Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
The Pharmacogenomics Journal (2007) 7, 386–394 & 2007 Nature Publishing Group All rights reserved 1470-269X/07 $30.00 www.nature.com/tpj ORIGINAL ARTICLE Fetal hemoglobin in sickle cell anemia: genetic determinants of response to hydroxyurea Q Ma1, DF Wyszynski1, JJ Farrell1, A Kutlar2, LA Farrer1, CT Baldwin3,1 and MH Steinberg1 1 Department of Medicine, Boston University School of Medicine, Boston, MA, USA; 2 Department of Medicine, Medical College of Georgia, Augusta, GA, USA and 3Center for Human Genetics, Boston University School of Medicine, Boston, MA, USA Correspondence: Dr MH Steinberg, Center of Excellence in Sickle Cell Disease, E248, Boston Medical Center, 88 E. Newton Street, Boston, MA 02118, USA. E-mail: [email protected] The increase in fetal hemoglobin (HbF) in response to hydroxyurea (HU) varies among patients with sickle cell anemia. Twenty-nine candidate genes within loci previously reported to be linked to HbF level (6q22.3–q23.2, 8q11–q12 and Xp22.2–p22.3), involved in metabolism of HU and related to erythroid progenitor proliferation were studied in 137 sickle cell anemia patients treated with HU. Three-hundred and twenty tagging single nucleotide polymorphisms (SNPs) for genotyping were selected based on HapMap data. Multiple linear regression and the nonlinear regression Random Forest method were used to investigate the association between SNPs and the change in HbF level after 2 years of treatment with HU. Both methods revealed that SNPs in genes within the 6q22.3–23.2 and 8q11–q12 linkage peaks, and also the ARG2, FLT1, HAO2 and NOS1 genes were associated with the HbF response to HU. Polymorphisms in genes regulating HbF expression, HU metabolism and erythroid progenitor proliferation might modulate the patient response to HU. The Pharmacogenomics Journal (2007) 7, 386–394; doi:10.1038/sj.tpj.6500433; published online 13 February 2007 Keywords: SNPs; association analysis; sickle cell; fetal hemoglobin; hydroxyurea Introduction Received 6 June 2006; revised 19 September 2006; accepted 6 November 2006; published online 13 February 2007 Fetal hemoglobin (HbF) inhibits the polymerization of sickle hemoglobin (HbS).1 As many of the complications of sickle cell anemia (homozygosity for HBB, glu6val), like osteonecrosis, acute chest syndrome and painful episode, are associated with the level of HbF, and, HbF is inversely associated with mortality, investigators have assiduously sought pharmacological means of increasing HbF production.2–6 Hydroxyurea (HU), a ribonucleotide reductase inhibitor, is one drug that increases HbF concentration in patients with sickle cell anemia7–10 and it is the sole FDA-approved agent for treating sickle cell anemia. Most, but not all patients respond to HU treatment with an increase in HbF, but as with the baseline HbF concentration, which varies widely among patients, the magnitude of the HbF response to HU is also variable.10–13 The regulation of HbF level might be a complex genetic trait governed by genetic elements linked to the b-globin gene-like cluster and quantitative trait loci (QTL) present on chromosomes 6, 8 and on the X-chromosome; other regulatory loci are also likely to exist and epigenetic and cellular factors could also have regulatory roles.14–27 It is possible that these and other regulatory elements also modulate the HbF response to HU. Accordingly, we hypothesized that single nucleotide polymorphisms (SNPs) in candidate genes or QTL with putative roles in the regulation of HbF SNPs, hydroxyurea and HbF in sickle cell anemia Q Ma et al 387 production might modulate the HbF response to treatment with HU. We therefore studied the association of SNPs in these loci with response to HU in patients who participated in the Multicenter Study of Hydroxyurea (MSH), a trial designed to evaluate the efficacy of this drug in sickle cell anemia. Results The distribution of HbF for 137 patients enrolled in this study is shown as Figure 1. The change of HbF after HU treatment did not follow a normal distribution and resembled a bimodal distribution with a large portion of people with minor or no change (mean ¼ 0) and a small portion of people with extreme change (mean 440). This distribution suggested that categorizing these data, like dividing subjects into