Main

Healthy People 2020, the federal public health agenda, has set a goal of “Increase(ing) the proportion of persons with hemoglobinopathies who receive disease-modifying therapies” (1). For the vast majority of people with sickle cell disease (SCD), the Healthy People goal will be reached through increased use of hydroxyurea (HU). Critical questions surrounding its use include how this drug works to ameliorate the clinical severity of SCD and what subpopulation of children with SCD benefit most from its use. This review addresses these questions from a translational science perspective.

SCD affects an estimated 90,000 people in the United States (2), with more than 1,900 newborns detected annually through universal newborn screening (2). Infant screening, early preventive therapy, and parental guidance have largely eliminated early child mortality from SCD (3,4,5). Moreover, specialized care and ongoing preventive services have prolonged average life expectancy (6). Despite these successes, multiorgan damage and mortality accumulate by early adulthood, resulting in shortened life span (6).

HU holds expanding promise for improved clinical outcomes. More than 2 decades ago, the seminal Multicenter Study of Hydroxyurea phase III trial for adults demonstrated the striking clinical impact of HU: 40% reduction in the incidence of acute pain episodes, acute chest syndrome, and hospitalization (7). These results led to approval in 1998 of HU for use in symptomatic SCD by the US Food Drug Administration (FDA). HU remains the only FDA-approved drug for SCD, but approval does not extend to pediatric use. The approval gap for children is partially attributed to the lack of a commercial pharmaceutical sponsor. Helping to span the gap is the FDA’s recent commissioning of a pediatric study of the pharmacokinetics of HU and its relative bioavailability of a liquid formulation (http://clinicaltrials.gov/show/NCT01506544).

Clinical efficacy of HU treatment varies among individuals, although most patients with severe phenotypes benefit from its use (7,8). This review describes newly identified mechanisms for the effects of HU, including genetic regulation of fetal hemoglobin (HbF) as a disease modifier and the biologic effects of HU on blood vessels and gene regulation. These recent advances improve the prospects for prospectively assessing efficacy of HU therapy, are inspiring clinical trials for additional salutatory effects of HU, and may guide future drug development.

Clinical Effects

The profound clinical effects of HU in children with SCD have been recently reviewed (9,10,11); these are summarized here in Table 1 . Much of the work on HU in children with SCD has come from phase II and III trials trials led by Ware et al., including pivotal studies such as phase I/II trial of HU in children with sickle cell anemia (HUG KIDS) (12,13,14), Hydroxyurea Safety and Organ Toxicity Trial (HUSOFT) (15), Pediatric Hydroxyurea Phase III Clinical Trial (BABY HUG) (16,17,18) and an early pediatric trial published in 1999 (12). French investigators have also contributed insights into the impact of HU (19,20). Randomized pediatric trials with HU have demonstrated decreased occurrences of pain episodes (18), acute chest syndrome, hospitalization (8,11,18), transfusion, and splenic autoinfarction (18), along with improved quality of life (21,22). Prolonged use sustains the laboratory effects of decreased anemia, markers of hemolysis, and counts of white blood cells and platelets, in addition to increased red cell mean corpuscular volume (23). Early HU use stabilizes renal hyperfiltration (24), hyposthenuria (25), and age-dependent decrease in HbF (18). Induction of HbF is described below.

Table 1 Clinical effects of hydroxurea on children with SCDa
Full size table

Of note, although the laboratory effects of HU apply across the pediatric ages tested, many of the various clinical improvements noted for one age range have not necessarily been assessed for other ranges. For example, reduced dactylitis, hyposthenuria, and transfusions were noted in the BABY HUG trial of children enrolled at age 9–13 mo (17,18). Improved transcranial Doppler blood flow through large cerebral arteries has been demonstrated in school-aged children (26,27). Despite positive findings, some of these trials had mixed results. For example, the primary end points of the BABY HUG study were not met (18). In the Stroke with transfusions changing to hydroxyurea (SWiTCH) study for secondary stroke prevention, continued chronic transfusion was advantageous relative to HU with phlebotomy (28). Moreover, HU reduces but does not eliminate the symptoms of and morbidity in SCD. For example, the SWiTCH trial demonstrated that chronic transfusions more effectively prevented pain episodes than HU with phlebotomy (29).

Two long-term studies demonstrated substantially improved life spans from prolonged use of HU in adults, including a study based on the Multicenter Study of Hydroxyurea in Sickle Cell Anemia trial (30,31). Prospective life span data for children taking HU are not yet available due to later uptake into pediatric trials. Nonetheless, a recent retrospective study from Brazil reported improved childhood mortality for those taking HU for up to 6 y (32). Collectively, these data are increasingly persuasive about the enduring impact of HU on SCD.

The pharmacokinetics of HU appears to follow a biphenotypic metabolism in children (33). Multiple single-nucleotide polymorphisms (SNPs) are associated with two apparent pharmacokinetic profiles of HU uptake and excretion. However, these genotypes do not correlate with response by the biomarker HbF.

Fetal Hemoglobin

The clinical severity of SCD is highly variable. Children experience multiple different clinical complications of differing severity levels and frequencies. HbF is of critical importance in the major sickle subtype homozygous sickle cell disease (HbSS; and sickle (HbS)-β-zero thalassemia, herein collectively referred to as HbSS). Lower HbF levels correlate with overall severer disease manifestations (34). Unlike “adult” hemoglobin A (HbA), HbF actively inhibits the polymerization of HbS, the underlying pathophysiology of SCD. In solution, HbF concentration higher than 15% prevents sickle globin polymerization (35). The cutoff for defining lower risk of severe complications has been estimated at 20% (36).

