Biventricular cardiac dynamic shape: genetics and cardiometabolic disease associations

This study demonstrates that a novel dynamic cardiac shape atlas derived from UK Biobank CMR imaging captures unique functional remodeling patterns and genetic loci not reflected in standard measures, thereby significantly improving the prediction of incident cardiometabolic diseases and offering new insights into the genetic architecture of cardiac function.

The Gist

Imagine your heart isn't just a pump that beats in a steady rhythm. Think of it instead as a living, breathing sculpture that changes its shape, stretches, and squeezes with every single beat.

For a long time, doctors and scientists have measured the heart like a carpenter measuring a wooden box: they check the volume (how much water it holds), the wall thickness (how thick the wood is), and the ejection fraction (how much water it squirts out). These are the "standard measurements."

But this new study suggests that looking at the heart only as a static box misses the magic. It's like judging a dancer only by their height and weight, ignoring how they move, stretch, and leap.

Here is the breakdown of this research, translated into everyday language:

1. The New "Heart Atlas"

The researchers used advanced MRI scans to create a 3D movie of the heart's shape. Instead of just looking at the heart when it's full of blood (relaxed) or empty (squeezed), they looked at the entire journey between those two moments.

They used a mathematical tool called "Principal Component Analysis" (think of it as a shape-shifting dial) to break down the heart's movement into 14 distinct "styles" of motion.

• PC1 is simply "Size": Some hearts are just bigger, some are smaller.
• PC2 is "The Squeeze": How far the top of the heart moves down when it pumps (like a piston).
• PC13 is "The Efficiency": How well the whole heart empties its blood.

2. The "Hidden Clues" in the Shape

The big discovery is that these 14 "motion styles" tell us things that standard measurements miss.

• The Analogy: Imagine two cars. Car A and Car B both have a 200-horsepower engine (standard measurement). But Car A has a suspension that bounces wildly, while Car B glides smoothly. If you only look at the horsepower, you think they are the same. But if you look at the ride quality (the dynamic shape), you can tell which one is about to break down.
• The Finding: The researchers found that specific ways the heart moves (like how much the valve rings stretch) were strong predictors of future heart disease, even when the standard "horsepower" numbers looked normal.

3. The Genetic Blueprint

The study then asked: "Why do some hearts move differently than others?"
They looked at the DNA of nearly 37,000 people. They found 75 specific genetic locations (like coordinates on a map) that control how the heart moves.

• The Surprise: 14 of these locations were brand new discoveries. We didn't know these genes existed before.
• The Metaphor: Think of the heart's shape as a house. We knew the genes that built the walls (size) and the roof (volume). But this study found the genes that control the hinges on the doors and the springs in the windows (the dynamic movement). Some of these new genes are like the "architects" of the heart's development, while others are the "mechanics" that keep the muscles contracting smoothly.

4. Predicting the Future (The Crystal Ball)

The researchers tested if knowing these "motion styles" could predict who would get sick.

• The Result: Yes! Adding these dynamic shape measurements to the standard tests made the prediction of Ischemic Heart Disease (clogged arteries) and Heart Failure much more accurate.
• The Takeaway: It's like a weather forecast. Standard tests tell you the temperature today. These new shape measurements tell you if a storm is brewing three days from now, allowing doctors to prepare earlier.

5. The "Cause and Effect" Mystery

Finally, they used a special statistical trick (Mendelian Randomization) to figure out if the heart shape causes the disease, or if the disease just changes the shape.

• The Verdict: It's a two-way street, but mostly the shape matters first. For example, if your genes make your heart pump less efficiently (a specific shape trait), you are genetically destined to have a higher risk of Heart Failure later in life. The heart shape isn't just a symptom; it's a warning sign written in your DNA.

Why Does This Matter?

• For Patients: In the future, a doctor might look at your heart MRI and say, "Your pump size is fine, but the way your heart stretches is slightly off. This puts you at risk, so let's start prevention now."
• For Science: We now have a new list of genes to study. This gives scientists new targets for drugs that could fix the "hinges" and "springs" of the heart before it fails.

In a nutshell: This study taught us that the heart is more than a static pump; it's a dynamic dancer. By learning the steps of that dance, we can predict heart trouble earlier, understand our genetic risks better, and potentially save lives by catching the problem before the music stops.

Technical Summary

1. Problem Statement

Cardiac morphology and disease have a bidirectional relationship: disease causes remodeling, and pre-existing morphology influences disease risk. While standard Cardiac Magnetic Resonance (CMR) measures (e.g., volumes, mass, ejection fraction) are clinically useful, they fail to capture the full complexity of 3D cardiac shape and motion. Previous genetic studies have focused on static morphology (specifically end-diastole, or ED) or standard scalar metrics. However, the dynamic cardiac shape—capturing the coupled geometry and function of both ventricles across the entire cardiac cycle (end-diastole and end-systole, or ED-ES)—remains largely unexplored in terms of its genetic architecture and its incremental value in predicting cardiometabolic disease.

2. Methodology

The study utilized data from 36,992 unrelated white European participants in the UK Biobank who underwent CMR imaging and had high-quality genetic data.

• Dynamic Shape Atlas Construction:

  • Automated segmentation of LV and RV endocardial/epicardial surfaces and valve landmarks was performed using deep learning (Circle cvi42).
  • A non-rigid diffeomorphic registration framework generated biventricular subdivision surface meshes at ED and ES.
  • A statistical shape atlas was constructed by aligning ED meshes (Generalized Procrustes alignment) and applying the same transformation to ES meshes to preserve individual displacement patterns.
  • Principal Component Analysis (PCA) was applied to the concatenated ED-ES vectors. Fourteen Principal Components (PCs) explaining >1% of variance each (cumulative 83.3%) were retained.

