Acute heart failure (AHF) is broadly defined as a rapid onset of new or worsening signs and symptoms of heart failure (HF) (McDonagh et al., 2021). These are due to an increase in left ventricular (LV) filling pressure with consequent pulmonary congestion but not necessarily associated with low cardiac output - both HF with reduced or preserved ejection fraction seem to have similar hospital admissions and readmissions (Metra and Teerlink, 2017). In fact, the economic burden of HF is becoming one of the most problematic public health issues with increasing prevalence and costs. Due to the ageing of the population and progressive treatment of HF with significant reduction of mortality (Benjamin et al., 2019; Groenewegen et al., 2020), it represents nowadays the most common diagnosis in patients over 65 years-old with unscheduled admission to hospital in high-income nations (Braunwald, 2015). It also reflects our poor pathophysiologic knowledge because no AHF clinical trials have shown to improve in-hospital symptoms and postdischarge clinical outcomes compared with placebo (Tomasoni et al., 2019), except maybe recently for sacubitril/valsartan where early titration before discharge was demonstrated to be associated with favourable reduction of natriuretic biomarkers, an accepted surrogate for prognostication (Velazquez et al., 2019; Wachter et al., 2019).

From what is known so far, pathophysiology in AHF is presently viewed as consisting theoretically of two phases - initiation and amplification phase (Sabbah, 2017) - in a spiral of multiple concurrent mechanisms contributing for worsening HF and end-organ damage (Harjola et al., 2017). A progressive chronic energetic exhaustion of the failing myocardium or a crescendo of vascular stiffness dependent on neurohormonal activation (Mentz and O'Connor, 2016) with inflammation (Reina-Couto et al., 2021) and endothelial disruption, fluid accumulation and cardiac vicious workload (Colombo et al., 2008) results in the clinical cardinal sign of congestion in most patients with AHF, but a smaller proportion presents with peripheral hypoperfusion or cardiogenic shock (CS) (Chioncel et al., 2017b). This most severe form of AHF remains with an unchanged mortality as high as 50% and also with few evidence-based effective therapeutics (Thiele et al., 2019). Interestingly, we still cannot conclude whether congestion is cause or consequence of endothelial dysfunction (Colombo et al., 2014; Colombo et al., 2015) which remains to be explored in AHF. Nevertheless, even though endothelial dysfunction prognostic significance is well recognized in chronic HF (Alem, 2019; Zuchi et al., 2020), the classical methods for its evaluation are technically challenging and have limited its clinical implementation (Shantsila et al., 2012). Also, the assessment of systemic endothelial biomarkers is not yet consolidated for clinical practice, namely for AHF.

Endocan, previously designated as endothelial cell-specific molecule-1 (ESM-1), is a soluble dermatan sulfate proteoglycan synthetized specifically by endothelial cells, probably mirroring inflammation-driven “endothelitis” (Lassalle et al., 1996). There has been an increasing interest in exploring endocan’s utility as a biomarker in a wide spectrum of pathological states, particularly in sepsis, acute respiratory distress syndrome and several cardiovascular diseases (De Freitas Caires et al., 2018; Bessa et al., 2020). Increased values independently related with prognosis and cardiovascular events have been recently reported for endocan in coronary and chronic HF patients (Kosir et al., 2019; Ziaee et al., 2019) and its therapeutic potential has just started to be explored for cancer (Zhang et al., 2021). As it can be easily detected in the human bloodstream and was proved to be stable, intact and reliable in critical patients (Gaudet et al., 2017), we aimed to explore serum and urinary endocan profiles in AHF and CS patients, where endothelial activation and congestion seem to be plausible and specific hallmarks. Furthermore, we analysed their correlation with prognostic parameters as well as with mechanistic biomarkers of inflammation, endothelial activation or cardiovascular dysfunction.

Our study highlights that inflammation-driven endothelitis might be a major pathophysiological mechanism across AHF spectrum, as evidenced by elevated concentrations of S-Endocan, S-VCAM-1, S-IL-6 and S-TNF-α both in AHF and CS group of patients. Of these parameters, S-Endocan seems the most promising biomarker since it not only raises with the hemodynamic severity of clinical presentation at admission but, most surprisingly, also increases with LVEF dysfunction. The positive correlation of S-Endocan with routine ICU prognostic scores and, particularly, its positive association with P-BNP in multivariate analyses suggest its relevance for inclusion in larger multimarker panels for AHF prognostication and eventually for future therapeutic targeting.