quartiles by the HbF change, and comparing patients in lowest quartile group vs those patients in the highest quartile of change, might be an alternative approach for analyzing these data. However, considering the relatively small sample size in our study, this approach provides very limited power for detecting genetic associations. Three-hundred and twenty tagging SNPs in 29 candidate genes (Table 1) were examined in 137 sickle cell anemia patients treated with HU. We considered an SNP to have a significant association with response to HU treatment when the P-value was p0.01, unless there was more than one Figure 1 Distributions of HbF change (a) in percentage (%) and (b) in grams (g/dl). SNP in a gene showing an association when the P-value for significance was set at 0.05. Tables 2 and 3 present the statistically significant results of the quantitative trait analysis for single SNP association with the change of HbF level after a 2-year treatment, expressed as percentage of total hemoglobin and as the absolute HbF level, expressed as g/dl, respectively. An analysis was also performed expressing the increment in HbF after HU treatment as F-cells. Seventeen SNPs were significantly associated with the change in percent HbF (Table 2). They included two in MAP3K5, five in TOX, two in NOS1, three in FLT1, two in ARG2 and two in NOS2A. The most significant association was observed with SNP rs2182008 (P ¼ 0.003) in FLT1 (Fms-related tyrosine kinase 1), a vascular endothelial growth factor, which is involved in cell proliferation and differentiation. Twenty SNPs were significantly associated with the change of absolute HbF (Table 3); a similar pattern of association was observed and the most significant association was in SNP rs10483801 (P ¼ 0.0013) in ARG2 (arginase type II, involved in drug metabolism of HU). Using F-cells as the outcome measure gave similar results (data not shown). For candidate genes with significant association in multiple SNPs, haplotype associations were explored using Haplo.stats (version 1.2.1) as given in the R library (available at http://cran.us.r-project.org).28 However, as these SNPs studied here are tagging SNPs and most of them are not in linkage disequilibrium (LD) with each other, we did not find improved association by haplotype analysis in any genes (data not shown). The results of joint analysis of all the SNPs and covariates (age, sex and the a- and b-globin gene cluster haplotypes) using Random Forest analysis are shown in Figures 2 and 3. The relative importance of one independent variable (a SNP) is measured by %IncMSE (see Materials and methods for details), and the larger the value of %IncMSE, the higher importance that variable has for correct prediction of HbF response to HU. This analysis revealed that the most important variables for predicting the change of HbF level matched most of the SNPs identified by SNP association analysis. Interestingly, SNPs within ASS (argininosuccinate synthetase) and ARG1 (arginase, liver) were observed to have strong effect on the change of HbF level, which was not detected by single SNP association analysis. This suggests that these two genes might be involved in interaction with other genes to regulate the response to HU treatment. SNP rs2182008 in FLT1 showed a strong effect on response to HU treatment. This SNP was significantly associated with the change in HbF under a dominant model (P-value ¼ 0.003 for HbF in percentage and 0.002 for HbF in g/dl) and it was also a highly ranked predictor for response to HU from the Random Forest analysis (second for HbF in percentage and third for HbF in g/dl). The A allele of this SNP was associated with increased HbF level after HU treatment; there is no difference between AA and AG genotypes and the increase in HbF in subjects with these genotypes was on average 5.9 times higher than that in subjects with the GG genotype (Figure 4). The Pharmacogenomics Journal SNPs, hydroxyurea and HbF in sickle cell anemia Q Ma et al 388 Table 1 Candidate genes selected Gene Chromosomes Function Cytochrome P450, family 4, subfamily A, polypeptide 11 (CYP4A11) Hydroxyacid oxidase 2 (long chain) (HAO2) Kinase insert domain receptor (a type III receptor tyrosine kinase) (KDR) Arginase, liver (ARG1) Phosphodiesterase 7B (PDE7B) Microtubule-associated protein 7 (MAP7) Mitogen-activated protein kinase kinase kinase 5 (MAP3K5) Peroxisomal biogenesis factor 7 (PEX7) NADPH oxidase 3 (NOX3) Met proto-oncogene (hepatocyte growth factor receptor) (MET) Nitric oxide synthase 3 (endothelial cell) (NOS3) Glutathione reductase (GSR) CCAAT/enhancer binding protein (C/EBP), delta (CEBPD) 1 1 4 6 6 6 6 6 6 7 7 8 8 Transcription elongation factor A (SII), 1 (TCEA1) 8 SRY (sex determining region Y)-box 17 (SOX17) 8 Thymus high mobility group box protein TOX (TOX) 8 Argininosuccinate synthetase (ASS) Cytochrome P450, family 2, subfamily C, polypeptide 9 (CYP2C9) Beta-globin gene cluster Nitric oxide synthase 1 (neuronal) (NOS1) Fms-related tyrosine kinase 1 (vascular endothelial growth factor/vascular permeability factor receptor) (FLT1) Arginase, type II (ARG2) Aquaporin 9 (AQP9) NADPH oxidase, EF-hand calcium binding domain 5 (NOX5) Nitric oxide synthase 2A (inducible, hepatocytes) (NOS2A) EGF-like-domain, multiple 6 (EGFL6) Glycoprotein M6B (GPM6B) C-fos induced growth factor (vascular endothelial growth factor D) (FIGF) Pirin (iron-binding nuclear protein) (PIR) 9 10 11 12 13 14 15 15 17 X X X X Drug metabolism Drug metabolism Cell differentiation Drug metabolism Chr6 QTL Chr6 QTL, cell differentiation Chr6 QTL Chr6 QTL Drug metabolism Cell proliferation NO production Drug metabolism Regulation of DNA transcription Chr8 QTL, regulation of DNA transcription Chr8 QTL, regulation of DNA transcription Chr8 QTL, regulation of DNA transcription NO production Drug metabolism Chr11 QTL NO production Cell proliferation and differentiation NO production Drug metabolism Drug metabolism NO production Chr X QTL Chr X QTL Chr X QTL Chr X QTL Tagging SNPs 4 5 17 3 17 3 10 6 17 5 4 6 2 3 3 57 24 4 8 23 32 7 8 6 6 9 21 6 4 Abbreviations: NO, nitric oxide; QTL, quantitative trait loci; SNPs, single nucleotide polymorphisms. Discussion HU can increase HbF concentration in most individuals with sickle cell anemia but some patients who take the drug exactly as directed by experienced physicians either fail to respond or have an HbF response that might not be clinically significant. Even among patients with an increase in HbF, the magnitude of this increase varies.10–13 The cause of this variability is poorly understood.11,12,29,30 HbF concentration in blood is determined by interactions among chromosome remodeling activities, transcription factors, genes modulating erythropoiesis, genetic elements linked to the b-globin gene cluster, the kinetics of erythroid cell differentiation and differential red cell survival (for reviews see Bank31 and Stamatoyannopoulos32). This complex regulatory environment provides ample opportunity for genetic modulation of HbF production. Because of evidence suggesting that heterogeneity in genetic elements modulate HbF baseline concentration, we reasoned that The Pharmacogenomics Journal similar heterogeneity might account for the varied response in HbF among sickle cell anemia patients treated with HU. We found statistically significant associations of the HbF response to HU with multiple SNPs in several genes. TOX (thymus high-mobility group box protein) is located within the 8q11–q12 linkage peak that Garner et al.,16,33 found to interact with the 158 CT 5’ Gg-globin gene SNP and that might also effect the g- to b-globin gene switch. A member of the high-mobility group (HMG) box protein family, TOX contains a single HMG box motif and binds DNA in a sequence-specific manner. All HMG box proteins are able to induce a sharp bend in DNA. Multiple SNPs in PDE7B (phosphodiesterase 7B) within 6q22–q23 QTL were also associated with the HbF response to HU. SNPs in this gene were previously reported to be associated with baseline HbF level in patients with sickle cell anemia.21 Nitric oxide (NO) binds and activates sGC, which increases cGMP production. cGMP interacts with transcription factors increasing the expression of the g-globin SNPs, hydroxyurea and HbF in sickle cell anemia Q Ma et al 389 Table 2 SNPs associated with a significant change in HbF% Genetic model SNP rs10494225 rs9376230 rs9483947 rs826729 rs765587 rs9693712 rs172652 rs380620 rs816361 rs7977109 rs9319428 rs2182008 rs8002446 rs10483801 rs10483802 rs1137933 rs944725 Chromosomes Gene a 1 6 6 8 8 8 8 8 12 12 13 13 13 14 14 17 17 HAO2 MAP3K5 MAP3K5 TOX TOX TOX TOX TOX