Regional analysis of inflammation and contractile function in reperfused acute myocardial infarction by in vivo ¹⁹F cardiovascular magnetic resonance in pigs

Experiments were performed in 22 adult female Aachen minipigs [27] with a mean body weight of 67 ± 9 kg, in accordance with the national guidelines on animal care and approval by the ‘Landesamt für Natur, Umwelt- und Verbraucherschutz’ (L84-02.04.2013.A437 and L84-02.04.2016.A322). The experimental protocol is displayed in Fig. 1. Myocardial infarction was induced in pentobarbital-isoflurane-anesthetized closed-chest pigs by 90 min of pressure-wire-controlled left anterior descending (LAD) coronary artery occlusion [33], for details see supplemental material & methods and supplemental Figure I. In 5 sham animals the LAD was instrumented but not subjected to balloon occlusion.

The perfluorooctyl bromide nanoemulsion (PFOB) was produced according to established protocols [4, 33]. A detailed description of the production, quality control, stability assessment, blood kinetics and phagocytosis of PFOB is provided in the supplemental Figures II and III. PFOB was infused intravenously on day 3 after AMI with 5 ml/kg body weight and a rate of 80 ml/h (Fig. 1) in all surviving animals (n = 10 AMI and n = 5 sham) according to previous protocols [4].

After proof of LAD patency 6 days after AMI by invasive angiography, CMR was performed using a whole-body 3.0-T Achieva X-series MR scanner (Philips Healthcare, Best, the Netherlands). In vivo CMR was performed according to previously established workflows and protocols for local signal precision with optimized ¹⁹F sequences [33]. The ¹H and ¹⁹F-sequence protocols and details are reported in supplemental Table I. After in vivo ¹H- and ¹⁹F-CMR acquisitions, the pigs were euthanized inside the scanner for in situ imaging with arrested hearts. Autopsy was performed, and organs (heart, liver, spleen and sternum) were explanted and stored in 4% paraformaldehyde (PFA) for analysis of histopathology and distribution of PFOB with high resolution ¹⁹F-CMR. A detailed description of the histopathological protocols and analysis, including the immunohistochemistry of TNFα, is provided in the supplemental material & methods and supplemental figure IV and V.

Dedicated software (Circle CVI 42, Circle Cardiovascular Imaging Inc., Calgary, AB, Canada) was used for full automatic delineation of ventricular contours, automatic calculation of volumes and analysis of parametric T1 and T2 maps. Due to misalignment of the automated contouring algorithm, significant manual correction of endocardial contours was necessary in 3/10 pigs after AMI. As described in detail in supplemental material & methods and shown in supplemental figure VI, volumes were indexed to body weight to measure end diastolic volume index (EDVi), end systolic volume index (ESVi), stroke volume index (SVi), cardiac index (CI) as well as ejection fraction (EF), global longitudinal strain (GLS), and early longitudinal diastolic strain rate (SRe) [34]. The sphericity volume index (ShVi) was calculated with EDV/ ((π/6) * L)³ [1]. Initial remodeling was defined as > 20% EDVi increase compared to the EDVi of sham-instrumented animals, adapted from studies investigating long-term remodeling after AMI [41]. T2 weighted images and T2 maps were analyzed for the extent of myocardial edema and expressed as % of LV mass or in absolute mass (g). T2* weighted images were analyzed for IMH and expressed as % of LV mass or in milliliters (ml). Contrast-enhanced images were analyzed at first pass for myocardial perfusion and late gadolinium enhancement (LGE) after 10 min for IS (> 5 SD threshold), border zone (BZ) (> 2 < 5 SD threshold) [20] and after 15 min for MVO and expressed as % of LV mass or in milliliters (ml) [34]. Volumes containing ¹⁹F signals were quantified using the 3D visualization software Amira 4.0 (ThermoFisher Scientific,Waltham,USA), and image fusion of ¹H/¹⁹F-datasets was performed with HOROS (Nimble Co LLC,USA). The fused images of all 10 animals which were included in the final analysis are shown in the supplemental figure VII. For quantification of ¹⁹F-signals, the primary signal in a respective region was corrected according to the coil sensitivity profile as outlined in supplemental figure VIII. The signal-to-noise-ratio (SNR) of respective regions was calculated from the ratio of the signal intensity mean of a region-of-interest (ROI) and the standard deviation of a background-noise ROI in the same slice located out of the thorax at comparable coil distance. ¹⁹F volumes were calculated by applying background subtraction with SNR 7, which was sufficient to subtract all unspecific technical background signals (outside the body). The ¹⁹F-integral was the product of SNR and volume.

