Adipose tissue dysfunction, inflammation, and insulin resistance: alternative pathways to cardiac remodelling in schizophrenia. A multimodal, case–control study

Cardiovascular diseases are the leading cause of death in schizophrenia. Patients with schizophrenia show evidence of concentric cardiac remodelling (CCR), defined as an increase in left-ventricular mass over end-diastolic volumes. CCR is a predictor of cardiac disease, but the molecular pathways leading to this in schizophrenia are unknown. We aimed to explore the relevance of hypertensive and non-hypertensive pathways to CCR and their potential molecular underpinnings in schizophrenia. In this multimodal case–control study, we collected cardiac and whole-body fat magnetic resonance imaging (MRI), clinical measures, and blood levels of several cardiometabolic biomarkers known to potentially cause CCR from individuals with schizophrenia, alongside healthy controls (HCs) matched for age, sex, ethnicity, and body surface area. Of the 50 participants, 34 (68%) were male. Participants with schizophrenia showed increases in cardiac concentricity (d = 0.71, 95% CI: 0.12, 1.30; p = 0.01), indicative of CCR, but showed no differences in overall content or regional distribution of adipose tissue compared to HCs. Despite the cardiac changes, participants with schizophrenia did not demonstrate activation of the hypertensive CCR pathway; however, they showed evidence of adipose dysfunction: adiponectin was reduced (d = −0.69, 95% CI: −1.28, −0.10; p = 0.02), with evidence of activation of downstream pathways, including hypertriglyceridemia, elevated C-reactive protein, fasting glucose, and alkaline phosphatase. In conclusion, people with schizophrenia showed adipose tissue dysfunction compared to body mass-matched HCs. The presence of non-hypertensive CCR and a dysmetabolic phenotype may contribute to excess cardiovascular risk in schizophrenia. If our results are confirmed, acting on this pathway could reduce cardiovascular risk and resultant life-years lost in people with schizophrenia.


Osimo et al, 2021
Adipose tissue dysfunction, inflammation, and insulin resistance: alternative pathways to cardiac remodelling in schizophrenia. A multimodal, cross-sectional study Supplementary materials

Cardiac MRI Acquisition
Ventricular function was assessed in the conventional manner using balanced steady-state free precession (bSSFP) cine images acquired in cardiac short-axis and long-axis planes (Kramer, Barkhausen et al. 2020) using a repetition time (TR), 3.3ms; echo time (TE), 1.43ms, flip angle, 40°; voxel size 1.5x1.5x8mm; bandwidth 962Hz/pixel; parallel imaging factor, 3; and temporal resolution, 45ms. Pulse wave velocity was assessed using a phase contrast (PC) sequence positioned perpendicular to both the ascending and descending aorta at the level of the bifurcation of the pulmonary trunk. The PC data were acquired with a retrospectively ECG-gated gradient echo sequence using a velocity encoding gradient of 180cm/s in the through-plane direction and a TR of 4.6ms; TE, 2.47ms, flip angle, 20°; voxel size 1.8x1.8x6mm; bandwidth 949Hz/pixel; parallel imaging factor, 2; and temporal resolution, 37ms. To calculate path length, a bSSFP single-shot sequence was acquired in a sagittal plane through the aortic arch using a TR of 3ms, TE, 1.23ms, flip angle, 60°; voxel size 1.3x1.3x8mm; bandwidth 849Hz/pixel; parallel imaging factor, 2. Myocardial tissue changes were assessed according to consensus guidelines (Messroghli, Moon et al. 2017) by measuring the longitudinal relaxation time constant (native myocardial T1 time). A Modified Look-Locker Inversion recovery (MOLLI) sequence, 5s (3s) 3s version, was acquired in a mid-ventricular short axis slice using a prospectively ECG-gated bSSFP single-shot sequence and a TR of 2.7ms; TE, 1.12ms, flip angle, 35°; voxel size 1.4x1.4x8mm; bandwidth 1085Hz/pixel; parallel imaging factor, 2.

Whole Body MR Acquisition
To assess whole-body fat, participants were positioned supine in the scanner and were scanned from the top of their head to their toes. Axial images were acquired using the integrated body coil in 10 contiguous sections of sixty slices with a Dixon volumetric interpolated breath-hold examination (VIBE) sequence and a TR of 6.53ms, TEs, 1.34ms and 2.57ms, flip angle, 10°; voxel size 2.5x2.5x3mm; bandwidth 620Hz/pixel. Sections through the neck, chest and abdomen were acquired at suspended expiration.

