Overview of the current animal models for heart failure

The second challenge for HF preclinical research is the selection of the experimental model. Several previous and recent reviews provide a framework for the present discussion (Gomes, Falcao-Pires, Pires, Brás-Silva, & Leite-Moreira, 2013; Houser et al., 2012; Patten & Hall-Porter, 2009; Riehle & Bauersachs, 2019; Y.-T. Shen et al., 2017; Silva & Emter, 2020). Building on these works, we introduced current preclinical animal models according to the most common triggers for HF, including ischemia, hypertension, dilated cardiomyopathy and valve disease. Since the process of generating various animal models has been described in many previous studies, we refrain from a detailed description of each modeling method. Instead, we emphasize the pathology features of HF that can be presented by the model and their application limitations (Figure 1, Table 2 ). Additionally, we summarized significant changes of hemodynamics and echocardiography derived parameters in each animal model (Table 2 ), which would provide practical guidance for researchers who considering preclinical studies for the treatment of HF.

3.1 HF induced by myocardial ischemia

Myocardial ischemia is one of the most common causes of HF, the imbalance between myocardial blood supply and demand can be spontaneous or precipitated by coronary artery disease (e.g., atherosclerosis or thrombosis) or can occur during periprocedural revascularization surgery performed to restore spontaneous ischemia (e.g., percutaneous coronary intervention or coronary artery bypass grafting) (Thygesen, Alpert, White, & Infarction, 2007). There are three main techniques in the model of HF induced by myocardial ischemia/infarction: permanent (irreversible) coronary occlusion, ischemia/reperfusion and coronary microembolization. Consistent with the progress of clinical patients, these three techniques combine the characteristics of HF induced by myocardial ischemia, including the initial ischemic event, followed by decreased cardiac output and ejection fraction, eccentric hypertrophy, focal fibrosis in the ischemic area, activation of neurohormone system and decrease of cardiac reserve (Houser et al., 2012). Correctly identifying the response of different types of myocardial infarction to ischemia is not only paramount for optimizing the management of patients, but also an important feature to simulate the transformation study of acute myocardial infarction induced HF.

