfunction, and high renin level
   exhibit the worst overall prognosis and lowest survival
   probability^[71]17,[72]18. Our goal was to establish a mouse model that
   exhibits both cardiac and extra-cardiac manifestations of HFpEF and
   meets the following criteria: (1) left ventricular (LV) hypertrophy
   (LVH) driven by a pathogenic driver relevant to the human condition,
   (2) obesity and insulin resistance, (3) LV diastolic dysfunction, (4)
   preserved LV systolic function, and (5) exercise intolerance. We also
   sought to develop a model with relatively balanced contributions of
   increased peripheral vascular resistance and obesity with insulin
   resistance. The model described here, termed “HFD-Renin” mice, involves
   modest, chronic hypertension achieved by adeno-associated virus
   overexpression of renin (AAV8-Renin) to activate the
   renin-angiotensin-aldosterone system (RAAS) in the setting of
   diet-induced obesity producing a modest but significant LVH, LV
   diastolic dysfunction with stable systolic function, exercise
   intolerance, and obesity-related glucose intolerance and insulin
   resistance. This model utilizes common pathogenic drivers resulting in
   features of human HFpEF and allows for assessment of the relative
   contribution of each driver independently^[73]18,[74]19. Given concerns
   of reproducibility with prior published models, we developed and
   conducted complementary studies at two sites in a coordinated manner to
   verify reproducibility. Renin was chosen as a tool for inducing modest
   renovascular hypertension, a critical driver of HFpEF. The relevance of
   this approach relates to the observation that the RAAS system is
   activated in most pathophysiological states that contribute to HFpEF.
   In contrast, L-NAME is an artificial pharmacologic intervention.
   Admittedly, AAV-mediated delivery of renin is not physiologic. No small
   animal model will be a perfect representation of this complex human
   disease, including the one presented here. However, the availability of
   multiple verified HFpEF models will be valuable to the field, allowing
   for more rigorous assessment of therapeutic interventions via
   orthogonal studies.

Results

Development and optimization of AAV-mediated renin delivery to achieve
hypertension-related LV hypertrophy and remodeling in mice

   We sought to use an AAV8 delivery system to achieve a mild degree of
   hypertension in mice. An AAV-Renin vector (details of the construct
   provided in Methods) or null backbone vector (control) were
   administered in order to constitutively express a cleavable form of
   renin in the liver^[75]20. Dose-ranging studies demonstrated a
   dose-dependent increase in circulating mouse renin concentration
   (Supplementary Fig. [76]S1A). No change in body weight was observed
   during the six-week study period (Supplementary Fig. [77]S1B).
   Implantable telemetry testing revealed a dose-dependent, modest
   (20-40mmHg) increase in systolic blood pressure (Supplementary Fig.
   [78]S1C), which was amenable to treatment with standard of care
   angiotensin converting enzyme (ACE) inhibitor enalapril. Consistent
   with the hypertensive state, increased normalized biventricular heart
   weight, left atrial enlargement, and modest lung congestion developed
   in the AAV-renin-treated mice (Supplementary Fig. [79]S1D). Gene
   markers of LV remodeling were also induced, including a decrease in the
   myosin heavy chain 6 (Myh6)/Myh7 ratio (Supplementary Fig. [80]S1E).

   The dose-ranging studies were repeated in HFD-treated mice. Mice were
   started on HFD at ~ 8 weeks of age prior to AAV administration at ~ 20
   weeks of age. Dose-dependent LVH was achieved in the context, but
   required higher viral doses presumably due to presumed decreased viral
   transduction in a fatty liver (Supplementary Fig. [81]S1F). Notably, at
   the highest doses, significant loss of body weight was observed
   (Supplementary Fig. [82]S1G). The dose-ranging studies allowed for the
   selection of similar, though not identical, doses at both institutions
   (lower for chow-fed animals, higher for HFD-fed animals) that resulted
   in the optimal 2-3x increase in circulating renin and LVH with minimal
   perturbance in body weight. Potential effects of viral-induced
   inflammation were controlled for by using a null virus in the control
   arms. Plasma aspartate aminotransferase (AST) and alanine
   aminotransferase (ALT) were measured 6 weeks post-AAV8 injection
   (Supplementary Fig. [83]S1H) and demonstrated no elevation in the
   chow-fed animals, and mild elevation (2-3x upper limit of normal) in
   HFD null and HFD-Renin conditions, suggesting that the transaminitis is
   a result of diet-induced steatohepatitis.

