Feasibility of detecting myocardial infarction in the sheep fetus using late gadolinium enhancement CMR imaging

Myocardial infarction (MI) is one of the leading causes of heart failure [1‐5]. Every year, at least one million people will have an MI in the United States [3]. Despite modern reperfusion and drug therapies, five year post-MI mortality remains over 20%, mainly due to heart failure from adverse left ventricular (LV) remodelling after MI [3‐5]. In the past two decades, the development of the late gadolinium enhancement cardiovascular magnetic resonance imaging (LGE CMR) technique has allowed for the accurate and reproducible non-invasive diagnosis of MI in adult humans and animals [6‐10]. LGE imaging, which uses a gadolinium chelate as an exogenous contrast agent to shorten myocardial T1, has enabled the accurate diagnosis of MI, evaluation of cardiac viability, and reliable quantification of even small areas of myocardial scar [8‐12]. Despite these advances in human and animal studies, to the best of our knowledge, the use of LGE CMR to assess myocardial injury in the fetus has not been investigated. Although not currently applicable to humans, the visualization of MI in the fetus in animal research could be particularly beneficial for comparing the physiological response to MI between fetal and adult hearts and determining proof of principle for mechanical and/or pharmaceutical interventions.

In contrast to the response to MI observed in human adults, the developing fetal heart is capable of cardiac regeneration after MI [13‐16]. This is consistent with the well-studied pattern of cardiac repair in neonatal rodents and in zebrafish throughout life [16, 17]. The difference between species in terms of the response to injury in relation to birth is due to the timing of the maturation of cardiomyocyte development [18‐21]. Thus, in humans and sheep, where cardiomyocytes undergo maturation before birth, cardiac regeneration occurs more readily in fetal life than in the postnatal situation [22]. A fetal sheep model of MI has allowed researchers to investigate the mechanisms involved in fetal myocardial regeneration. The fetal response to MI is characterized by a distinct gene expression profile, elevated myocardial proliferation, diminished inflammation, lack of fibrosis, and better restoration of cellularity and cardiac function within 30 days when compared to adults [13‐15]. Developing a method of detecting and monitoring the post-MI fetal heart could therefore help researchers study the mechanisms of cardiac regeneration in fetuses post-MI by promoting or inhibiting repair. Serial CMR could aid these investigations by allowing the visualization of cardiac tissues and the tracking of changes in infarct size and cardiac function.

Therefore, in the current study, we aimed to investigate the feasibility of LGE CMR to detect MI in the sheep fetus. We hypothesized that LGE CMR would reveal enhancement in the infarcted area of the fetal heart and allow us to reliably identify MI in utero.

Sheep are one of the most commonly used large animals for research into perinatal cardiac physiology because of the similarities with humans in terms of the timing of development relative to birth, fetal size and the number of foetuses [13‐15, 22]. Unlike the adult heart, the fetal heart responds to MI with a different gene expression profile, better recovery of cardiac function, diminished inflammatory response and no fibrosis [13‐15]. Understanding the mechanism of fetal cardiac regeneration could help us develop ways to regenerate the myocardium in the injured adult heart. However, to date, no method has yet been developed to visualize the MI site in utero. It is therefore important to develop a non-invasive tool that monitors the response to MI in the fetus over time. Such a tool would allow us to visualize and potentially quantify the regenerative capacity of cardiac tissue over gestation through quantification of infarct size and cardiac function. Herein, we have shown that CMR offers a unique solution that allows us not only to monitor the infarcted area with LGE imaging, but also to assess the functional capacity of the infarcted heart with cine imaging. To the best of our knowledge, this is the first study that employs contrast enhanced imaging to detect MI in the fetus.

The implementation of CMR imaging in the fetal heart has traditionally been challenging, in part due to the difficulty in detecting the fetal heart rate for cardiac triggering. To solve this problem, several groups have developed non-invasive methods of cardiac triggering suitable for human fetuses [34‐41]. Here, we employed an invasive technique similar to a method previously described by Yamamura et al. [28], and found that gating to the carotid pulse pressure was reliable. Another potential challenge to performing LGE CMR in the fetal heart arises from possible differences in gadolinium contrast handling. In normal adult cardiac tissues, the contrast agent distributes in the extracellular space, while remaining excluded from the intracellular space [42]. However, after acute irreversible myocardial injury, contrast enters the intracellular space due to disruptions in cell membranes and becomes trapped for a period of time in reperfused cardiac tissue. The fetal myocardium has different cardiomyocyte characteristics, extracellular matrix components, and response to MI, compared with the adult heart [13‐15, 43]. For example, fetal cardiomyocytes are mononucleated and proliferative [44, 45]. The extracellular matrix volume is lower, and there is a reduction in wound healing and extracellular matrix deposition gene activation [15]. We were therefore uncertain about whether the gadolinium contrast agent would behave in a predictable way. However, our results show that despite known differences in microstructure and response to injury, gadolinium contrast agent is similarly trapped in the infarcted myocardium of fetal and postnatal hearts. Thus we have demonstrated, for the first time, that gadolinium contrast agent can be used in the fetus for LGE CMR.