3.1.1 Permanent coronary artery occlusion

Coronary artery ligation that simulates myocardial infarction (MI) is the most commonly used model of HF in small animals, which was first described in mice by Zolotareva et al (Zolotareva & Kogan, 1978). After that, the pioneering study in rats by Pfeffer et al.(J. M. Pfeffer, Pfeffer, & Braunwald, 1985) proved that infarct size, post-MI LV chamber dilatation and LV function are correlated. Their subsequent findings regarding the improvement of Captopril (an ACE inhibitor) on the contractile function and survival rate of rats (M. Pfeffer, Pfeffer, Steinberg, & Finn, 1985) and patients (M. A. Pfeffer et al., 1992) after myocardial infarction, contribute to the establishment of the pharmacological ACE inhibitor treatment strategy, which has become a standard first-line treatment for HF.
As is well known, the infarct size is closely related to the duration of coronary artery occlusion. In most species, the shortest time required for coronary artery occlusion to induce MI should not be less than 20-30 minutes, which provides a definite and smallest size of MI. In the study of coronary artery ligation in small animals, the surviving mice gradually developed into HF during the 2.5 - 9 week period post-surgery (Gao, Dart, Dewar, Jennings, & Du, 2000). The infarct size varies significantly from 15.2% to 55.2% of the left ventricle mass, which is directly related to the degree of left ventricular dysfunction and affects the time course of HF development (Gao et al., 2000; M. A. Pfeffer et al., 1979). Rats with infarctions greater than 46% had congestive heart failure, with elevated filling pressures, reduced cardiac output, and a minimal capacity to respond to pre- and after load stresses. In addition, other factors, including age (Gould et al., 2002), sex (Wu, Nasseri, Bloch, Picard, & Scherrer-Crosbie, 2003) and genetic background (van den Borne et al., 2009) of animals also exert complex integrated effects on the time course of HF development induced by this model.
In large animals, such as pigs and sheep, irreversible coronary occlusion can be performed in the left anterior descending coronary artery (LAD) or left circumflex coronary artery (LCx) by acute ligation(Van der Velden et al., 2004), or thrombogenic coil (CHARLES, ELLIOTT, NICHOLLS, RADEMAKER, & RICHARDS, 2000), hydraulic occlude (Y. T. Shen & Vatner, 1996) and ameroid constrictor (Chekanov et al., 2003; O’konski, White, Longhurst, Roth, & Bloor, 1987) placement. The former is an immediate approach to develop acute MI, whereas hydraulic occlude, thrombogenic coil or ameroid constrictors placement can mimic MI resulting from coronary stenosis due to progressive atherosclerotic plaque formation. These methods can produce phenotypes similar to those of small animals, including the initial observation of MI and the subsequent induction of a significant decrease of ejection fraction by 25% to 35% (J. Zhang et al., 1996). Besides, this model can also increase LV weight and LV end-diastolic area. It is worth noting that the location of coronary artery occlusion or ligation is related to the infarct size. In pigs or dogs, compared with LCx ligation, LAD occlusion leads to a relatively larger infarct size(Becker, Schuster, Jugdutt, Hutchins, & Bulkley, 1983). Additionally, infarct size differs among species and among lines of the same species in this model. For instance, beagle dogs have a larger infarct size than mongrel dogs(Uemura et al., 1989). Other studies that examined obese Ossabaw swine by LCx ameroid constrictor placement resulted in a model of chronic ischemia as opposed to MI-induced HF because no or minimal subendocardial signs of infarction were observed (Elmadhun, Lassaletta, Chu, & Sellke, 2013).
One potential limitation of the irreversible coronary occlusion animal model is permanent occlusion of myocardial vascular flow, which is contrary to the clinical features of patients with progressive non-occlusive coronary artery occlusion. Patients have infarct-related arteries due to spontaneous thrombolysis or reperfusion by fibrinolytic enzyme or balloon angioplasty, which cannot be reproduced in this model. In addition, this model has a high mortality rate due to acute heart failure or LV rupture within the first week post-MI (Gao et al., 2000).

3.1.2 Ischemia-reperfusion model

To overcome the limitations of irreversible coronary artery occlusion in animal model, ischemia/reperfusion (I/R) models have been established in different species. Ischemia/reperfusion will block the coronary blood flow to the myocardium and then reintroduce the blood flow into the ischemic area. This is helpful to study the molecular mechanisms and tissue damage after temporary occlusion of LAD. Compared with permanent left anterior descending branch occlusion, ischemia/reperfusion model shows less tissue damage compared to permanent LAD occlusion. Advantages of this model include a high survival rate, adjustable MI size according to coronary artery occlusion site, reproducible cardiac dysfunction and relatively low invasive methods (Bikou, Watanabe, Hajjar, & Ishikawa, 2018).
Lots of works has been carried out on I/R model in small animals (Michael et al., 1995; Podesser et al., 2002; Reitz et al., 2019; Shimizu et al., 2018), from which left coronary artery occlusion for 45-60 minutes followed by reperfusion usually leads to HF 4 weeks post-surgery in mice (Reitz et al., 2019; Shimizu et al., 2018), with signs of HF including increased left ventricular (LV) internal diastolic (LVIDd) and systolic (LVIDs) dimensions, along with decreased ejection fraction (% EF) and fractional shortening (% FS) (Reitz et al., 2019). In large animals, I/R is commonly used in left anterior LAD or LCx, which is produced by reversible ligation or inflation of an intracoronary angioplasty balloon (Bikou et al., 2018). Although dogs have been used for I/R research in history, the extensive coronary collateral circulation in canine hearts limits the use of this model. As pigs and sheep share the similar coronary arteries as humans (including gross anatomical structure and absence of existing collateral vessels), the I/R approach can be used to induce infarcts of predictable size and location in these species (Y.-T. Shen et al., 2017). Moreover, the infarct size produced by I/R may vary with different surgical sites, in general, using the I/R protocol of proximal occlusion of LAD would lead to more serious diseases, which indicates that it may be a better preclinical model (Ishikawa et al., 2014). The pathological features of HF induced by I/R include the acute increase of natriuretic peptide level and cardiac troponin T, the progressively increased LV volume and the decreased EF (CHARLES et al., 2000). Limitations of pig and sheep I/R models include severe acute susceptibility to arrhythmia, and difficulty in imaging the heart with ultrasound techniques due to differences in ruminant dependence in gastrointestinal anatomy.

