Right ventricular myocardial deoxygenation in patients with pulmonary artery hypertension

Background In pulmonary arterial hypertension (PAH), progressive right ventricular (RV) dysfunction is believed to be largely secondary to RV ischaemia. A recent pilot study has demonstrated the feasibility of Oxygen-sensitive (OS) cardiovascular magnetic resonance (CMR) to detect in-vivo RV myocardial oxygenation. The aims of the present study therefore, were to assess the prevalence of RV myocardial ischaemia and relationship with RV myocardial interstitial changes in PAH patients with non-obstructive coronaries, and corelate with functional and haemodynamic parameters. Methods We prospectively recruited 42 patients with right heart catheter (RHC) proven PAH and 11 healthy age matched controls. The CMR examination involved standard functional imaging, OS-CMR imaging and native T1 mapping. An ΔOS-CMR signal intensity (SI) index (stress/rest signal intensity) was acquired at RV anterior, RV free-wall and RV inferior segments. T1 maps were acquired using Shortened Modified Look-Locker Inversion recovery (ShMOLLI) at the inferior RV segment. Results The inferior RV ΔOS-CMR SI index was significantly lower in PAH patients compared with healthy controls (9.5 (– 7.4–42.8) vs 12.5 (9–24.6)%, p = 0.02). The inferior RV ΔOS-CMR SI had a significant correlation to RV inferior wall thickness (r = – 0.7, p < 0.001) and RHC mean pulmonary artery pressure (mPAP) (r = – 0.4, p = 0.02). Compared to healthy controls, patients with PAH had higher native T1 in the inferior RV wall: 1303 (1107–1612) vs 1232 (1159–1288)ms, p = 0.049. In addition, there was a significant difference in the inferior RV T1 values between the idiopathic PAH and systemic sclerosis associated PAH patients: 1242 (1107–1612) vs 1386 (1219–1552)ms, p = 0.007. Conclusion Blunted OS-CMR SI suggests the presence of in-vivo microvascular RV dysfunction in PAH patients. The native T1 in the inferior RV segments is significantly increased in the PAH patients, particularly among the systemic sclerosis associated PAH group.


Introduction
Pulmonary arterial hypertension (PAH) is a progressive disorder that affects both the pulmonary vasculature and the heart [1]. It is a rare but serious disease, affecting up to 50 per million population [2]. Although the initial insult in PAH involves the pulmonary vasculature, prognosis of patients with PAH is most dependent on right ventricular (RV) size and function [3].
While there are several putative causes, RV ischemia is thought to be a significant driver [4]. The increasing pulmonary artery pressure decreases the pressure gradient between the aorta and the RV, thereby reducing myocardial blood flow in the right coronary artery which, leads to myocardial ischemia [5,6]. In addition, the RV adapts to the increased afterload by increasing wall thickness and contractility. However, RV hypertrophy without a corresponding increase in the cross-sectional area of the right coronary artery can exacerbate ischemia of the RV [7]. RV adaptive changes due to the elevated pulmonary artery pressures affects the integrity of the collagen network of the cardiomyocyte and interstitium leading to RV fibrosis [8]. Whether the adaptive changes in the myocardial interstitium has any causal relationship to myocardial ischaemia in PAH patients remains unknown.
Oxygen-sensitive (OS) cardiovascular magnetic resonance (OS-CMR), also known as blood oxygen level dependent (BOLD) CMR enables the in-vivo assessment myocardial oxygenation at the tissue level [9][10][11][12][13]. OS-CMR utilises the natural paramagnetic properties of haemoglobin [14]. Following vasodilator stress, the alterations to the deoxygenated haemoglobin concentration act as an endogenous contrast agent leading to signal intensity (SI) changes in OS-CMR sequence. CMR native T1 mapping can be used to characterise changes in the interstitial myocardium of the left ventricle (LV) [15]. T1 mapping quantifies the T1 relaxation time of each voxel of an image being quantified. Different tissues each have distinctive T1 relaxation times and this can vary with a change in tissue composition. CMR T1 mapping has been used for inflammation/oedema imaging and serves as a surrogate for diffuse interstitial fibrosis in the absence of an alternative cause of interstitial expansion (oedema, infiltrations/fibre disarray) [16]. Previous studies have demonstrated higher T1 values in the RV of patients with PAH [17], however to our knowledge no study has looked at the relationship between interstitial myocardial changes in PAH and myocardial ischaemia. Furthermore, in a recent small proof of concept study, we have demonstrated the feasibility of OS-CMR technique in detecting myocardial oxygenation abnormalities in the RV of 20 patients with PAH [18]. However, the sample size was too small to look at the relationship between myocardial oxygenation and functional or haemodynamic indices. Hence the aim of the present study was to assess the prevalence of RV myocardial ischaemia and its relationship with changes in the myocardial interstitium of patients with known PAH and with non-obstructive coronaries using OS-CMR and native T1 mapping.

