Fetal hemoglobin (HbF) is the major genetic modulator of sickle cell anemia. Candidate gene-based and genome-wide association studies (GWAS) have provided strong evidence that single nucleotide polymorphisms (SNPs) linked to genes on chromosome 6q (HBS1L-MYB) and 2p (BCL11A), along with elements in cis to HBG help determine HbF concentration in untreated patients with sickle cell anemia and β thalassemia, and in normal individuals. The HbF response to hydroxyurea (HU) varies considerably among treated patients, even when compliance with treatment is good and patients are treated under controlled conditions. This suggests that genetic factors might affect the response to treatment with this agent. In the Multicenter Study of Hydroxyurea, 299 patients were randomized to receive either HU titrated to maximum tolerated doses, or a placebo, and HbF levels were measured before and at the completion of the randomized phase of the study. In 123 HU-treated patients, we completed GWAS using Illumina 370K chips that include approximately 350,000 haplotype tagging SNPs, and studied the association of SNPs with the change in HbF from baseline levels to levels measured at the end of the active treatment portion of the study. We conducted a GWAS using the analytical program PLINK, of approximately 273K SNPs with minor allele frequency >0.05, using linear regression and an additive model of inheritance. We selected for further investigation those SNPs with association that reached 0.05 significance, after we adjusted for sex. Because of the limited sample size that results in relatively large p-values, no single SNP reached so-called genome-wide significance after correcting for multiple comparisons using a Bonferroni correction (p-value <10-7) or 5% false discovery rate. Two SNPs had an association with p value <10–6 and 27 SNPs reached at least 10–5 significance. Noticeably, the SNP rs6899351 in FABP7 in 6q22.31 was associated with the largest increment in HbF after treatment with HU (6.9% change per copy of allele G, p-value 4 ×10–5). We also identified 2 SNPs in PDE7B (6q23.3) that were significantly associated with positive changes of HbF and 3 SNPs in MAP7 (6q23.3) that were significantly associated with a reduction of HbF after treatment. Using candidate gene association studies, we had previously shown that PDE7B and MAP7 were significantly associated with differential expression of HbF in sickle cell anemia. These new GWAS results suggest a regulatory role for these genes, or this region of chromosome 6q, in the HbF response to HU in sickle cell anemia. Analysis of the distribution of significant SNPs per chromosome also showed that chromosome 20 had a larger number of significant SNPs than expected at random, especially in CST9, one of a family of protease inhibitors. CST9 is tagged by 3 SNPs in the 370K array (rs2983639, rs2983640, rs10485646), 2 of which were associated with significant positive changes in HbF after treatment with HU and one with significant negative changes of HbF. Specifically, the average increase in HbF was 1.5% for each copy of allele G for SNP rs2983639 (p = 0.025), and 1.6% for each copy of allele A for SNP rs2983640 (p = 0.047), while the level of HbF decreased by approximately 1.5% for each copy of allele A for SNP rs10485646 (p = 0.035). The SNP rs2983640 is an exon variant that produces the amino acid change F-L. Although these SNPs do not individually reach genome-wide significance, cumulatively they provide strong evidence of association, as the probability that they are all simultaneously associated by chance is 10-4. Furthermore, we identified significant variants in other genes that belong to the same family of type 2 cysteine protease inhibitors, specifically 2 SNPs in CTS3 and 1 SNP in CTS5. Although the small sample size and the large number of SNPs tested suggest caution until these results are replicated in independent patient treatment groups, these preliminary findings suggest that type 2 cystatin genes and pseudogenes are associated with the HbF response to HU. If confirmed, it might be possible to use results like these to build a prognostic model of the HbF response to HU in sickle cell anemia.

Disclosures: No relevant conflicts of interest to declare.

Author notes

Corresponding author