Our results indicate that LGE CMR was successful at identifying the site of myocardial injury 3 days after ligation of a branch of the LAD coronary artery, but was unsuccessful immediately after the surgery (Day 0). This is perhaps not surprising, as LGE hyperenhancement occurs at the infarction site when the gadolinium contrast agent passively diffuses into and becomes trapped in the intracellular space after the loss of myocyte membrane integrity following necrosis [46]. We suspect that the lack of enhancement at the infarct site immediately after ligation (on Day 0) is likely because necrosis had not yet occurred at a cellular level. As a result, insufficient gadolinium had entered the intracellular space due to a lack of membrane rupture. We were therefore unable to see hyperenhancement on the LGE images. This finding is consistent with the original landmark study by Kim et al. and a later study in adult canine heart which only showed irreversible injury 4 h after artery ligation [47, 48]. On the other hand, three days after the ligation of the LAD artery, massive cell death as a result of the infarction allowed the infarcted area to trap sufficient gadolinium in the intracellular space, which eventually enabled us to see the hyper-enhanced infarct site on Scan 2.

Unlike LGE, EGE CMR is clinically useful in the setting of acute myocardial infarction, to delineate microvascular obstruction or left ventricular thrombus, which prevent the distribution of gadolinium to the infarct zone and therefore gives rise to low T1 signal [49]. Therefore, we would expect EGE CMR to show hypoenhancement in both Scan 1 and Scan 2 of the infarcted fetuses. However, in our study, EGE CMR showed microvascular obstructions in only half of the infarcted fetuses on Day 0 and none on Day 3. One possible explanation could be our compromised spatial resolution, which was necessary to achieve adequate signal to noise ratio within a reasonable scan time. Other alternative explanations are that the size of the infarcts we induced was small, whereas a microvascular obstruction is typically seen in large infarcts in humans. Finally, it is possible that we performed the second scan after resolution of microvascular occlusion, i.e. the timing of the second scan was too late to demonstrate hypo-enhancement on EGE. To fully understand the best timeline to perform LGE and EGE CMR after fetal MI, a more comprehensive study including a series of CMR scans would be needed to define the precise timeline of injury and repair, and their associated imaging characteristics.

We performed cine imaging to investigate whether regional wall motion abnormalities could be observed in the infarcted fetal hearts. We saw clear regional wall motion abnormalities in two of the five infarcted hearts. We therefore conclude that performing cine imaging is helpful in examining whether regional wall motion abnormality is present in the fetal heart. Although the infarct sizes were similar in all fetuses, we suspect the subjects with infarcted hearts that showed no regional wall motion abnormality on cine images may have had more subtle regional wall motion abnormalities that were not appreciable given the limitations of our technique.

We were also interested to see whether cine imaging could be used to quantitatively evaluate cardiac function of the infarcted fetal heart. Cine CMR is recognized as the gold standard for functional imaging and assessment of the both ventricles postnatally. However, its utility in the fetus has not been established. Yamamura et al. have previously demonstrated that cine CMR is feasible in sheep fetuses [28, 41]. They used cardiotocography as a trigger and performed cine CMR in 4 sheep fetuses (119–120 days gestation) [41]. They determined an average LVEF of 60.53 ± 4.1% and an average unindexed LVSV of 2.87 ± 0.31 ml. The LVEF and the unindexed LVSV we obtained in the sham fetuses (58.2 ± 0.4% and 2.79 ± 0.85, respectively) were similar to this prior work [41]. While we contoured the LV and RV short-axis images, we found that the most basal slice was difficult to define. This is a well-recognized problem with cine short-axis imaging [50, 51]. However, since the fetal heart was extremely small (~11g), it was particularly difficult to obtain and define the most basal slice in the short axis view with the given spatial and temporal resolution. Thus, the inclusion or exclusion of basal slices could give rise to variability in our measured ventricular volumes [50, 51]. Hawkins et al. determined a lower LVSV of 0.80 ± 0.1 ml/kg than the MRI stroke volumes using an electromagnetic flow probe at around 127 days gestation [52]. Morton et al. (137 ± 3 days) and Anderson et al. (127–133 days) also found lower LVSV using a flow probe [53, 54]. Alonso et al. used a flow probe to determine RVSV and reported a mean RVSV of 1.142 ± 0.042 ml/kg in fetal sheep of around 115 days gestation [55]. These differences in SV could be due to the variations in fetal heart size, the difference in gestational age, the lower heart rate and/or our use of anesthesia at the time of the scan, as well as variation in the accuracy of the measurement techniques. Although the mean LVSV and RVSV that we determined were higher, the heart rate we measured was lower than those reported [52‐55]. Therefore, our mean fetal cardiac output (= SV x heart rate) was similar to those determined by Hawkins et al. (236 ± 4 versus 265 ± 63 ml/min/kg) and Alonso et al. (198 ± 13 versus 204.5 ± 36.8 ml/min/kg) for the left and right ventricles, respectively.

