Lipidomics Identifies HFpEF Phenogroups and a High-Risk Metabolic Signature - The BElgian and CAnadian MEtabolomics in HFpEF (BECAME-HF) project.

The BECAME-HF project utilized plasma lipidomics to identify distinct, clinically relevant HFpEF phenogroups across Belgian and Canadian cohorts, revealing a specific high-risk metabolic signature characterized by severe organ dysfunction and reduced survival.

The Gist

Imagine Heart Failure with Preserved Ejection Fraction (HFpEF) as a massive, chaotic traffic jam in a city. The cars (blood) are moving, and the engine (the heart) is still strong enough to pump, but the roads are clogged, the traffic lights are broken, and the city is gridlocked. For years, doctors have treated all these traffic jams the same way, but it hasn't worked very well because every jam has a different cause. Some are caused by too many cars (obesity), others by bad road construction (aging), and some by mysterious accidents we can't see.

This paper is like a team of super-smart detectives who decided to stop looking at the traffic from the outside and instead started analyzing the fuel the cars are burning. They realized that by looking at the specific "chemical fuel" (lipids/fats) floating in the blood, they could figure out exactly which type of traffic jam a patient has.

Here is the story of their discovery, broken down simply:

1. The Big Discovery: It's Not Just One Disease

The researchers took blood samples from two groups of people: those with HFpEF and those with healthy hearts. They used a high-tech scanner (lipidomics) to look at hundreds of different types of fats in the blood.

The Analogy: Imagine you have a bag of mixed Lego bricks. Before, doctors just saw "a bag of bricks." This study showed that the bag actually contains three very different sets of instructions.

• Set A: Bricks for a rusty, broken-down factory.
• Set B: Bricks for a sleek, aging car.
• Set C: Bricks for a car overloaded with too much cargo.

They found that HFpEF patients naturally fall into three distinct groups based on their fat profiles, not just their symptoms.

2. The Three "Traffic Jam" Types

The study identified three specific "phenogroups" (types of patients):

• The "High-Risk Factory" Group (Cluster B1):

  • The Vibe: This is the most dangerous group. Their blood showed signs that their internal "factories" (liver and heart) were under massive stress and breaking down.
  • The Clues: Their blood was full of "rusty exhaust" (specific fats called acylcarnitines and oxidized sphingomyelins) and lacked the "clean fuel" needed for energy.
  • The Result: These patients had the worst outcomes. They were more likely to be hospitalized or pass away. It's like a car engine that is overheating and about to seize up.

• The "Aging Car" Group (Cluster B2):

  • The Vibe: These patients were mostly older and had issues with their heart rhythm (atrial fibrillation), but their internal organs were actually doing okay.
  • The Clues: Their blood looked relatively "clean," with good levels of "protective oil" (HDL cholesterol) and healthy kidney function.
  • The Result: They had the best survival rates. They were like an old car that still runs smoothly, even if it's not new.

• The "Overloaded Cargo" Group (Cluster B3):

  • The Vibe: This group looked like the classic "obese metabolic" patient. They had high blood sugar, high triglycerides, and were carrying a lot of extra weight.
  • The Clues: Their blood was full of "cargo" (triglycerides and glucose).
  • The Result: They were distinct from the others, driven by metabolic overload rather than organ failure.

3. The "Magic 10" List

The researchers realized that to spot the dangerous "High-Risk Factory" group, you don't need to scan the whole bag of Legos. You only need to look at 10 specific bricks.

They created a 10-lipid signature. If a patient's blood has high levels of these 10 specific fats, it's a giant red flag. It's like a smoke detector that goes off specifically when the kitchen is on fire, ignoring the toast that's just slightly brown. This signature was strongly linked to liver damage, heart stress, and poor survival.

4. The "Copycat" Problem

The team tried to find these same groups in a second group of patients from Canada.

• The Good News: They found the "Aging Car" and "Overloaded Cargo" groups easily.
• The Bad News: The dangerous "High-Risk Factory" group was almost missing in the Canadian group.
• Why this matters: It proves that not every hospital or city has the same mix of traffic jams. If you only look at one type of patient, you might miss the most dangerous ones entirely. This is why personalized medicine is so important.

