Heart failure with reduced ejection fraction (HFrEF) is characterized by progressive dysfunction of left ventricular muscle, myocyte remodeling, and activation of autonomic and hormonal systems (Jackson et al., 2000; Kemp and Conte, 2012). Despite the advances in therapies and prevention, patients with HF have high rates of morbidity and mortality (Falk et al., 2005; Savarese and Lund, 2017). The balance between the sympathetic and parasympathetic systems, in relation to cardiovascular function, is altered in response to ventricular dysfunction (Amorim et al., 1981; Benedict et al., 1996; Floras, 2009). Specifically, it is recognized that there is abnormal hyperactivity of the sympathetic nervous system (Watson et al., 2007; Ramchandra et al., 2009b) and disinhibition of the parasympathetic nervous system that leads to worsening of the condition (Eckberg et al., 1971; Kinugawa and Dibner-Dunlap, 1995; Bibevski and Dunlap, 1999; Dunlap et al., 2003).

Numerous studies have been conducted on small animal models of HF and these have been instrumental in understanding the pathogenesis of HF. However, there are significant differences between small animal models of HF and human HF beyond scale. These include differences in baseline heart rate, oxygen consumption, and excitation-contraction coupling in cardiac tissue (Haghighi et al., 2003; Dixon and Spinale, 2009). Therefore, a clinically relevant large animal model of HF with similar anatomy and physiology of the heart is crucial. In this context, a number of different large animal models of HF have been studied.

The primary methods to induce HFrEF in large animals include high rate pacing, cardiotoxin infusion, and coronary ligation. The pacing-induced HF model has been used in sheep (Timek et al., 2003; Byrne et al., 2004), pigs (Tanaka et al., 1993; McMahon et al., 1996), and dogs (Prabhu and Freeman, 1995; O’Rourke et al., 1999). The technique of rapid supraphysiological pacing of either the atrium or the ventricle for approximately 4 weeks results in a reproducible model of HFrEF. This model shows similar hemodynamic and mechanical phenotypes as dilated cardiomyopathy seen in human patients with HF. However, one disadvantage of this model is the absence of tissue fibrosis if the pacing is not continued for a prolonged period of time. In addition, left ventricle dysfunction tends to reverse when pacing is terminated, especially if the pacing was performed for short durations (Spinale et al., 1991; McMahon et al., 1996).

Intracoronary or intravenous infusion of the cardiotoxin doxorubicin in dogs (Bristow et al., 1980; Toyoda et al., 1998) and sheep (Chekanov, 1999; Borenstein et al., 2006) leads to cardiac myocyte injury, cell loss, and HF. A major limitation of this method is the variable degree of LV dysfunction that can result, with no reliable dosing strategy to provide a stable model. A further drawback is the potential for systemic side effects and the management of these.

Finally, coronary ligation has been used to induce HF in large animals (Hood et al., 1967). Infarcts larger than 25% resulted in a significant increase in left ventricular end-diastolic pressure and a reduction in stroke volume index. However, mortality using this model can be more than 50%, often as a result of fatal arrhythmias, despite the use of anti-arrhythmic agents or due to severe symptomatic heart failure. Additionally, when the infarction is induced by external coronary ligation rather than percutaneously, post-operative adhesions and fibrosis may make further surgical dissection for subsequent experimental work more difficult with higher rates of complications (Oizumi et al., 1990; Yim et al., 1998; Getman et al., 2006). The combination of extent and impact of infarction with animal loss and surgical access limits the usefulness of this model in the research setting.

Recently, the microembolization model has been used in sheep to develop a chronic model of HF over a period of 3 months. The embolizations have been conducted using injection of either microspheres (Schmitto et al., 2009) or gelfoam (Devlin et al., 2000). The advantage of the microembolization model is that the changes are irreversible and the study with the infusion of microspheres has reported a high success rate of induction of HF; however, mortality rates were high in the gelfoam study. Until now, studies that have examined this model have focused on structural changes in the heart (Huang et al., 2004; Monreal et al., 2004) and putative alterations in neural control have not been examined. Thus, the aim of this study was to further characterize changes in heart function after repeated microembolizations in sheep and examine whether baroreflex control of heart rate is altered in the conscious state.