3.1.3 Coronary microembolization

Repeated coronary microembolization not only avoids the risk of severe acute myocardial infarction, but also mimics the chronic ischemic process caused by repeated ischemic injury. Contrary to coronary artery ligation or ameroid constrictor placement, HF with different degrees of left ventricular dysfunction can be produced in closed chest model by repeated coronary microembolization. This protocol has been investigated in very distinct species, but predominantly apply to the practices of translational study in large animals, by injecting glass (Franciosa, Heckel, Limas, & Cohn, 1980) or polystyrene microspheres(Sabbah et al., 1991) (with various diameters depend on materials) into the LAD or LCx coronary artery. The regional contractile dysfunction induced by microspheres is proportionate to the number of injected particles (Erbel & Heusch, 2000). In a study on canine model three months post-surgery (Sabbah et al., 1991), reductions in cardiac output and function were observed and were associated with a transmural MI distributed throughout the LV, septum, and right ventricle. Signs of HF include decreased EF, LV dilatation (increased end diastolic volume), increased pulmonary artery wedge pressure, systemic vascular resistance and plasma atrial natriuretic peptide and/or norepinephrine. This model has been successfully used in many pharmacological studies aimed at preventing the remodeling process of LV failure. In a separate study, Sabbah et al. showed dopamine β -hydroxylase treatment can reduce the continuous increase of end diastolic and end systolic volumes, thus preventing progressive LV dysfunction and remodeling (Sabbah et al., 2000). In another similar study, long-term treatment with bosentan (a mixed endothelin-1 receptor antagonist A and B) can prevent the progress of LV dysfunction and weaken LV chamber remodeling in dogs with moderate HF induced by multiple consecutive intracoronary embolization (Mishima et al., 2000).
Repeated coronary microembolization can mimic the chronic effect of myocardial ischemia increasing with time. However, this process requires an intracoronary injection technique, which is usually performed under the anesthesia of a special catheter inserted into a specific area or implanted into a coronary artery for a long time. In addition, it is difficult to find a consistent protocol for the number of injections and the amount of microembolization (Y.-T. Shen et al., 2017). Therefore, this model is not easy to be repeated by most laboratories.

3.2 HF induced by systemic hypertension

In the United States, 46% of adults suffer from hypertension, which is the main risk factor for brain, heart and kidney events (Virani et al., 2020). Moreover, 17% of patients with HF were induced by hypertension (Virani et al., 2020). In understanding that hypertension pathophysiology is complex and not definitive, it is difficult to establish an animal model simulating clinical hypertension.