The HFD-Renin model results in features of HFpEF, including systemic
metabolic derangements, LV hypertrophy, diastolic dysfunction, and exercise
intolerance

   Having established that AAV-Renin produces modest hypertension and
   cardiac hypertrophic remodeling, this effect was overlayed on HFD-fed
   animals. For the HFpEF model, 8- to 10-week-old mice fed chow (control)
   or HFD for 10 weeks were injected with AAV-Renin or null vector
   (control) (Fig. [84]1A). Renin dosing was calibrated to achieve a 2-3x
   increased in circulating protein levels (Fig. [85]1B). HFD feeding
   reliably resulted in obesity, though AAV-Renin dosing at week 10 caused
   a modest flattening in weight gain (Fig. [86]1C). The obese mice had
   impaired glucose tolerance and insulin resistance (Fig. [87]1D and
   Supplementary Fig. S2A). This corresponded with fasting hyperglycemia,
   hyperinsulinemia, and hypercholesterolemia (Supplementary Fig. S2B).
   There were no observed metabolic effects due to AAV-Renin (see
   chow + Renin cohort).

Fig. 1.

   [88]Fig. 1
   [89]Open in a new tab

   Novel “two-hit” HFpEF mouse model driven by renin overexpression and
   HFD. (A) Schema of HFD-Renin model, where C57BL/6NJ mice are
   preconditioned on HFD for 10 weeks prior to injection of AAV8-Renin
   with endpoints at week 16–18. Created with Biorender.com. (B) Serum
   renin levels one week post-injection. (C) Weight trend (left) and
   terminal weights (right). (D) Glucose tolerance test with associated
   area-under-curve (right) in four-hour fasted male littermate mice. (E)
   Echocardiographic measures at 16 weeks: LV ejection fraction (EF) and
   interventricular septal thickness in diastole (IVSd). (F) Tibia
   length-normalized gross biventricular weight (BiV/TL). (G)
   Representative m-mode echocardiograms, red bar notes LV posterior wall
   dimension in diastole. N = 10–15 per conditions. Data displayed as
   mean ± SEM. Statistical analyses used one-way analysis of variance
   (ANOVA) followed by multiple-comparisons test. P values < 0.05
   displayed on graphs.

   All cohorts had stable LV systolic function at six weeks post-injection
   (Fig. [90]1E and Supplementary Table S2). AAV-Renin resulted in 30–40%
   increase in LV intraventricular septal wall thickness (IVSd) and LV
   mass index (LVMI) as determined by echocardiography (Fig. [91]1E-G, and
   Supplementary Table S2). Normalized ventricular weights (biventricular
   weight to tibia length, BiV/TL) similarly increased, with an additive
   effect between HFD and renin expression (Fig. [92]1E). Despite the LVH,
   there was no significant change in ventricular volumes (Supplementary
   Table S2).

   To assess diastolic function, multiple complimentary endpoints were
   utilized. HFD or AAV-Renin alone did not produce significant changes in
   echo-based parameters. An additive effect of HFD and AAV-Renin was
   observed in mitral valve inflow and tissue doppler imaging parameters
   (increased E/e’, a trend in Tei index, no change in E/A ratios,
   Fig. [93]2A and B) as well as global longitudinal and radial strain and
   strain rates (Fig. [94]2C and Supplementary Table S2). Radial strain
   appeared to be a more sensitive marker of dysfunction in this model. As
   observed in the human HFpEF condition, the mouse left atria (LA) was
   impacted, including significantly increased in normalized LA weight
   relative to the control (Fig. [95]2D) and trended toward increased LA
   volume measured by echo (Fig. [96]2E), consistent with LV diastolic
   dysfunction. LA strain, a standard human clinical parameter, was
   reduced. Similarly, pulmonary vein doppler flow was significantly
   decreased in the HFD Renin cohort, indicative of decreased LV
   compliance (Fig. [97]2F). Expression of elevated natriuretic peptides
   Nppb (e.g., BNP, brain natriuretic peptide) and Nppa, a diagnostic
   criterion in human disease, were significantly elevated in heart tissue
   from HFD Renin animals compared to other treatment groups
   (Fig. [98]2G).

Fig. 2.

   [99]Fig. 2
   [100]Open in a new tab

   HFD-Renin HFpEF model exhibits LV diastolic dysfunction, abnormal left
   atrial (LA) function, and exercise intolerance. (A) Representative
   echocardiographic mitral valve inflow and tissue doppler images with
   (B) measurements of diastolic function including E/A and E/e’ ratios,
   and Tei index (n = 5–15). (C) Echocardiographic left ventricular global
   longitudinal and radial strain (n = 4–8). (D) Tibia length-normalized
   LA gross weight. Representative (E) LA minimum volume and reservoir
   strain, and (F) Doppler echo-measured pulmonary vein flows during
   systole (S) and diastole (D). (G) QT-PCR from bulk ventricle RNA for
   indicated genes. (H) Treadmill exercise testing to exhaustion. Data
   displayed as mean ± SEM. One-way analysis of variance (ANOVA) followed
   by multiple-comparisons test. P values < 0.05 displayed on graphs.