Experiments were conducted on conscious, adult female Romney ewes weighing 50–80 kg, housed in individual crates, and acclimatized to laboratory conditions (18°C, 50% relative humidity, and 12-h light-dark cycle) and human contact before any experiments. All experiments and surgical procedures were approved by the Animal Ethics Committee of the University of Auckland. The sheep were fed 2 kg/day (Country harvest pellets) and had access to water ad libitum.

We have established a reliable model of heart failure with reduced ejection fraction by repeated microembolizations of left ventricular coronary arteries. Our study has a number of new findings not reported previously: (1) there is a depression in the gain of the baroreflex in this model of HF; (2) diaphragmatic electromyography showed increased respiratory rate in the HF animals compared to the control animals, suggesting increased inspiratory effort; (3) in addition to an increase in respiratory rate, there was an increased incidence of apneas in this animal model. Taken together, our findings suggest that this model of HFrEF replicates the neural and respiratory instabilities seen in human HF.

One of the main clinical signs of a myocardial infarct is changes in the ECG, namely changes in the S-T segment and the T-wave. For the S-T segment, in most cases, patients that have had a suspected myocardial infarct present with an elevation in this segment (Members et al., 2012). In contrast to clinical findings, we observed a consistent depression in the S-T segment of the embolized animals. Clinically, patients exhibiting an S-T depression had adverse long-term outcomes (Schechtman et al., 1989; Krone et al., 1993; Hyde et al., 1999). This depression in our study may reflect the recordings being done in a supine cradled position for these sheep. In the absence of pre-cordial leads, it makes it difficult to draw conclusions as to which region of the left ventricle received most ischemic damage.

Several other studies, in patients, have shown collagen deposition in the myocardium, suggestive of ventricular remodeling (Weber et al., 1993; Martos et al., 2007; Mewton et al., 2011). The gross and microscopic changes of myocyte hypertrophy and increased collagen fibrosis have also been demonstrated in other models of HF (Falk et al., 2005). After an episode of myocardial infarction, there is structural remodeling that is initiated by an inflammatory response (Sun, 2008; Suthahar et al., 2017). Scarring then forms at the infarcted area. While the scarring and fibrosis at the site of infarct maintain the structure and integrity of the heart, this remodeling can have consequences for ventricular contractility (Brilla and Weber, 1992).

One of the ways to assess the degree of HF is the construction of pressure-volume loops. One of the advantages of constructing pressure-volume loops is the measurement of parameters independent of pre- and after load (Burkhoff, 2013). A hallmark of systolic dysfunction with reduced ejection is an increase in left ventricular end-diastolic pressure (LVEDP), which was observed in the present study (Table 2). Since LVEDP is a reflection of ventricular performance, the increases in LVEDP may be associated with the size of the infarct or the amount of damage to the left ventricle (Mielniczuk et al., 2007). Furthermore, we also see a significantly higher end-diastolic volume in the HF animals similar to findings in other studies (Sabbah et al., 2000; Morita et al., 2002; Falk et al., 2005). Additionally, there was a significant increase in both end-diastolic (EDV) and end-systolic volume (ESV) in the HF animals. In HF, the severity of an increase in both these volumes at time of referral is a prognostic indicator of mortality in patients (Diaz et al., 1987).

In the present study, we show significant increases in plasma levels of brain natriuretic peptide (BNP) and norepinephrine (Table 1). These increases in plasma levels of BNP (Sakurai et al., 2003) and norepinephrine (Cohn et al., 1984) have been shown previously to relate to the severity of HF. Our findings are in agreement with other models of HF, which have also found elevated plasma levels of norepinephrine and brain natriuretic peptide (Sabbah et al., 2000; Morita et al., 2002).

