Heart failure with preserved ejection fraction (HFpEF) is characterized by impaired relaxation of the left ventricle (LV) during diastole and accounts for over 50% of all patients with heart failure (HF) (Redfield et al., 2003; Yancy et al., 2006). Both the proportion of HFpEF-patients and morbidity, mortality, and healthcare costs associated with this disease are rising (Bhatia et al., 2006; Liao et al., 2006; Owan et al., 2006; Steinberg et al., 2012). Multiple processes including cardiomyocyte hypertrophy, interstitial fibrosis, impaired calcium handling, and increased passive cardiomyocyte stiffness contribute to the left ventricular stiffening characteristic for HFpEF (Borlaug, 2014; Gladden et al., 2014; Sharma and Kass, 2014). Although ejection fraction is still normal, systolic dysfunction is present in HFpEF, as measured by tissue Doppler or strain imaging (Borlaug, 2014; Tadic et al., 2017). In large population studies, the majority of the HFpEF patients are women (Masoudi et al., 2003). Whereas men have more coronary artery disease indicative of macrovascular disease, women typically present with obesity, left ventricular hypertrophy, diastolic dysfunction and more often have microvascular angina (Gori et al., 2014a; Crea et al., 2017).

The current paradigm for HFpEF proposes that commonly present comorbidities such as diabetes mellitus (DM), obesity, and hypertension lead to a systemic pro-inflammatory state. This pro-inflammatory state causes coronary microvascular dysfunction, evidenced by an imbalance between nitric oxide (NO) and reactive oxygen species (ROS) leading to stiffening of the LV (Paulus and Tschope, 2013; Gladden et al., 2014). Excessive ROS-production in the endothelium of the coronary microvasculature lowers NO bioavailability through scavenging of NO. Loss of NO reduces soluble guanylate cyclase (sGC) activity in the cardiomyocytes, thereby lowering cGMP levels and decreasing PKG activity. The latter results in hypophosphorylation of titin and induces cardiomyocyte hypertrophy (Paulus and Tschope, 2013; Franssen et al., 2016). Given the proposed central role for disruption of the NO pathway in pathogenesis of HFpEF, it is rather surprising that all large clinical trials, which targeted the NO-cGMP-PKG pathway failed to date. Organic and inorganic nitrates are therapeutic agents that can be metabolized to NO systemically and thus act as NO-donors. However, the NEAT-HFPEF trial showed that isosorbide mononitrate, a long working organic nitrate, tended to reduce physical activity and did not improve quality of life and exercise capacity (Redfield et al., 2015). Inhaled nebulized inorganic nitrate, also did not improve exercise capacity, as recently shown in the INDIE-HFpEF trial (Borlaug et al., 2018). The phase 2b SOCRATES-PRESERVED trial showed no reduction of NT-pro-BNP or left atrial dimensions at 12 weeks after treatment with the sGC stimulator Vericiguat. However, Vericiguat was well tolerated and increased quality of life, warranting further research (Pieske et al., 2017). Inhibition of the cGMP-degrading enzyme phosphodiesterase 5 with Sildenafil did not improve clinical status rank score or exercise capacity (Redfield et al., 2013), and failed to improve vascular and cardiac function (Borlaug et al., 2015). Therefore, new therapeutic targets need to be identified that can interfere with the development and progression of HFpEF.

It is important to note that the impact of microvascular dysfunction on cardiac structure and function is not limited to dysfunction of the NO-cGMP-PKG pathway. Indeed, upregulation of VCAM-1 and E-selectin on the coronary microvascular endothelium induces transendothelial leucocyte migration and activation, increased transforming growth factor β (TGF-β) levels, thereby promoting pro-fibrotic pathways and differentiation of fibroblast to myofibroblasts (Westermann et al., 2011; Paulus and Tschope, 2013) and increasing interstitial fibrosis (van Heerebeek et al., 2012; Sharma and Kass, 2014). Secretion of autocrine and paracrine factors, such as apelin, TGF-β, and endothelin-1, by dysfunctional coronary microvascular endothelial cells can also directly induce left ventricular hypertrophy (Kamo et al., 2015). Finally, capillary rarefaction and inadequate angiogenesis could contribute to a decreased oxygen supply and subsequent left ventricular myocardial stiffening (Gladden et al., 2014).

