Background
Acute Myocardial Infarction (AMI) occurs due to cessation of coronary blood flow to the heart muscle and the extent of injury follows a wavefront pattern of necrosis in proportion to the duration of occlusion [
1]. While prompt reperfusion is favorable for myocardial salvage, ischemia-reperfusion injury (IRI) may often present itself as an adverse consequence. IRI has been associated with expansion of the prior ischemic damage and greater inflammation along with oxidative stress, apoptosis, stunning, free radicals and reperfusion arrhythmias [
2,
3]. Lethal IRI may further result in myocardial hemorrhage [
4] which, in association with microvascular no-reflow or obstruction (MVO), has been recognized to be a new independent predictor of adverse outcomes [
5‐
8].
The presentation of myocardial hemorrhage as a consequence of IRI has been well documented in both humans [
9] and animal models of AMI [
10‐
12]. With recent advances in quantitative Cardiovascular Magnetic Resonance (CMR) techniques, hemorrhage-sensitive relaxation parameters - T2 and T2* have been instrumental in the in-vivo detection of hemorrhagic byproducts [
7,
8,
13,
14], and have also made it possible to study the associated long-term consequences. A study by Mather et al. [
6] demonstrated that hemorrhage was associated with large infarct size, reduced salvage, greater MVO and lower ejection fraction; it was also the strongest independent predictor of adverse left ventricular (LV) remodeling with an increased risk of arrhythmia and predictive power even greater than MVO. Recently, a much larger study in 286 patients by Carrick et al. [
15] has shown that hemorrhage occurred in 40% of ST-segment elevation myocardial infarction (STEMI) patients and that it was independently responsible for adverse remodeling at 6 months. It was further found that hemorrhage was more associated with adverse remodeling and N-terminal pro-brain natriuretic peptide (surrogate outcomes) and cardiovascular death and heart failure following discharge (health outcomes) than MVO.
Although some of these observational clinical studies have reported the prognostic implications of myocardial hemorrhage utilizing the predictive power of CMR, the findings have not been consistent in the literature. Several other investigators have found hemorrhage to only be associated with infarct size and presence of MVO, but not offering prognostic significance beyond these parameters [
7,
16]. Given these clinical observations, a systematic and mechanistic understanding of the underlying pathophysiological implications of hemorrhage are not fully understood, particularly as an individual and an additive phenomenon over an ischemic insult.
A critical question is whether hemorrhage is simply a bystander and a marker of severe microvascular injury or does it actively contribute to downstream myocardial and microvascular damage beyond the initial ischemic insult. Furthermore, hemorrhage has always been associated with MVO [
6,
17]; however, the interaction and causality between the two is less understood. Our study was designed to probe the pathophysiological consequences of hemorrhage in a large animal model of AMI with CMR as the investigational tool. Our hypothesis was that hemorrhage is a mediator of inflammation, directly contributing to myocardial and microvascular damage in the setting of AMI.
Discussion
This is the first study to systematically and mechanistically demonstrate that hemorrhage is a critical component of ischemia-reperfusion injury. The novel preclinical model developed in this study was able to create three distinct patterns of ischemic injury – hemorrhage only, infarction without hemorrhage and infarction with hemorrhage. This offered a means of probing the contribution of myocardial hemorrhage in AMI in a progressive manner. Key findings of our study were as follows: 1) hemorrhage alone can induce a mild inflammatory response in the heart; 2) hemorrhage when combined with a mild ischemic insult, can further aggravate inflammation and result in greater myocardial injury beyond the initial ischemic insult; 3) hemorrhage can contribute towards the occurrence of an MVO; and 4) edema is correlated with the degree of hemorrhage and the hemorrhagic sites are targeted by infiltrating macrophages. Overall, these observations suggest that a significant interaction occurs between hemorrhage, the inflammatory response and ischemic injury. Thus, hemorrhage is not simply a bystander, but an active modulator of cardiac injury and the reparative processes that follow mechanical reperfusion in AMI.