NOS1 NOS1 FLT1 FLT1 FLT1 ARG2 ARG2 NOS2A NOS2A Function Untranslated Intron Intron Intron Intron Intron Intron Intron Intron Intron Intron Intron Intron Intron Intron Synonymous Intron Codominant Dominant Recessive 0.015 NS NS NS NS 0.019 0.049 NS NS NS NS 0.012 0.033 0.026 0.015 NS NS NS NS NS 0.045 0.031 0.0098 NS 0.047 NS NS NS 0.003 NS NS NS NS 0.02 0.0039 0.036 0.034 NS NS NS NS NS 0.045 0.029 0.047 NS 0.011 0.0075 0.0038 0.031 NS Abbreviations: HbF, fetal hemoglobin; NS, nonsignificant; SNPs, single-nucleotide polymorphisms. a See Table 1 for full names. NS: P-value 40.05. Bold and italics: SNP with the most significant P-value. Table 3 SNPs associated with a significant change in HbF (g/dl) Genetic model SNP rs10494225 rs2327669 rs11154849 rs9376173 rs1480642 rs487278 rs2693430 rs765587 rs12155519 rs9693712 rs380620 rs7309163 rs7977109 rs3751395 rs9319428 rs2182008 rs2387634 rs8002446 rs10483801 rs10483802 Chromosomes Gene a Function 1 6 6 6 6 6 8 8 8 8 8 12 12 13 13 13 13 13 14 14 HAO2 PDE7B PDE7B PDE7B PDE7B PDE7B TOX TOX TOX TOX TOX NOS1 NOS1 FLT1 FLT1 FLT1 FLT1 FLT1 ARG2 ARG2 Untranslated Intron Intron Intron Intron Intron Intron Intron Intron Intron Intron Intron Intron Intron Intron Intron Intron Intron Intron Intron Codominant Dominant Recessive 0.0078 NS NS NS NS NS NS NS NS 0.048 0.038 NS NS NS NS 0.0085 NS 0.029 0.0052 0.014 NS NS 0.05 NS NS NS 0.049 0.044 0.037 0.02 0.016 NS NS NS NS 0.0021 NS NS NS NS 0.0018 0.041 NS 0.049 0.044 0.017 NS NS NS NS NS 0.038 0.023 0.039 0.044 NS 0.037 0.01 0.0013 0.0037 Abbreviations: HbF, fetal hemoglobin; NS, nonsignificant; SNPs, single nucleotide polymorphisms. a See Table 1 for full names. NS: P-value 40.05. Bold and italics: SNP with the most significant p-value. genes.34 ASS (argininosuccinate synthetase) is an enzyme that catalyzes the penultimate step of the arginine biosynthetic pathway. Arginine is the substrate for the NO synthases. SNPs in ASS gene were strong predictors for HbF response to HU. Significant associations were also found with SNPs in ARG2 and ARG1, genes coding for enzymes The Pharmacogenomics Journal SNPs, hydroxyurea and HbF in sickle cell anemia Q Ma et al 390 Figure 2 Joint analysis of multiple SNPs with change in HbF in percentage (%): only predictors that have relative importance value (%IncMSE) 43.0 are shown here; *: SNPs in x-QTL. that hydrolyze arginine to ornithine. SNPs in NOS1 (12q24.2-q24.31, neuronal NO synthase) and NOS2A, an NO synthase expressed in liver, were also associated with HbF response to HU. FLT1 (vascular endothelial growth factor (VEGF)/vascular permeability factor receptor) is a receptor for VEGF. SNP rs2182008 in this gene was found to be strongly associated with the HbF response to HU. This gene has tyrosine protein kinase activity that is important for the control of cell proliferation and differentiation.35 Most SNPs associated with HbF response to HU treatment were in either untranslated portions of the genes or in introns. Most likely, the associations we found are in LD with the actual functionally important elements. Although most associations have biologically plausible mechanisms by which they might influence the expression of the g-globin genes or the concentration of HbF, our studies are hypothesis generating rather than mechanistic, and we were unable to do functional studies because of the The Pharmacogenomics Journal nature of our study samples. Nevertheless, these SNPs were associated with HbF concentration, for example, the increase of HbF in subjects carrying A allele of SNP rs2182008 was almost six times higher than that in subjects with GG genotype. To identify genetic factors regulating the expression of HbF in sickle cell anemia at baseline and in the absence of treatment with HU, we studied a sample of 327 subjects and an independent group of 987 subjects. We replicated in part our previous work21 using this expanded sample of patients and denser SNP coverage, by finding multiple tagging SNPs in genes abutting the 6q22.3–q23.2 QTL. We also found a significant association of HbF with SNPs in TOX and within EGFL6, GPM6B and FIGF in the Xp22.2–p22.3 QTL (unpublished data). According to the common disease–common variant hypothesis, many of the genetic variants causing complex diseases or phenotypes are expected to have only a small effect