Statistical analysis was performed using SPSS software (version 24.0, IBM, Armonk, NY, US). Unless otherwise stated, continuous variables are presented as the mean ± standard deviation (SD). Differences between the two groups (sham and AMI) were analyzed by Student’s 2-sided unpaired t test for normally distributed data and Mann–Whitney U test for non-normally distributed data. Pearson’s or Spearman’s correlation were used to assess the relationships between different CMR parameters. Separate linear regression models in pairwise combination with edema, IS, MVO, and IMH were created to adjust the univariate correlations of ¹⁹F with LV function and volumes for these parameters to test whether ¹⁹F would independently predict LV function and geometry. ROC analysis was used to identify optimal cut-off values for prediction of initial remodeling (> 20% EDVi increase compared to the sham animals mean) (AUC 0.87) or impaired remote systolic wall thickening (< cohort median of diastolic wall thickening) (AUC 0.78). The respective p values are reported in the figures.

Twenty-two pigs were included in the study. Baseline and procedural characteristics of sham pigs (n = 5), AMI survivors (n = 10) and AMI non-survivors (n = 7) are reported in supplemental Tables II and III. There were no differences in baseline characteristics between survivors and non-survivors. The cause of death (n = 7) was ventricular fibrillation refractory to defibrillation in all pigs. No pig died beyond 90 min of ischemia. There was no serious adverse side reaction when pigs received PFOB on day 3 after AMI (supplemental Table IV). The blood kinetics of PFOB are given in supplemental Figure III and revealed an almost complete clearance from the circulation 3 days after injection. Ten pigs had a patent LAD on day 6 after AMI on coronary angiography and were included in the CMR study.

Edema extent was 35 ± 8% of LV mass and IS was 22 ± 6% of LV mass. Six of 10 surviving pigs had MVO (3.3 ± 1.6% of LV mass), and the same animals also had IMH (1.9 ± 0.8% of LV mass). Figure 2 shows systemic hemodynamic, volumetric and functional data 6 days after AMI. Compared to sham pigs, pigs with AMI had a slight increase in heart rate, a trend toward lower arterial pressure and higher systemic vascular resistance. SVi, CO, EF and LGE were reduced; ESVi, EDVi and ShVi were increased. Thus, the present model provided robust infarction with early LV dysfunction and initial remodeling of varying inter-individual severity.

The average acquisition time of the 3D-whole heart ¹⁹F-information was 18 ± 2 min. Figure 3 displays the correlation of local ¹⁹F-signal intensity and monocyte/macrophage density. ¹⁹F-signals were restricted to infarcted tissue and its border zone, as delineated by infarct size imaging with LGE, while remote myocardial areas displayed no ¹⁹F-signal (Fig. 3A, left). The colocalization of the ¹⁹F-label with monocytes/macrophages was confirmed by histology (Fig. 3A, right and B). The infarct area displayed patchy infiltration with monocytes/macrophages, mirrored by the respective local ¹⁹F signals. An equal number of M1 (CD68^high CD163 ^low) and M2 (CD68^low CD163^high) macrophages was present within the areas of ¹⁹F signal (supplemental Figure V). The mean ¹⁹F-signal intensity was quantified considering the coil sensitivity profile (supplemental Figure VIII) and expressed as myocardial ¹⁹F-SNR. The calculated local mean ¹⁹F SNR acquired in vivo correlated with the mean monocyte/macrophage density determined by histology (Fig. 3C; R² = 0.6075, p < 0.0001). The individual localization of the analyzed regions of interest is given in the supplemental Table V. The ¹⁹F detection threshold was 70,000 cells/mm³. In sham pigs, no myocardial ¹⁹F signal was observed, and immunohistochemistry did not show signs of acute myocardial inflammation.

