Proteogenomic analysis of 5,411 plasma proteins in sickle cell disease patients

This study presents a proteogenomic analysis of 5,411 plasma proteins in 343 sickle cell disease patients, identifying 560 protein quantitative trait loci (including 58 novel ones) and using Mendelian randomization to prioritize five proteins that may increase fetal hemoglobin production as potential therapeutic targets.

The Gist

Imagine the human body as a bustling, high-tech city. In this city, Sickle Cell Disease (SCD) is like a major traffic jam caused by a specific error in the city's blueprint (the DNA). This error forces the red blood cells to become sticky and crescent-shaped, clogging the streets, causing pain, and damaging the city's infrastructure.

For a long time, doctors have had very few tools to fix this traffic jam. They can clear the roads temporarily (blood transfusions) or slow down the construction crews causing the mess (Hydroxyurea), but they can't permanently fix the blueprint without expensive, complex surgery (gene therapy).

This paper is like a massive detective investigation into the city's "supply chain" (the proteome) to find new, cheaper, and easier ways to fix the traffic.

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

1. The Great Protein Census

The researchers went into the "city" of 343 SCD patients and took a snapshot of 5,411 different proteins floating in their blood. Think of proteins as the city's workers, messengers, and construction crews. Some build things, some clean up, and some send warnings.

They wanted to see: Which genetic "blueprint errors" cause which "worker" to be too busy, too lazy, or missing entirely?

2. Finding the "Switches" (pQTLs)

They discovered 560 genetic switches (called pQTLs) that control how much of these proteins are made.

• The Good News: Most of these switches work the same way in SCD patients as they do in healthy people. It's like finding that the light switch for the streetlights works the same in a flooded city as it does in a dry one.
• The New Discovery: They found 58 brand-new switches that no one knew about before. These are like discovering hidden levers in the city hall that nobody knew existed.

3. The "Haptoglobin" Glitch

One of the most interesting findings was about a protein called Haptoglobin.

• In a healthy city: Haptoglobin is a "cleanup crew" that picks up broken red blood cells.
• In the SCD city: Because the cells are breaking constantly, the cleanup crew is overwhelmed and runs out.
• The Twist: The researchers found that a genetic switch that usually controls Haptoglobin in healthy people doesn't work in SCD patients. The "traffic jam" (hemolysis) is so severe that it overrides the genetic switch.
• The Lesson: You can't just copy-paste medical advice from healthy people to SCD patients. Sometimes, the disease changes the rules of the game entirely.

4. The "Fetal Hemoglobin" Magic Potion

The ultimate goal of this research is to find a way to boost Fetal Hemoglobin (HbF).

• The Analogy: Imagine the city has two types of fuel. The "Adult Fuel" causes the traffic jams (sickling). The "Fetal Fuel" is a super-clean, smooth-burning fuel that prevents jams.
• The Problem: As we grow up, our bodies stop making the "Fetal Fuel." Hydroxyurea is a drug that tries to trick the body into making more of it, but it doesn't work for everyone.
• The Breakthrough: The researchers used a method called Mendelian Randomization (think of it as a "genetic time machine") to see which proteins act as the "gas pedal" for making Fetal Fuel.

They narrowed down thousands of possibilities to five top candidates that might be the keys to unlocking more Fetal Fuel:

1. ENPP5: An enzyme involved in energy metabolism. The researchers suspect it might be the most promising "gas pedal."
2. PTX3: A marker of inflammation.
3. ZP3, LBP, and NAAA: Other proteins that might help turn on the Fetal Fuel switch.

Why This Matters

Think of this paper as a map to new treasure.

• Before, we only had a few keys (drugs) to open the door to better health for SCD patients.
• Now, the researchers have found five new potential keys (the proteins ENPP5, LBP, NAAA, PTX3, and ZP3).
• Because proteins are easier to target with drugs than DNA, these findings could lead to new, affordable pills that help patients make more of their own "Fetal Fuel," reducing pain and saving lives without needing expensive gene therapy.

In short: The scientists mapped the genetic controls of the blood in SCD patients, found some new switches, realized some old rules don't apply in this specific disease, and identified five new "magic buttons" that could be pressed to make the blood healthy again.

Technical Summary

1. Problem Statement

Sickle Cell Disease (SCD) is a severe monogenic blood disorder affecting over 7 million people globally, characterized by extreme clinical heterogeneity and limited therapeutic options. While curative interventions (transplants, gene therapy) exist, they are inaccessible to most. Current pharmacological treatments, such as Hydroxyurea (which increases Fetal Hemoglobin, HbF), do not work equally for all patients. There is an urgent need to identify new, affordable drug targets.

The authors posit that proteogenomics—integrating human genetics with high-throughput proteomics—can identify causal drug targets. While protein Quantitative Trait Loci (pQTL) have been mapped in healthy populations, it remains unclear if these genetic associations hold true in the complex, hemolytic environment of SCD patients, where chronic inflammation and cell-free hemoglobin might alter protein levels or genetic effects.

2. Methodology

The study utilized a multi-step proteogenomic approach involving the GEN-MOD cohort (343 SCD patients with HbSS genotype, African ancestry) and comparative datasets.