Study population
Patients undergoing treatment in the PAH clinics at two South Australian hospitals were invited to participate in this study. Inclusion criteria included right heart catheter (RHC) proven PAH (defined as pulmonary capillary wedge pressure (PCWP) ≤ 15 mmHg, and mean pulmonary artery pressure (mPAP) ≥ 25 mmHg). Exclusion criteria included severe RV dysfunction on echocardiography (determined by tricuspid annular plane systolic excursion < 1.7), echocardiographic LV ejection fraction < 50% and/or coronary artery disease (defined as > 70% luminal stenosis in an epicardial coronary artery at angiography or prior myocardial infarction) as well as contraindications to CMR and/or adenosine (second or third degree heart block, obstructive pulmonary disease or dipyridamole use). Eleven healthy subjects also underwent the same CMR imaging procedures and served as a control group. The healthy subjects had no known cardiac/respiratory disease or symptoms and were free of any cardiac risk factors, including hypertension, smoking and diabetes. This study was approved by the Southern Adelaide Clinical Human Research Ethics Committee (HREC/15/SAC/397), and all participants provided written and informed consent to participate in the study.

CMR protocol
The full protocol has been described elsewhere [18]; in brief, CMR was performed using a 3 T clinical CMR scanner (MAGNETOM Skyra, Siemens Healthineers, Erlangen, Germany) with 18 channel torso phased array coil in conjunction with a spinal coil posteriorly). Electrographically (ECG) gated balanced steady-state free precession (bSSFP) sequence was used to acquire cine images in vertical and horizontal long-axis, and ten short-axis images covering the entire RV and LV. For OS imaging, a single mid-ventricular slice was acquired at mid-diastole using a single-shot T2-prepared ECG-gated bSSFP sequence (FOV of 340 × 340 cm, matrix 168 × 192 m, slice thickness 6 mm, producing a voxel size 2.02 × 1.77 mm. The repetition time (TR) was 256 ms and with echo time (TE) of 1.21 ms, flip angle 44˚). Each OS-CMR image was acquired during a single breath-hold over six heart beats. Prior to adenosine infusion, four resting OS-CMR images were acquired. Stress OS-CMR stress images acquisition commenced 2 min after adenosine infusion. Four to six OS-CMR images were acquired over the stress period with the typical duration of stress OS-CMR image acquisition was 3 min with the total adenosine infusion duration of about 5 min (Fig. 1).
Native T1 mapping were acquired using Shortened Modified Look-Locker Inversion recovery (ShMOLLI) on a single mid-ventricular short-axis slice as previously described [19]. The ShMOLLI sequence was performed with prospective ECG triggering performed until optimal image quality was obtained. This sequence has a FOV of 360 × 360 cm, matrix 192 × 144, slice thickness 8 mm, producing an interpolated voxel size 0.9 × 0.9 mm. The TR was 379.40 and with TE of 1.07. Typically, the flip angle was set to 35 degrees and IPAT (GRAPPA) factor of 2, with inversion time (TI) of 260 ms. The T1 maps were acquired after the resting OS-CMR images prior to adenosine infusion (Fig. 1). The mid-ventricular slice location selected was matched to the selected mid-ventricular resting OS-CMR images. Stress heart rate and blood pressure were obtained every minute of adenosine infusion. Patients were monitored throughout the study by ECG, sphygmomanometry and pulse oximetry.