The pathophysiological mechanisms leading to reperfusion hemorrhage in humans are currently unclear. A comprehensive review on intramyocardial hemorrhage in AMI has been recently published by Betgem et al. [
25]. In as early as 1960, Jennings et al. [
26] identified hemorrhage as one of the key histological features of IRI in a canine model of myocardial infarction; hemorrhage always presented itself within the infarct core [
24]. The primary mechanism for the development of hemorrhage appears to be via hypoxia-induced endothelial disruption. The absence of coronary blood flow might lead to increased endothelial stress, swelling and blebbing, further causing loss of interendothelial junction integrity. Beyond ischemia, reperfusion is thought to aggravate the endothelial junction leakiness and damage resulting in extravasation of red blood cells into the tissue interspaces. It has been also been demonstrated that permanent occlusions have resulted only in cellular swelling without hemorrhage and that reperfusion is a necessary condition for the occurrence of hemorrhage [
4,
27‐
29]. Experimental studies have indicated that hemorrhage progressively increases with the duration of coronary occlusion; hemorrhage appears after 50-60 min occlusion in canines [
26] and after 45 min occlusion in pigs [
10,
19]. Hemorrhage is thus considered a hallmark of severe microvascular damage.
Clinical studies have shown that patients with durations of ischemia >4 h are at significantly high risk of experiencing major adverse cardiovascular events, such as re-infarction, repeat revascularization, heart failure and death [
30]. These adverse outcomes appear to be highest in hemorrhagic infarcts [
5,
6], although, it is still not clear whether hemorrhage has prognostic significance beyond infarct or MVO size. In a study involving STEMI patients, Beek et al. [
16] found hemorrhage to be closely associated with infarct size, presence of MVO and cardiac function, however with no prognostic significance beyond MVO. In a study with reperfused AMI patients, Bekkers et al. [
7] further concluded that MVO and hemorrhage together only represent a consequence of a severely damaged microvasculature in large infarction and that only infarct size is an independent predictor of adverse remodeling. Thus, the true role of hemorrhage remains less understood.
To understand the consequences of hemorrhage, animal models have been utilized to show differences between hemorrhagic and non-hemorrhagic infarcts using either fixed or variable occlusion times [
14,
19]. For example, in the porcine model, it has been shown that 45 min coronary occlusions result in a non-hemorrhagic infarct [
19] and that 45 min appears to be a threshold for the presentation of intramyocardial hemorrhage [
10]; hence this model was our choice to serve as the control group. However the presence or absence of hemorrhage in these models is dependent on the physiological response of the animal to ischemia coupled with coronary architecture. Moreover, routine occlusion models cannot assess the effect of hemorrhage alone as they are always associated with ischemic injury. This is the case with clinical studies as well – clinical studies are able to determine outcomes of hemorrhage in association with infarction and MVO but cannot isolate the contribution of hemorrhage post-AMI or assess its active interaction with the healing process. In contrast, our proposed collagenase model offers a more controlled and systematic framework to interrogate the effect of hemorrhage in a progressive manner.
Collagenases are a group of metalloproteinases that can disrupt the interstitial and basal membrane collagen matrix [
31]; when injected in a blood vessel this results in an increase in wall permeability and thus spillage of blood into the extravascular space. Collagenase-induced bleeding has been widely used in small animal models of intracerebral hemorrhage and the method has been shown to be reproducible without causing tissue necrosis [
32]. Osherov et al. [
18] were the first to show the effect of collagenase in the heart. They demonstrated that a single bolus injection of bacterial collagenase into a normal coronary artery was able to induce localized myocardial hemorrhage in a dose-dependent manner. For the +HEM group, we utilized an 8 min occlusion model, similar to that used by Osherov et al., to avoid any backflow of the infusate and to ensure the collagenase effects were confined to the territory of occlusion. At high dose levels, moderate sized hemorrhages were present in both epicardial and intramyocardial tissue. In our study we chose a high dose regimen of 3.2 mg bolus to ensure the consistent occurrence of intramyocadial hemorrhage in all animals. Utilizing this model, the pattern of distribution of myocardial hemorrhage observed in our study closely approximated previously observed changes following a 90 min LAD occlusion model [
33].