The so-called cardio-renal syndrome describes the co-existence of HF and chronic kidney disease (CKD). Approximately 50% of the patients with HFpEF also suffer from CKD (Ter Maaten et al., 2016). Although this co-existence is partially due to shared risk factors, such as hypertension, DM and obesity, it has also been proposed that HF directly impacts kidney function, and vice versa, CKD worsens cardiac function (Brouwers et al., 2013). Interdependence of the heart and kidneys, similarities between their microvascular networks, and the coexistence of CKD and HF further imply a role for microvascular dysfunction in development and progression of both diseases (Ter Maaten et al., 2016).

Given the co-incidence of HFpEF and CKD, the present review aims to provide a mechanistic link between CKD and HFpEF, by describing potential pathways through which CKD can induce or aggravate coronary microvascular dysfunction and thereby contribute to the development and progression of left ventricular hypertrophy and diastolic dysfunction. These include mechanical effects, neurohumoral activation, systemic inflammation, anemia and changes in mineral metabolism as induced by CKD (Figure 1). As some of these CKD-induced effects may induce HFpEF and contribute to cardiovascular disease in general, they may provide targets to intervene with the development of diastolic dysfunction and/or its progression towards HFpEF. Hence, this review will also describe the (potential) druggable therapeutic targets within these pathways, and where applicable, clinical trials intervening with these pathways.

CKD is defined as a progressive decline of renal function and is associated with hypertension, proteinuria, and the loss of nephron mass (Noone and Licht, 2014). CKD is an independent risk factor for the development of HF, with increasing cardiovascular risk and mortality as renal function declines (Tonelli et al., 2006; Ronco et al., 2008). Additionally, HF is the major cause of death among patients with CKD (Kottgen et al., 2007; Bansal et al., 2017). Although renal dysfunction is present in about half of the patients with HF in general (Hillege et al., 2006; Smith et al., 2013), and is an important prognostic marker for adverse outcomes (Yancy et al., 2006; Ahmed et al., 2007; McAlister et al., 2012), particularly the association between HFpEF and CKD is very strong. In a cohort comparing patients with heart failure with reduced ejection fraction (HFrEF), HF with mid-range ejection fraction and HFpEF, renal dysfunction was associated with increased mortality in all HF subtypes, but was most prevalent in HFpEF (Streng et al., 2018). Gori et al. showed that 62% of the patients with HFpEF display abnormalities in at least one marker of renal insufficiency, with different markers correlating with different HFpEF phenotypes (Gori et al., 2014b). Further evidence for a causal relationship between CKD and HFpEF comes from a rat model, in which CKD was mimicked by nephrectomy of one whole kidney and two-third of the remaining kidney. Loss of nephron mass in these rats resulted in a cardiac HFpEF-like phenotype, with LV hypertrophy and diastolic dysfunction, but critical HFpEF features such as lung congestion and exercise intolerance were not reported (Sarkozy et al., 2019). In accordance with CKD as a causative factor for HFpEF, the majority of patients on hemodialysis display diastolic dysfunction and left ventricular hypertrophy, whereas overt systolic dysfunction and HFrEF are visible in only a minority of these patients (Hickson et al., 2016; Antlanger et al., 2017). In a prospective cohort study, 74% of the patients admitted for dialysis displayed left ventricular hypertrophy. In contrast, systolic dysfunction and left ventricular dilatation were present in only 15% and 32% of the patients, respectively (Foley et al., 2010). Left ventricular hypertrophy is not restricted to end stage CKD, but is already highly prevalent in the general CKD population (Collins, 2003). Indeed, the first visible myocardial alteration in patients with CKD is left ventricular hypertrophy (London, 2002), developing early in the progression of kidney dysfunction (Levin et al., 1996; Pecoits-Filho et al., 2012) and often co-occurring with myocardial fibrosis and diastolic dysfunction (Silberberg et al., 1989). Hypertension is an important predictor for development of left ventricular hypertrophy and HFpEF in patients with CKD (Levin et al., 1996; Thomas et al., 2008), while blood pressure reduction is associated with a lower cardiovascular risk (Blood Pressure Lowering Treatment Trialists Collaboration et al., 2013).