After many negative or neutral AHF trials, newer therapeutic targets are demanded once they are validated in pilot mechanistic studies where surrogate markers prove a solid and comprehensive inference for clinical benefit (Gheorghiade et al., 2005b). Endothelial dysfunction is prevalent and seems to be a predictor of adverse events in HF patients, also implicated in HF development and progression (Alem, 2019; Zuchi et al., 2020). Of note, therapies that have shown consistently to improve HF survival (e.g. angiotensin-converting enzyme inhibitors, spironolactone, beta blockers, etc) were demonstrated to simultaneously improve endothelial function (Marti et al., 2012). Therefore, a new interest in endothelial activation is re-emerging not only for risk stratification but also as a potential therapeutic target (De Keulenaer et al., 2017; Premer et al., 2019). The most validated theory considers that nitric oxide (NO) regulation of vascular tone contributes to the hemodynamic status in acute HF. Its imbalance with vasoconstrictors and reactive oxygen species determines increased vascular stiffness in pulmonary and systemic circulation and consequently ventricular workload and neurohormonal activation (Marti et al., 2012). Numerous clinical trials explored this therapeutically, either directly through drugs that modulate NO release, such as nebivolol (SENIORS trial) (Flather et al., 2005), or indirectly through drugs acting on cGMP-signalling axis, such as sildenafil (RELAX trial) (Redfield et al., 2013), riociguat (LEPHT study) (Bonderman et al., 2013) or vericiguat (SOCRATES trial) (Gheorghiade et al., 2015) and, ultimately, via the neprilysin inhibitor sacubitril that inhibits natriuretic peptide degradation causing amplification of the intracellular level of cGMP via natriuretic peptide receptor-A (McMurray et al., 2014). However, endothelial dysfunction extends beyond NO-mediated effects in the endothelium and its role in HF pathophysiology is more complex than previously anticipated. Endothelial dysfunction is now viewed as a common and important feature of all circulatory beds in HF, regardless of LVEF spectrum (Triposkiadis et al., 2019).

Although several non-invasive techniques have been developed for endothelial function testing, including flow-mediated vasodilation and finger plethysmography, these techniques are hardly implemented in the clinical daily practice. Due to these difficulties, some studies have quantified circulating endothelial biomarkers in HF patients, and, therefore, the measurement of many circulating endothelial biomarker candidates is becoming promising (Walczak et al., 2015). Circulating endothelial cells in the peripheral blood as well as E-selectin, von Willebrand factor and soluble thrombomodulin were shown to be significantly higher in patients with HF although they did not present significant differences between AHF and chronic HF and did not correlate with P-BNP (Chong et al., 2006a), ejection fraction or New York Heart Association (NYHA) class (Chong et al., 2006b). Also, VCAM-1 and ICAM-1 were shown to be associated with the development of new post-acute myocardial infarction HF symptoms (Lino et al., 2019) and increased subset-specific monocyte expression of their receptors were observed in AHF patients, with potential prognostic value of VCAM-1R (Wrigley et al., 2013) and sVCAM-1 after ST-elevation myocardial infarction (Hayek et al., 2021). But of all these, serum endocan has been one of the most qualified for the cardiovascular arena (Bessa et al., 2020).

In fact, in our work, most endothelial biomarkers, but not P-BNP, seem to differentiate AHF phenotype of presentation, since admission concentrations of S-ICAM-1, S-VCAM-1 and S-Endocan were significantly higher in CS patients. Among them, S-Endocan and S-VCAM-1 were also significantly increased in AHF patients, besides being linearly associated with the hemodynamic presentation of AHF. However, only S-Endocan increased in line with the degree of LVEF impairment evaluated by echocardiography. Thus, it may potentially be the first biomarker correlated with the severity of presentation of AHF phenotype and LVEF impairment, ultimately being anticipated as a better surrogate for prognostication in this entity if we consider hypoperfusion phenotype as the worst indicator for in-hospital mortality (Chioncel et al., 2019). This is corroborated by its positive association with P-BNP in repeated measures multivariate analysis adjusted for age and gender, as well as by its positive correlations with APACHE II and SAPS II. P-BNP (Metra et al., 2007) and LVEF (Solomon et al., 2007; O'Connor et al., 2008) are well established surrogates for prognosis in AHF and CS patients, although with some particularities and limitations (Chioncel et al., 2017a; Pang et al., 2019; Salah et al., 2019), whereas APACHE II and SAPS II are mostly validated for ICU patients in general. Accordingly, in our study, CS patients have longer ICU length of stay and tended to have a higher 1-year mortality, but there were no significant differences in in-hospital mortality between the AHF or CS groups. This might have resulted from a selection bias of AHF patients, which were only recruited from ICU, implicating selection of most severe patients, for example with need of non-invasive ventilation support. Moreover, the significantly older age of AHF patients and the higher prevalence of myocarditis (non-coronary etiologies) in CS group might also have contributed to these outcomes (Harjola et al., 2015). Noteworthy, we found that non-survivors had higher admission values of S-Endocan compared to survivors, although this did not reach statistical significance probably due to the aforementioned reasons and also to the exploratory nature of this small study, not designed to detect differences between survivors and non-survivors. We further assessed the performance of S-Endocan, U-Endocan, S-VCAM-1, P-BNP, APACHE II and SAPS II at admission to discriminate in-hospital mortality using ROC analysis. As already established, APACHE II and SAPS II are the best models for predicting mortality and this was confirmed again by their higher and significant AUC values. This was expected since they are a composite of clinical and analytical variables. However, when comparing confidence intervals, their performance did not appear to be significantly different from that of S-Endocan. This biomarker also seemed to perform, at least, as well as P-BNP and S-VCAM-1 in relation to in-hospital mortality, with the clinical advantage of being linearly associated with both hemodynamic presentation and ventricular dysfunction, which no analytical parameter demonstrated until now. These results are in line with previous studies, where for example, in ventilator-associated pneumonia, higher endocan concentrations were seen in non-survivors at day 1 and 7 (El Halim and Sayed, 2015) and also in patients with acute respiratory distress syndrome (ARDS) (Tang et al., 2014). In severe sepsis, a cut-off point was determined at days 1, 4, and 7, where higher endocan values were associated with poor prognosis (Hsiao et al., 2018) and with the need for mechanical ventilation (Mangat et al., 2017), thus representing a better biomarker than procalcitonin (Pauly et al., 2016; Zhao and Dong, 2017). Also remarkably, U-Endocan was even more evidently higher in CS non-survivors but again not statistically significant. Furthermore, its performance to discriminate in-hospital mortality did not appear to be significantly different from the other endothelial biomarkers tested.