Although the occurrence of myocardial hemorrhage following AMI has been known from human autopsy studies from the 1980s and preclinical tissue specimens [
9,
10,
34], this feature was neglected due to lack of sensitive imaging techniques for in vivo detection. The in vivo identification of hemorrhage with CMR is a relatively new finding, especially in humans. Shortening of T2 and T2* relaxation times have been exploited to detect and quantify hemorrhage degradation products post-AMI [
5,
6,
8,
13,
14,
17,
35,
36]; these parameters have further been validated against histopathological ground truth as well as tissue iron stores [
14,
17,
35]. Visualization of hemorrhage with CMR has established the link between hemorrhage and compromised function, major cardiac events, re-hospitalization and death [
37,
38]. Thus, there has been renewed interest in reperfusion hemorrhage due to its adverse consequences presented in the clinic.
The sole effect of hemorrhage in tissue has been extensively studied in the brain for understanding the implication of intracerebral hemorrhage [
39‐
41]. CMR relaxation parameters (T1, T2, T2*) have also been well characterized in the case of brain hemorrhage along with the alterations during hemorrhagic transformation [
42]. Animal models of brain hemorrhage have been instrumental in correlating the effects of hemorrhage alone on inflammation and adverse neurological effects [
32]. Our study is the first to utilize this collagenase model to demonstrate that hemorrhage itself may have similar detrimental effects in the heart as well. Our findings indicate that the presence of hemorrhage attracts abundant macrophages (MAC387 positive cells) to the site of injury and interestingly, the recruitment is qualitatively many fold greater than that seen with a mild ischemic insult where macrophage infiltration is scarce, even in the acute setting. This indicates that hemorrhagic byproducts themselves are inherently pro-inflammatory in the heart and that greater hemorrhage (lower T2*) is associated with a larger inflammatory response (higher T2). Note that the low inflammatory status offered by the 45 min occlusion model (I-HEM), as demonstrated earlier [
19], provided a reference for the other two groups, with which we could observe that if a small ischemic insult were coupled with hemorrhage, then the inflammatory response is further amplified. This overdriven inflammatory response may possibly be the first-stage mechanism for downstream adverse consequences in hemorrhagic infarcts.
Following successful reperfusion, the ‘no-reflow’ phenomenon is often encountered, characterized by ischemia-induced microvascular obstruction (MVO) and has been correlated with adverse left ventricular (LV) remodeling and poor patient outcome [
43]. Recent studies have indicated that the clinical presentation of MVO and myocardial hemorrhage can be as high as 25%-63% in patients with ST-segment elevation myocardial infarction (STEMI) [
5‐
8], and represents a high-risk group. It is well known now that hemorrhage and MVO are related; however, the causal relationship between the two has been less explored. It is uncertain whether MVO causes endothelial damage resulting in blood leakage, making hemorrhage simply a marker of severity, or whether pools of hemorrhagic sites compress the vessels leading to or worsening an MVO. Kumar et al. have demonstrated in a dog model that degree of hemorrhage increases with the size of infarction and also with extent of MVO [
44]. Recent findings in STEMI patients by Carrick et al. [
15,
45] demonstrate that MVO almost always precedes hemorrhage in the evolution of reperfused ischemic injury. It was shown that the temporal evolution of hemorrhage was different than that of MVO, where MVO volume peaked maximally between 4-12 h post-AMI and decreased thereafter while hemorrhage volume gradually increased from 4-12 h post-AMI, peaked at 72 h and then decreased thereafter. This study concluded that hemorrhage is a downstream consequence of MVO. This study also indicated that it is possible to observe an MVO without hemorrhage in some patients. However, it is important to note that the degree of hemorrhage is highly variable spatially within the MVO region as well as across patients i.e. it is possible to have a very small hemorrhagic region with in a very large MVO. In this regard, Carrick et al. utilized only three T2* slices in their study to detect hemorrhage, which has been acknowledged as a limitation, as a result of which regions of hemorrhage could have been missed in patients identified as having MVO but no hemorrhage.