on disease outcome.36 The power to obtain SNPs, hydroxyurea and HbF in sickle cell anemia Q Ma et al 391 Figure 3 Joint analysis of multiple SNPs with change in HbF in grams (g/dl): only predictors that have relative importance value (%IncMSE) 43.0 are shown here; *: SNPs in x-QTL. typical thresholds of P-value significance after applying multiple testing corrections is limited for such markers, because significance is a function of sample size, allele frequency and effect size. Considering our relatively small sample size, which nevertheless constitutes the largest number of patients with sickle cell anemia taking HU under controlled conditions, we validated our findings by using a different analytical method but did not apply multiple testing corrections for reporting our results. Also, as we selected haplotype tagging SNPs that tag different LD blocks and almost all our associations are with more than a single SNP in a gene, it is unlikely that the result with one SNP merely reflects the same association signal with other SNPs. Therefore, out of the 320 SNPs tested, it is less likely that multiple hits with SNPs in the same gene is a chance event. Further replication with fewer candidate genes and denser SNP coverage may be a more practical way to reduce false discovery and confirm our findings. Materials and methods Patients DNA samples and laboratory data from adult AfricanAmericans with sickle cell anemia who participated in the MSH were analyzed. This was a randomized, placebocontrolled, double-blind trial designed to test whether HU could reduce the number of vasoocclusive events in adults with moderate to severe sickle cell anemia.37 HU was given daily, in a single dose starting at 15 mg/kg, and increased by 5 mg/kg every 12 weeks up to a maximum-tolerated dose of 35 mg/kg, unless toxicity developed. When toxicity occurred, treatment was stopped until blood counts recovered. HU was then resumed at a dose 2.5 mg/kg lower than the toxic dose. The MSH enrolled 299 patients, of which 152 were randomly assigned to HU. The minimum length of follow-up evaluation for patients with HbF measured at the end of the study was 21 months (maximum, 38 months; mean, 28). DNA samples were available and HbF was The Pharmacogenomics Journal SNPs, hydroxyurea and HbF in sickle cell anemia Q Ma et al 392 study population have sickle cell disease.43–48 SNPs that had more than 25% missing genotypes or less than 5% minor allele frequency were not considered in the analysis. This resulted in 280 SNPs being tested for association. Figure 4 HbF change for SNP rs2182008 (a) in percentage (%) and (b) in grams (g/dl). median; D: mean. measured in 137 of the 152 patients randomized to receive HU. Although pill counts and measurements of serum HU levels were used to assess compliance with treatment, the rapid clearance of HU from the circulation prevented us from knowing with certainty if a patient took all of the medication prescribed. The studies reported here were approved by the IRB of Boston Medical Center. Laboratory studies HbF was measured by alkali denaturation at least twice pretreatment and twice post-treatment.10 The laboratory methods used to define the hemoglobin phenotype and the haplotypes of the b-like and a -globin gene cluster were described previously.11,38,39 SNP selection and genotyping The selection of genes was based on previously reported linkage peaks (6q22–q23, 8q11–q12 and Xp22.2–p22. 3)16,19,40 genes involved in metabolism of HU and genes related to erythroid progenitor proliferation. Thirty-three candidate genes were chosen in our study, and 320 tagging SNPs were selected based on genotype data from the HapMap project (Phase I, Yoruba sample)41 using Haploview (v3.2)42 (Table 1). Genotyping was carried out using a custom 384 multiplex design using an Illumina platform. For quality control purposes, about 3% of the DNA samples were re-genotyped, and Hardy–Weinberg equilibrium was assessed for each SNP. Hardy–Weinberg equilibrium was determined before analysis and was performed for quality control purposes rather than to evaluate if the genotypes met Mendelian expectation because all members of our The Pharmacogenomics Journal Statistical analysis Single SNP association was investigated via multiple linear regression analysis (SAS v9.1) with simultaneous