Multiparametric MRI identifies subtle adaptations for demarcation of disease transition in murine aortic valve stenosis

AS was induced as described previously [40]. In brief, mice were anesthetized by intraperitoneal injection with a mixture of ketamine (100 mg/kg BW) and xylazine (10 mg/kg BW). Thereafter, anaesthesia was maintained by 1.5–2.0% isoflurane. After endotracheal intubation, mice were mechanically ventilated and placed in a supine position on a warming pad. Thereafter, the right carotid artery was exposed, and a coronary wire (Universal; Abbott Cardiovascular, Plymouth, MN, USA) was inserted, advanced beyond the valve level and rotated as described before for induction of AS [40]. Afterwards, the right carotid artery was ligated. For sham operation, an identical surgery was performed, but the wire was not passed over the valve and rotations were performed above valve level. For post-surgical analgesia, 0.05–0.1 mg/ml buprenorphine in a maximum volume of 10 ml/kg body weight were applied subcutaneously every 6–8 h for 72 h following surgery.

For validation of MRI-derived flow velocity data, a Fujifilm Visualsonics Vevo 3100 Ultra high frequency imaging platform (Toronto, ON, Canada) was utilized. Mice were anesthetized with 2% isoflurane under continuous monitoring of ECG, respiratory rate and body temperature. The chest of all mice was depilated and pre-warmed bubble-free ultrasound gel was applied to allow artefact-free image acquisition. Measurements were carried out in suprasternal view with a pulsed wave Doppler using angle correction between 45° and 55° as described in detail previously [40].

Animals were sacrificed after MRI by exsanguination in deep anaesthesia with xylazine (10 mg/kg BW) and ketamine (100 mg/kg BW), injected intraperitoneally. Blood collection for determination of circulating troponin levels was carried out via transthoracic heart puncture. Thereafter, hearts were excised and immediately washed with ice-cold buffer. Hearts were then prepared for Langendorff perfusion (see below) or transferred to a cryomold filled with embedding medium (Tissue-Tek; Sakura Fineteck, Staufen, Germany), frozen in liquid isopentane and stored at -80 °C for at least 1 week.

High-resolution MRI was carried out 4 weeks after induction of AS. Using a segmented FISP sequence with prospective gating, cine loops at the atrio-ventricular level were acquired for assessment of valve morphology with accurate assignment to early and late diastole and systole, respectively, allowing an exact planimetry of the valve in all phases of the cardiac cycle (Fig. 1, Supplementary Movie 1). Sham-operated (sham) and age-matched untreated mice (con) showed a homogenous and unrestricted opening of the valve resulting in an opening area of 80.1 ± 8.6% and 77.8 ± 6.4%, respectively, as normalized to the total inner supravalvular aortic area in end-diastole (beginning of the QRS complex). As expected, in mice subjected to wire injury we observed a more triangular orifice and an impaired valve opening of only 60.3 ± 9.8% (P < 0.001, n = 9 each group), which is comparable to human morphology in moderate degenerative AS (Fig. 1A–C, Supplementary Movie 1). Besides detection of the restricted orifice, high-resolution long axis slices acquired perpendicular to the aortic valve also revealed an altered shape and thickening of the valve as consequence of the remodelling processes induced by the surgical procedure (0.18 ± 0.02 vs. 0.12 ± 0.02 mm², P < 0.001, n = 9 each group; Fig. 1D + E; Supplementary Movie 2). Subsequent histology confirmed the in vivo findings of valve thickening and revealed that this was accompanied by incipient fibrosis as indicated by enhanced vimentin and α-smooth muscle actin staining (Supplementary Fig. 7).

