The relationship between myocardial extracellular matrix remodeling and ventricular function^☆

Summary

Elevations in myocardial stress initiate structural remodeling of the heart in an attempt to normalize the imposed stress. This remodeling consists of cardiomyocyte hypertrophy and changes in the amount of collagen, collagen phenotype and collagen cross-linking. Since fibrillar collagen is a relatively stiff material, a decrease in collagen can result in a more compliant ventricle while an increase in collagen or collagen cross-linking results in a stiffer ventricle. If continued elevations in wall stress exceed the ability of the heart to compensate, then the ventricular wall thickness is disproportionately reduced compared to chamber volume and diastolic and systolic dysfunction ensues. This review describes the structural organization of collagen within the myocardium, discusses its effect on ventricular function and considers whether therapy aimed at reducing fibrosis is efficacious in heart failure. The evidence indicates that chamber stiffness can clearly be affected by alterations in both collagen quantity and quality, with the effect of changes in collagen concentration being modified by the extent of collagen cross-linking. The limited evidence available regarding the effects of collagen on systolic function indicates that pharmacological attempts to reduce interstitial collagen have a negative impact. Accordingly, a shift in treatment strategies directed more specifically at affecting collagen cross-linking, rather than reducing the concentration of collagen, may be warranted in the prevention of the adverse impact of collagen alterations on myocardial remodeling.

Fibrosis, Collagen, Cross-linking, Collagen types, Myocardial remodeling, Hypertrophy

1 Introduction

The elevation in myocardial stress that results from cardiac injury and/or persistent elevations in ventricular developed pressure or volume initiates a structural remodeling of the myocardium in an attempt to return the tissue stress to its normal value. This process consists of progressive remodeling of the muscular, vascular and extracellular matrix (ECM) components of the heart that manifests as changes in ventricular wall and chamber dimensions. Depending on the cause, duration and magnitude of the increase in myocardial stress, the surviving myocytes undergo hypertrophy through in-series and/or parallel addition of sarcomeres and the ECM undergoes alterations in its interstitial collagen properties (e.g., fibrillar collagen concentration, types and cross-linking). In general, the stressed ventricle is considered to be compensated if the remodeling process results in near normal diastolic and systolic function reflected by attainment of an above normal ventricular mass-to-volume ratio sufficient to normalize myocardial wall stress. However, the heart will eventually fail if the sustained increase in wall stress exceeds the compensatory ability of the heart, exhausting the hypertrophic reserve without normalizing myocardial stress. At this stage of dysfunction, ventricular wall thickness is disproportionately reduced relative to an increase in chamber volume, ventricular geometry is becoming spherical, and there is significant myocardial fibrosis.

As fibrillar collagen is a relatively stiff material that is in intimate contact with all other components of the myocardium, it plays a crucial role in the maintenance of ventricular shape, size and function [1,2]. Degradation of myocardial collagen typically results in ventricular dilatation and a decrease in ventricular stiffness, while an increase in interstitial collagen concentration and/or cross-linking results in a stiffer myocardium and ventricular diastolic dysfunction. While the effects of changes in myocardial collagen on stiffness have been investigated extensively over the past 25 years, the influence of specific alterations in collagen properties on diastolic, systolic and lusitropic function have not been clearly delineated. This review will begin with a brief description of the organization and mechanical properties of myocardial collagen, continue with a synopsis of what is known regarding ECM remodeling and myocardial function, and conclude with a consideration of whether targeted therapy directed at reducing fibrosis in the failing heart would be efficacious.

2 Myocardial fibrillar collagen structural organization and mechanical properties

Interspersed between cardiomyocytes, nerves and blood vessels, the cardiac interstitium is comprised of ground substance and connective tissue which is predominantly collagen with relatively small amounts of fibronectin, laminin and elastin. The fibrillar collagens found in the myocardium include types I, III and V. These fibrillar collagens are comprised of three peptide chains intertwined to form a right-handed super-helix. Once interstitial collagen is deposited in mature form and subsequently cross-linked, it is extremely stable and resistant to degradation. However, interstitial collagen in the myocardium is not a static protein, and changes in the balance between synthesis and degradation may lead to alterations in the composition of the collagen network in the heart [3]. While the relative proportions of these collagens in the myocardium are dependent on both species and the type of digestion utilized to characterize solubility (i.e., pepsin or cyanogen bromide), type I collagen is the predominant type (normally >50%), followed by type III (between 10 and 45%) and type V (≪5%) [4–9]. The organization of the fibrillar collagen matrix has been described in terms of three major subdivisions, consisting of (1) the endomysial component which forms intricate networks of fibers that surround and interconnect individual myocytes, including the collagen struts that connect myocytes to neighboring myocytes and capillaries; (2) perimysial fibers that arborize to form weaves of collagen surrounding groups of myocytes and collagen fibers referred to as strands which join adjacent groups of myocytes to one another, as well as tendinous extensions spanning interconnections with the epimysium; and (3) the epimysium which is a sheath of collagenous connective tissue that encompasses muscle bundles [9–12].