adjustment for age, sex and the a- and b-globin gene cluster haplotypes. Three genetic models (codominant, dominant and recessive) governing modulation of response to HU treatment were tested. In the codominant system, an allele causes the homozygous form to look different from the wild-type and the heterozygous form (all three look different from each other). In dominant transmission, an allele causes the homozygous and the heterozygous forms to look the same as each other, but different from the wild type. In a recessive model, an allele affects the phenotype if it is present in the homozygous state only. One subject exceeding the top third standard deviation of all the values of the change in HbF levels was omitted from the analysis to prevent the extreme values from biasing the results. We used the nonlinear regression Random Forest method to identify additional predictors of response to HU treatment and to validate our results from single SNP association.49 Random Forest consists of a collection of regression trees, each regression tree itself being a regression function. Each of these trees predicts a real value by querying a set number of variables and instances within the regression model. Each regression tree is thus trained on a different bootstrap sample of both training instances and features. The Random Forest then averages the predictions made by the trees in the forest to produce the final output. Random Forest is a variance reduction technique and has provable properties with regard to resisting over fitting. Additionally, Random Forest is very efficient to train and test, and has built-in mechanisms for estimating test error and confidence in each prediction made. This procedure is nonparametric, not model based, and identifies those independent variables that best segregate subgroups as important predictors and identifies interactions among independent variables. The relative importance of the independent variables (e.g. SNPs in our case) is defined as follows. The mean of squared residuals (MSE) is computed on the out-of-bag data for each tree, and then the same is computed after permuting a variable. The differences are averaged over all trees and normalized by the standard error, and this calculation is repeated for each independent variable. A large difference in MSE (%IncMSE) indicates that the independent variable is important for correct prediction, whereas a small difference indicates that the independent variable is less important for correct prediction. Relative importance provides a measure by which predictors can be ranked with respect to each other. Our analyses were performed using the Random Forest (v 4.5–16) package as given in the R library (available at http://cran.us.r-project.org), except that, as the number of trees to be grown in the SNPs, hydroxyurea and HbF in sickle cell anemia Q Ma et al 393 forest was set at 2500, we used default parameters of this program. 16 Acknowledgments 17 We thank the investigators of the MSH who obtained blood samples for DNA-based studies and analyzed data from these studies for the study publications cited in the text of this paper. This study was supported by NHLBI Grant HL R01 70735 (MHS). 18 Duality of Interest The study sponsor had no involvement in study design, data collection, analysis or interpretation, writing of the paper or the decision to submit the paper for publication. References 1 Poillon WN, Kim BC, Rodgers GP, Noguchi CT, Schechter AN. Sparing effect of hemoglobin F and hemoglobin A2 on the polymerization of hemoglobin S at physiologic ligand saturations. Proc Natl Acad Sci USA 1993; 90: 5039–5043. 2 Milner PF, Kraus AP, Sebes JI, Sleeper LA, Dukes KA, Embury SH et al. Sickle cell disease as a cause of osteonecrosis of the femoral head. N Engl J Med 1991; 325: 1476–1481. 3 Castro O, Brambilla DJ, Thorington B, Reindorf CA, Scott RB, Gillette P et al. The acute chest syndrome in sickle cell disease: incidence and risk factors. The cooperative study of sickle cell disease. Blood 1994; 84: 643–649. 4 Platt OS, Thorington BD, Brambilla DJ, Milner PF, Rosse WF, Vichinsky E et al. Pain in sickle cell disease-rates and risk factors. N Engl J Med 1991; 325: 11–16. 5 Platt OS, Brambilla DJ, Rosse WF, Milner PF, Castro O, Steinberg MH et al. Mortality in sickle cell disease. Life expectancy and risk factors for early death. N Engl J Med 1994; 330: 1639–1644. 