It should be noted however, that in addition to decreased diastolic function, both hemodialysis and pre-dialysis CKD patients show impaired regional systolic function measured by longitudinal, circumferential, and radial strain while ejection fraction was preserved (Yan et al., 2011). Similarly, patients with HFpEF can also display signs of systolic dysfunction defined by decreased global longitudinal strain and S′ velocity measured with tissue Doppler. Unger et al. showed in a large group of HFpEF patients that not only diastolic dysfunction, but also the severity of systolic dysfunction and mortality increased in parallel with CKD stage (Unger et al., 2016).

Arterial remodeling in CKD patients is characterized by arterial stiffening, increasing pulse pressure, as a consequence of premature aging, and atherosclerosis of the arteries (Laurent et al., 2006; Briet et al., 2012). Premature vascular aging is common in both CKD and HFpEF. Increased aortic stiffness has been strongly associated with both left ventricular dysfunction, and markers of renal dysfunction (Bortolotto et al., 1999; Borlaug and Kass, 2011), which precede and increase cardiovascular risk in patients with CKD (Kendrick et al., 2010; Middleton and Pun, 2010). Stiffer arteries result in an increased pulse pressure, as well as an increased pulse wave velocity, which cause the increased pulsatility to be transmitted into the microvasculature (Mitchell, 2008). Renal and coronary microvascular networks are very vulnerable to pulsatile pressure and flow, thus failure in decreasing pulsatility can result in damage of the capillary networks (Mitchell, 2008; Safar et al., 2015), and thereby contribute to coronary microvascular dysfunction. Fukushima et al. showed an impaired global myocardial flow reserve in CKD patients, even with a normal regional perfusion and function of the LV (Fukushima et al., 2012). Furthermore, coronary microvascular dysfunction was shown to be present in patients with end stage CKD (Bozbas et al., 2009), and was associated with an increased risk of cardiac death in patients with renal failure (Murthy et al., 2012).

Hypertension in CKD is thought to be mainly a consequence of volume overload due to increased sodium reabsorption by the kidneys (Charra and Chazot, 2003; Judd and Calhoun, 2015). Increased sodium loading might also contribute to HFpEF development independent of hypertension, through inducing a systemic pro-inflammatory state, which is detrimental to the coronary microvasculature (Yilmaz et al., 2012). Indeed, empagliflozin, a sodium glucose co-transporter-2 (SGLT2) inhibitor, initially developed as an anti-diabetic drug, resulted in decreased cardiovascular mortality in an initial type 2 diabetes cohort (Zinman et al., 2015). Interestingly, these effects seem to, at least for some part, be specific for empagliflozin as canagliflozin protected less against cardiovascular death (Neal et al., 2017). Although the mechanisms of action have not completely been elucidated yet, multiple pre-clinical studies are being conducted to investigate the myocardial effects of SGLT2-inhibitors (Uthman et al., 2018a,b). Currently, three mechanisms have been proposed to contribute to reduced cardiovascular mortality in patients receiving SGLT2-inhibitors in general and/or empagliflozin in particular (Bertero et al., 2018); (1) osmotic diuresis and natriuresis lower blood pressure and subsequently reduce left ventricular afterload; (2) empagliflozin may instigate a shift to cardiac ketone body oxidation, increasing mitochondrial respiratory efficiency and reducing ROS production; (3) empagliflozin can lower intracellular Na⁺ by inhibition of the cardiac Na⁺/H⁺ exchanger (NHE) and induce coronary vasodilation (Uthman et al., 2018a). The latter effect is especially promising as increased intracellular Na⁺, as present in failing cardiomyocytes, results in altered mitochondrial Ca²⁺ handling and subsequent ROS production, which may be ameliorated by SGLT2-inhibitors (Bertero et al., 2018). SGLT2-inhibitors, therefore, seem promising in the cardiorenal field as they are both cardio- and reno-protective (Butler et al., 2017). The effect of empagliflozin on cardiovascular mortality in HFpEF specifically, regardless of diabetic status, is being investigated in the ongoing EMPEROR-Preserved trial (ClinicalTrials.gov NCT03057951).