In the normal adult heart, morphometric assessments indicate that approximately 2–4% of the myocardium is collagen. However, because collagen is a relatively stiff material with a high tensile strength, small changes in its concentration have been shown to exert marked effects on the passive mechanical properties of the heart [13]. In addition to the concentration of collagen, the passive behavior of the myocardium may also be dependent on the relative proportions of the types of collagen, the diameter of the collagen fibers and their spatial alignment, and the degree of cross-linking. Accordingly, tissue containing predominantly type I collagen, large diameter collagen fibers, and/or a high degree of cross-linking will be stiffer than tissue composed of greater concentrations of type III fibers, relatively small diameter collagen fibers and essentially no cross-linked collagen [14]. Examples of both extremes would be tendinous tissue which is highly resistant to elongation versus relatively compliant skin tissue. Another characteristic of fibrillar collagen which influences passive tissue properties is related to the coiled, spring-like nature or crimp of the collagen fibers [13,15]. Initially as the tissue is stretched, collagen fibers uncoil with minimal resistance, but once uncoiled, the stress required to produce further elongation increases exponentially, effectively restricting further stretch of the tissue.

3 Alterations in ECM fibrillar collagen properties and left ventricular diastolic function

3.1 Increased collagen concentration

An interest in the effects of increased collagen on left ventricular (LV) diastolic function was sparked in the late 1970s and early 1980s by the following seminal studies: Bing et al. [16] reported an increase in myocardial hydroxyproline in hypertrophied LV; Caulfield and Borg [17] reported the first description of the cardiac ECM ultrastructure; and Pearlman et al. [14] documented an increase in myocardial collagen concentration in hypertrophied human hearts from postmortem patients with and without heart failure. Since then there have been numerous studies with experimental (i.e., Goldblatt and abdominal aortic coarctation induced renovascular hypertension, aortic arch constriction, and perinephritic hypertension) or genetic (i.e., spontaneously hypertensive and Dahl salt-sensitive rats) hypertension to indicate a strong relationship between increased LV collagen concentration and chamber stiffness [2,18]. While these models of chronic pressure overload also result in significant myocyte hypertrophy, it has been shown that myocyte enlargement is not the cause of the increased myocardial stiffness. For example, Narayan et al. [19] were able to prevent myocyte hypertrophy with hydralazine but not the abnormal accumulation of collagen in spontaneously hypertensive rats (SHR). A comparison between hydralazine treated and untreated, hypertrophied SHR indicated that the passive myocardial stiffness was similar in both and significantly greater than the corresponding stiffness obtained in the genetic control. Others found no change in myocardial collagen concentration and LV diastolic chamber compliance in dogs with experimental hypertension despite significant hypertrophy [20,21]. Finally, in the athlete with a significant increase in LV mass and presumably physiologic hypertrophy, diastolic function is normal at rest and enhanced during exercise [22,23]. This physiologic hypertrophy is in contrast to hypertensive patients who typically have significant diastolic dysfunction despite an increase in LV mass that is less than that occurring in athletes [24]. In view of the fact that collagen concentration is increased in humans with systemic hypertension [14] and that diastolic abnormalities in hypertensive patients are not related to increases in LV mass [25], it is reasonable to assume that hypertension-related diastolic dysfunction is the result of excessive myocardial collagen.