Hence, mechanistic studies are still lacking, particularly related to complex interaction of hemorrhage with the ischemic tissue bed and vasculature. Specifically, once hemorrhage is present, does it contribute to worsening the ischemic injury and aggravating the MVO (causality)? Our study was designed to address this aspect and we speculate that vascular compression by hemorrhage and the associated inflammatory response (edema) could potentially worsen an MVO. In the + HEM group, we did not observe the classic MVO, however histology confirmed the presence of a perfusion deficit with the Thioflavin S stain. A possible explanation for this could that the regions of perfusion abnormality were too small and subtle to be detected on CMR given that this group did not present with a classical infarct. However, our findings indicate that ischemia in association with hemorrhage (I+HEM group) leads to a greater incidence of MVO suggesting a potential causal relationship between the two. Furthermore, a secondary inflammatory response arising from hemorrhage could also contribute to the no-reflow phenomenon as suggested in the literature [
46]. We would also like to note that in the clinical situation, hemorrhage is not seen in the absence of an MVO post-AMI, as observed in animals in the +HEM group and some in the I+HEM group. Our animal model was not aimed to directly replicate the clinical scenario but rather identify the pathophysiological implications of hemorrhage itself, given that it is now confirmed to be a better predictor of adverse outcome than MVO.
We did not note any significant difference in cardiac function parameters, particularly EF at 24 h post-intervention across the three groups. This could potentially be attributed to the relatively small infarct sizes observed in the I-HEM and I+HEM groups with mean infarct size of 2% and 12% (of LV mass) respectively; +HEM group did not show an infarction on CMR. Based on our previous study utilizing a 90 min occlusion model [
19], an infarct size greater than 20% might be necessary to see a significant alteration in cardiac function parameters in the early sub-acute phase. Nevertheless, further chronic studies are necessary to evaluate cardiac remodeling as a consequence of the hemorrhage induced by the collagenase model.
Limitations
We acknowledge several limitations. Firstly, the collagenase-based model for artificially induced hemorrhage is not a true representation of reperfusion hemorrhage resulting from injured/leaky microvasculature. Nevertheless, the model was able to achieve a similar microvascular response. Histological findings clearly identified the extravasation of blood in the interstitial spaces, creating a blood distribution pattern identical to that seen in a 90 min LAD occlusion hemorrhagic model of porcine AMI [
17,
19]. Thus the goal of the collagenase model was not to mimic all aspects of reperfusion hemorrhage, but rather to introduce hemorrhage in a manner similar to that seen in IRI, i.e. through injured microvasculature.
Our study also cannot address or distinguish the effects (eg. inflammatory response) of collagenase itself from that arising from hemorrhage, which occurs as a result of its administration; this is an inherent limitation of the model. Nevertheless, the collagenase model does offer an elegant methodology for creating hemorrhagic tissue, similar to that utilized in brain hemorrhage studies [
47,
48]. Our findings are in good agreement with the extensive studies investigating collagenase-induced bleeding in the brain; these studies have demonstrated that the features and consequence (edema, inflammation) of this model closely represent that seen in the case of human intracerebral hemorrhage. The collagenase model was able to achieve the intended result of blood spilling into the interstitial spaces of the myocardium. Furthermore, we would like to highlight that irrespective of the other possible effects of collagenase (if any), our study does demonstrate that hemorrhage itself induces an inflammatory response as evidenced by the presence of macrophages near red blood cells under histology, identical to that observed in the collagenase-based studies in the brain [
49].
Despite a fixed dose of intracoronary collagenase, the amount of myocardial hemorrhage induced was variable across animals; in some animals hemorrhage was predominantly epicardial with some right ventricular involvement. This would be expected in a cohort of animals since the degree of microvascular response to enzymatic action and ischemia is not predictable. Nevertheless, the selection of a high dose collagenase regimen ensured a consistent induction of myocardial hemorrhage within a given group.
Lastly, T2 measurements can be influenced by both edema and hemorrhage; T2 is elevated in edematous conditions whereas it is lowered by hemorrhage and its by products [
17,
33,
42]. Hence our study may have underestimated the intensity and extent of edema, as measured by the T2 parameter, in the hemorrhagic cases. In addition, given the absence of an infarct and uneven distribution of hemorrhage in the +HEM group, the T2-based regions of edema appeared to be non-contiguous given the nature of the injury. Yet, histology patterns agreed well with T2 in representing the inflammatory response with areas of wide interstitial spaces and inflammatory cell aggregation correlating well with regions of elevated T2 values; this has also been verified in our prior porcine study [
17].