• Phenotypic Associations:

  • Correlations: PCs were correlated with standard CMR measures (volumes, mass, EF, strain) and risk factors.
  • Disease Association: Multivariable logistic regression (prevalent disease) and Cox proportional hazards models (incident disease) were used to test associations with six cardiometabolic diseases: Ischemic Heart Disease (IHD), Heart Failure (HF), Atrial Fibrillation (AF), Significant Atrioventricular Block (AVB), Stroke, and Type 2 Diabetes (T2D).
  • Prediction Modeling: Penalized Cox regression (Elastic Net) compared models using standard CMR measures vs. models adding dynamic shape PCs to predict incident disease (assessed via C-index).

• Genetic Analyses:

  • GWAS: Performed for each of the 14 PCs using BOLT-LMM (linear mixed model), adjusting for covariates and genetic principal components.
  • Functional Annotation: Candidate genes were prioritized using VEP, eQTL colocalization, TWAS, Hi-C chromatin interactions, and Polygenic Priority Score (PoPS). Pathway enrichment analysis was conducted via FUMA.
  • Rare Variants: Gene-burden testing (ReGenie) was performed to assess rare variant contributions.
  • Polygenic Risk Scores (PRS): PRS were constructed for each PC and tested against disease outcomes in independent cohorts.
  • Mendelian Randomization (MR): Two-sample MR was used to infer causal relationships between dynamic shape PCs and cardiometabolic diseases.

3. Key Contributions

• Novel Phenotyping: Introduction of a dynamic biventricular shape atlas that captures both structural geometry and functional deformation (systolic motion) simultaneously, moving beyond static ED-only models.
• Genetic Discovery: Identification of 75 genome-wide significant loci associated with dynamic cardiac shape, including 14 variants previously unreported for any cardiac trait.
• Clinical Utility: Demonstration that dynamic shape features provide incremental predictive value for incident Ischemic Heart Disease (IHD) beyond standard CMR measures and clinical covariates.
• Causal Inference: Establishment of causal links between specific dynamic shape features (e.g., reduced ejection fraction) and Heart Failure using Mendelian Randomization.

4. Key Results

• Phenotypic Characterization:

  • The 14 PCs captured distinct biological patterns. PC1 represented overall heart size. PC2 reflected mitral/tricuspid annular plane systolic excursion (MAPSE/TAPSE). PC13 captured global biventricular ejection fraction.
  • Dynamic PCs showed only modest correlations with standard CMR measures (e.g., PC2 correlated weakly with LVEF, $r=0.18$), indicating they capture complementary functional information not represented by standard metrics.

• Disease Associations:

  • All 14 PCs were associated with at least one incident cardiometabolic disease.
  • Strongest Associations: Incident IHD, HF, and AVB.
  • Notable Finding: Increased systolic annular excursion (PC2) was protective against prevalent IHD but strongly associated with increased risk of incident IHD (HR 3.01), suggesting a complex relationship between compensatory mechanisms and future risk.
  • Reduced ejection fraction (PC13/PC14) was strongly linked to incident HF and IHD.

• Predictive Modeling:

  • Adding PC2 (annular excursion) to a model with standard CMR measures and clinical covariates significantly improved the prediction of incident IHD (C-index increased from 0.63 to 0.67).
  • Similar improvements were seen for incident Type 2 Diabetes prediction.

• Genetic Architecture:

  • Heritability: All PCs were heritable ($h^2$ range: 12.3%–36.9%).
  • Loci Discovery: 75 loci identified. 14 were novel to cardiac traits.
  • Candidate Genes: Enrichment in pathways related to cardiac development, muscle structure, and contractile function. Key genes included BAG3, TTN, FHOD3, WNT5A, and PPIP5K2.
  • Rare Variants: Gene-burden testing implicated established cardiomyopathy genes (e.g., TTN, CSRP3) and novel genes (RBM20, SMARCAD1).

• PRS and Causality:

  • PRS derived from dynamic shape PCs were significantly associated with incident HF and AF.
  • Mendelian Randomization: Confirmed a causal effect of reduced biventricular ejection fraction (PC13) on Heart Failure risk ($\beta$ 0.492, $P=6\times10^{-7}$). Causal relationships were also found between T2D and heart size (PC1) and between IHD and basal systolic excursion (PC9).

5. Significance

This study establishes dynamic cardiac shape atlasing as a powerful framework for integrating cardiac structure, function, and genetics.

1. Beyond Standard Metrics: It proves that capturing the motion of the heart (dynamic shape) reveals genetic and phenotypic signals missed by static volumes or standard ejection fraction.
2. Risk Stratification: The ability of dynamic shape PCs to improve the prediction of incident IHD suggests potential clinical utility in refining risk assessment for patients undergoing CMR.
3. Biological Insight: The identification of novel genetic loci and pathways (e.g., WNT5A in development, PPIP5K2 in contractile function) provides new targets for understanding the mechanisms of cardiac remodeling and cardiometabolic disease.
4. Causal Understanding: By distinguishing between disease-induced remodeling and shape-mediated susceptibility via MR, the study clarifies the directionality of risk factors in cardiometabolic disease progression.

In conclusion, the authors demonstrate that dynamic shape phenotyping is a superior approach for uncovering the genetic basis of cardiac function and for predicting future cardiometabolic events compared to traditional imaging metrics.