   Exercise intolerance is the most consistent symptom in patients with
   HFpEF, though the etiology is debated^[101]21,[102]22. Mice underwent
   exercise testing to exhaustion to determine the combinatorial effects
   of DIO and hypertension. HFD Renin animals displayed greatly diminished
   exercise tolerance in both time and distance to exhaustion as well as
   total work, which adjusts for body mass differences (Fig. [103]2H).
   Chow AAV-Renin mice did not perform differently from control mice.
   Accordingly, the dominant effect was due to HFD, as has been
   demonstrated in prior studies^[104]7.

   HFpEF is more prevalent in human females, yet as a group, they are
   relatively understudied in clinical trials^[105]23. We tested the
   cumulative effects of HFD and AAV-Renin in a cohort of female mice
   using the same protocol as that for males (Supplementary Fig. S3A and
   S3B). Female mice fed HFD became obese, but to a lesser extent than
   male mice, consistent with the pre-clinical rodent
   literature^[106]24–[107]26. The HFD-fed groups developed insulin
   resistance (data not shown). Renin-treated groups had preserved
   systolic function (Supplementary Fig. S3C). Both HFD and AAV-Renin
   resulted in mild LVH with an additive degree (20–30%) in the HFD-Renin
   cohort (Supplementary Fig. S3D). Diastolic parameters were mixed: Echo
   parameters for diastolic dysfunction, including E/A, E/e’, and
   isovolumic relaxation time (IVRT), were unchanged in the HFD-Renin
   group compared to controls (Supplementary Fig. S3E). However, decreased
   pulmonary vein doppler flow and LA enlargement, consistent with
   decreased ventricular compliance, were observed in the HFD Renin group
   (Supplementary Fig. S3F and S3G). Lastly, female HFD-Renin mice had
   decreased exercise performance in an additive manner (Supplementary
   Fig. S3H). Taken together the HFpEF phenotype driven by HFD-Renin was
   milder in female mice.

Two-hit HFpEF model results in cardiac fibrosis

   Cardiac fibrosis is a hallmark of human HFpEF, and variably
   demonstrated in current models^[108]6,[109]9,[110]14. Analysis of
   histologic sections of LV free wall myocardium stained with picrosirius
   red demonstrated a trend toward increased collagen deposition in the
   HFD-Renin group (Fig. [111]3A-B). Perivascular changes were more
   apparent than interstitial fibrosis, similar to the known effects of
   angiotensin II in rodents^[112]9. These observations correlated with
   increased expression of markers associated with fibrosis, fibroblast
   activation, and extracellular matrix production by QT-PCR
   (Fig. [113]3C).

Fig. 3.

   [114]Fig. 3
   [115]Open in a new tab

   HFD-Renin HFpEF results in increased cardiac fibrosis. Representative
   histology with picrosirius red and fast green counterstain of (A) whole
   heart (scale bar 3 mm) and (B) 20x magnification (scale bar 200 μm)
   demonstrating perivascular fibrosis, quantified from two separate
   regions in n = 4 independent samples. (C) QT-PCR from cardiac tissue
   for gene markers of fibrosis and myofibroblast activation, n = 6. Data
   displayed as mean ± SEM. One-way analysis of variance (ANOVA) followed
   by multiple-comparisons test. P values < 0.05 displayed on graphs.

Transcriptomic and metabolomic changes mirror observations in human HFpEF

   Targeted gene expression changes suggested cardiac remodeling largely
   predicated on fibrotic and hypertrophic pathways (Figs. [116]2H and
   [117]3C). To further understand the gene expression signature of the
   HFD-Renin model, we performed bulk RNA sequencing (RNA-seq) of LV
   samples from control, DIO, hypertensive, and HFpEF samples. Principal
   component analysis and heat mapping suggested that the null-treated
   chow and HFD groups were notably similar, whereas the chow Renin group
   was quite distinct (Fig. [118]4A and B). The addition of HFD to renin
   caused a left shift, repositioning the HFpEF group to overlay with
   control and HFD only groups. We analyzed the effects of the individual
   drivers and combined effect on differentially expressed genes (DEGs).
   DIO primarily drove downregulation of genes associated with metabolic
   pathways and cell signaling (839 downregulated DEGs with adjusted
   p-value < 0.01) (Supplementary Fig. S4A). Conversely, AAV-Renin was
   responsible for a large upregulation of genes (927 DEGs) involved with
   proliferative growth and hypertrophic signaling pathways (Supplementary
   Fig. S4B). Interestingly, several pathways moved in opposing directions
   (e.g., cAMP signaling, calcium signaling, and circadian entrapment
   downregulated by HFD and upregulated by Renin). Consequently, some of
   these pathways were not present in the gene ontology and KEGG pathway
   analyses in the HFD-Renin condition. Compared to Chow null, the
   HFD-Renin cohort resulted in a unique set of 498 upregulated and 472
   downregulated genes (Fig. [119]4C), with a separate set of
   differentially regulated pathways from either the DIO or hypertensive
   components. This suggests an additive effect, resulting in
   downregulated metabolic pathways and upregulation of ECM and cardiac
   hypertrophic growth (Fig. [120]4D). These signatures were consistent
   with published transcriptomic data from HFpEF patients^[121]19. Top
   DEGs are listed in Supplementary Figure S4C.