MR Image Analysis
All MR image analysis was performed using pseudonymised participant codes, which preserved the operator's blinding to diagnosis.

Cardiac Mass and Volume
Volumetric analysis of the cine images was performed using CMRtools (Cardiovascular Imaging Solutions, London, UK) by an experienced user (10 years of CMR experience), blind to diagnosis. Subjects whose images were degraded by respiration or ECG synchronisation artefacts to the extent where cardiac contours could not be clearly identified were excluded from the analysis. After manual segmentation of epi-and endocardial borders at end-diastole and end-systole, semi-automated thresholding was used to identify the papillary muscles; these were included in the left ventricular mass and excluded from volumetric measurements. To increase the accuracy of measurements, the valve positions were identified on the long-axis images, allowing the valve planes to be tracked through the cardiac cycle. Volumes and mass were indexed to body surface area to give the indexed volumetric data: left ventricular mass (LVMi) and left ventricular end-diastolic volumes (LVEDVi). The end-diastolic volume was calculated using the Simpson's method of discs (Schiller, Shah et al. 1989).

Whole body fat MRI
Fat content and distribution were determined as previously described(O'donovan, Thomas et al. 2009). Images were analysed using SliceOmatic (Tomovision, Montreal, Quebec, Canada), and regional volumes were recorded in litres (L), including total adipose tissue and visceral fat (Thomas, Parkinson et al. 2012).

Pulse Wave Velocity
PWV analysis was performed using the ARTFUN software (ART-FUN; Inserm, Paris, France) (Dogui, Redheuil et al. 2011) by an experienced CMR user (6 years CMR experience). Using a magnitude PC image, the contours of the ascending and proximal descending aortic were semiautomatically delineated and then propagated throughout the cardiac cycle. To define the path length, six to eight markers were defined across the aortic arch to create a three-dimensional Bezier curve through the centre of the aorta that intersected the plane where the flow measurements were acquired. The arch-PWV was calculated as the ratio between this 3D length of the aortic arch, and the transit time (Dt) between the velocity waveforms in the ascending and descending aorta.

Native T1
The MOLLI sequence was analysed using Circle Cardiovascular Imaging (CVI), Calgary, Canada, version 5.12.1 T1 mapping software. A Siemens recommended correction factor of 1.035 was applied. Each slice was divided into 6 segments, as per the American Heart Association model, with the 2 septal segments defined by the anterior and inferior right ventricular insertion points. An epicardial and endocardial erosion offset of 10% was applied to the contours to ensure only myocardium was included. Septal T1 was derived by taking a mean of the two septal segments; septal T1 was used in place of the global T1 of the mid-ventricular short axis slice, to reduce partial volume artefact.

Statistical analysis
For measures such as CRP, which have an established risk cutoff, odds ratios (ORs) and 95% confidence intervals (CIs) for "high" vs "low" values in schizophrenia vs controls were calculated and reported separately. All multivariable analyses use continuous values. All analyses were performed in R (R Core Team 2021). Significance was taken as p <0.05 (two-tailed). Difference in pathways between patients and HCs, were tested using a one-way multivariate analysis of variance (one-way MANOVA or Wilk's test (Todorov and Filzmoser 2009)) for each group of continuous dependent variables, with diagnosis as the independent variable, to determine for group differences. The cardiac measures pathway included LV concentricity, native septal T1, PWV. LVMi and LVEDVi are reported in Table 2 but not included in group testing as already included as their ratio, LV concentricity. The fat measures pathway included total, visceral body fat and their ratio. The hypertensive pathway included systolic and diastolic blood pressure, PWV, active renin, NT-proBNP and Troponin I. The non-hypertensive pathway measures included leptin and adiponectin, ALP, GGT, ALT, triglycerides, HDL and LDL, hsCRP, insulin, glucose, HOMA and endothelin-1. If the Wilk's test was significant, post-hoc tests for each variable were conducted. P values outside of pathways, i.e., linear regression results, were adjusted using the Benjamini & Hochberg method.

Supplementary Figure 3: regression of log-transformed triglycerides over concentricity in schizophrenia vs healthy controls.
Left panel: whole sample. Right panel: schizophrenia (SCZ) in orange, healthy controls (HC) in green. Figure 4: regression of log-transformed endothelin-1 over concentricity in schizophrenia vs healthy controls.