6 Steinberg MH. Therapies to increase fetal hemoglobin in sickle cell disease. Curr Hematol Rep 2003; 2: 95–101. 7 Platt OS, Orkin SH, Dover G, Beardsley GP, Miller B, Nathan DG. Hydroxyurea enhances fetal hemoglobin production in sickle cell anemia. J Clin Invest 1984; 74: 652–656. 8 Dover GJ, Humphries RK, Moore JG, Ley TJ, Young NS, Charache S et al. Hydroxyurea induction of hemoglobin F production in sickle cell disease: relationship between cytotoxicity and F cell production. Blood 1986; 67: 735–738. 9 Charache S, Dover GJ, Moore RD, Eckert S, Ballas SK, Koshy M et al. Hydroxyurea: effects on hemoglobin F production in patients with sickle cell anemia. Blood 1992; 79: 2555–2565. 10 Charache S, Terrin ML, Moore RD, Dover GJ, Barton FB, Eckert SV et al. Effect of hydroxyurea on the frequency of painful crises in sickle cell anemia. N Engl J Med 1995; 332: 1317–1322. 11 Steinberg MH, Lu ZH, Barton FB, Terrin ML, Charache S, Dover GJ. Fetal hemoglobin in sickle cell anemia: determinants of response to hydroxyurea. Multicenter study of hydroxyurea. Blood 1997; 89: 1078–1088. 12 Bakanay SM, Dainer E, Clair B, Adekile A, Daitch L, Wells L et al. Mortality in sickle cell patients on hydroxyurea therapy. Blood 2005; 105: 545–547. 13 Zimmerman SA, Schultz WH, Davis JS, Pickens CV, Mortier NA, Howard TA et al. Sustained long-term hematologic efficacy of hydroxyurea at maximum tolerated dose in children with sickle cell disease. Blood 2004; 103: 2039–2045. 14 Labie D, Pagnier J, Lapoumeroulie C, Rouabhi F, Dunda-Belkhodja O, Chardin P et al. Common haplotype dependency of high Gg-globin gene expression and high HbF levels in b-thalassemia and sickle cell anemia patients. Proc Natl Acad Sci USA 1985; 82: 2111–2114. 15 Steinberg MH, Voskaridou E, Kutlar A, Loukopoulos D, Koshy M, Ballas SK et al. Concordant fetal hemoglobin response to hydroxyurea in siblings with sickle cell disease. Am J Hematol 2003; 72: 121–126. 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 Garner CP, Tatu T, Best S, Creary L, Thein SL. Evidence of genetic interaction between the beta-globin complex and chromosome 8q in the expression of fetal hemoglobin. Am J Hum Genet 2002; 70: 793–799. Thein SL, Sampietro M, Rohde K, Rochette J, Neatherall DJ, Lathrop GH et al. Detection of a major gene for heterocellular hereditary persistence of fetal hemoglobin after accounting for genetic modifiers. Am J Hum Genet 1994; 54: 214–228. Chang YC, Smith KD, Moore RD, Serjeant GR, Dover GJ. An analysis of fetal hemoglobin variation in sickle cell disease: the relative contributions of the X-linked factor, b-globin haplotypes, b-globin gene number, gender, and age. Blood 1995; 85: 1111–1117. Chang YP, Maier-Redelsperger M, Smith KD, Contu L, Ducroco R, de Montalembert M et al. The relative importance of the X-linked FCP locus and b-globin haplotypes in determining haemoglobin F levels: a study of SS patients homozygous for bS haplotypes. Br J Haematol 1997; 96: 806–814. Craig JE, Rochette J, Sampietro M, Wilkie AO, Barnetson R, Hatton CS et al. Genetic heterogeneity in heterocellular hereditary persistence of fetal hemoglobin. Blood 1997; 90: 428–434. Wyszynski DF, Baldwin CT, Cleves MA, Amirault Y, Nolan VG, Farrell JJ et al. Polymorphisms near a chromosome 6q QTL area are associated with modulation of fetal hemoglobin levels in sickle cell anemia. Cell Mol Biol (Noisy -le-grand) 2004; 50: 23–33. Blobel GA, Weiss MJ. Nuclear factors that regulate erythropoiesis, In: Steinberg MH, Forget BG, Higgs DR, Nagel RL, (eds). Disorders of Hemoglobin Genetics, Pathophysiology, and Clinical Management, 1st edn. Cambridge University Press: Cambridge, 2001, pp. 72–94. Forrester WC, Takegawa S, Papayannopoulou T, Stamatoyannopoulos G, Groudine M. Evidence for a locus activation region: the formation of developmentally stable hypersensitive sites in globin-expressing hybrids. Nucleic Acids Res 1987; 15: 10159–10177. Grosveld F, Antoniou M, Berry M, de Boer E, Dillon N, Ellis J et al. Regulation of human globin gene switching. Cold Spring Harb Symp Quant Biol 1993; 58: 7–13. Tanimoto K, Engel JD. In vivo modulation of human beta-globin gene switching. Trends Cardiovasc Med 2000; 10: 15–19. Tuan D, Solomon W, Li Q, London IM. The "beta-like-globin" gene