Fig. 4.

   [122]Fig. 4
   [123]Open in a new tab

   Cardiac transcriptomic and metabolomic profiling of HFpEF hearts shows
   strong signatures of reduced mitochondrial fuel (BCAA) and activation
   of ECM/fibrosis. (A) Principal component (PC) analysis and (B) heatmap
   of ventricular RNAseq comparing control and HFD-Renin groups (n = 5–6).
   (C) Volcano plot of differentially expressed genes in HFpEF (vs.
   control) hearts (n = 5–6). (D) KEGG pathway analysis of upregulated
   (yellow bars) and downregulated (blue bars) pathways. (E) Targeted
   metabolomics from heart tissue displaying short-chain acylcarnitine
   species (n = 5–7). Data displayed as mean ± SEM. Unpaired, 2-tailed
   Student’s t-test, with p values as displayed or * <0.05.

   Given the observed changes in mitochondrial pathways, including
   downregulation of branch chain amino acid metabolism and oxidative
   phosphorylation in the HFD-Renin state, we performed targeted
   metabolomic analysis in control and HFpEF hearts. Short-chain
   acylcarnitines, including C03-DC and C04-OH isobutryl, intermediates in
   BCAA catabolism, were decreased, consistent with the observed reduction
   in BCAA degradation pathways in the RNA-seq dataset (Fig. [124]4E).
   There were few changes in organic acids or medium- and long-chain
   acylcarnitines (Supplementary Fig. S5).

Human HFpEF standard-of-care therapy improves a subset of pathophysiologic
parameters in the HFD-Renin model

   Since the HFD-Renin model demonstrated changes in metabolism, cardiac
   remodeling, and exercise tolerance endpoints that capture critical
   translatable features in HFpEF patients, we next assessed the effects
   of the newly approved HFpEF therapy sodium glucose cotransporter 2
   inhibitor (SGLT2i) empagliflozin (empa) in this model^[125]27. Empa was
   administered concurrent with AAV injection as a preventive treatment
   intervention modality for a total of 14 weeks at the Pfizer site
   (Fig. [126]5A). Treatment with SGLT2i resulted in a slight but
   significant improvement in body weight, a trend toward reduced blood
   glucose, with significant reduction in hyperinsulinemia/HOMA-IR
   (Fig. [127]5B). No differences were observed in systolic function
   (Fig. [128]5C). There was a reduction in echocardiographic LVH and
   normalized ventricular gross weight following empa therapy
   (Fig. [129]5D). Radial strain and echocardiographic measures of
   diastolic dysfunction were largely unchanged (e.g., IVRT,
   Fig. [130]5E), but there was a trend toward diminished normalized LA
   mass, which correlated with reduced liver congestion. Exercise testing
   demonstrated significant improvement in maximal distance run prior to
   exhaustion (Fig. [131]5F). Confirmatory studies conducted in parallel
   at Penn with a shorter time course of treatment and lower dose
   demonstrated similar findings (Supplementary Fig. S6). These results
   suggest that SGLT2i may exert a positive benefit through effects on
   improved metabolic state, cardiac hypertrophic remodeling, and exercise
   tolerance and underscore the importance of a model that exhibits
   multiple pathophysiological responses to determine the impact of a
   given intervention. Importantly, this study design also demonstrated
   that the HFpEF characteristics do not degenerate into HF with reduced
   ejection fraction over > 12 weeks of HFD-Renin exposure (e.g., systolic
   function remains stable, albeit slightly lower than control animals).

Fig. 5.

   [132]Fig. 5
   [133]Open in a new tab

   SGLT2 inhibitor treatment improves HFpEF in mice. (A) Schema for empa
   treatment regimen with HFD-Renin HFpEF model conducted at two
   independent research sites. (B) Biophysical parameters of Pfizer
   cohort, n = 12–13. (C) Echocardiographic measures of systolic function
   and strain. (D) LV hypertrophy measures by echo and normalized
   biventricular weight. (E) Diastolic parameters including IVRT, atrial
   enlargement, and liver congestion. (F) Maximal exercise testing. Data
   displayed as mean ± SEM. One-way analysis of variance (ANOVA) followed
   by multiple-comparisons test. P values < 0.05 displayed on graphs. RWT
   regional wall thickness, IVRT isovolumic relaxation time, SR strain
   rate.