CMR analysis
CMR analysis was performed using cvi 42 (Circle CVI, Calgary, Canada). The short-axis bSSFP cine images were used to quantify LV and RV systolic function. Standard function parameters were calculated including LV ejection fraction (LVEF), RV ejection fraction (RVEF), cardiac index, and body surface area (BSA) indexed enddiastolic (EDVI), end-systolic (ESVI) volumes for LV/RV and LV indexed myocardial mass. End-diastolic RV mass and volumes was used to calculate RV mass/volume ratio.
The OS-CMR assessment of the RV and LV was performed using cvi 42 (Circle Cardiovascular Imaging, Calgary, Canada) as previously described [18]. In brief the RV was segmented into RV anterior, RV free-wall and RV inferior segments. OS-CMR analysis of RV anterior and RV free-wall were limited by thinned myocardium and partial volume effects [18]. Reliable OS signal was only obtained in RV inferior [18], hence a region of interest (ROI) contour was traced manually in this RV segment. For the LV OS-CMR assessment, the epicardium and endocardium borders were manually traced and the LV was sub-divided into 6 equiangular segments based on a standard American Heart Association (AHA) segmentation of the mid-ventricular slice as previously reported [20]. The mean myocardial Signal Intensity (SI) within the RV and LV segments was obtained as previously described [18]. As it is a cardiac gated sequence, the mean signal intensity at rest and stress was corrected for the variations of heart rate and its effects on T1 relaxation as previously described [21]. The relative SI change was calculated as ΔSI (%) = (SI stress-SI rest)/SI rest × 100. Reproducibility of the OS-CMR of the RV has been previously published [18].
The T1-values of the RV were assessed using cvi 42 (Circle Cardiovascular Imaging). As previously described [22], a ROI contour was traced manually on the T1 maps of inferior RV region avoiding the RV insertion point (Fig. 2); attempts were made to draw ROI in RV anterior and RV free-wall regions but this was not feasible. The ROI was carefully drawn within myocardial walls to avoid partial volumes and epicardial fat (Fig. 2). For the LV, the endocardial and epicardial contours were manually traced and was sub-divided into 6 equiangular segments based on a standard AHA segmentation [20]. A mean myocardial T1 value was obtained within each segment.

Statistical analysis
All analyses were performed using the Stata statistical software version 15.1 (StataCorporation, College Station, Texas, USA). Categorical data are described using frequencies and percentages and continuous data using median and range. Differences between groups in the mean change in myocardial oxygenation response and native T1-values were assessed using Mann-Whitney test. Changes in these same outcomes within patient groups were assessed using the Sign-rank test. Reproducibility of measurements for inter and intra-observers was assessed by using the coefficient of variation and by assessing the absolute and proportional bias. Absolute bias was assessed using a t-test of the differences in the 2 measures versus zero and proportional bias by regressing the differences in the 2 measures on the mean for the Fig. 1 CMR Imaging Protocol. Cine, Oxygen sensitive cardiovascular magnetic resonance (OS-CMR) and native T1 mapping imaging protocol for pulmonary artery hypertension (PAH) and healthy control groups 2 measurements. Associations between test parameters were assessed using the Spearman's rho (ρ) correlation coefficient. A 2-sided Type 1 error rate of alpha = 0.05 was used for assessing statistical significance.

Myocardial oxygenation response (OS-CMR) in PAH patients
Of the total 42 PAH patients recruited, 40 (95%) patients and all the healthy controls completed the OS-CMR study protocol. The mean RV ΔOS-CMR SI change in the inferior RV segments was significantly lower in PAH group compared to the controls (9.5 (-7.4 -42.8) vs 12.5 (9 -24.6)%, p = 0.02) (Fig. 3a). The global LV ΔOS-CMR SI change was also significantly lower in PAH group compared to the controls (11.0 (-2.7 -33.2) vs 21.0 (2.5 -35.7)%, p = 0.004) (Fig. 3b). The ΔOS-CMR SI in the inferior RV wall of PAH patients was comparable to both the LV septal wall

Right ventricular native T1-values in PAH patients
Good quality and analysable RV native T1 maps were obtained from 71% (30/42) PAH patients and 73% (8/11) of controls. Reasons for excluding T1 maps were image artefacts and partial volume effects. Compared to healthy controls, patients with PAH had higher native T1 values in the inferior RV wall: 1303 (1107-1612) vs 1232 (1159-1288)ms, p = 0.049 (Fig. 3c). The LV native T1 for the healthy controls were within previously published ranges [23]. In PAH patients, the native T1 values in the inferior RV wall was comparable to both the LV septal (1303
Among the 5 CTEPH patients, 2/5 (40%) did not complete the OS-CMR protocol and 3/5 (60%) had poor quality RV native T1 maps, hence no comparison was made in this group.