3.2.1 Transgenic animal models

The most widely used model of HF induced by hypertension is transgenic animal models, including spontaneously hypertensive rat (SHR), spontaneously hypertensive HF (SHHF) rat and DS rat, etc.
As early as 1960s, Okamoto et al. (Okamoto & Aoki, 1963) have produced spontaneously hypertensive rat (SHR), which is a natural model of pressure overload. Systemic hypertension in SHR mice gradually attacks with age and leads to obvious CHF at the age of 18–24 months, which is characterized by increased end-diastolic and end-systolic volumes and depressed EF in LV (Bing et al., 1995). With these features, this model is suitable for studying the transition from hypertrophy to CHF and replicating CHF induced by hypertension in human. On the basis of SHR mice, a further developed subspecies, SHR stroke prone (SHR-SP), was developed, which would quickly develop severe hypertension with salt diet and has a strong tendency to die of stroke within a few weeks (Yamori, 1991). Another commonly used Transgenic lines for HF study is Spotaneously Hypertensive HF (SHHF) rat, which carry the facpobesity gene that encodes a defective leptin gene, to enable them develop noninsulin-dependent diabetes mellitus, obesity, hypertension and CHF (McCune, Baker, & Stills Jr, 1990; McCune, Park, Radin, & Jurin, 1995). Of notice, the time course for the development of HF depends on the dosage of facp gene (McCune et al., 1995). Besides, hypertension, plasma renin activity and male gender are independent factors contributing to cardiac hypertrophy and HF in this rat model (Holycross, Summers, Dunn, & McCune, 1997). When compared to SHR rats, SHHF rats would have suffer from HF earlier with loss of cardiac function begins at age of 15 months (Heyen et al., 2002). And similar to SHRs, male animals are more likely to suffer from HF than female animals (Park, Liu-Stratton, Medeiros, Mccune, & Radin, 2004). During the same period SHR rat emerged, a model of salt sensitivity, namely the Dahl-salt-sensitive (DS) rat was also developed, characterized by susceptibility to develop different degrees of hypertension depending on the variations in sodium intake (Dahl, Heine, & Tassinari, 1962). DS rats fed with high diet at age of 7 or 8 week would develop different phenotypes of hypertensive heart failure (Doi et al., 2000), which is suitable for studying the transition from compensatory hypertrophy to failure. Generally speaking, the main advantage of these models is that they avoid the complications related to surgery or drug intervention, and can imitates the clinicopathologic features of HF induced by human essential hypertension.

3.2.2 Renal artery stenosis

In addition to genetic animal models, the most direct way to mimic the pathological changes of patients with hypertension is to perform surgery through kidney clipping or wrapping. As early as 1934, since Goldblatt et al. introduced an elevation of blood pressure by constriction of both main renal arteries of dog (Goldblatt, Lynch, Hanzal, & Summerville, 1934), the Goldblatt Hypertension method (including Two-kidney, one clip [2K1C] and one-kidney, one clip [1K1C] hypertension) has been extensively studied and was successfully applied to the renal induced models of hypertension in rats (Junhong et al., 2008), rabbits (Kumagai, Suzuki, Ryuzaki, Matsukawa, & Saruta, 1990) and sheep(Lau et al., 2010). Animal undergo 2K1C or 1K1C surgery would induce clinical signs of HF. In a rat model study, 2K1C can induce systemic hypertension and LV concentric remodeling within 8 weeks (Junhong et al., 2008). In another, 1K1C on sheep for 3-11 weeks showed elevated blood pressure, enlarged LA and reduced LAEF (Lau et al., 2010). Another model named ”Page model”, which can be produced by wrapping one or both kidneys in cellophane, has been historically used for dogs to develop systemic hypertension (Page, 1939). Systemic hypertension induced by bilateral renal wrapping on aged dogs can model several aspects of structural and functional characteristics described in the limited studies of human diastolic HF, including LV hypertrophy, myocardial fibrosis, and isolated diastolic dysfunction, which provides insight into the pathogenesis of diastolic HF (Munagala, Hart, Burnett Jr, Meyer, & Redfield, 2005).

3.3 HF failure induced by dilated cardiomyopathy

The phenotype of Dilated cardiomyopathy (DCM) is induced by many primary and secondary causes(Houser et al., 2012). Primary diseases only affect myocardium (idiopathic DCM), whereas secondary causes are extensive, with most commonly clinical signs being coronary artery disease, myocardial infarction (ischemic cardiomyopathy) and long-term hypertension. Other causes include myocarditis (especially viral), chemotherapy drugs (such as anthracyclines), persistent and inappropriate tachycardia, etc. Consistent with these clinical incentives, DCM can be induced by the above-mentioned ischemic injury/myocardial infarction and systemic hypertension, as well as toxic injury, rapid ventricular pacing, pressure overload and volume overload described below. These models can be used to define hemodynamic, mechanical, neurohormonal, cellular and molecular changes during HF, and to evaluate the potential efficacy of new therapies.