Discussion

   Understanding the underlying pathophysiology and identifying new
   treatment strategies for HFpEF has been hindered, in part, by a lack of
   mouse models of HFpEF that faithfully recapitulate the human disease
   state. This challenge is further complicated by the existence of
   multiple human HFpEF phenogroups necessitating multiple pre-clinical
   models. The ease of genetic manipulation in mice allows rigorous
   testing of new targets using genetic loss-of-function approaches
   complemented by pharmacologic interventions. The two most common
   drivers in human HFpEF are obesity and hypertension—each to varying
   extent—ranging from obesity with dominant hypertension to modest
   chronic hypertension/renal and vascular dysfunction. We developed a
   “two-hit” HFpEF model that combines persistent, mild hypertension
   through renin overexpression with metabolic dysfunction from DIO to
   simulate the human phenotype. A unique feature of these studies is that
   two separate facilities assessed the models in parallel, increasing
   confidence in the results and reproducibility for the model. Following
   careful dose assessment, AAV-Renin could be administered to reliably
   effect a 20-40mmHg increase in systolic blood pressure and ~ 30% LV
   hypertrophic response without weight loss and preserved systolic
   function. Activation of the RAAS was the not the goal of this model, as
   many trials of RAAS antagonists have not shown benefit in HFpEF
   treatment;^[134]28–[135]30 rather, we were co-opting a physiologic
   pathway to induce hypertension, a clear component of the disease state.

   We comprehensively phenotyped this new HFpEF model for cardiac and
   extra-cardiac physiologic endpoints, confirming obesity, insulin
   resistance, hypertension, preserved systolic function, concentric
   cardiac hypertrophy, left atrial enlargement, diastolic dysfunction,
   elevated cardiac natriuretic peptide gene expression, and impaired
   exercise capacity. To assess diastolic dysfunction, we relied on a
   plurality of endpoints, including novel left atrial and classic mitral
   valve echocardiographic measures, gross and normalized tissue weights,
   exercise intolerance, and natriuretic peptide expression. Multiple
   metrics of abnormal diastology were present (ventricular strain, E/e’,
   increased trend in Tei index and IVRT), as well as novel
   echocardiographic parameters quantifying LA strain, LA volume, and
   pulmonary vein doppler patterns, which have been shown to robustly
   correlate with diastolic dysfunction^[136]9. Increased hepatic and
   pulmonary congestion and elevated BNP, a marker of cardiac congestion,
   mirror the exercise intolerance critical to the human disease. The need
   for a plurality of markers is critical since E/A or E/e’ individually
   can be an unreliable representation of diastolic dysfunction in
   mice^[137]31,[138]32, requiring technical expertise to obtain heart
   rates slow enough to see E and A separation but not so slow as to
   affect ventricular function and loading conditions. In addition,
   multiple pathophysiological endpoints also allow for assessment of
   therapies that may only impact a subset as we observed with
   intervention of SGLT inhibitor therapy. We acknowledge the model is
   limited by modest sex differences, where male mice have greater
   evidence of HFpEF compared to female mice. This observation is
   disparate from the human condition but is similar to other mouse HFpEF
   models^[139]11,[140]33, and highlights a need in the field for better
   female HFpEF modeling.

   This model allows for assessing how individual pathophysiological
   interventions that are known drivers (LV pressure overload, caloric
   excess, abnormal neurohormonal axis) contribute to the development of
   HFpEF. The ability to isolate component drivers is a powerful technical
   aspect to the model, querying the specific and additive contributions
   of hypertension and metabolic dysfunction to the overall HFpEF
   phenotype. For some metrics, the hypertensive component was the
   dominant driver, such as measures of LA size and LV hypertrophy. A
   primary dietary effect was seen in assessing exercise capacity, where
   HFD largely determined response to exercise. However, for many
   endpoints, including hypertrophy, diastolic dysfunction, and fibrosis,
   a significant additive effect was observed beyond either constituent
   effect.

   Ventricular transcriptomic interrogation (RNAseq) was used to assess
   component effects, identify signature transcriptional changes in the
   HFpEF state, and allow for comparison with recent published human data
   to further assess relevance^[141]19. Gene expression changes caused by
   the individual drivers had incomplete similarity to the patterns seen
   in the HFpEF state, further emphasizing the “syndrome” aspect to this
   disease—multiple, multiorgan, chronic inputs are required. Among the
   top upregulated genes in both datasets, many are implicated in
   extracellular matrix remodeling, cardiac fibrosis, and inflammation
   (e.g., Acta1, Edn3, Col8a1, Postn, Adamtsl2). For example, Acta1 has
   previously been demonstrated to be upregulated in mouse HFpEF
   LV^[142]34. Cartilage intermediate layer protein 1 (CILP1), reported as
   a novel sensitive biomarker for cardiac fibrosis^[143]35–[144]37, and
   disintegrin-like and metalloproteinase domain with thrombospondin type
   1 motifs-like (Adamtsl2), an extracellular matrix glycoprotein
   upregulated in human heart failure, act to modulate TGFβ signaling in
   cardiac fibrosis^[145]38. There was also strong evidence for altered
   mitochondrial function, including diminished branched chain amino acid
   catabolism. Targeted metabolomics from cardiac tissue bore out the
   decrease of BCAA metabolites in the HFpEF state, which is similar to
   observed changes in human HFpEF samples^[146]39.