Correlations between RV myocardial oxygenation, RV T1 values, CMR RV volumetric/functional and RHC haemodynamic indices
There was a moderate-good correlation between the inferior RV ΔOS-CMR SI and the global LV ΔOS-CMR SI (r = 0.7, p < 0.001) (Fig. 5a). Compared with the CMR volumetric/functional and haemodynamic indices (Additional file 1: Table S1), the inferior RV ΔOS-CMR SI had a moderate-good inverse correlation to RV inferior wall thickness (r = -0.7, p < 0.001) (Fig. 5b) and moderate inverse correlation to RHC mPAP (r = -0.4, p = 0.02) (Fig. 5c). This suggests that increased RV hypertrophy is associated with increased myocardial deoxygenation. However, compared to native T1 values, there was no correlation between RV ΔOS-CMR SI and RV native T1 values (r = 0.02, p = 0.91) (Fig. 5D). Furthermore, no correlation between RV ΔOS-CMR SI and RV native T1 values was demonstrated in IPAH and SScPAH subgroups. RV native T1 had a moderate inverse correlation to RV mass/volume ratio (r = -0.41, p = 0.036). However, there was no other significant correlation between RV native T1 and CMR volumetric/functional or haemodynamic indices.

Discussion
The principal finding from this study is that myocardial oxygenation response to adenosine stress is blunted in patients with PAH compared to healthy controls. The blunted oxygenation response was demonstrated in PAH patients who were stable on pulmonary vasodilator therapy and at an adaptive stage of PAH with non-obstructive epicardial coronaries, and was correlated inversely with RV wall thickness and RHC mPAP. Although we had demonstrated changes to the myocardial interstitium, no clear interaction with myocardial oxygenation was noted. These findings highlight the potential contribution of myocardial ischemia to the pathophysiology of RV dysfunction in PAH, and could subsequently lead to new potential treatment targets, in such patients.

Right ventricular myocardial deoxygenation in PAH
In PAH, RV myocardial ischemia is acknowledged to be a significant contributor to adverse remodelling and progressive RV dysfunction [4]. Reduced right coronary artery perfusion pressure [24] and microvascular ischemia leads to progressive RV dysfunction causing RV failure and death. Abnormal pressure gradient between the aorta and the RV due to escalating pulmonary vascular resistance, reduces myocardial blood flow in the RV branch of the right coronary artery causing ischemia [7,25]. RV myocardial ischemia is further exacerbated by inadequate compensatory increase in the myocardial vasculature owing to progressive RV hypertrophy (RVH) [7,26,27]. Occlusive microvascular damage and impairment in angiogenesis (known as capillary rarefaction) has been demonstrated in animal models and SScPAH, is thought to cause microvascular ischemia [5,28]. Furthermore, a study by Meloche et al. proposes common pathobiology mechanisms between PAH pulmonary vasculature and coronary arteries [29]. These pathobiology mechanisms trigger coronary vascular remodeling in both the LV and RV leading to microvascular ischemia. In this study, we have demonstrated a blunted myocardial oxygenation response to vasodilator stress in the inferior RV segment on PAH patients. Specifically our study demonstrates the presence in-vivo myocardial deoxygenation in the adaptive stage of PAH. The findings of in-vivo RV myocardial deoxygenation in PAH are novel because abnormalities in myocardial blood flow or microvascular dysfunction may not necessarily lead to myocardial deoxygenation [11]. Arnold et al. had demonstrated that in coronary artery disease, 50% of hypoperfused segments on quantitative myocardial perfusion demonstrate no evidence of deoxygenation [11]. Furthermore in dilated cardiomyopathy, Dass et al. had demonstrated disassociation between microvascular dysfunction and oxygenation [30]. Although we had not performed any direct comparison between myocardial blood flow and myocardial oxygenation, these findings provide further mechanistic insight into pathophysiology of myocardial ischemia in the RV in PAH.
The OS-CMR SI changes in the RV had good correlation with global OS-CMR SI changes seen in the LV, an area of myocardium in which there is good validation with the OS-CMR technique. In our study, the OS-CMR technique has demonstrated significant changes in the inferior RV segment. Thinned myocardium (causing partial volume effects), off-resonance artefacts and image artefacts are known limitations of the OS-CMR technique in the RV [18]. However, these limitations have previously been known to affect the accuracy of OS-CMR SI in LV myocardium [31]. Consistent voxel-size was applied during OS-CMR imaging, hence the measured inferior RV wall thickness provided sufficient coverage to avert partial volume effects. Additionally, it has been shown that the inferior RV segment has the greatest end-diastolic wall thickness once the afterload, serum biomarkers and RV structural adaptation in pre-and post-capillary pulmonary hypertension have been taken into consideration [32].
Furthermore, our study has demonstrated a moderate inverse correlation between RV OS-CMR SI and mPAP. The moderate correlation highlights that there are other pathophysiological factors (such as microvascular ischemia) that affect the RV besides elevated mPAP. Patients with microvascular ischemia often present with typical effort-induced angina, but also with atypical symptoms such as dyspnea on exertion [33]. Exertional limitation is the dominant symptom in PAH, however it is contributed by many factors such as respiratory mechanics/ventilation, cardiovascular response and psychological/emotional aspects [34]. Even though RHC resting hemodynamic measures such as mPAP and cardiac output correlate with severity of symptoms, there are considerable inter-individual variability that is not explained by RHC hemodynamic severity [35]. The findings of myocardial deoxygenation provide further mechanistic insights into the pathophysiology of myocardial ischaemia in PAH population.