CKD is associated with hyperactivation of the renin-angiotensin-aldosterone system (RAAS) in response to renal hypoxia resulting in volume overload (Nangaku and Fujita, 2008), which may contribute to the development and/or progression of HFpEF. Interestingly, testosterone can increase, whereas estrogen can lower renin concentrations (Fischer et al., 2002). Such protective effects of estrogen would especially be relevant in pre-menopausal women, and be lost in the typically older, post-menopausal female HFpEF population. Consistent with a detrimental effect of RAAS activation on HFpEF progression, RAAS activation can increase myocardial workload, by elevating systemic vascular resistance and left ventricular afterload, through vasoconstriction of systemic blood vessels in response to angiotensin II or by causing volume expansion due to increased sodium and water reabsorption in response to increased aldosterone levels (Brown, 2013; Forrester et al., 2018). It is not clear if angiotensin II can also induce myocardial cell hypertrophy and fibrosis independently of hypertension. Although in vitro studies have shown that there is a hypertension-independent effect of angiotensin-II on cardiomyocytes, multiple in vivo studies could not confirm these findings, suggesting that the effect of angiotensin II is blood pressure-dependent (Reudelhuber et al., 2007; Qi et al., 2011). Furthermore, RAAS-activation induces coronary microvascular endothelial dysfunction, through NADP(H)-oxidase activation and subsequent ROS formation (Bongartz et al., 2005; Wong et al., 2013). Myocardial perfusion might also be impaired by the vasoconstrictor effects of angiotensin II. During prolonged exercise, vasoconstriction occurs within metabolically less active tissues, mediated by angiotensin II and endothelin-1. Such response is inhibited in metabolically active tissues by NO and prostanoids, resulting in an efficient distribution of blood (Merkus et al., 2006). In a state of systemic inflammation, locally decreased NO bioavailability in the coronary microvasculature might result in disinhibition of angiotensin II-mediated vasoconstriction, resulting in reduced blood delivery to the heart.

Downstream from angiotensin II in the RAAS, aldosterone regulates blood pressure and sodium/potassium homeostasis through the mineralocorticoid receptor in the kidneys, by enhancing sodium reabsorption, thereby contributing to hypertension and high plasma sodium levels. Besides the renal effects, aldosterone has been shown to directly promote myocardial fibrosis, left ventricular hypertrophy, and coronary microvascular dysfunction, acting through endothelial and myocardial mineralocorticoid receptors, independently of angiotensin II (Brown, 2013).