The extent of stenosis is usually estimated by determination of peak flow velocity across the valve by echocardiography. Here, we used MRI velocity maps to reconstruct cine flow profiles above the valve over the entire cardiac cycle (Fig. 2, Supplementary Movie 3). While sham-operated mice showed a bell-shaped flow profile in early systole, AS resulted not only in significantly enhanced peak flow velocities (204.8 ± 38.9 vs. 111.9 ± 26.4 cm/s; P < 0.001, n = 9 each group; Fig. 2C), but also in fragmented turbulent patterns of the flow profile which included isolated spikes of strongly accelerated velocities and as well areas with negative flow peaks (Fig. 2A + B). Note, that due to the increased resistance for blood output through the reduced orifice, peak velocity is achieved later in mice with AS compared to sham controls as shown in representative time courses in Fig. 2D. Parallel echocardiographic examinations corroborated the magnitude of flow increase in AS animals determined by MRI (~ 100 cm/s, Supplementary Fig. 8), while absolute values acquired by Doppler echocardiography tended to be higher in all groups—a phenomenon already reported earlier in comparison to 2D phase-contrast measurements [44]. Of note, dispersion of data and SDs were quite similar for MRI and echocardiography in all groups.

Aortic regurgitation (AR) might occur as consequence of the remodelling process after induction of AS or due to the surgical intervention itself. In MRI cine loops, the emerging backward flow is reflected by black jet artefacts in early diastole immediately after valve closure (Fig. 3A + B + G, Supplementary Movie 4). However, this is a rather qualitative measure and the magnitude of this artefact strongly depends on the imaging plane and its orientation to the backward flow. Thus, we used the created cine flow profiles for a more quantitative evaluation of the degree of AR (Fig. 3C + E + F + H + I). With this, diastolic backward flow could be precisely determined and mice with AS only easily discriminated against the combined case with AR. Depending on the individual remodelling process after induction of AS the resulting valve morphology led to asymmetric opening and regurgitation patterns. In an interesting example, functionally bicuspid AS (Fig. 3D–F) was associated with AR exactly at the same position where valve opening was restricted (Fig. 3G-I). Here, we also observed an accelerated increase in mean flow velocity caused by enhanced slew rate due to pendular volume (Fig. 3C). Nevertheless, the majority of stenotic valves in this model (> 90%) exhibited no or only a mild component of AR. Only two animals showed a persisting mean diastolic backward flow ≥ 5 cm/s and were excluded from analysis of AS due to severe AR.

While an unrestricted valve opening led to a homogenous blood flow coaxial to the vessel wall (Fig. 4A + B; Supplementary Movie 2), aortic valve stenosis resulted in turbulent flow above the valve as reflected by irregular black jet artefacts in the ascending aorta (Fig. 4D + E; Supplementary Movie 2). The altered ejection patterns were associated with an increased distension of the vessel as reflected by significantly enhanced systolic diameters and circumferential strain in the aortic root of AS mice (2.49 ± 0.11 vs. 2.36 ± 0.12 mm and 23.1 ± 7.2 vs. 16.0 ± 4.6%, respectively; P < 0.05, n = 9 each group; Fig. 4G + H). As a consequence of the raised long-term burden within the outflow tract, 4 weeks after surgery aortic wall thickness was significantly increased in AS mice as compared to sham-operated controls (0.16 ± 0.02 mm vs. 0.12 ± 0.02 mm; P < 0.001, n = 9 each group; Fig. 4C + F + I).

To assess the impact of AS in the present model on the LV, short axis cine loops were acquired for analysis of morphological and functional parameters (Fig. 5+6A Supplementary Movie 5). Of note, global cardiac function in terms of EF and cardiac output was unaffected 4 weeks after surgery (Fig. 5A + B). End-diastolic and -systolic volumes remained unchanged excluding mice presenting with severe AR (data not shown). The regional analysis of the LV (see Supplementary Fig. 2) indicated an increased fractional shortening in all sectors except for the inferior wall but without reaching the level of significance (Fig. 5C). However, for diastolic wall thickness, we observed a mild but significant increase in AS compared to sham-operated controls (1.02 ± 0.05 vs. 0.95 ± 0.04 mm; P < 0.05, n = 9 each group) indicating a mild hypertrophy of the LV wall (Fig. 6A). In line with this, LV mass showed a trend to increase but without reaching significance (data not shown).

