DETAILED CARDIAC FUNCTIONAL AND ELECTRICAL PHENOTYPING IN PROPIONIC ACIDEMIA

PI: Bart Bjinens, ICREA Research Professor, Universitat Pompeu Fabra, Barcelona, Spain

“Detailed Cardiac Functional and Electrical Phenotyping in Propionic Acidemia”

In Winter 2023, PAF awarded a $46,000 Initial Research Grant for 3/1/24-2/28/25. Summary of Project

Experience has shown that individuals with Propionic Acidemia (PA) can get Heart Failure (HF) or experience a Sudden Cardiac Death (SCD), induced by cardiac arrhythmias. Given the rarity of PA, little is known about how having the disease, or being a carrier of PA, affects the heart, so treating cardiologists often rely on general guidelines for Heart Failure, associated with the enlargement of the heart and potential decrease in Ejection Fraction; or SCD, often associated with a prolongation of the QT interval on the electrocardiogram (ECG).

However, preliminary assessment of more detailed echocardiographic studies, and individual experience about the lack of improvement using standard therapies, suggests that the PA heart might not behave in a similar way as the majority of HF and SCD patients. Therefore, we have started with the collection of an extended cohort of PA patients, carriers and non-carrier family members, in which we performed comprehensive echocardiograms and ECGs, as well as extended genetic testing.

In this study, we propose to analyze the available imaging and electrical information in detail to study if indeed there is a specific PA cardiomyopathy, that explains both HF and SCD, and how this is expressed in different individuals. To ensure the most comprehensive quantification, we will use contemporary Machine Learning, besides classical cardiological parameters.

If successful, this could open possibilities for a better and more personalized assessment of the PA heart, as well as improved follow-up and (medical and dietary) management to reduce the risk of HF and SCD.

Propionic acidemia (PA) is a rare autosomal recessive metabolic disorder caused by deficiency of
propionyl-CoA carboxylase, resulting in toxic accumulation of propionic acid. Although metabolic
complications are well characterized, the cardiovascular phenotype across different genotypes
remains incompletely understood. PA has classically been associated with dilated cardiomyopathy
and QT prolongation, but the prevalence and spectrum of these abnormalities are unclear. This
study aimed to comprehensively characterize cardiac involvement in homozygous patients and
heterozygous carriers compared with non-carriers, and to define disease phenotypes using machine
learning–based phenomapping.
A total of 114 individuals were enrolled, including 37 homozygous patients, 61 heterozygous carriers,
and 16 controls. To support automated analysis and interpretation, a modular cloud-based platform
was developed.

• Image Analysis Platform (IAP): A tool for extracting features from clinical images.
• ECG Viewer: Dedicated to waveform segmentation, this tool uses Deep Learning algorithms
  to automate and speed up the analysis of QRS complexes, T-waves, and P-waves.
• TransCOR Explorer: A specialized exploration tool that assists clinicians in interpreting
  complex datasets and dimensionality reduction algorithms.
• PA Platform: Provides access to individual patient data, including clinical timelines and risk
  assessment calculators track patient trajectories through different risk phenogroups.
  The initial analysis was restricted to the pediatric cohort and included 31 homozygous individuals, 28
  heterozygous carriers and 11 controls. Cardiac assessment was performed using standard 12-lead
  ECG and echocardiographic measurements. Age-adjusted z-scores were used to account for
  physiological differences, and genotype-associated differences were assessed statistically. For
  phenomapping, 44 features, including ECG and echocardiographic variables, were analyzed using
  Unsupervised Multiple Kernel Learning (UMKL), followed by K-means clustering to identify four
  phenogroups.
  The study offers novel insights into the cardiac phenotype associated with PA across different
  genotypes:
• PA Phenotype: While PA has traditionally been associated with dilated cardiomyopathy and
  prolongation of QT interval, this study suggests that these severe manifestations are not
  generalized. Instead, observations point toward a distinct clinical spectrum. As a hallmark for
  PA, and present in all ages, the left ventricle is clearly more spherical. LV dilatation is present
  in a more varying extent, alongside progressive QT prolongation, but very few individuals
  strictly meet the diagnostic thresholds for dilated cardiomyopathy or prolonged QT, as
  defined by standard guidelines. Similarly, ejection fractions are only abnormal in very few
  patients, and are therefore unlikely to be helpful for management in most individuals.
  Furthermore, there is a tendency towards increased heart rates.
• Carrier Phenotypes: Contrary to current understanding, this pathological spectrum appears
  to extend toward heterozygous carriers. Carrier individuals, previously thought to be
  unaffected, demonstrated an intermediate phenotype characterized by enlarged ventricular
  dimensions and reduced BMI z-scores relative to controls.
• Phenotypic Continuum: Machine Learning-based phenomapping effectively distinguished
  genotypes but also revealed shared cardiac features between certain heterozygous and
  homozygous clusters. This overlap supports the existence of a phenotypic continuum,
  suggesting carriers may be at risk for subclinical or mild cardiac disease manifestations.
• Risk Stratification: The progressive increase in QTc intervals, ventricular volumes, and
  sphericity index across clusters shows the potential of advanced phenotyping tools for
  improved risk stratification in all PA-associated genotypes.