RAAS inhibition is the preferred therapeutic strategy to slow down progression of renal failure and reduce proteinuria in CKD (Levin and Stevens, 2014). Despite the fact that most data show RAAS overactivation in HFpEF, clinical trials in HFpEF with drugs acting on the RAAS, have failed to improve (all-cause) mortality so far (Pitt et al., 2014; Zhang et al., 2016). It is, however, important to note that AT₁-blockade with Irbesartan reduced mortality and improved outcome on cardiovascular endpoints in patients with natriuretic peptides below the median, but not in patients with higher natriuretic peptide levels (Anand et al., 2011), suggesting that RAAS inhibition may be beneficial in early HFpEF. Furthermore, post hoc analysis of the TOPCAT trial demonstrated geographically different effects of the mineralocorticoid receptor blocker spironolactone, with small clinical benefits in patients from America (Pfeffer et al., 2015). However, these patients were generally older, had a higher prevalence of atrial fibrillation and diabetes, were less likely to have experienced prior myocardial infarction, had a higher ejection fraction and had a worse renal function (Pfeffer et al., 2015), suggesting that a benefit of spironolactone was associated with a more HFpEF-like phenotype. A more recent post hoc analysis of this trial further showed that spironolactone did show an improvement in primary endpoints in patients with lower levels of natriuretic peptides and hence less advanced disease (Anand et al., 2017). Consistent with this suggestion, a recent meta-analysis showed that mineralocorticoid receptor antagonists do improve indices of diastolic function and cardiac structure in HFpEF patients (Kapelios et al., 2019). Interestingly, treatment of DM type 2 with mineralocorticoid receptor antagonists also improved coronary microvascular function (Garg et al., 2015). Altogether, these data suggest that intervening with the RAAS is beneficial in patients with less advanced HFpEF, whereas beneficial effects are lost in patients with more advanced disease. Therefore, clinical studies investigating HFpEF progression and clinical trials focusing on reducing or preventing progression of early HFpEF into advanced HFpEF need to be conducted.

Another approach intervening with the RAAS is the use of Entresto, an angiotensin receptor and a neprilysin inhibitor (ARNI), which is a combination of valsartan (AT₁ receptor blocker) and sacubitril (neprilysin inhibitor). Neprilysin inhibition exerts its beneficial effects through inhibition of the breakdown of natriuretic peptides. Entresto was superior to the standard therapy, enalapril, in patients with HFrEF in reducing mortality and number of hospitalizations for HF (McMurray et al., 2014). In hypertensive rats with diabetes, ARNI reduced proteinuria, glomerulosclerosis, and heart weight more strongly than AT₁ receptor blockade, and this occurred independently of blood pressure (Roksnoer et al., 2015, 2016). In a phase 2 double-blind randomized controlled trial in HFpEF patients, Entresto reduced NT-pro-BNP plasma levels and left atrial diameters to a greater extent than valsartan (Solomon et al., 2012). These findings led to the ongoing PARAGON-HF trial (ClinicalTrials.gov NCT01920711), which investigates the long-term effect (26 months) of Entresto compared to valsartan in HFpEF (Solomon et al., 2017).

Both CKD and HFpEF are accompanied by autonomic dysregulation (Salman, 2015). Sympathetic hyperactivity has a detrimental effect on both the heart and the kidney and aggravates hypertension and proteinuria. Furthermore, HFpEF patients show attenuated withdrawal of parasympathetic tone and excessive sympathoexcitation during exercise that cause β-adrenergic desensitization, chronotropic incompetence, and may thereby contribute to the limited exercise tolerance of these patients (Phan et al., 2010). A critical role for CKD in this process was suggested by Klein et al. (Klein et al., 2015), showing a clear correlation between CKD, decreased heart rate variability, chronotropic incompetence in HFpEF, and decreased peak VO₂. Unfortunately, neither the SENIORS trial (van Veldhuisen et al., 2009), nor the OPTIMIZE-HF registry (Hernandez et al., 2009) showed a beneficial effect of beta-adrenoceptor blockade on all-cause mortality or cardiovascular hospitalizations. Furthermore, beta-adrenoceptor blockade failed to improve LV systolic or diastolic function in patients with ejection fraction >35%, as measured in the SENIORS echocardiography sub-study (Ghio et al., 2006). It should be noted that in the SENIORS trial ejection fraction cutoff was set at 35%, which is lower than current consensus about the cutoff of reduced and preserved ejection fraction. Additionally, in these studies, beta-adrenoceptor blockade was administered on top of existing medication, which often included RAAS-inhibitors. Conversely, in patients with treatment resistant hypertension, renal sympathetic denervation did improve diastolic function and reduce left ventricular hypertrophy, besides reducing blood pressure (Brandt et al., 2012), suggesting that there is indeed an interaction between CKD, sympathetic hyperactivity and diastolic cardiac function.