domain in human erythroid cells. Proc Natl Acad Sci USA 1985; 82: 6384–6388. Weiss MJ, Orkin SH. GATA transcription factors: key regulators of hematopoiesis. Exp Hematol 1995; 23: 99–107. Schaid DJ, Rowland CM, Tines DE, Jacobson RM, Poland GA. Score tests for association between traits and haplotypes when linkage phase is ambiguous. Am J Hum Genet 2002; 70: 425–434. Maier-Redelsperger M, de Montalembert M, Flahault A, Neonato MG, Ducrocq R, Masson MP et al. Fetal hemoglobin and F-cell responses to long-term hydroxyurea treatment in young sickle cell patients. The French Study Group on Sickle Cell Disease. Blood 1998; 91: 4472–4479. Ware RE, Eggleston B, Redding-Lallinger R, Wang WC, Smith-Whitley K, Daeschner C et al. Predictors of fetal hemoglobin response in children with sickle cell anemia receiving hydroxyurea therapy. Blood 2002; 99: 10–14. Bank A. Regulation of human fetal hemoglobin: new players, new complexities. Blood 2006; 107: 435–443. Stamatoyannopoulos G. Control of globin gene expression during development and erythroid differentiation. Exp Hematol 2005; 33: 259–271. Garner C, Silver N, Best S, Menzel S, Martin C, Spector TD et al. Quantitative trait locus on chromosome 8q influences the switch from fetal to adult hemoglobin. Blood 2004; 104: 2184–2186. Ikuta T, Ausenda S, Cappellini MD. Mechanism for fetal globin gene expression: role of the soluble guanylate cyclase-cGMP-dependent protein kinase pathway. Proc Natl Acad Sci USA 2001; 98: 1847–1852. Wang D, Donner DB, Warren RS. Homeostatic modulation of cell surface KDR and FLT1 expression and expression of the vascular endothelial cell growth factor (VEGF) receptor mRNAs by VEGF. J Biol Chem 2000; 275: 15905–15911. Reich DA, Lander ES. On the allelic spectrum of human disease. Trends Genet 2001; 17: 502–510. Charache S, Terrin ML, Moore RD, Dover GJ, McMahon RP, Barton FB et al. Design of the multicenter study of hydroxyurea in sickle cell The Pharmacogenomics Journal SNPs, hydroxyurea and HbF in sickle cell anemia Q Ma et al 394 38 39 40 41 42 43 anemia. Investigators of the multicenter study of hydroxyurea. Control Clin Trials 1995; 16: 432–446. Dozy AM, Kan YW, Embury SH, Mentzer WC, Wang WC, Lubin B et al. Alpha-globin gene organisation in blacks precludes the severe form of alpha-thalassaemia. Nature 1979; 280: 605–607. Steinberg MH, Rosenstock W, Coleman MB, Adams JG, Platica O, Cedeno M et al. Effects of thalassemia and microcytosis on the hematologic and vasoocclusive severity of sickle cell anemia. Blood 1984; 63: 1353–1360. Game L, Close J, Stephens P, Mitchell J, Best S, Rochette J et al. An integrated map of human 6q22.3-q24 including a 3-Mb high-resolution BAC/PAC contig encompassing a QTL for fetal hemoglobin. Genomics 2000; 64: 64–76. The International HapMap Consortium. The International HapMap Project. Nature 2003; 426: 789–796. Barrett JC, Fry B, Maller J, Daly MJ. Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics 2005; 21: 263–265. Xu J, Turner A, Little J, Bleecker ER, Meyers DA. Positive results in association studies are associated with departure from Hardy–Weinberg equilibrium: hint for genotyping error? Hum Genet 2002; 111: 573–574. The Pharmacogenomics Journal 44 45 46 47 48 49 Nielsen DM, Ehm MG, Weir BS. Detecting marker-disease association by testing for Hardy–Weinberg disequilibrium at a marker locus. Am J Hum Genet 1998; 63: 1531–1540. Deng HW, Chen WM, Recker RR. QTL fine mapping by measuring and testing for Hardy–Weinberg and linkage disequilibrium at a series of linked marker loci in extreme samples of populations. Am J Hum Genet 2000; 66: 1027–1045. Goring HH, Terwilliger JD. Linkage analysis in the presence of errors IV: joint pseudomarker analysis of linkage and/or linkage disequilibrium on a mixture of pedigrees and singletons when the mode of inheritance cannot be accurately specified. Am J Hum Genet 2000; 66: 1310–1327. Terwilliger JD. Inflated false-positive rates in Hardy–Weinberg and linkage-equilibrium tests are due to sampling on the basis of rare familial phenotypes in finite populations. Am J Hum Genet 2000; 67: 258–259. Weinberg CR, Morris RW. Testing for Hardy–Weinberg disequilibrium using a genome single-nucleotide polymorphism scan based on cases only. Am J Epidemiol 2003; 158: 401–403. Breiman L. Random forests. Machine Learning 2001; 45: 5–32.