The Bottom Line

This paper tells us that HFpEF isn't just one disease; it's three different diseases wearing the same mask.

• Before: Doctors treated everyone the same, hoping for the best.
• Now: We have a "chemical map" that can tell us who is in the "High-Risk Factory" group.

The Takeaway: By looking at the specific fats in a patient's blood, doctors can soon identify who is in immediate danger (the "Factory" group) and treat them differently than the "Aging Car" or "Overloaded Cargo" groups. It's like moving from giving everyone the same generic map to giving each driver a GPS that knows exactly where their specific roadblock is.

This is a huge step toward precision medicine, where treatment is tailored to the specific "fuel mix" of the patient, potentially saving lives by catching the most dangerous cases early.

Technical Summary

1. Problem Statement

Heart Failure with Preserved Ejection Fraction (HFpEF) is a heterogeneous syndrome with high morbidity and mortality, yet effective therapies remain limited. A major barrier to progress is the lack of understanding of the underlying biological heterogeneity among patients. While previous studies have used clinical or echocardiographic data to define phenogroups (e.g., "metabolic obese," "older vascular"), these approaches often fail to capture the full molecular complexity of the disease. Specifically, the role of systemic lipid dysregulation in defining distinct HFpEF subtypes remains largely unexplored, particularly using untargeted lipidomics rather than targeted analyses of a few metabolites.

2. Methodology

The study employed a multi-cohort, untargeted lipidomics approach combined with machine learning to stratify HFpEF patients.

• Cohorts:
  • Discovery Cohort (BECAME-HF1/Belgian): 105 HFpEF patients and 72 age/sex-matched non-HF subjects.
  • Validation Cohort (BECAME-HF2/Canadian): 74 HFpEF patients and 103 non-HF subjects.
  • Both cohorts underwent deep clinical phenotyping, including biomarkers for congestion, fibrosis, liver/kidney function, and metabolic syndrome.
• Lipidomics Workflow:
  • Platform: Untargeted semi-quantitative Liquid Chromatography-Mass Spectrometry (LC-MS).
  • Data Processing: 1,666 lipid features were initially detected. After quality control and annotation, 235 unique lipids spanning 19 subclasses were retained for the Belgian cohort.
  • Differential Analysis: Linear regression adjusted for age and sex identified lipids significantly associated with HFpEF status.
• Statistical & Machine Learning Approaches:
  • Unsupervised Learning: Principal Component Analysis (PCA) and Hierarchical Clustering (Ward's method) were used to identify distinct lipidomic subgroups (phenogroups) within the HFpEF population.
  • Cross-Cohort Validation: A Random Forest classifier trained on the Belgian cohort was applied to the Canadian cohort using a shared subset of 158 lipids to test reproducibility.
  • Feature Selection: A resampling-based LASSO (Least Absolute Shrinkage and Selection Operator) strategy was used to derive a minimal lipid signature (10 lipids) capable of discriminating the high-risk group.
  • Survival Analysis: Kaplan-Meier curves and multivariable Cox regression models were used to assess the prognostic value of the identified phenogroups.

3. Key Contributions

• First Untargeted Lipidomic Phenomapping: This is the first study to use comprehensive untargeted plasma lipidomics to define metabolically driven HFpEF phenogroups, moving beyond targeted metabolite panels.
• Identification of a High-Risk "Cardio-Hepatic" Phenotype: The study uncovers a specific, previously unrecognized HFpEF subgroup characterized by severe multisystem dysfunction (heart and liver) and poor survival, distinct from the classic "obese-metabolic" phenotype.
• Derivation of a Minimal Biomarker Signature: The authors identified a parsimonious set of 10 lipids that can effectively stratify patients into high-risk vs. lower-risk categories, offering a potential tool for clinical translation.
• Mechanistic Insights: The study links specific lipid alterations to pathophysiological mechanisms, including mitochondrial dysfunction, peroxisomal failure, oxidative stress, and ferroptosis.