Figure 4 shows the individual whole heart quantitative ¹⁹F-analysis of all surviving pigs. For inter-individual image comparison, ¹⁹F-signals of all 10 pigs are shown in supplemental Figure VII. The monocyte-dependent ¹⁹F-signals were also observed in the bone marrow of the sternum and ribs (Fig. 4A and supplemental Figure IX). These signals could easily be distinguished from AMI signals and did not disturb the signal quality or quantification. After cardiac segmentation, the average volume covered by ¹⁹F-signals per heart was 15.8 ± 9.5 ml, and it was always within and smaller than the respective individual edema extent (31.3 ± 8.7 ml) (Fig. 4A, B) and IS (20.9 ± 6.5 ml) (Fig. 4A, C). The mean myocardial ¹⁹F SNR was 13.3 ± 4.7 with a considerable inter-individual and segment-wise variation. As shown in Fig. 4D, there was a correlation of ¹⁹F-SNR with ¹H derived T1 and T2 times taken from the infarcted segments of all pigs. Along with the systemic inflammatory response, we identified a correlation (R² = 0.66) of the individual increase in circulating leukocytes from baseline until day 3 and the ¹⁹F integral (supplemental Figure III D).

As shown in Fig. 5A + B, not all myocardial segments with edema and infarction were positive for the ¹⁹F-CMR derived monocytes/macrophages signals. The correlation of edema extent and IS with the ¹⁹F-integral was only fair. Pigs with high ¹⁹F integral displayed only minor volume of MVO with concomitant IMH (Fig. 5 D + E). Apparently, the ¹⁹F-CMR derived monocytes/macrophages signal was of patchy nature with only fair correlation to edema extent or IS, and the ¹⁹F integral was reduced in pigs with MVO and IMH. As shown in the supplemental figure X, hemorrhage itself had no impact on ¹⁹F SNR.

As shown in Fig. 6A, regional wall thickening displayed a gradient from infarct core to remote myocardium, while ¹⁹F SNR decreased in most pigs from infarct core to remote myocardium. Interestingly, the ¹⁹F-signal in the infarct border zone correlated inversely with the neighboring remote myocardial wall thickening (Fig. 6B). Histological analysis revealed significant TNF-α signals in remote cardiomyocytes of those pigs with high border zone ¹⁹F SNR (6F + 6G). Increased ¹⁹F-integrals over the entire heart correlated with enlarged EDVi and ShVi in univariate analysis. Even when adjusted for edema, IS, MVO and IMH, respectively, in multivariate models, the macrophage/monocyte derived ¹⁹F-signal was independently associated with initial LV remodeling (supplemental Table VI).

As demonstrated in Fig. 7, a cut-off ¹⁹F-integral value of 200 combined with a border zone SNR > 8 predicted impairment of remote myocardial contractile function and LV remodeling on day 6 after AMI. Pigs with whole heart ¹⁹F-integral > 200 and border zone SNR > 8 (n = 5) had less remote myocardial systolic wall thickening (29.4 ± 6 vs. 59 ± 16%, p = 0.005), larger EDVi (1.4 ± 0.1 vs. 1.0 ± 0.1 ml/kg body weight, p = 0.005) and larger ShVi (0.7 ± 0.1 vs. 0.4 ± 0.1, p = 0.005) than pigs with whole heart ¹⁹F-integral < 200 and border zone SNR < 8 (n = 5).

Our preliminary experiments were conducted to test and prove the applicability of various components and processes necessary for the planning, initiation and conduction of the present study. First, we tested optimal ¹⁹F containing nanoemulsions with respect to safety, dose and timing of the emulsion administration [4]. ¹⁹F measurements were done in explanted hearts to assure easy and sufficient ¹⁹F signal acquisition [4]. Thereafter, we tested and improved the in vivo CMR approach-applicability with ¹⁹F signal distance > factor 100 compared to explanted hearts and respiratory as well as cardiac movement [33]. After these experiments, it was clear, that a bFFE sequence with an off-set frequency of 58 ppm would be the best compromise for signal precision and SNR [33].

The findings of these preliminary studies (dose, timing, optimized sequences, work flows) were necessary to efficiently conduct the present study. The specific novelty of the present approach is the comparison of state-of-the-art CMR tissue markers as well as local and global myocardial function after AMI with the in vivo ¹⁹F signal, which was also validated by histology.