Cardiac tissue characterization by T1 and T2 mapping surprisingly revealed reduced values for both parameters in AS as compared to sham controls (T1: 791 ± 136 vs. 981 ± 88 ms; P < 0.01, n = 9 each group; T2: 16.8 ± 2.1 vs. 19.4 ± 0.9 ms; P < 0.01, n = 9 each group; Fig. 6B + C). Concomitantly, we found no signs for myocardial LGE after application of Gd-based contrast agents (data not shown). Additional acquisition of post-contrast T1 maps also provided no evidence for any alterations in cardiac extracellular volume of AS mice (ECV 23.3 ± 6.9 vs. 25.3 ± 4.9% in sham controls, n = 9 each group). This clearly argues against development of significant myocardial oedema or necrosis at this time point but indicates subtle structural alterations in the hearts of AS mice owing to the increased afterload. In a final step, cine arterial spin labeling (cine-ASL [54]) was used to assess in the present model the impact of AS on myocardial perfusion. As can be recognized in Fig. 6D, baseline myocardial blood flow was similar in both groups, but upon Regadenoson stimulation we found a small, but significantly lower perfusion increase in AS mice (Fig. 6D, Supplementary Movie 6) despite the drug induced in both groups a comparable increase in heart rates (~ 40 bpm, Fig. 6E). Of note, a similar restriction in coronary perfusion reserve as observed in vivo (Fig. 6F) was also detected ex vivo upon reactive hyperemia in Langendorff-perfused hearts of AS and sham mice (Fig. 6G).

Subsequent histology demonstrated that the moderate thickening of the LV wall in AS observed in ¹H MRI cine loops was related to an increased cardiomyocyte size (∅ 16.4 ± 1.1 vs. 13.7 ± 1.4 µm; P < 0.01, Fig. 7A). This was associated by a trend towards higher capillarization in AS mice (Fig. 7B, P = 0.08), but without indications of interstitial or perivascular fibrosis and/or remodelling of the extracellular matrix (ECM), respectively (Fig. 7C–F, n  = 5–9 each group). In parallel, we found the subtle morphological alterations described above to be accompanied by approximately doubled troponin levels in the blood of AS mice as compared to sham controls (1.15 ± 0.55 vs. 0.48 ± 0.17 mg/ml; P < 0.05).

Furthermore, we observed that myocardial oxidative capacity and electron transfer system capacity did not differ between animals with and without AS as determined by high-resolution respirometry (Fig. 8A). However, succinate-dependent oxygen flux increase, when added after malate, glutamate, and ADP, was higher in AS than in sham (162.8 ± 45.2 vs. 123.2 ± 45.9 pmol⋅s⁻¹⋅mg⁻¹; P < 0.05; Fig. 8B). In addition, AS animals exhibited a higher respiratory control ratio (a marker of mitochondrial coupling) as compared to shams (1.99 ± 0.19 vs. 1.72 ± 0.23 a.u.; P < 0.01; Fig. 8C; n = 10–11 each group) suggesting further intracellular adaptations to the altered hemodynamic state. Electron microscopy indicated that this was associated with a beginning disorganization of the cardiomyocyte ultrastructure (Fig. 8D). While total mitochondrial area was unchanged compared to sham controls (43.8 ± 4.7 vs. 42.8 ± 5.9% of the field-of-view, n = 9 each group), we noted a tendency to less but larger mitochondria (Fig. 8E + F, P = 0.12 + 0.11) with incipient distributional imbalance of mitochondria and a significantly increased distance between cristae (28.4 ± 5.0 vs. 18.2 ± 1.8 nm; P < 0.001). Finally, western blot analysis was employed to monitor alterations in proteins involved in mitochondrial biogenesis (COX4), respiratory chain (ATP5A1), energetics (OPA1), and mitophagy (ratio LC3B-II/I; Fig. 9A–D, n = 7 each group). While no differences were detected for COX4, ATP5A1 was significantly decreased in AS vs. sham (2.54 ± 0.15 vs. 3.30 ± 0.48 a.u.; P < 0.01). On the other hand, both LC3B-II/I ratio (sham vs. AS: 0.36 ± 0.29 vs. 1.39 ± 0.57 a.u.; P < 0.01) and OPA1 (sham vs. AS: 1.04 ± 0.49 vs. 1.52 ± 0.13 a.u.; P < 0.05) were significantly increased in AS.