4. Key Results

A. Lipidomic Heterogeneity and Phenogroup Discovery

In the Belgian cohort, unsupervised clustering of the 235 annotated lipids revealed three distinct HFpEF phenogroups (B1, B2, B3) with divergent clinical and biochemical profiles:

1. Cluster B1 (High-Risk / Cardio-Hepatic):

   • Lipid Profile: Markedly elevated long-chain acylcarnitines (LC-CAR), ether phosphatidylcholines (ether PC), and oxidized sphingomyelins (OxSM).
   • Clinical Features: Severe atrial dysfunction (high Right Atrial Volume Index), elevated cardiac injury/fibrosis markers (Troponin T, FGF-23, sST2), and significant liver dysfunction/fibrosis (high FIB-4, NAFLD score, hypoalbuminemia).
   • Outcomes: Significantly worse survival. Kaplan-Meier analysis showed B1 had the lowest event-free survival (p=0.0037) and overall survival (p=0.00084). In multivariable Cox models, B1 remained an independent predictor of mortality/HF hospitalization (HR 1.98).
   • Note: This group did not exhibit the classic "obese-metabolic" features (BMI was lower than B3), suggesting a frailty or sarcopenia-associated metabolic state.

2. Cluster B2 (Aging / Renal-Preserved):

   • Lipid Profile: Intermediate patterns; highest levels of Sphingomyelins with very-long-chain fatty acids (VLCFA) and lysophosphatidylcholines (LPC).
   • Clinical Features: Higher prevalence of atrial fibrillation but preserved renal function (high eGFR) and higher HDL-cholesterol.
   • Outcomes: Favorable survival outcomes, similar to non-HF subjects.

3. Cluster B3 (Obese-Metabolic):

   • Lipid Profile: Highest levels of Triglycerides (TGs); lowest levels of ether PCs.
   • Clinical Features: Hallmarks of Metabolic Syndrome: highest BMI, diabetes prevalence, sleep apnea, and insulin resistance (high TyG index).
   • Outcomes: Intermediate survival, better than B1 but distinct from B2.

B. Cross-Cohort Validation

• When the Canadian cohort was analyzed using the shared 158 lipids, the unsupervised clustering identified analogous lower-risk groups (C1, C2, C3) that aligned with Belgian clusters B2 and B3.
• Crucial Finding: The high-risk B1 phenotype was underrepresented in the Canadian cohort (only ~8% predicted as B1-like). This highlights that while the biological mechanism exists, its prevalence may vary by population, and targeted detection strategies are needed to find these patients.

C. The Minimal Lipid Signature

A 10-lipid signature was derived to discriminate the high-risk B1 group. It includes:

• Increased: Long-chain acylcarnitines (CAR 18:1, 26:1), oxidized sphingomyelin (SM(d18:0/16:0(O))), and ether PCs.
• Decreased: Polyunsaturated phosphatidylcholines (PCs), ether phosphatidylethanolamines (PEs), triglycerides, and specific VLCFA-containing sphingomyelins.
• Correlation: This signature strongly correlates with markers of cardiac necrosis, liver fibrosis, and congestion, confirming its link to the multisystem failure seen in B1.

5. Significance and Implications

• Pathophysiological Insight: The results suggest that the high-risk HFpEF phenotype (B1) is driven by mitochondrial dysfunction (accumulation of acylcarnitines), peroxisomal dysfunction (altered ether lipids), and oxidative stress (oxidized lipids). The specific lipid pattern suggests a failure in lipid droplet remodeling and an inability to buffer oxidative stress, potentially leading to ferroptosis (iron-dependent cell death).
• Clinical Stratification: The study demonstrates that standard clinical variables (age, BMI, NT-proBNP) fail to distinguish the high-risk B1 group. Lipidomics provides a necessary layer of resolution to identify patients with "cardio-hepatic" failure who are at imminent risk of death.
• Therapeutic Targets: The identified lipids point to specific metabolic pathways (e.g., peroxisomal metabolism, ferroptosis regulation) that could be targeted by future therapies.
• Future Directions: The underrepresentation of the B1 phenotype in the Canadian cohort suggests the need for larger, diverse cohorts and the development of the minimal lipid signature as a screening tool to identify these high-risk patients for targeted clinical trials.

In summary, this paper establishes that HFpEF is not a monolithic disease but comprises distinct metabolic subtypes. It identifies a specific, high-risk "cardio-hepatic" lipidomic signature that predicts poor survival and offers a mechanistic framework for understanding the metabolic drivers of HFpEF heterogeneity.