3.2 Decreased collagen concentration

As stated earlier, collagen is extremely stable and highly resistant to degradation by all proteinases except specific collagenases [26]. While copious amounts of collagenase were first demonstrated in the heart by Monfort and Pérez-Tamayo [27], fortunately only 1–2% of this total myocardial collagenase is normally present in its active form [28,29]. However, most of the latent collagenase is closely associated with the collagen matrix, such that a sudden activation of this ECM-bound enzyme would result in rapid and extensive collagen degradation. Indeed studies have demonstrated significant collagen breakdown occurring within 24 h following an experimentally induced myocardial infarction. Specifically, Takahashi et al. [30] reported degradation of approximately 50% of myocardial collagen in the infarcted area within 3 h of infarction coinciding with a two- to three-fold increase in tissue collagenase activity. A similar increase in collagenase activity and rapid collagen degradation was observed in ischemic reperfused, stunned myocardium [31,32].

Matrix metalloproteinase (MMP) activation and consequent collagen degradation have also been reported in the cardiomyopathic and failing heart. Gunja-Smith et al. [33] reported a 30-fold increase in collagenase activity in explanted hearts from patients with documented dilated cardiomyopathy. Subsequently, others have reported similar increases in MMP activity in patients with heart failure secondary to dilated cardiomyopathy [34–36]. While these findings suggest an association between increased MMP activity and ventricular remodeling (i.e., ventricular dilatation, sphericalization and increased compliance), they are ‘after-the-fact’ findings in that dilatation had already occurred, and did not necessarily reflect a cause and effect relationship. However, a direct cause and effect relationship between increased MMP activity and ventricular remodeling was evident from the findings in a temporal study comparing cardiomyopathic and golden Syrian hamsters between the ages of 150 and 300 days [37]. In contrast to control hearts, in which the collagen volume fraction (CVF) and MMP activity were age-invariant, there was a continual increase in MMP activity in the cardiomyopathic Syrian hamster, such that collagen degradation eventually exceeded synthesis, producing a decrease in CVF which coincided with the occurrence of significant ventricular dilatation and wall thinning.

Further insight into the relation between MMP activity, collagen degradation and adverse ventricular remodeling was demonstrated in a rat aorto-caval (AV) fistula model of chronic biventricular volume overload [38]. Within 12 h of creating the fistula, a significant increase in MMP activity occurred which was sustained for approximately the first week. As a consequence of this mast cell-mediated MMP activation, CVF was significantly decreased by the third day [39]. The onset of progressive ventricular dilatation and hypertrophy was apparent at one week post-fistula [38], and this myocardial remodeling in the initial phase of volume overload was considered to be deleterious in that it resulted in a significant depression in chamber contractility and the ventricle was clearly more compliant. As further evidence of a cause and effect relation between collagen degradation and ventricular dilatation, there are several reports indicating the ability of MMP inhibitors to attenuate ventricular dilatation using this and other experimental models of heart failure [40–43]. From these findings, one could conclude that elevations in MMP activity during the early stages of injury or elevated wall stress and the consequent degradation of fibrillar collagen are responsible for the initiation of a progressive remodeling process that eventually leads to heart failure.

3.3 Altered ratio of collagen types

The ratio of collagen types I/III has been reported to increase in the following disease states: experimentally induced or genetic hypertension [6,8,44]; in patients with dilated cardiomyopathy [45]; and in experimental myocardial infarction [46]. Whereas, in experimental animals and patients with diabetes the type I/III ratio was reported to be below that of a non-diabetic control group [47,48]. However, the functional consequences of altered amounts of type I and/or type III collagens, if any, have not been rigorously determined. Burgess et al. [44] reported a correlation (i.e., r = 0.91) between an index of LV relaxation (i.e., maximal rate of fall of LV pressure divided by mean arterial pressure) and the type I/III ratio indicating a slower relaxation rate as the amount of collagen type I increased relative to type III. Weber et al. [9] reported an early but transient increase in type III collagen relative to type I in hypertensive primates, which corresponded to an increase in myocardial stiffness. An increase in type III collagen, but not type I, was also documented in the myocardium of diabetic patients [47]. However, in addition to alterations in the ratio of collagen types in these studies there were also significant increases in total collagen concentration, which could have influenced the LV relaxation rate, myocardial stiffness or both. This was supported by a study myocardial biopsy specimens and LV function in patients with hypertension, where a significant increase in the collagen type I/III ratio was not associated with an increase in diastolic stiffness when overall collagen content was unchanged [49]. Given the variable relationship between the ratio of collagen types and myocardial stiffness, no definitive conclusions can be drawn regarding the functional consequences of such alterations.

