Characterising heart failure associated with diabetes using cardiac MRI and blood transcriptomics

Presenting Author(s)

• 

  Marilena Giannoudi, PhD, MSc

  Academic Clinical Lecturer in Cardiology
  University of Leeds
  Leeds, England, United Kingdom

Primary Author(s)

• 

  Marilena Giannoudi, PhD, MSc

  Academic Clinical Lecturer in Cardiology
  University of Leeds
  Leeds, England, United Kingdom

Co-Author(s)

• MC

  Marcella Conning-Rowland

  Phd Candidate
  University of Leeds, United Kingdom
• WJ

  Wasim Javed, MBChB, MRes

  MRI Fellow
  Leeds Institute of Cardiovascular and Metabolic Medicine, United Kingdom
• PT

  Patrick Thompson

  Cardiac Imaging Fellow
  Leeds Institute of Cardiovascular and Metabolic Medicine
  Leeds, England, United Kingdom
• OB

  Oliver Brown

  Cardiology Research Fellowgy
  University of Leeds, United Kingdom
• MD

  Michael Drozd

  Academic Clinical Lecturer in Cardiology
  University of Leeds, United Kingdom
• CC

  Chew Cheng

  Post Doc in cardiovascular research
  University of Leeds, United Kingdom
• 

  Sven Plein, MD, PhD

  Professor of Cardiology
  Leeds Institute of Cardiovascular and Metabolic Medicine
  Leeds, England, United Kingdom
• EL

  Eylem Levelt

  Academic Clinical Lecturer
  Leeds Institute of Cardiovascular and Metabolic Medicine, University of Leeds
  Leeds, England, United Kingdom
• 

  Peter P. Swoboda, PhD

  Consultant Cardiologist
  Leeds Institute of Cardiovascular and Metabolic Medicine
  Leeds, England, United Kingdom
• RC

  Richard Cubbon, PhD

  Associate Professor
  University of Leeds
  Leeds, England, United Kingdom

Background: Type 2 diabetes (T2D) increases the incidence and worsens prognosis of heart failure with reduced ejection fraction (HFrEF). We aimed to define blood transcriptomic correlates of cardiovascular magnetic resonance imaging (CMR) parameters in patients with and without T2D.

Methods: We prospectively recruited patients with HFrEF (left ventricular ejection fraction (LVEF) ≤40% on echocardiography) who subsequently underwent CMR. The CMR protocol consisted of cine imaging, pre- and post-contrast T1 mapping (MOLLI), quantitative dual sequence adenosine stress and rest perfusion imaging with automated inline myocardial blood flow (MBF) analysis, and late gadolinium enhanced (LGE) imaging.Patients with a known history of coronary artery disease ( >70% on invasive coronary angiography), myocardial infarction, coronary revascularisation or anginal symptoms were excluded.





At the time of CMR, patients had whole blood preserved using PAXgene tubes for subsequent RNA extraction. Bulk RNA sequencing (RNA-seq) of single-end reads was performed using Illumina NextSeq 2000, followed by alignment to the human genome. Principal component (PC) analysis was used to extract the largest signals in overall RNA-seq data, allowing us to correlate CMR variables with RNA-seq. These data guided differential gene expression (DEG) analyses to identify specific genes of interest.

Results: A total of 74 patients were recruited, (25% female; mean age 66 [63, 68] years). Their clinical and CMR characteristics are shown in Table 1.




Over 40,000 genes were quantified by RNA-seq. We extracted their first 10 PCs, which accounted for >65% of gene expression variance between patients. To illustrate the strength of associations between these PCs and CMR characteristics, correlation (Eigencor) plots were generated. Figure 1 illustrates this for CMR data of the HFrEF patients. Here, PC1 demonstrated strong positive correlations with left ventricular (LV mass), modest positive correlations with indexed left ventricular end-diastolic and end-systolic volumes (LVEDVi & LVESVi) and the presence of inducible ischaemia. Strong negative correlations were seen with stress myocardial blood flow (MBF). DEG analyses were used to explore these associations in more detail. For example, expression of 30 genes differed between patients in the upper and lower quartiles of stress MBF.

Conclusion: Whole blood gene expression patterns correlate with CMR phenotypes offering insights into myocardial remodelling in response to a new diagnosis of HFrEF. These data can be used to define specific genes associated with CMR metrics, further exploration of which may help to understand CMR traits and develop blood biomarkers.

Table 1: Table of the Clinical Characteristics, Biochemistry of recruited HFrEF patients

Variable	Patients with HFrEF N=74
Age, y	66 (63, 68)
Female, n (%)	25 (34)
BMI, kg/m2	29 (28, 30)
Heart rate, bpm	67 (63, 71)
Systolic BP, mmHg	119 (114, 124)
Diastolic BP, mmHg	76 (74, 79)
Haemoglobin, g/L	147 (143, 151)
HbA1c, mmol/mol	44 (41, 46)
Cardiovascular Past Medical History
Stroke TIA, n(%)	5 (7)
PAF, n(%)	19 (26)
Hyperlipidemia, n (%)	30 (41)
Hypertension, n (%)	38 (51)
CMR Results
LV end-diastolic volume indexed to BSA, mL/m2	121 (108, 134)
LV end-systolic volume indexed to BSA, ml/m2	79 (67, 92)
LV mass, g	145 (133, 156)
LV mass indexed to BSA, g/m2	70 (64, 77)
LV mass to LV end-diastolic volume, g/mL	0.62 (0.59, 0.66)
LV ejection fraction, %	37 (34, 40)
Native T1, (ms)	1346 [1315, 1385]
Extra cellular volume fraction, (%)	32 (31, 33)
Ischaemic LGE, n(%)	22 (30)
Non-Ischaemic LGE, n(%)	26 (35)
Visual Inducible ischaemia, n(%)	9 (12)
Stress MBF, ml/min/g	1.5 (1.3, 1.5)
Rest MBF, ml/min/g	0.57 (0.52, 0.61)
% change in RPP (stress - rest), %	16 (10, 21)

 

HFrEF indicates heart failure with reduced ejection fraction; y, years; BMI, body mass index; bpm, beats per minute; BP, blood pressure; HbA1c, glycemic hemoglobin; TIA, transient ischemic attack; PAF, paroxysmal atrial fibrillation; LV, left ventricle; BSA, body surface area; LGE, late gadolinium enhancement; MBF, myocardial blood flow; RPP, rate pressure product. Normally distributed continuous variables are expressed as mean (±95% confidence intervals); nonparametric continuous variables are expressed as median [IQR]; and categorical variables are expressed as counts (percent). P signifies p value for comparisons across the groups with t-test for normally distributed datasets and Mann-Whitney test for non-parametric tests.

Figure 1: Eigencor plot showing principal components 1-10 correlated against selected CMR parameters. The scale denotes R values with asterisks representing p<0.05.