Native T1-values in the right ventricle
In PAH, the RV adaptation extends to the cardiomyocytes and interstitium altering the cardiac structure and function [8]. Although the initial adaptive changes of myocardial collagen accumulation serve as a favorable response helping the RV withstand higher pressures, over time maladaptive alteration ensues leading to fibrosis and progressive RV dysfunction [36]. Native CMR T1 parametric mapping has been used as a surrogate for diffuse interstitial fibrosis in the absence of an alternative cause of interstitial expansion (oedema, infiltrations/fibre disarray) [16]. Previous studies of RV native T1 mapping in PAH using Modified Look-Locker Inversion recovery (MOLLI) sequence has produced varying results [17,22,37]. One possible explanation is that each of these studies have adapted an altered version of the MOLLI sequence thus producing differing results. Utilising the ShMOLLI sequence, our study demonstrated increased native T1-values in the inferior RV wall compared to healthy controls. The native T1-values in the inferior RV wall were comparable to the LV septum and LV free wall in PAH patients, similar to the findings previously demonstrated by Spruijt et al. [22]. Unexpectedly in our study, we found that the RV native T1-values did not correlate with inferior RV OS-CMR SI changes. This is pertinent in PAH as the pressure overload leads to disruption of the healthy extracellular matrix in the RV due to the excess collagen formation and myocardial interstitial fibrosis. Changes in RV extracellular matrix could potentially disrupt the coronary microvasculature and hence myocardial oxygen supply. However, the lack of association between RV native T1-values and RV OS-CMR SI in our study suggests that the changes in RV extracellular matrix does not directly impact myocardial oxygenation.
Interestingly our study is the first to have demonstrated a significant difference in the native T1-values between the IPAH and SScPAH. We have used the ShMOLLI sequence in contrast to other studies that have used the MOLLI sequence [17,22,38]. The ShMOLLI sequence is heart rate independent [39] and as such is able to estimate long T1s which is important for assessing oedematous tissue [40]. Previous ShMOLLI T1 mapping studies of the LV in systemic sclerosis patients suggested the presence of interstitial fibrosis and low-grade inflammation [41]. Therefore, the higher T1-value SScPAH seen in our study could signify a combination of diffuse interstitial fibrosis and myocardial inflammation, a theory that fits with the pathophysiology of systemic sclerosis.

Study limitations
Our study only performed a single mid-ventricular slice on OS-CMR imaging rather than the entire ventricle hence there was incomplete coverage of the full RV myocardium. This is unlikely to be of major relevance in RV assessment of PAH which is a global process. While our study had demonstrated in-vivo myocardial deoxygenation in the RV of PAH, we had not performed any direct comparison with myocardial blood flow, an area for future studies. Late-gadolinium enhancement or extracellular volume assessment was not performed due to concerns of patient safety and tolerability during CMR research scan. While OS-CMR sequence has been studied in many conditions [9], to our knowledge it has not been studied in PAH. The multiple breath-hold during rest/stress OS-CMR image acquisitions could potentially be challenging for PAH patients as dyspnoea is a common symptom in this cohort of patients. The number of subjects in the healthy control groups were relatively small especially with female controls, compared to the PAH group. However the control group were age-matched, minimizing any significant bias. This is important, especially in T1-values whereby age is known to influence native myocardial T1-values [39]. Measuring RV T1 on a curved RV myocardium can be challenging. RV T1 mapping with ShMOLLI sequence at 8 mm slice thickness can be a potential substrate for partial volume effects affecting T1-values. However, every effort was taken to ensure the ROI was drawn in the myocardium. Furthermore the adequate reproducibility and the comparable RV T1-values to the LV T1-values in PAH patients suggest accurate interpretation of RV T1. Although this study was prospective, the PAH patients were stable on pulmonary vasodilator therapy. This could have mitigated against the finding of RV ischemia. A larger study would help to determine the clinical and prognostic utility of these novel CMR techniques in PAH.

Conclusions
Blunted OS-CMR SI suggest the presence of in-vivo microvascular dysfunction in the RV of PAH patients. Future studies are essential to determine the utility of OS-CMR of the RV in predicting progressive RV dysfunction and mortality.