Introduction: Hydroxyurea is a potent therapeutic agent for sickle cell anemia (SCA), and treatment at maximum tolerated dose (MTD) is becoming the standard of care. Hydroxyurea exerts its disease-modifying effects primarily through induction of fetal hemoglobin (HbF), although the cellular and molecular mechanisms by which hydroxyurea increases HbF expression remain unclear. Children with SCA treated with hydroxyurea at MTD have substantial phenotypic variation, however, as some have higher HbF responses than others. We hypothesized that unknown quantitative trait loci modulate the pharmacological induction of HbF, so we performed a large genome wide association study (GWAS) of hydroxyurea-associated HbF responses for children with SCA treated prospectively with dose escalation to MTD.

Methods: We analyzed genomic DNA from 831 children with SCA enrolled in pediatric research trials from the US (HUSTLE, SWiTCH, TWiTCH), the Caribbean (EXTEND, SACRED) and sub-Saharan Africa (REACH, NOHARM); all of these trials reported robust treatment responses with average HbF >20%. Study participants received hydroxyurea with dose escalation to MTD based on mild myelosuppression. Whole blood DNA was genotyped using the H3Africa SNP array (Illumina) with whole exome sequencing (WES) using NimbleGen VCRome 2.1 capture reagents and the Illumina HiSeq2500 platform. A transformed z-score for each study cohort gave a standardized measure of HbF induction relative to their steady-state level and their treatment HbF level at MTD. These standardized z-score HbF values were then used as a continuous variable for association testing using single-locus mixed model (EMMAX) adjusted for population stratification, using age, hydroxyurea dose at MTD, and sex as co-variates. We first performed an initial GWAS discovery using hydroxyurea response data from four distinct African populations (n=377). Single nucleotide variants (SNVs) with nominal significance (p<0.001) in the discovery step were then selected for replication using an additional African cohort (n=168). Variants that were significant in both the discovery and replication cohorts were then verified using a cohort of US (n=200) and Caribbean (n=86) children with SCA, identifying genomic loci with consistent associations for HbF induction across all cohorts.

Results: In the discovery GWAS step, no variant passed genome wide significance (p<10-8) for the MTD HbF phenotype, including no significant associations with known genetic modifiers of endogenous HbF (BCL11A, HBS1L-MYB, HBG2). A total of 2057 low frequency and common SNVs had at least nominal association (p<0.001) with the hydroxyurea treatment responses, of which 44 were also significant (p<0.05) and with the same direction of association with HbF induction in the replication cohort. In the final verification step, these 44 significant variants were then tested in additional independent SCA cohorts with at least three demonstrating a strong effect (Table 1). The rs10978155 variant in the PTPRD gene and the rs55695413 variant in the RPH3AL gene were both consistently associated (p<0.05) with lower HbF treatment responses. Another variant (rs75442556) near the ELL2 gene approached statistical significance (p=0.08) in the verification cohort and was also associated with lower HbF expression. The allele frequencies for these PTPRD, RPH3AL, and ELL2 variants were 0.32, 0.017, and 0.25, respectively, and did not affect baseline HbF levels. Children with these PTPRD, RPH3AL, and ELL2 genetic variants still had substantial HbF induction, but achieved lower hydroxyurea MTD HbF levels on average by 2.9%, 9.8%, and 2.7%, respectively.

Conclusions: This large GWAS using global cohorts of children with SCA and robust prospective HbF phenotype data has identified genetic predictors of HbF hydroxyurea treatment responses. Three novel genetic loci, PTPRD, RPH3AL, and ELL2 have SNVs associated with lower HbF responses. PTPRD is a protein tyrosine phosphatase receptor involved in cellular processes such as cell growth and differentiation, while RPH3AL, a rabphilin 3A like protein, is known to be involved in calcium-ion-dependent exocytosis. ELL2 is an elongation factor for RNA polymerase II and could modify RNA processing under the cytostatic effects of hydroxyurea. These genes and variants will be investigated to determine how they impact individual HbF responses to hydroxyurea treatment.


Aygun:National Heart, Lung, and Blood Institute: Research Funding; bluebird bio: Membership on an entity's Board of Directors or advisory committees, Research Funding; National Institute of Nursing Research: Research Funding; Patient-Centered Outsomes Research Institute: Research Funding.

Author notes


Asterisk with author names denotes non-ASH members.