High-resolution MRI enabled reliable analysis of valve opening in controls and in mice with AS by accurate planimetric delineation of the aortic valve orifice over the entire cardiac cycle. The provided image quality of valvular structures and orifice area facilitated correct plane recording of all valve cusps with high reproducibility despite diminutive proportions of the valve. Consistent with human valvular disease [48], we identified with this approach functionally bicuspid stenotic aortic valves (e.g. Figure 3). In this case, accompanying AR resembled a coaptation defect due to an immobile rather than a ruptured cusp, which cannot be clearly distinguished by echocardiography due to limits in spatial resolution. Besides assessment of valve morphology, we were also able to delineate valve thickening as a consequence of tissue remodelling, which was confirmed by subsequent histology. However, we have clearly to admit that evaluation of valve thickness is somewhat restricted by pixel size. Compared to histology, leaflet area determined by MRI was overestimated by approximately 30% (~ 0.12 vs 0.09 mm² in AS) which is most likely caused by (i) imprecise orthogonal orientation of the longitudinal MR imaging plane to the valve and (ii) averaging its area over a slice thickness of 1 mm (and not only 6–8 µm as in histologic sections) and thereby along the valve curvature over the leaflets to the cusp. Despite these limitations, the magnitude of valve thickening (~ 0.1 mm²) found by MRI and histology was quite similar in AS mice, which offers the option for non-invasive longitudinal and serial monitoring of subtle changes in valve morphology of murine disease models.

As expected, mice with AS were characterized by an increase in maximum peak velocity comparable to echocardiographic assessment. In line with previous observations in humans [44], absolute values assessed by 2D phase-contrast measurements tended to be lower in all groups, but the extent of flow increase in AS animals (~ 100 cm/s) almost matched the Doppler findings. Furthermore, we identified fragmented turbulent flow patterns which were shifted to substantially higher peak flow velocities as consequence of valve orifice narrowing. The present approach could further be extended to segmented analysis of flow patterns in the entire aorta, which permits full visualization of valves and their inflow/outflow tracts. In contrast, quantification of transvalvular flow by echocardiography is challenging because velocities must be measured in line with the transducer beam, making it susceptible to errors caused by misalignment of the transducer beam to the direction of the blood flow. This is highly relevant in eccentric and dynamic flow jets and essential for reliable quantification of AS [5].

Accurate flow quantification furthermore allows a precise measure of the degree of regurgitation as well as cardiac shunt volumes/ratios and differential flow volumes [39]. In our model, it was possible to identify regurgitation flow as black jets in cine loops and to quantify the resulting backflow across the valve in corresponding velocity maps. With these techniques, we were able to easily differentiate between mice presenting with AS only and AS combined with AR. By future implementation of 4D-flow techniques, also eccentric regurgitation jets could be assessed by multidirectional acquisition encoding all components of the velocity vector [8, 32, 51].

As a consequence of the altered flow patterns in AS, we observed enhanced circumferential strain in mice with AS which was associated with an increase in aortic wall thickness at the level of the aortic root. This implies relevant structural and functional aortic changes in our AS model. Aortic strain in tricuspid aortic stenosis has been assumed to be a diagnostic measure of aortic stiffness in patients [57], but in contrast to findings in humans with bicuspid aortic valve [34], we observed an enhanced circumferential strain in the ascending aorta in animals with functionally bicuspid aortic valve (Fig. 4H). This is most likely due to the pronounced turbulent flow pattern with strongly enhanced radial velocity components and a still intact windkessel function in this early phase of AS.