   To test the translational utility of this HFpEF model, mice were
   treated with empa, a potent sodium-glucose co-transporter 2 inhibitor
   (SGLT2i) which has demonstrated improvement in reducing heart failure
   hospitalization in HFpEF^[147]40,[148]41. Preclinical data has shown
   variable benefit of SGLT2i in mouse models of HFpEF, including
   improvement of fibrosis^[149]42, improved hypertrophy with^[150]43 or
   without improved diastolic dysfunction^[151]7, and improved
   electrophysiologic parameters^[152]11. We demonstrate consistent,
   albeit subtle improvement in a subset of cardiac and extra-cardiac
   endpoints, including body weight, insulin sensitivity, LA and LV
   hypertrophy, and exercise performance, with trends toward improved
   strain. The benefit of a model that exhibits multiple phenotypic
   features of HFpEF is emphasized by these results given that therapeutic
   interventions will likely impact only a subset of the many drivers of
   this disease. The results are also reproducible in two separate
   experimental settings using testing protocols that are similar, not
   identical; for instance, exercise testing that measured sprint (Pfizer)
   versus high intensity endurance (Penn). These results highlight the
   value of this model for translational and mechanistic studies including
   the potential for genetic interventions that can provide insight into
   the relevant pathogenic origins of HFpEF.

   In summary, we present a tractable, consistent model of mouse HFpEF
   that recapitulates multiple concerted endpoints observed in the human
   condition. Compared to the models currently available, this model
   utilizes AAV-Renin to create a stable, consistent hypertension not
   affected by food or water intake, or time-limited drug dosing via
   implantable minipumps^[153]6,[154]8,[155]9,[156]14,[157]44. Moreover,
   renin-driven hypertension is a well-known driver of human hypertension
   and cardiac remodeling. Additionally, there is no artificially imposed
   vascular dysfunction from an exogenous chemical, such as L-NAME.
   Hijacking the renin-angiotensin-aldosterone system as a tool to induce
   hypertension is physiologically relevant to human HFpEF patients,
   mimicking the significant increase of plasma renin levels observed in
   the TOPCAT HFpEF clinical trial phenogroup 3 patients^[158]17.
   Accepting the limitation of using rodents to model human disease, we
   believe this model, used in conjunction with others previously
   developed, will allow for critical investigation into a complex and
   poorly understood form of heart disease.

Methods

In vivo studies

   Mouse studies were performed in parallel at two separate institutions:
   University of Pennsylvania (Penn) and Pfizer. Penn utilized male and
   female C57BL/6NJ mice (Jackson Laboratory, Stock 005304) age 8–10 weeks
   old, and randomized to chow or 60%kcal HFD (D12492i, Research Diets)
   arms for 10 weeks. Pfizer utilized 8-week-old C57BL/6 N male mice fed
   chow or 60% HFD-fed (Taconic, DIO-B6NTac) for 12 weeks. Mice were
   housed in a facility under a 12-hour light/12-hour dark cycle, at 72
   degrees Fahrenheit and 40–50% humidity. For tissue collection, mice
   were fasted for four hours prior to anesthesia with intraperitoneal
   pentobarbital (Sagent, 100 mg/kg) and euthanasia by cardiectomy. Tissue
   weights were normalized to tibia length. Lungs and livers were excised
   and weighed “wet” and, for lungs only, following 48 h at 37°C “dry.”
   For empagliflozin (empa) studies, mice were randomized to treatment
   with drug (Advanced ChemBlocks Inc, Cat#: G-7261) admixed at 500 mg/kg
   (Pfizer) or 50 mg/kg (Penn) in 60%kcal HFD, validated to match the
   human pharmacologic effects^[159]45.

AAV8-Renin generation and treatment

   Mouse Ren1d cDNA was purchased from Horizon Discovery (Clone ID
   30313659, Sequence [160]BC061053) and mutated at F61R and P65S as
   previously described (to enable prorenin cleavage to renin in non-renal
   locations, see Supplementary Fig. [161]S1A) using the In-Fusion HD
   system (Clontech, #639650)^[162]20. Construct was subcloned in the
   pAAV.TBG.PI.Null.bGH (Addgene #105536) plasmid provided by the Vector
   Core at Penn. All cloning and mutagenesis primer sets are listed in
   Supplementary Table [163]S1. pAAV.TBG.PI.mRen1(F61R)(P65S).bGH
   (AAV8-Renin) or null constructs were packaged at the Vector Core. For
   the Pfizer model, AAV was produced at the University of Massachusetts
   Medical Center Gene Therapy Center and Vector Core. Eighteen to
   20-week-old lean or obese mice were randomized to receive AAV8-Renin or
   AAV8-null virus. Mice were anesthetized using 2.5ppm isoflurane prior
   to injection via the medial canthus into the retro-orbital venous
   plexus.