A pro-inflammatory state is already present in early stages of CKD (Stenvinkel et al., 2002), and is likely an important risk factor for cardiovascular morbidity and mortality on the long term (Ruggenenti et al., 2001; Sarnak et al., 2003). In HFpEF, a systemic pro-inflammatory state has been proposed to be a critical causal factor in coronary microvascular dysfunction as inflammatory cytokines can directly induce endothelial cell dysfunction, cause upregulation of adhesion molecules on coronary microvascular endothelial cells, and reduce NO bioavailability, resulting in impaired vasodilation and pro-fibrotic signaling (Figure 2; Rosner et al., 2012; Paulus and Tschope, 2013).

Targeting this pro-inflammatory state with 14 days of treatment with the recombinant human IL-1 receptor antagonist Anakinra, increased peak VO₂, which correlated with a reduction in C-reactive protein (CRP) in the D-HART trial including 12 patients (Van Tassell et al., 2014). Unfortunately, prolonged treatment (12 weeks) in the follow-up D-HART2 trial in 28 patients did not increase VO₂, despite small improvements in exercise duration and quality of life, as well as reductions in CRP and NT-pro-BNP compared to baseline values (Van Tassell et al., 2018).

It is possible that targeting systemic inflammation in general to ameliorate HFpEF is too broad to be successful. In the subsequent paragraphs, the contribution of the individual systemic factors: anemia, proteinuria, and reduced excretion of so-called uremic toxins as consequences of renal dysfunction and possible contributors to systemic inflammation, development of microvascular dysfunction, and HFpEF will be considered in more detail.

The kidneys and heart are interdependent organs that are highly connected through multiple systems on both macrovascular and microvascular level. Unfortunately, many studies on the cardiorenal connection have not been conducted in specific HFpEF populations. Pathological processes which are present in CKD, such as vascular changes, deficiencies in kidney produced factors, and impairments in renal filtration can cause and/or contribute to development of HFpEF via several processes, as summarized in Figure 2. Elevated levels of phosphate, PTH, FGF-23, AGEs and uremic toxins, but also anemia and proteinuria can induce a systemic pro-inflammatory state. This state can lead to left ventricular stiffening and coronary microvascular dysfunction by initiating endothelial cell dysfunction, oxidative stress, and vascular smooth muscle cell proliferation. Arterial stiffening, volume expansion, hypertension and RAAS activation, as consequences of CKD, increase left ventricular workload and hypertrophy.

The complexity and multitude of connections between the heart and kidney make it unlikely that there is a single causal contributor for progression from CKD to HFpEF. In addition, although HFpEF is more prevalent in women, and the effect of sex on cardiovascular disease is increasingly recognized (Regitz-Zagrosek, 2006), the specific role of sex in HFpEF pathology still needs to be identified. Multiple large trials have been conducted with treatments for HFpEF, targeting different pathophysiological processes, but unfortunately failed to show clinical benefit. Therefore, current guidelines on treatment of HFpEF focus on lifestyle interventions and the management of comorbidities such as diabetes mellitus, hypertension, obesity and CKD. In addition, it has been proposed that different HFpEF phenotypes exist that should be targeted with different therapeutic strategies. Both male and female CKD patients are interesting and easily identifiable subgroups of HFpEF patients, warranting further investigation both in pathogenesis, as in clinical trials to further investigate cardiorenal connection in HFpEF specifically, and to identify the unique mechanistic pathways involved in various phases of the disease.

This review was designed by JW, MB, and DM. JW and MB have written the largest body of text. OS, JJ, MV, DD, AD, and DM have made a substantial, direct, and intellectual contribution to specific topics. All authors approved the work for publication.