Transcriptomic Profiling Supports the Phenotypic Classification of Stable-State Sickle Cell Disease Patients

Background and Aim

Sickle cell disease (SCD) has significant pathophysiological complexity and phenotypic heterogeneity that need to be considered for therapeutic decision-making and the development of new treatments. Here we derived transcriptomic profiles from an unbiased whole-blood gene expression analysis to investigate their associations with SCD clinical phenotypes.

Methods

36 patients aged ≥18 years diagnosed with SCD (HbSS and S/β-thalassemia) were included and compared to 36 age, sex, and ethnicity-matched healthy controls in a prospective cross-sectional observational study in two Swiss university hospitals between June 2020 and February 2023. Current and past (≤2 years) hematological parameters and clinical information were used to characterize SCD clinical phenotypes. Blood of SCD patients and healthy controls was collected and mRNA was sequenced on an Illumina system. Fragment counts were normalized as transcripts per million (TPM), and transcriptomic profiles were defined by dimensionality reduction using a Principal Coordinates Analysis (PCoA). Gene set enrichment analysis (GSEA) was performed against the GSEA Hallmark gene sets, based on the correlation of gene expression to PCoA dimensions. For the correlation of hematological and clinical variables with PCoA dimensions, Pearson correlation statistics are reported for continuous variables and one-way ANOVA statistics for factorial variables.

Results

Whole-blood transcriptomic profiles of SCD patients diverged from healthy controls in an unsupervised PCoA analysis, resulting in a clear separation of these two groups. Focusing on SCD profiles, we found substantial variability with increasing dissimilarity from controls captured by the first dimension of the PCoA solution (PCoA1). A strong correlation between PCoA1 and heme metabolism-related gene expression (GSEA adjusted p-value <0.001) indicated higher hemolytic activity with increasing distance from controls on PCoA1. In line with the expression of heme metabolism-related genes, blood values confirmed increased hemolysis in patients with more distant transcriptomic profiles. We found strong positive correlations between PCoA1 and increased LDH (R=0.52, adj.p=0.008), total bilirubin (R=0.47, adj.p=0.013), and reticulocyte counts (R=0.75, adj.p<0.001). Correspondingly, PCoA1 was negatively correlated with hemoglobin concentration (R=-0.42, adj.p=0.025) and erythrocyte counts (R=-0.61, adj.p<0.001). PCoA1 also correlated with blood values of the same patients from the 2 previous years (LDH R=0.46, adj.p=0.038; bilirubin R=0.50, adj.p=0.038; reticulocytes R=0.63, adj.p=0.007; erythrocytes R=-0.48, adj.p=0.035), providing longitudinal evidence for the association between transcriptomic profiles and the hemolytic phenotype. We investigated the correlation of transcriptomic profiles with other SCD-relevant demographic and clinical parameters. None of the examined characteristics (e.g., SCD genotypes, current hydroxyurea use, previous blood transfusions, hospitalizations, or complications such as vaso-occlusive crisis, acute chest syndrome, or pulmonary hypertension) were significantly associated with transcriptomic profiles.

Conclusion

Using an unbiased gene expression approach, our results confirm the fundamental role of hemolysis in the pathophysiology and phenotypic variability of SCD. While transcriptomic profiles strongly reflect hemolytic intensities in SCD patients, we did not find evidence for an association with specific complications related to the hemolytic phenotype, probably due to the limited size of the study population. However, there is great potential in further analyzing unbiased SCD gene expression profiles to better understand the transcriptomic basis of phenotypic heterogeneity in SCD.