Of note, the same considerations mentioned above for measurement of valve thickness apply for determination of the dimensions of the aortic wall. Again, the restrictions caused by pixel size and partial volume effects lead to a systematic overassessment as compared to histology [56]. Nevertheless, our results were quite robust for the individual groups investigated and the magnitude of the observed increase in wall thickness (~ 30%) is in the same order as in histologic examinations of a hypertensive mouse model with pathological remodelling of the aorta [28]. Furthermore, also circumferential strains derived from those measures were in pretty good agreement with a previous reference study on morphometry and strain distribution of the C57BL/6 mouse aorta by ultrasound [18].

While we observed a gradual wall thickening of the LV myocardium, LV function was interestingly still preserved 4 weeks after induction of AS without any signs of necrosis (no LGE) or alterations in the ECV. By trend fractional shortening was enhanced in all LV wall areas of AS mice (except for the inferior wall), revealing an early compensatory mechanism together with the mild LV hypertrophy. At the same time, we observed reduced T1 and T2 relaxation times in our AS model, which preclude the presence of myocardial fibrosis and oedema at this stage, but rather reflect early adaptation processes. Histology confirmed the absence of any interstitial/perivascular fibrosis and/or remodelling of the ECM in AS mice and revealed an increased cardiomyocyte size as underlying cause of the observed wall thickening. Concomitantly, while all groups displayed similar maximum oxidative capacity, we observed both an enhanced complex II-dependent respiration and respiratory control ratio in mitochondria of AS mice. This indicates an increased relative ATP yield via ATP synthase for the myocardium [11, 25] and, thus, most likely represents another intracellular compensatory mechanism in response to the extended energy demand for pumping against the elevated afterload [1]. On the other hand, raised activation of complex-II-associated respiration may be accompanied by enhanced release of reactive oxygen species [21] which could be causative for the observed incipient distributional and microarchitecture disorders in the mitochondrial apparatus. Of note, these were also accompanied by adaptions at the protein level impacting on respiratory chain (down-regulation of ATP5A1), energetic efficiency (up-regulation of OPA1) and mitophagic processes (increased ratio of LC3B-II/I) [7, 17, 29]. Importantly, these alterations in tissue texture can provide a simple explanation for the observed decrease in both relaxation times: The beginning disorganization of the usually tightly packed cardiomyocyte ultrastructure leads to increased local tissue inhomogeneity, which is well known to cause faster proton dephasing and affects T1 and T2 in the same way [27, 59]. Thus, it is tempting to speculate that the uncommon simultaneous drop of cardiac T1 and T2 might be useful as a premature in vivo readout for mitochondrial disorders in the heart, which could be even more sensitive than conventional parameters of myocardial damage, such as ECV, LGE, etc. [33].

Together, the induced myocardial adjustments seem to be sufficient to keep LV function at this moment in a compensated state reflecting the latent asymptomatic period of AS also known from human pathophysiology [9]. Nevertheless, despite a trend to an enhanced myocardial capillarization in AS mice, we observed in vivo and ex vivo a significant restriction in coronary flow reserve well known to be associated with AS in humans [23, 35, 60] and as well a mild increase in circulating troponin levels. Obviously, the current time point represents a period of transition, where a variety of compensation mechanisms still keeps cardiac function in balance but the persisting increased afterload already leads to incipient impairment of coronary flow and cardiomyocyte ultrastructure.

Further investigations are required to define more precisely the point in time when the present model turns into the onset of severe symptoms, such as depressed LV function and its sequelae [26]. Thereafter, progressive fibrosis and oedema are expected to reverse the somewhat surprising results of reduced T1 and T2 in our AS mice [4, 5]. However, the early observation of these subtle alterations clearly highlights the potential of the present MR approach to identify early changes in cardiac tissue texture and premature transition points in the adaptive process to AS thereby fostering future mechanistic studies in genetic mice models.