3.4 Altered collagen cross-linking

Cross-linking is a process whereby collagen fibers are covalently linked to one another resulting in increased material stiffness and greater resistance to degradation. The formation of these cross-links occurs naturally with aging [50] and is also exacerbated in chronic diseases such as hypertension [51,52], chronic ventricular volume overload [53,54] and diabetes [48,55]. While there are several types of cross-linking, the most extensively characterized and predominant type is thought to be associated with advanced glycation end (AGE) products [56]. In most studies, the degree of cross-linking has been determined by the amount of insoluble collagen relative to soluble collagen content in the heart. Pertinent to this review, several studies have investigated the influence of collagen cross-linking on diastolic function. Norton et al. [51] concluded that increased myocardial stiffness secondary to hypertension in SHR is the result of enhanced myocardial collagen cross-linking rather than increased collagen concentration. Similarly, Woodiwiss et al. [52] demonstrated that a reduction in collagen cross-linking results in reduced myocardial stiffness and ventricular dilatation, irrespective of collagen concentration or type, in two rat models of pressure-overload induced heart failure (i.e., suprarenal abdominal aortic banding and chronic isoproterenol administration). Finally, Avendano et al. [55] reported aminoguanidine was able to prevent collagen cross-linking and the increase in LV chamber stiffness in diabetic dogs, without preventing diabetes-induced fibrosis. From these studies, it is obvious that the degree of collagen cross-linking has a strong impact on diastolic function, which is independent of myocardial collagen concentration. However, this should not be interpreted as suggesting myocardial stiffness is not influenced by alterations in myocardial collagen concentration. Rather, as Badenhorst et al. [57] elegantly demonstrated using experimental models of pressure overload, myocardial and LV chamber stiffness is affected by changes in both collagen quantity and quality (i.e., cross-linking), with the effect of changes in collagen concentration being modified by collagen quality.

4 Alterations in ECM fibrillar collagen and left ventricular systolic function

In contrast to the preponderance of myocardial ECM studies focused on diastolic function, there have been relatively few which have attempted to assess the effects of ECM alterations on myocardial systolic function. Baicu et al. [1] studied isolated papillary muscle from normal rat and feline hearts, as well as from felines with pressure overload induced hypertrophy and fibrosis, that were treated with plasmin to produce an acute activation of MMPs and disruption of the ECM. This decrease in collagen in the papillary muscles was accompanied by a marked decrease in systolic function (i.e., reduced amount and rate of shortening, reflected by developed tension (T) and +dT/dt) without affecting isolated cardiomyocyte contractility, leading the authors to conclude that myocardial fibrillar collagen is a major contributor to normal myocardial systolic performance. In contrast to these findings, Lamberts et al. [58] reported myocardial function was improved following specific degradation of collagen type I (i.e., an increase in developed tension and the rate of relaxation). Unfortunately, relaxation rates were not reported in the Baicu study, precluding comparison of this parameter. When considering this disparity, however, it should be noted that the preparation used by the Lamberts group was perfused, and consequently significant edema, evident by a significant increase in the diastolic tension-length curve, accompanied the disruption of collagen. In their discussion the authors concluded that an edema-induced Starling response may have masked the functional effects of collagen removal. Recent studies in our laboratory using an isolated, beating rat heart preparation, resulted in significant decreases in the peak isovolumetric pressure (25%) and in the maximal rates of LV pressure development (80%) and relaxation (90%) following an acute decrease of approximately 50% in collagen concentration (unpublished results). While edema also occurred in these studies, collagen degradation resulted in significant dilatation without a change in the overall compliance of the myocardium [59,60]. Thus, our findings are in agreement with the conclusion of Biacu et al. [1] that normal amounts of interstitial collagen contribute to the maintenance of systolic function.