   Based on extensive dose-response testing, doses were selected to
   generate a 2-3x increase over basal renin expression but without
   affecting weight gain (Supplementary Fig. [164]S1). Given viral
   production occurred in two independent facilities, empiric doses were
   as follows: Penn lean mice received 3.0 × 10^10 and HFD-fed mice
   received 1.0 × 10^11 viral genomes (vg) diluted in sterile saline,
   injected in a volume of 100ul. Pfizer lean mice received 1.0 × 10^10
   and HFD-fed 3 × 10^11 vg. Higher doses of AAV-Renin were required to
   raise plasma renin levels in obese mice, potentially due to lower
   transduction efficacy of AAVs in the fatty liver of the HFD model.
   Blood samples were obtained one week post-injection and assayed for
   circulating renin using the mRen1 DuoSet ELISA (Bio-techne, #DY4277)
   per manufacturer protocol. Samples from mice receiving AAV8-Renin were
   diluted 1:50. Aspartate aminotransferase (AST) and alanine
   aminotransferase (ALT) assays were performed from blood collected in
   EDTA tubes and placed on ice. Plasma was collected after centrifugation
   at 10,000 RPM for 10 min at 4°C, and stored at -80°C until use. AST/ALT
   assays were performed on the Siemens ADVIA Chemistry XPT system (ALT
   #07501976, and AST #07499718).

Telemetry study

   Radiotelemetry recording of arterial pressure was performed in 8- to
   10-week-old male C57BL/6 N mice, implanted with telemeters (PA-C10,
   Data Science) by Envigo (Invito, Indianapolis, IN). After one week,
   telemetry units were activated, and arterial pressure was recorded for
   baseline recording. Measurements were recorded for five seconds every
   minute for the length of the study. Mice were randomized and dosed with
   either 1 × 10^10 AAV null or 3 × 10^9, 1 × 10^10 or 3 × 10^10
   AAV-Renin. Two weeks post-AAV dosing, mice were treated with enalapril
   (200 mg/L; Sigma) in the drinking water for three days, followed by
   switching back to normal drinking water.

Mouse echocardiography

   Ultrasound examination of the left ventricle was performed
   independently at both study sites using a Fujifilm VisualSonics
   Ultrasound System. At the Rodent Cardiovascular Phenotyping Core of the
   University of Pennsylvania Cardiovascular Institute, adult mice were
   anesthetized with an intraperitoneal injection (0.05 mg/g) of 2%
   avertin (to maintain a heart rate close to 600 beats per minute or
   higher for the evaluation of LV systolic function). Hair was removed
   from the anterior chest using chemical hair remover, and the animals
   were placed on a warming pad in a left lateral decubitus position to
   maintain normothermia. Ultrasound gel was applied to the chest. Care
   was taken to maintain adequate contact while avoiding excessive
   pressure on the chest. 2D long-axis and short-axis M-mode images were
   obtained. Diastolic function–related parameters were evaluated using a
   modified four-chamber view. Transmitral inflow velocities, tissue
   Doppler, pulmonary vein Doppler, and left atrial strain measurements,
   as previously described^[165]9, were obtained following intraperitoneal
   injection of zatebradine (funny current inhibitor, 0.008 mg/gm body
   weight) to reduce heart rate to 425–475 bpm (15–25% heart rate
   reduction compared with baseline) when needed. After completion of the
   imaging studies, mice were allowed to recover from anesthesia and
   returned to their cages. At Pfizer, animals were anesthetized at 2%
   isoflurane (chow) or 3% isoflurane (HFD) with heart rate goals of
   300-450 bpm for diastolic parameters and > 400 bpm for functional
   parameters. Images were obtained similarly. Imaging and image analysis
   was performed by blinded sonographers and analyzed using Vevo Lab
   software (VisualSonics).

Exercise testing

   Mice were exercised to exhaustion utilizing similar protocols. At Penn,
   mice underwent three consecutive days of exercise on a five-lane
   treadmill with shock grid (Maze Engineering) starting with two days of
   an acclimation protocol of nine minutes at 10 m/min followed by one
   minute at 20 m/min at a grade of 6–8 degrees. Testing was performed on
   a 25-degree treadmill, starting at 10 m/min for three minutes followed
   by a ramp over three minutes to a final speed of 18 m/min, continuing
   to exhaustion. At Pfizer, animals were acclimated to the Exer3/6 Animal
   Treadmill (Columbus Instruments) for two days prior to testing (Day 1:
   five minutes off and five minutes at 2.5 m/min; Day 2: five minutes at
   2.5 m/min followed by five minutes at 5 m/min). Testing was performed
   at a 5 degree incline starting at 5 m/min for two minutes, increased to
   8 m/min for two minutes and then increased an additional 2 m/min for
   every two minutes until the animal was exhausted. Exhaustion was
   defined by more than five seconds in contact with shock pad, or more
   than 50% of the time off the treadmill.