There are rare reports about MRI with focus on aortic valve disease in mice [37, 43, 55], but the comprehensive and quantitative characterization of transvalvular aortic flow profiles together with structural/functional/perfusion changes in aortic valve, LV, and the ascending aorta over the entire cardiac cycle for a holistic analysis of the impact of AS in mice is unique and has not been carried out so far. With this, it is feasible to treat myocardium, valve, ascending aorta and aortic arch together as a mechanistical unit (Fig. 10) and to study pathophysiological changes in the context of complex aortovalvular diseases, such as AS, in all compartments in a serial and simultaneous manner. While echocardiography is established to determine the manifest grade of AS via assessment of aortic flow velocity, our MRI approach allows not only to delineate alterations in flow patterns and valve anatomy but also subtle and early adaptations in tissue texture (T1/T2 mapping), perfusion patterns, and functional feedback loops in parallel (see above). The reliable characterization of structural components such as the annulus, left ventricular outflow tract, aorto-coronary coupling, and valvular morphological changes itself are key features of our imaging approach and highly relevant for future optimization of diagnostic and therapeutic options in aortovalvular diseases.

There are a number of animal models of calcific aortic valve disease (CAVD) recapitulating distinct human hallmarks in diet-induced, genetic, congenital, and developmental mouse as well as rabbit and porcine models, each with its specific restrictions [22, 50]. While diet-induced models often resemble changes in metabolic disorders with mild valvular impairment, some mice and rabbit models acquire hemodynamically significant calcific aortic valve disease, but the same has yet to be shown in pig for long-term follow up [2]. Moreover, it has to be noted, that some hemodynamic situations from mice are not transmissible to human parameters and need to be considered during analysis, e.g. high resting heart rate of about 600 bpm and, therefore, reduced heart rate reserve. On the other hand, while a translational approach with large animal models is closer to human dimension, it entails other restrictions, e.g. only few options for genetic manipulation. In each case, the animal needs to be immobilized during the acquisition phase and general anesthetic regimens have the potential to affect cardiac function [3]. However, to avoid cardiodepressant side effects [30], in our case isoflurane levels were kept at the lowest possible concentration for a stable sleep and a valid assessment of cardiac function and perfusion.

Concerning the reliability of echocardiography and MRI to determine the severity of aortic valve stenosis, many comparisons have been conducted and the results from these clinical studies support the assumption, that both techniques are in good agreement [13, 51, 58]. Beyond altered morphology and flow patterns, the major pathophysiological features of AS comprise inflammation, fibrosis, and calcification. While evaluation of the latter is clearly the domain of CT and ¹⁸F-NaF-PET, MRI offers also the possibility to detect more subtle events which precede the overt myocardial and valvular phenotype of AS, that are inflammatory and fibrotic processes. Additional application of Gd-based contrast agents (CAs) would allow detection of both replacement and interstitial fibrosis by late gadolinium enhancement and determination of extracellular volumes via T1 mapping, respectively. Furthermore, chemical exchange saturation transfer techniques have recently been applied to monitor remodelling of the extracellular matrix, while ¹⁹F MRI in combination with fluorinated CAs permits the background-free visualization of infiltrating immune cells [16, 42]. This bridges the gap to analysis of pathophysiologic pathways and elucidation of potential therapeutic targets in transgenic mouse models. In the clinical setting, this may be helpful to delineate and predict (bioprosthetic) valve degeneration which is expected to become a progressively important issue in the near future [10]. To this end, MRI may be combined with complementary methods (micro-CT, PET-CT, etc.) for a premature identification of disease initiation via altered flow profiles and micro-calcification allowing an early therapy of incipient degenerative processes.

Given that the described MRI approach has the potential to elucidate more subtle changes of the valve itself, the aortic wall and the affected myocardium, this can be expected to provide valid markers for the early diagnosis of incipient adverse myocardial and aortic remodelling allowing a timely initiation of the required therapy. Furthermore, combination of a reproducible animal model of AS with high-end longitudinal imaging will facilitate a better mechanistical understanding of AS which ultimately could lead to identification of critical transition points in disease development at the valvular, myocardial, and aortic level, and in turn to generation of novel pharmacological treatment options in humans. Thus, the further refinement of MRI techniques in well-defined preclinical longitudinal AS studies is a critical prerequisite for the translation of future imaging approaches to guide treatment of disease in affected patients.