Human endocan is synthetized by the vascular endothelium, namely by the pulmonary and kidney endothelial cells (Lassalle et al., 1996) but, unlike the other proteoglycans of the endothelial glycocalyx, it circulates freely in the bloodstream (Gaudet et al., 2017). The catabolism of endocan is not well known, resulting probably from proteolytic degradation and hepatic metabolism (De Freitas Caires et al., 2013; Nault et al., 2013). Urinary endocan is probably derived from kidney endothelial cells and in our work it was markedly higher in CS, suggesting locally increased renal endothelial dysfunction in these patients. Although it was positively correlated with proteinuria, which might indicate a relationship with glomerular lesion, we did not observe significant differences in renal function parameters between AHF and CS patients. Interestingly, urinary endocan was also positively correlated with inflammatory biomarkers such as IL-6 and TNF-α, reinforcing endocan association with inflammatory status. In contrast to our work, in populations with community acquired pneumonia, circulating endocan did not correlate with IL-6 or VCAM-1 at admission (Smart et al., 2018). This probably suggests that, in AHF, endocan is more timely linked to the inflammatory response and could potentially better reflect the complex multitude of pathophysiological pathways of AHF patients or even add incremental value to a multi-marker multi-point strategy of risk stratification (Lassus et al., 2013; Demissei et al., 2016).

We did not observe a reduction of endocan values or of other endothelial and inflammatory biomarkers during the first week of hospitalization in both patient groups. In other studies, namely on post-operative cardiac surgery patients, endocan was shown to peak very quickly at 6 h and slowly decline but it did not return to baseline at day 5 (Madhivathanan et al., 2016). Furthermore, higher endocan concentrations were observed in patients with the longest duration of norepinephrine support (Bougle et al., 2018) and could predict nosocomial pneumonia earlier and better than CRP (Perrotti et al., 2018). Endocan was also shown to increase during sepsis, worsening into multiple organ dysfunction syndrome (MODS), as well as to decrease when sepsis improves (Ioakeimidou et al., 2017). Severe sepsis with endocan concentrations remaining above 6.28 ng/ml at days 1, 4, and 7 was already demonstrated to be associated with poor prognosis (Hsiao et al., 2018). Of note, in our study, CS patients showed median S-Endocan values higher than 8 ng/ml throughout hospitalization. We hypothesize that in our patient groups, endocan kinetics could be more delayed than in the cases of sepsis already described, being specifically related to cardiovascular pathophysiology as evidenced by its linear association with ventricular dysfunction, but presenting similar prognostic implications.