The mean EF we determined (58.2% ± 0.4%) was similar to that determined by Yamamura et al. using cine CMR and cardiotocography gating (60.53 ± 4.1%) [41]. However, these values differ from other studies that used echocardiography [13, 31]. For example, Xiong et al. used 2D echocardiography and determined a mean LVEF of 72 ± 7% in 7 control fetuses at 113 days gestation (term, 145 days), slightly higher than our results (infarction: 54.8 ± 11.9, sham: 58.2 ± 0.6) [31]. Consistent with our findings, Herdrich et al. found similar ejection fractions between fetal sheep pre-MI and post-MI (53 ± 8.1% and 54 ± 9.6%, respectively) using 2D echocardiography at 65–76 days gestation with anesthesia [13]. This variability in EF is likely due to the imaging modalities used. Measuring ventricular dimensions with 2D echocardiography is prone to error. It has been shown in previous studies that cine-MRI is superior to 2D echocardiography for determining ventricular function, demonstrating higher reproducibility and lower intra- and inter-observer variability [56‐58]. Finally, an important limitation of our study was the low sample size for the sham fetuses (n = 2) and our results do not represent a valid set of reference ranges for normal fetal ventricular volumetry.

Major challenges to our LGE, EGE and cine imaging techniques were the fast heart rates and small size of the fetal hearts, which inevitably led to long scan times. The imaging methods we employed here are therefore not suitable for non-sedated human fetuses that move frequently. In addition, the long scan times used in this study may have also caused problems for the GE imaging relating to changes in the gadolinium concentration in the blood pool and in the myocardium during the LGE and EGE sequences. However, from a practical perspective, such variation did not appear to affect the quality of the LGE images as we were able to obtain reasonable contrast between the infarct site and the normal myocardium for each slice. While the correct inversion time for myocardial nulling was also likely to be changing during these long acquisitions, our imaging parameters resulted in good contrast between the infarct and normal myocardium. Contrast between normal and diseased myocardium was enhanced by the PSIR reconstruction used in our LGE imaging, which reduced background noise and improved the contrast-to-noise ratio [59]. The EGE sequence, however, was perhaps more affected by the long scan time, compromised spatial resolution and worse contrast-to-noise ratio, and may have contributed to its poor sensitivity. Another limitation of the study is the lack of quantification of the infarct volume in the GE images. We did not attempt to determine the infarct volume because the slices we obtained were rather thick (~ 5 mm), which was important to achieve adequate signal-to-noise ratio. As a result, the infarct was only present on one or two slices, and this precluded a reliable estimation of the volume of the infarct. The image quality was also insufficient for reliable estimation of the infarct length. However, we did estimate the epicardial infarct length seen on the heart at post mortem, and it was as we had anticipated.

We performed an analysis of our fetal body volumes obtained from the 3D–SSFP acquisitions and found excellent correlation with fetal weight at post mortem. Thus, our results support the use of fetal CMR volumetry for monitoring fetal volume over late gestation. To our knowledge, there is no established method for converting fetal volume to weight in the sheep. However, calculations of this kind have been derived for human foetuses, and a similar approach may be possible in the sheep, which would then allow fetal hemodynamic measurements to be indexed to fetal weight as they have been for invasive measurements in the past [60‐62].

Our results showed that LGE CMR imaging is highly feasible for the detection and monitoring of MI while EGE imaging as implemented here was less informative. Cine CMR imaging allows the visualization of regional wall motion abnormalities associated with fetal MI. However, fast fetal heart rates and the small size of the structures of interest made cine cardiac function measurements challenging and compromised their accuracy relative to adult hearts. In conclusion, our study provides evidence for the utility of a new tool for monitoring the cardiac response to MI in the fetus. However, further investigation into the optimal timing of LGE and EGE scanning and improvement of the sequences may be helpful.