Since ¹⁹F-CMR is essentially background-free, perfluorocarbon nanoemulsions are the primary source of the ¹⁹F-signal, and they are preferentially taken up by monocytes/macrophages [3, 26]. In the present study, the nanoemulsion was administered on day 3 after AMI to allow for sufficient uptake into monocytes/macrophages [26] and to induce a maximum monocyte/macrophage signal in the injured myocardium on day 6 after the insult. Untargeted perfluorocarbon nanoemulsions are taken up by M1 and M2 tissue macrophages to an equal extent [13]. Since both cell types were present in equal numbers in our study on day 6 after reperfused AMI, the ¹⁹F-signal observed reflected most likely a mixture of a M1 and a M2 macrophage infiltration. Additionally, the ¹⁹F-signal was closely associated with altered tissue relaxation properties indicating early tissue reorganization in infarcted myocardium with a high volume and density of monocytes/macrophages.

As an alternative to ¹⁹F labeling of immune cells, ultrasmall iron oxide nanoparticles (USPIOs) enhanced ¹H-based CMR has been used. USPIOs also preferentially label monocytes/macrophages and significantly affect ¹H relaxation properties. This technique, leading to signal voids in the area of monocytes/macrophages, has provided initial, promising diagnostic results in patients after AMI, but its specificity has been hampered by interference with IMH [29, 46]. Beyond direct visualization of immune cells, the metabolic signature of inflammation can be assessed with positron-emission tomography (PET) or hyperpolarized CMR [23, 24, 31, 32].

PET MRI using 18F-fluorodeoxyglucose (18F-FDG) has proven significant value in the detection of atherosclerosis-related inflammation and in predicting restenosis in peripheral artery disease [5, 8]. There are specific nuclear PET tracers for imaging monocytes/macrophages after AMI [30, 31, 39]. Indeed, combined PET-MRI approaches have identified certain cell types and inflammatory activity using optimized image resolutions of 4 × 4 × 4 mm³ [8]. However, a direct quantification of the signal in terms of local cell abundance is difficult due to tracer/image acquisition timing and complex post-processed image reconstructions. The need for cost intensive specialized equipment (linear accelerator) and staff (chemist) might hamper the applicability of PET/MRI in clinical routine. The ¹⁹F principle easily works on the widely available CMR platform, with an image resolution (1.35 × 1.35.1.5 mm³) sufficient for imaging inflammatory foci in the AMI border zone with a directly quantifiable signal, without the need for additional specialized equipment and personnel.

With the present CMR approach, the images of monocyte/macrophage infiltration were generated in vivo with a resolution and acquisition time comparable to clinical standard sequences and protocols (supplemental Table I). The pig model of reperfused AMI in the present study exhibited a similar extent of IS and frequency of MVO or IMH as typical patients with an anterior STEMI [6, 29]. Moreover, the monocyte/macrophage density was comparable to histopathological studies in patients [42], rendering the present model suitable for translational perspectives. Thus, ¹⁹F-CMR is a highly promising imaging technique for clinical trials aiming to further evaluate the effects on monocyte/macrophage infiltration in patients early after AMI.

In vivo myocardial ¹⁹F-signals revealed not only considerable inter-individual variation, which correlated to the systemic increase in leukocytes (supplemental figure IIID), but also a patchy appearance in myocardial infarction, which was validated by histology (Fig. 3). The spatial heterogeneity of monocytes/macrophage infiltration was reflected in features of the ¹⁹F-signal: Even those pigs with low ¹⁹F whole heart volumes (< 5 ml) had a mean myocardial ¹⁹F-SNR of > 10, indicating hot spots of monocyte/macrophage infiltration as known also from human autopsy studies [42]. The area of myocardial edema and IS was not entirely invaded by monocytes/macrophages as evidenced by ¹⁹F-CMR. Moreover, the incidence of MVO and IMH was even reduced in segments positive for ¹⁹F-CMR. Since MVO and IMH coincided in our model (day 6 after AMI), in good agreement to clinical observations [29], the attenuating effect of MVO on monocyte/magrophage infiltration, which has already been shown by other groups [45], might have counterbalanced the pro-inflammatory effect of IMH at day 6 [2].

Our present study provided novel information on the regional relationship between monocyte/macrophage infiltration in the infarct border zone with systolic wall thickening in the adjacent remote viable myocardium (Fig. 6A + B) and with LV remodeling early after AMI (Fig. 6G + H). The infarct border zone derived ¹⁹F-signal correlated inversely with remote myocardial systolic wall thickening. On histology, TNF-α was identified in remote cardiomyocytes (Fig. 6F). A spillover of cytokines with a negative inotropic action such as TNF-α from the border zone into the remote myocardium might thus explain the impaired wall thickening possibly mimicking stunned myocardium [17]. It is unclear to what extent such increased TNF-α levels are derived from monocytes/macrophages or the cardiomyocytes in the inflamed border zone, as shown in microembolized myocardium [7, 18].