Significant fibrosis is known to occur in human hypertensive, hypertrophied hearts and in the dilated failing heart regardless of etiology [14,61–63]. This has led to the notion that this fibrosis is at least partially responsible for the diminished ventricular function. Accordingly, one could hypothesize that reducing the amount of fibrosis would be beneficial to the failing heart. Thus far, the impact of ECM alterations on functional deterioration in heart failure has been primarily derived from studies of left ventricular assist device (LVAD) performance as a bridge to recovery/transplantation. There is currently intensive interest in identifying those patients who can be successfully weaned from LVAD support [64–66]. The supposition has been that reversal of the deleterious remodeling in heart failure is responsible for the observed clinical improvement in LVAD patients. Some findings appear to corroborate this expectation, with several studies reporting reverse remodeling characterized by significant reductions in the extent of myocardial fibrosis and cardiomyocyte size [67–71]. However, an equivalent number of studies have found improvement in functional parameters despite significant increases in myocardial collagen [72–75]. The improvement in systolic function observed in both cases suggests that the benefit of LVAD support was independent of effects on myocardial fibrosis. This was a possibility raised by Li et al. [76], who found minimal differences in total collagen, but significant increases in collagen cross-linking in LVAD supported hearts. Thus, a definitive cause and effect relationship between fibrosis and depressed systolic function has not been established.

Furthermore, there is evidence to indicate that a reduction in the myocardial collagen matrix could further exacerbate the severity of heart failure. As mentioned above, Baicu et al. [1] reported a decrease in systolic function in papillary muscles from pressure overloaded right ventricles following ECM disruption. Another study indicating that inducing collagen degradation may be contraindicated utilized the spontaneously hypertensive heart failure (SHHF) rat [41]. In the SHHF model, significant elevations in myocardial MMP activity occur between 9 and 13 months of age. During this interval, clinical signs of congestive heart failure become evident and the heart becomes functionally depressed, markedly enlarged and abnormally compliant, such that it can no longer maintain the elevated pressures. However, treatment with a broad-spectrum MMP inhibitor starting at 9 months of age, which presumably prevented collagen degradation, was cardioprotective in that the transition to a decompensated state did not occur [41]. The disconnect between fibrosis and impaired systolic function is also reflected by our findings in the AV fistula model, where LV contractility progressively declined despite myocardial collagen concentration remaining at normal levels prior to the development of heart failure [39]. Finally, if myocardial fibrosis is primarily due to cardiomyocyte necrosis (i.e., reparative or replacement fibrosis), then it is doubtful that collagen degradation would be beneficial. In this case, the heavily cross-linked scar tissue is less vulnerable to degradation than the normal endomysial collagen responsible for maintaining integrity of the myocardial tissue. Significant degradation of this ultrastructural collagen coinciding with the development of ventricular dilatation was noted in the cardiomyopathic hamster even though total collagen remained above normal due to the extensive scarring resulting from myocyte necrosis [77].

While therapy to reduce fibrosis may be contraindicated, the use of pharmacological agents to reduce the degree of cross-linking warrants further clinical investigation. Studies in older animals with and without experimental diabetes using such agents have reported marked improvement in both diastolic and systolic function [48,50,55,56]. The cross-link breaker, alagebrium chloride (ALT-711) has been tested in a clinical study using 21 aged patients with diastolic dysfunction. In this 16-week, open-label study, LV diastolic filling, peak oxygen consumption, aortic distensibility and LV ejection fraction and mass before and after treatment was non-invasively assessed. The results indicated a small but significant reduction in mass and a slight improvement in some of the early filling parameters with no change in the remaining variables. Despite the limited improvement during this relatively short study, these clinical results together with the encouraging animal findings justify additional larger clinical trials of longer duration.

In summary, the properties of the interstitial collagen matrix influence the stiffness of the myocardium, thereby contributing to the development of diastolic dysfunction. While changes in the relative distribution of collagen types occur during the course of myocardial remodeling secondary to sustained elevations in pressure or volume, the functional consequences remain to be determined. In contrast, chamber stiffness can clearly be affected by alterations in both collagen quantity and quality, with the effect of changes in collagen concentration being modified by the extent of collagen cross-linking. The limited evidence available regarding the effects of collagen on systolic function indicates that pharmacological attempts to reduce interstitial collagen have a negative impact. Therefore, a shift in treatment strategies for heart failure directed more specifically at affecting collagen cross-linking, rather than reducing the concentration of collagen, may be warranted in the attenuation of the adverse impact of collagen alterations on myocardial remodeling.

☆

Presented at Cardiodynamics: Ventricular Wall Structure and Cardiac Function, Institute of Biomedical Engineering, University and ETH Zurich, Switzerland, November 4–5 2005.

Acknowledgements

This work was supported in part by grants from: NHLBI (RO1-HL-59981, HL-62228, HL-73990 and F32 HL072566) and from the American Heart Association (0435298N).