• Proteomics: Plasma levels of 5,411 proteins were measured using the Olink Explore 5K platform (antibody-based). Samples were collected at a steady state (no Hydroxyurea or transfusion in the last 90 days).
• Genomics: Participants underwent genome-wide genotyping and imputation (TOPMed haplotypes).
• pQTL Mapping:
  • Associations between 4.6M imputed genetic variants and normalized protein levels were tested using an additive linear model (correcting for age, sex, and 20 principal components).
  • Significance Thresholds: $P_{cis} < 5 \times 10^{-8}$ and $P_{trans} < 7.7 \times 10^{-11}$.
  • Conditional Analysis: Stepwise conditional regression was performed to identify independent sentinel variants for each protein.
• Comparative Analysis:
  • Replication: Results were compared against the Jackson Heart Study (JHS) (1,054 Black adults, non-SCD) and the UK Biobank.
  • Heterogeneity Testing: Wald tests were used to detect differences in effect sizes between SCD (GEN-MOD) and non-SCD (JHS) cohorts.
  • Phenome-Wide Association Study (PheWAS): GEN-MOD pQTL variants were queried against 159 laboratory traits in the Million Veteran Program (MVP) (121,177 African Americans).
• Mendelian Randomization (MR):
  • Two-sample MR was performed to infer causality between plasma proteins and Fetal Hemoglobin (HbF) levels.
  • Instrumental variables were selected from independent cis-pQTLs ($P < 0.001$, distance < 500kb from TSS).
  • Methods included Inverse-Variance Weighted (IVW), Simple Median, and Cochran's Q test for heterogeneity.

3. Key Contributions

• First SCD-Specific Proteogenomic Map: Generated the largest pQTL map specifically for SCD patients, identifying 560 independent pQTLs (529 cis, 31 trans) affecting 467 unique proteins.
• Discovery of Novel Targets: Identified 58 novel pQTLs (10% of total), largely driven by the inclusion of ~2,000 proteins in the latest Olink HT platform not previously assayed in large cohorts.
• Validation of Generalizability: Demonstrated that the vast majority of pQTL effect sizes are consistent between SCD and non-SCD African-ancestry populations, validating the use of non-SCD pQTLs as instruments for SCD research in most cases.
• Identification of Context-Specific Interactions: Uncovered specific instances where the SCD genotype alters genetic effects on protein levels (e.g., APOL1, Haptoglobin), highlighting the necessity of disease-specific validation for MR instruments.
• Prioritization of HbF Regulators: Used MR to nominate five plasma proteins that causally increase HbF production, offering new therapeutic avenues.

4. Key Results

A. pQTL Discovery and Replication

• Replication: Of 930 overlapping protein-variant pairs tested between GEN-MOD and JHS, 484 were replicated with consistent effect directions, confirming high data quality.
• Novelty: 58 novel pQTLs were identified. 51 were novel because the proteins were not measured in previous studies; 7 were novel associations for proteins already measured (e.g., ANXA11, CEACAM20), suggesting SCD-specific genetic regulation or unique LD patterns.
• Allele Frequency: Novel pQTL variants were more common in African-ancestry populations (gnomAD AFR) than in European populations, emphasizing the importance of diverse cohorts.

B. Heterogeneity in Effect Sizes

While most pQTLs were concordant between SCD and non-SCD cohorts (Pearson's $r = 0.90$), seven variant-protein pairs showed significant heterogeneity:

• APOL1: The effect size in SCD was nearly double that in JHS. The authors suggest this may reflect differential epitope-antibody binding due to protein-altering variants (G1/G2 alleles) rather than true biological changes in protein concentration.
• Haptoglobin (HP) Locus: Variants upstream of HP (e.g., SERPIND1, GALNT2) showed heterogeneous effects. The authors hypothesize that in SCD, constant hemolysis depletes Haptoglobin, masking the genetic regulation seen in healthy individuals.
• Implication: This highlights that while general pQTLs are transferable, specific loci involved in hemolysis or protein structure may require disease-specific instruments for Mendelian Randomization.

C. Clinical Associations

• SCD Complications: No significant associations were found between the 560 pQTLs and major SCD complications (stroke, priapism, leg ulcers, etc.) in the combined GEN-MOD/CSSCD cohort.
• PheWAS (MVP): 58 pQTL variants were associated with laboratory traits in non-SCD African Americans. Notable findings included associations between Transferrin and red blood cell traits, and ANGPTL4 with lipid levels. A novel pQTL for Charcot-Leyden Crystal (CLC) protein correlated with eosinophil counts, reinforcing CLC's role in eosinophil biology.

D. Mendelian Randomization for HbF

The study prioritized plasma proteins that could modulate HbF levels:

• Five proteins associated with increased HbF:
  1. ENPP5: An ecto-nucleotide pyrophosphatase. The authors hypothesize it may regulate HbF by modulating the nucleotide pool (dNTPs), a mechanism relevant to Hydroxyurea's mode of action.
  2. PTX3: A marker of inflammation/ischemia.
  3. LBP: Lipopolysaccharide-binding protein (innate immunity).
  4. NAAA: N-acylethanolamine acid amidase.
  5. ZP3: Zona pellucida glycoprotein 3.
• Negative Association: Genetically controlled levels of HBZ (embryonic hemoglobin zeta subunit) were negatively associated with HbF, suggesting a potential regulatory switch mechanism.

5. Significance and Conclusion

This study provides a critical resource for precision medicine in SCD. By confirming that most pQTLs are robust across disease states, the authors validate the use of large non-SCD biobanks for drug target discovery in SCD. However, they also caution that specific loci (like APOL1 and Haptoglobin) require careful validation due to disease-specific interactions.

The identification of ENPP5 and other proteins as potential HbF regulators offers a new class of therapeutic targets. Unlike gene editing, which is complex and expensive, plasma proteins are often "druggable" via small molecules or biologics. These findings lay the groundwork for functional validation and the development of novel, affordable therapies to increase HbF and reduce SCD severity.