3.3.1 Chemical induced model

Both small animals (e.g. mice and rats), or large animals (e.g. dogs, sheep and pig) can be used to as toxic DCM model induced by intracoronary adriamycin or isoprenaline. These chemical agents can produce dose-dependent dilated phenotype associated with arrhythmias, myocyte loss, and fibrosis, with progression to heart failure. Sheep undergo sequential intracoronary injections of adriamycin causing a dilated phenotype with dramatically reduced EF and increased LV volumes, resulting in decrease of LV wall thickness and eventually in HF (Borenstein et al., 2006). These models are particularly attractive as it easy to perform in common laboratories without the need of surgical procedures. However, the disadvantages of these methods are also obvious, including the variability of response to doxorubicin, the degree of LV dysfunction, animal mortality caused by arrhythmia, and engenders several systemic side effects, such as bone marrow suppression, gastrointestinal discomfort.

3.3.2 Rapid ventricular pacing model

DCM induced by persistent and inappropriate tachycardia can be simulated by rapid ventricular pacing model, which is commonly applied to large animal studies, mostly in the canine model (Byrne et al., 2002; Shannon, Komamura, Shen, Bishop, & Vatner, 1993; Spinale et al., 1997; Takagaki et al., 2002). Rapid pacing can be performed by pacing leads implanted in the atrium, right atrium or left atrium, to lead to a rapid and significant decline in cardiac function, thus meeting the hemodynamic conditions of HF. The severity of cardiac dysfunction depends on the two factors: pacing rate and duration, of which the pacing rate is controlled by programmable internal or external pacemakers in the range of 200-300 bpm according to specific conditions, while the duration of pacing varies from a few days to several weeks (Y.-T. Shen et al., 2017). This model can cause high output HF, accompanied by biventricular dilatation, a deterioration of systolic and diastolic function and activation of neurohormone axis. Along with progressive HF, plasma catecholamine levels increase. At more advanced stages, the levels of endothelin and renin increased and the density of β receptor decreased, all of which closely resembling the phenotype of CHF in human. However, a major drawback of rapid ventricular pacing model is that the cardiovascular hemodynamics and biochemical alterations would progressively revert to near baseline levels after pacing stopped (Byrne et al., 2002).

3.3.3 Transgenic animal models

In small animals, two mouse strains were established in rodents to simulate the development, progression and regression of human DCM. A muscle LIM protein (MLP)-/-mouse developed by the deletion of actin-related cytoskeleton protein (Arber et al., 1997), which gradually decreases heart function from 4 to 6 months after birth and continues to develop into dilated cardiomyopathy with hypertrophy and HF. The other is calcium binding protein calsequestring (CSQ) mice produced by cardiac-restricted overexpression of the CQS in the (Sato et al., 1998) heart. By echocardiography, CSQ mice showed mild LV enlargement at 7 weeks, and mild decreased fractional shortening with increased wall thickness. At age of 14 weeks, the phenotype developed into obvious LV enlargement and severe systolic function decline (Cho et al., 1999). In addition, these two mouse strains replicate other aspects of human dilated cardiomyopathy, including abnormal β -adrenergic receptor (β-AR) signal transduction.

3.4 HF induced by dilated cardiomyopathy

Typical symptoms of HF, including shortness of breath, peripheral and pulmonary edema, and low exercise tolerance, can be caused by two types of valve lesions: aortic stenosis (AS) (abnormally high ejection resistance prevents the valve from fully opening) or mitral regurgitation (MR) (a failure of complete coaptation of the leaflets and adequate closure) (Houser et al., 2012). Although these two types of valve disease will eventually lead to HF events such as LV diastolic/atrial pressure, fluid retention and fatigue, they have different underlying pathophysiological characteristics. The former can lead to significant LV pressure overload, whereas the latter causes a significant LV volume overload.