Metabolic measurements

   Glucose and insulin tolerance testing (GTT and ITT, respectively) was
   conducted after a five-hour fast. A bolus of glucose (1 g/kg) or
   insulin (0.75U/kg) was administered via intraperitoneal injection. A
   blood sample was obtained from the tail tip for the measurement of
   baseline glucose using a handheld glucometer (Accu-Chek Aviva Plus,
   Roche or AlphaTrak2 #71681-01) at 0, 15, 30, 45, 60, and 90 min
   following the injection. At Penn, plasma insulin was measured using
   Mouse Insulin ELISA kit (Mercodia). Cholesterol and triglycerides were
   measured using Infinity colorimetric assays (ThermoScientific). Plasma
   samples at Pfizer were analyzed using Insulin Rodent (Mouse/Rat)
   Chemiluminescence ELISA kit (ALPCO, #80-INSMR-CH01). HOMA-IR was
   calculated using the insulin and glucose measurements, using the
   formula: insulin (µU/ml)*glucose (mmol/L)/22.5. Targeted metabolomics
   was performed at the Metabolomics Core in the Penn Cardiovascular
   Institute as previously described, from flash frozen mouse
   hearts^[166]46.

Histology

   The tissues were prepared as previously described^[167]46 including
   overnight fixation in 4% paraformaldehyde in PBS and dehydration
   through sequential ethanol washes. The tissues were then embedded in
   paraffin and sectioned as 6 mm. Picro Sirius Red staining was performed
   by the Histology and Gene Expression Co-Op at the University of
   Pennsylvania and imaged by the Children’s Hospital of Philadelphia
   Pathology Core. Collagen staining was quantified with Aperio ImageScope
   (Leica Biosystems).

RNA isolation and qRT-PCR

   Total RNA was isolated using the RNeasy Mini Kit (QIAGEN) and the
   RNase-Free DNase Set (QIAGEN) according to the manufacturer’s
   instructions. cDNA was synthesized using the Affinity Script cDNA
   Synthesis Kit (Agilent Technologies) with 0.25 µg total RNA. PCR
   reactions were performed using Brilliant III Ultra-Fast SYBR Green QPCR
   Master Mix (Agilent Technologies) on a QuantStudio 6 Flex Real-Time PCR
   System (Applied Biosystems) with specific primers for each gene. The
   Penn primer sets are listed in Supplementary Table [168]S1. The
   expression of target mRNAs was normalized by that of Rplp0 (36B4).
   Pfizer cardiac gene expression was measured by RT-PCR with TaqMan
   probes (Myh6: Mm00440359_m1, Myh7: Mm00600555_m1) normalized to Hprt
   (Mm03024075_m1). All experiments were repeated independently, with
   samples in either triplicate or quadruplicate.

RNA-seq library preparation, sequencing, and analysis

   RNA library preparations and sequencing reactions were conducted at
   GENEWIZ, LLC. (South Plainfield, NJ), per their standard operating
   procedure and as previously described^[169]47,[170]48. We used Salmon
   v1.10.2 to count reads in GRCm38 (mm10) reference transcripts^[171]49.
   The counts were summarized at gene level and differentially expressed
   genes were identified with DESeq2 1.40.2^[172]50. Genes with an FDR
   below 0.05, when comparing conditions of interest, and at least a
   1.2-fold expression change (up or down) were considered differentially
   expressed. We performed pathway enrichment analysis, using gProfiler
   web service with FDR < 0.05 cutoff and enrichment at least
   2^[173]51,[174]52.

Statistics

   For two-group comparisons, a Student’s t test was performed when the
   data were normally distributed based on a Shapiro-Wilk test (α = 0.05).
   For multiple comparisons, two-way ANOVA with Tukey’s post hoc test was
   performed. GraphPad Prism 7.04 or 8.03 (GraphPad Software) was used for
   graphing and statistical analysis.

Study approval

   All animal experiments were performed in accordance with NIH and ARRIVE
   guidelines for the humane treatment of animals and approved by the
   Institutional Animal Care and Use Committees of the Perelman School of
   Medicine at the University of Pennsylvania and Pfizer.

Electronic supplementary material

   Below is the link to the electronic supplementary material.
   [175]Supplementary Material 1^ (1.8MB, pdf)

Acknowledgements