Our results might indicate a perpetuation of endothelial dysfunction and inflammation in AHF and CS patients (Reina-Couto et al., 2021), reflecting the “non-resolving” course typical of this condition with high incidence of re-hospitalization and worse outcomes of AHF, as verified also in our population with median survival time of 13 months. This is in accordance with previous evidence in AHF patients showing that neurohormones and inflammatory biomarkers remain elevated 48h to 5 days after the acute event or even for longer periods (Milo et al., 2003; Gheorghiade et al., 2005a; Cotter et al., 2008). The persistence of these neurohormonal and inflammatory responses may be responsible for the high rehospitalization rates of this syndrome. So far, no therapeutic drug is well established as securely modifying prognosis in AHF (McDonagh et al., 2021) but current guidelines recommend decongestive therapy and eventually vasodilators for AHF patients who present with pulmonary edema or inotropes/vasopressors for CS patients (McDonagh et al., 2021). In fact, AHF or CS pharmacotherapy may be part of the problem. AHF pharmacological treatment may contribute to heightened neurohormonal activation (Goldsmith et al., 2018). Furthermore, the use of catecholamines as inotropes and/or vasopressors for the management of CS patients may further worsen cardiac and renal function (Amado et al., 2016; Tarvasmaki et al., 2016). Sympathetic activation is known to be associated with both inflammation and endothelial dysfunction and may underlie the unresolved inflammatory and endothelial responses in AHF and CS patients (Marvar and Harrison, 2012; Johansson et al., 2015; Ziegler et al., 2018). The therapeutic modulation of endocan has been explored in some studies as recently reviewed by our group (Bessa et al., 2020). However, as far as we know, no studies analysed the effects of mainstay pharmacotherapies for AHF (diuretics, vasodilators) or CS (inotropes, vasopressors) specifically on endocan. Regarding other drug treatments used in patients with cardiovascular disease, there is some evidence that RAAS inhibitors, statins and P2Y₁₂ receptor antagonists can reduce endocan (Celik et al., 2015; Wei et al., 2017; Gao et al., 2018; Tuncez et al., 2019). However, in our study, only prior RAAS inhibitor treatment significantly decreased serum endocan and this was just observed for CS patients at admission. Treatment with statins or antiplatelet agents throughout hospitalization did not seem to affect serum or urinary endocan values in AHF or CS patients.

Endothelial dysfunction is poorly characterized in humans on mechanical circulatory support. In our study, we performed an extensive evaluation of endothelial and inflammatory biomarkers. Nevertheless, although the time frame selected in our study seemed adequate based on a previous study (Frerou et al., 2021), a more prolonged evaluation period might be required for VA-ECMO patients, ideally comparing pre- and post-cannulation. Indeed, during the first week of hospitalization, we could not observe any differences in the endothelial or inflammatory biomarkers in VA-ECMO patients compared to the conventional treatment group, with the exception of a tendential rise in S-CRP in the VA-ECMO group at days 5–8. These results are in contrast to a recent study describing that CS patients on VA-ECMO had higher values of plasma IL-6 and TNF-α within 24 h of VA-ECMO initiation compared to CS patients without VA-ECMO. Nevertheless, similarly, the authors could not find differences in these parameters between groups on day 4 after VA-ECMO initiation (Frerou et al., 2021). So far, from the scarce evidence found, there is still no consensus on the impact of ECMO on endothelial dysfunction and inflammation. Persistently high IL-6 concentrations in VA-ECMO patients were shown to be associated with poor prognosis (Risnes et al., 2008; Al-Fares et al., 2019), but in studies conducted in experimental models to investigate endothelial cell dysfunction associated with prolonged contact of blood components with synthetic surfaces, plasma from the ECMO experiments did not induce ICAM-1 expression in human umbilical vein endothelial cells during the 8 h of exposure (Graulich et al., 2000).

Major strengths of our study include an ICU population sample in “real-world conditions”, particularly CS and VA-ECMO patients, which are scarcely studied in the literature, as well as an extensive panel of endothelial and inflammatory biomarkers and the most commonly used clinical and echocardiographic prognostic indicators, including ICU APACHE II and SAPS II scores. Furthermore, our cohort appears to be representative of the mortality rate described in the literature for both AHF and CS (Chioncel et al., 2017b; Thiele et al., 2021). The small size of sample population and the unicentric character of this study are important limitations to point out that might limit the ability to generalize our results in terms of prognosis. Nevertheless, our results provide important clues of endocan usefulness as a prognostic marker in AHF and CS. We believe that endocan relation with mortality deserves to be explored in larger studies for its potential impact on disease management and prognosis. A further long-term prospective study should also be conducted after hospital discharge in patients presenting to ambulatory HF clinic since the results obtained throughout 1 week of hospitalization suggest that endothelitis might be a valuable therapeutic target in these patients.

In conclusion, admission concentrations of serum and urinary endocan significantly increase across AHF spectrum, but there is no reduction in the values of endocan or of other endothelial and inflammatory markers throughout hospitalization, suggesting a perpetuation of endothelial dysfunction and inflammation in these patients that could be related with the poor prognosis of this condition. Importantly, serum endocan appears to be a potential new biomarker of endothelitis and a putative therapeutic target in AHF and CS, being closely associated with LVEF impairment, BNP and prognostic scores.