We report a significant decrease in the MAP of the animals with HF. One mechanism that serves to maintain BP is the arterial baroreflex and previous studies have reported a dampening of the sensitivity of the arterial baroreflex control of heart rate in HF (Eckberg et al., 1971; Ferguson et al., 1992; Grassi et al., 1995) as well as in clinical patients (Mortara et al., 1997). We hypothesized that this model would be associated with a blunted baroreflex and observed an impaired HR response to changes in MAP. This impaired baroreflex is similar to previous reports by us in a pacing-induced model of HF (Watson et al., 2007; Abukar et al., 2016) as well as other models (Higgins et al., 1972; Greenberg et al., 1973; Zucker et al., 1977). In addition to depressed baroreflex function, we also examined sympathetic control of heart rate in this model. We observed a significant decrease in heart rate in the HF animals after beta blockade. This suggests there is a significant contribution in HF animals from the sympathetic nervous system, more so than the control animals in which no significant decrease in heart rate was observed after propranolol infusion. These findings are consistent with our previous studies suggesting low levels of cardiac sympathetic drive and cardiac norepinephrine spillover in normal animals. Both these variables are elevated in animals with pacing-induced HF (Ramchandra et al., 2009a, 2018).

In addition to changes in hemodynamic parameters, we also observed an increase in the breathing rate and an increase in the incidence of apneas. Chronic hyperventilation has been reported in HF patients (Sullivan et al., 1988) and pulmonary congestion is suspected to be the main cause of this hyperventilation (Sullivan et al., 1988). The animals in the HF group had significantly increased body weight compared to the control animals (Figure 1), which points toward fluid congestion. We also saw signs of fluid congestion in the heart and ascites in the HF group at the time of post-mortem although this was not systematically investigated. We speculate that this fluid congestion contributed to the hyperventilation in these animals. In addition to peripheral causes, central apneas are also common in heart failure, with around 40–50% of patients being affected (Javaheri et al., 1995; Vazir et al., 2007). Central sleep apnea can be a marker of HF severity, with central apnea patients showing more advanced symptoms (Oldenburg et al., 2007). Together, these results suggest that the microembolization model replicates the respiratory imbalance seen in HF.

We report successful induction of heart failure in sheep using repeated percutaneous injection of microspheres into the coronary arteries. This approach targets the coronary circulation and thereby circumvents potential systemic effects if a toxin is administered intravenously. Since the HF is induced by a durable ischemic insult, interventions with novel pacing paradigms can be undertaken in pacing naïve hearts, avoiding the potential confounder of pacing-induced changes. Practically, repeated access to the arterial circulation is an important consideration. Previous studies have constructed carotid artery loops to allow repeated access. We percutaneously accessed the femoral arteries and following embolization, digitally compressed the puncture site for hemostasis without the use of a closure device, returning the animal to the crate/pen without any indwelling cannulae retained for future access. This has a number of advantages over other methods of surgical approach to perform the microembolization. Cardiac catheterization using these techniques requires access to fluoroscopy to guide cannulation and confirm successful and stable engagement in the coronary vessels.

There are a few limitations to this approach as well. One important limitation of this approach is the diffuse pattern of blockage and fibrosis that is observed. As such, it is difficult to quantify the region of infarcted tissue adequately as would be possible with coronary ligation. We utilized ejection fraction and fractional shortening from the echocardiographic data to quantify the decline in heart function. In addition, three out of the nine sheep did not survive the first embolization as they developed pulmonary congestion that could be resolved using diuretics. In addition to these three animals, two of the animals studied also needed administration of diuretics after the embolization procedure, and in these two cases, the congestion was resolved and the animals continued in the protocol. While the presence of congestion makes this model clinically relevant, this does necessitate increased animal numbers to complete a cohort of animals for a study.

In conclusion, our study indicates that chronic heart failure can be successfully induced in sheep using repeated injection of microspheres into the coronary arteries. After 2 months, the animals develop clinical signs of heart failure. There is marked increase in heart weight with histological evidence of ventricular fibrosis. There is a decrease in mean arterial pressure and an increase in incidence of apneas. Furthermore, pressure-volume loops show altered left ventricle dynamics in the heart failure sheep and the baroreflex challenge under conscious conditions showed a significant decrease in the gain. We conclude that this is a good model of HF to test changes in either neural control or respiratory function after interventions.

The data gathered in this study are available upon request to the corresponding author.

The animal study was reviewed and approved by The University of Auckland Animal Ethics Committee.

YA, NL, MP, IL, and RR collected the ECG data for this study. All other data were collected by RR, MP, and YA. RR supervised the study. YA completed the data analysis and wrote initial manuscript. All authors contributed in revising the manuscript toward the final version.