On the level of a global analysis over the entire heart, the global ¹⁹F integral was associated with LV remodeling on univariate analysis. In multivariate analysis, the correlation of the global ¹⁹F integral was independent from edema, IS, MVO, and IMH, respectively. The greater the monocytes/macrophages infiltration, as indicated by ¹⁹F-CMR, the more enlarged was EDV. Our data revealed threshold values: when the ¹⁹F-derived monocyte/macrophage signal exceeded 200 SNR × ml over the entire heart, it was associated with LV remodeling; when the border zone ¹⁹F SNR exceeded 8, it was associated with impaired remote myocardial wall thickening.

PFOB nanoemulsions have already been evaluated in clinical phase III studies as blood substitutes with similar application doses as used in the present study [35]. They are taken up by monocytes or macrophages irrespectively of their M1/M2 polarization in a time-dependent manner, while preserving their function [13, 26], which opens the translational perspective to develop ¹⁹F-based CMR imaging for future clinical applications. Due to the preferential uptake of PFOB nanoemulsions, ¹⁹F-based monocyte/macrophage imaging is rather cell-specific and directly quantifiable, as shown in the present study, and it might be even capable of being more specific when using functionalized nanoemulsions [11]. Intravenous administration of the nanoemulsion on day 3 after AMI enabled sufficient accumulation in monocytes/macrophages for sensitive detection by ¹⁹F-CMR within the infarcted myocardium on day 6 and was not associated with major side effects. Thus, ¹⁹F-CMR is capable to deliver in vivo information about early inflammatory infiltration after AMI. Inflammation can thus be visualized and followed as a potential additional target for subacute cardioprotection after AMI—beyond the already well established concept of preservation of coronary microcirculation [16].

Inflammation is a driver of adverse LV remodeling after AMI [44]. There are large inter-individual differences with respect to monocyte/macrophage quantity and localization in post-AMI inflammation [42]. The peri-infarct border zone, as identified by the specific gadolinium contrast agent uptake kinetics, is characterized by inflammatory infiltration, involved in the pathogenesis of arrhythmias [40] and associated with mortality [20].

For patients after AMI, it is currently unclear whether local hot spots of inflammation, a patchy pattern of inflammation in the infarct border, or the overall inflammatory volume in the heart drive clinical endpoints. The ¹⁹F platform might add novel information in this respect, since it specifically images monocytes/macrophages and thus can also monitor specific modulation of the inflammatory response by pharmacotherapy [21, 22, 28]. Compared to techniques using radiotracers, CMR is widely available and has better spatial resolution. In the future, the nanoemulsion delivery and the imaging workflow can easily be adopted to the management of patients after AMI in the hospital.

¹H CMR-derived tissue markers for improved prediction of cardiovascular endpoints are not always present after AMI. Major markers of mortality after AMI, such as MVO and IMH, appear in only 40% of all patients. The ¹⁹F signal will most probably be detectable in every singular individual myocardium post-AMI within the infarct region and border zone, thus enabling quantification of inflammatory patterns and extent. For patients after AMI, it is currently unclear whether local hot spots of inflammation, a patchy pattern of inflammation in the infarct border, or the overall inflammatory volume in the heart drive clinical endpoints. The ¹⁹F platform might add novel information in this respect, since it specifically images monocytes/macrophages and thus can also monitor specific modulation of the inflammatory response by pharmacotherapy [22]. However, whether or not these additional information further enhance the risk prediction after AMI needs to be tested in clinical trials.

Since the present study aimed to mimic a clinical imaging scenario at day 6 with subsequent histological validation, the predictive value of the respective ¹H- or ¹⁹F-derived signals on more long-term remodeling remains unclear. In mice, increased ¹⁹F-derived monocyte/macrophage signals predicted worsening of LV function also 28 days after AMI [3]. The time point of CMR analysis at 6 days post-AMI was chosen in line with recent expert consensus regarding the dynamic nature of infarct development and repair [19]. Our imaging study does not establish a cause-effect relationship between early inflammatory infiltration and impaired remote contractile function or LV remodeling early after reperfused AMI, but demonstrates a close association.