3.4.1 Transverse aortic constriction

In small animals, the most widely used model of aortic stenosis is the Transverse aortic constriction (TAC), which is characterized by an initial compensatory phase with concentric LV hypertrophy, followed by enlargement of the cardiac chamber with further deterioration of LV function, and eventually leading to HF. These characteristics are similar to the progression of human HF, particularly in patients with aortic stenosis. Therefore, TAC is often used to induce chronic LV pressure overload and hypertrophy in rodents (Boluyt et al., 2005; Li et al., 2012; Litwin et al., 1995). Several TAC surgical techniques have been developed, including a minimally invasive approach by making a small incision in the proximal sternum, and placement of surgical clips, sutures, or O-rings to prevent blood flow through the aortic arch (Riehle & Bauersachs, 2019). Besides constriction at TAC, the surgical constriction can also be performed at other locations, including ascending aortic constriction (AAC), or abdominal aorta. These models result in varied effects depending on the anatomic location of the constriction, in general, constriction at AAC is used to study the effects of early injury due to pressure overload, while TAC and suprarenal aortic coarctation appear as more gradual increases in pressure leading to hypertrophy and HF (Gomes et al., 2013). Moreover, factors that affect the hypertrophy response and progression in HF include sex, body weight, age, and genetic background of the species used (Barrick, Rojas, Schoonhoven, Smyth, & Threadgill, 2007). Several studies have also described in large animal models with progressive aortic contraction on supravalvular position (Moorjani et al., 2006; Tagawa et al., 1998; Ye, Gong, Ochiai, Liu, & Zhang, 2001). These animal models replicate many key characteristics of human aortic stenosis, including the gradual increase of LV aortic pressure gradient and compensatory LV remodeling reaction, accompanied by myocardial hypertrophy, cellular hypertrophy with evidence of diastolic HF and abnormal myocardial matrix. The shortcomings of TAC model are that they do not possess some of the key features of human disease, including the inability to easily induce slowly progressive stress overload, and other limitations include the long duration of the experimental protocol, variability in the individual response to stress overload, and a reduction in the high proportion of constriction due to internalization of the contractile knot.

3.4.2 Chordae tendineae cutting

Mitral valve insufficiency leads to chronic volume overload, resulting in compensatory LV hypertrophy, followed by dilatation with pulmonary edema. In large animals (most commonly the dog or sheep), significant clinical phenotype of chronic mitral regurgitation can be induced by cutting chordae tendineae (Leroux, Moonen, Pierard, Kolh, & Amory, 2012; Nielsen et al., 2003). MR occurs in canine model leads to LV dilatation and eccentric LVH pattern, which is accompanied by the increase of muscle cell length, the change of myofilament structure and the decrease of myofibril content. Different from LVH in large AS animal model, chronic MR in dogs can lead to severe LV contractile dysfunction at both levels of chamber and myocyte (Spinale et al., 1993). This chronic MR model has been successfully used to test the role of angiotensin II receptor pathways in the progression of HF (Perry et al., 2002). Although the use of angiotensin II type-1 (AT1) receptor blocker (AT1RB) in this study reduced systemic vascular resistance and local expression of the renin-angiotensin system in the MR dog model, it did not prevent adverse left ventricular and myocardial cell remodeling in the early myocardial adaptive phase of MR.
To date, there is no universally accepted model for MR in small animals. Some such attempts have been made by Min et al. (Kim et al., 2012), who develop a rat MR model by inserted a fine needle (0.36 mm in diameter) into LV through the apex of LV, followed by pushing a needle into the mitral valve to puncture and/or tear the mitral leaflets to form MR. In this model, the time course of LV remodeling and functional change is similar to that of human. Since this MR model was created by perforating the mitral leaflet, it does not represent MR in human, particularly ischemic MR, which is an inherent limitation of animal models of valvular disease. However, it can still be used as an inexpensive model for evaluating the effects of drug therapy on MR-induced LV remodeling, which is important for clinical translational research.