Sharp declines of HbF during infancy occur as HbF-producing γ-globin is replaced by β-globin. This switch leads to the predominant expression of either HbA or HbS. The F-to-S switch in children affected by HbSS (37) occurs more gradually than the F-to-A switch in nonanemic children. HbF levels in toddlers with SCD stabilize by age 3 or 4 y and are generally constant throughout childhood. Despite bearing the same β-globin sickle variant, affected populations with African ancestry exhibit wide variations in HbF levels (37,38,39,40). In the United States, pediatric levels vary from 3 to 20% of total hemoglobin, compared with only 0.5–2% for nonanemic individuals. The average HbF level in the US SCD pediatric population is ~10% (36).

Use of HU

Mechanism of Action

The physiology of the HU effect is complex and can generally be generally categorized into two overlapping pathways: effects on HbF production and improved blood flow through reduced intercellular adhesion ( Figure 1 ). HU is a short-acting cytotoxic drug that induces a state of “stress erythropoiesis.” Enhanced HbF production from intermittent mild marrow toxicity is believed to stem from the steady shifting of marrow physiology to the stressed state. The marrow responds to the repetitive pharmacological injury of daily use by enhanced erythropoiesis and increased HbF production (34,41). Paradoxically, the net effect of marrow toxicity is induced HbF and stabilization of cellular hemoglobin solubility. These effects lead to decreased levels of red blood cell (RBC) membrane damage and hemolysis (34,41).

Figure 1
figure 1

Physiological effects of hydroxyurea on sickle cell disease (SCD). Hydroxyurea has pleiotropic effects in ameliorating SCD, with complex and interacting effects of vascular and red blood cell (RBC) components. Hb, hemoglobin; HbF, fetal Hb; HbS, sickle hemoglobin; WBC, white blood cells.

PowerPoint slide

Full size image

HbF induction usually occurs within the first few months after initiating HU and is reversible on cessation or diminution of dosing ( Figure 2 ). Relevance of HU induction of HbF was demonstrated through a proof-of-principle murine model for SCD. Lack of expression of human HbF precluded HU induction in those mice. In the murine model, HU itself had no effect on improving anemia or protecting organs from SCD damage. In contrast, HbF gene therapy markedly improved the blood smear, microscopic, and organ-level pathological effects of SCD (42).

Figure 2
figure 2

Fetal hemoglobin (HbF) levels of a teenager with homozygous sickle hemoglobin (HbSS) on hydroxyurea (HU). Before HU use, this teenager had two to three hospitalizations for pain each year. She had no admissions for 1.7 y after beginning HU. Her baseline HbF was 2.4%, and maximum recorded HbF level was 16.9%. She acknowledged intermittent adherence in the years 2 and 3, during which time she had two admissions for acute pain episodes. Blue diamonds refer to HbF data points.

PowerPoint slide

Full size image

HU appears to influence RBC–endothelial interactions. Decreased expression of RBCs, white blood cells, and endothelial integrins and other adhesion molecules probably improves microvascular blood flow and reduces proinflammatory cell–cell interactions (43,44). Microvascular effects of SCD and HU appeared to be replicated using an interesting microfluidic model of blood flow and endothelialized microfluidic channels (44). Whole-blood samples from SCD lead to microvascular occlusion and thrombosis. Blood samples from patients with SCD had diminished velocity and greater tendency to obstruct in the microchannels. These effects nearly normalized using blood from patients on HU (44). HU may be associated with reduced generation of microparticles, suggesting a reduction in markers of inflammation and thrombosis (45).

HU may reduce cellular adhesion in general and/or adhesion provoked by infection or inflammation. Integrins and other cell surface glycoproteins regulate neutrophil migration and RBC flow through endothelial interactions. In a murine model for SCD and pneumococcal pneumonia and sepsis, HU provided some protection by decreasing the recruitment of neutrophils into infected lungs. Mice genetically engineered to lack E-selectin were not protected by HU (46). This finding strengthens the view that HbF-independent effects of HU include decreasing leukocyte–endothelial adhesion.

HU may also stimulate nitric oxide (NO) production as an NO donor or through stimulation of intermediates (discussed below). As a potent vasodilator, NO repletion contributes to improved vascular health in SCD ( Figure 1 ) (47). Along with decreased “sticky” interaction between blood cells and the endothelium, enhanced NO-induced local vasodilation may also benefit blood flow ( Figure 1 ) (48,49). However, questions have arisen regarding these effects of NO (50). In all, decreased pathology from damaged RBCs and pathological interactions between RBCs and endothelial cells appear to synergistically reduce clinical signs, symptoms, and morbidities of the disease ( Table 1 ). The ameliorative effects of HU appear to persist for as long as it is taken and the pharmacokinetics are maintained.

HbF Response to HU

The individual extent of HU-induced HbF is highly variable. Standard pediatric dosing of HU adjusts for dose-dependent myelotoxicity (14,33). Under these conditions, HU generally induces HbF by an additional 8–18% relative to the baseline levels (14,33,51,52). In contrast with the biomarker glycosylated Hb (HbA1c) for diabetes, no absolute HbF target exists. Nonetheless, peak attained HbF levels remain fairly constant in childhood (23). No absolute limit to the therapeutic amount of HbF induction has been described. For example, people of Southeast Asia or Saudi Arabia with SCD have baseline HbF levels averaging 16–20%. HU induction raises their levels 1.5- to 2-fold, associated with further diminution of their already tempered clinical symptoms (53).

Children with SCD generally have higher baseline HbF levels than adults and more pronounced HbF response to HU (14,54). Factors responsible for differences may include the need for highly regenerative marrow RBC precursors and, for HU, normal renal function for prompt excretion. Adults normally experience age-dependent decreased marrow cellularity. In SCD, disease-related marrow infarcts and other age-related physiological effects could exacerbate normal marrow regression. Age-related diminution of response to HU increases the likelihood that genetic studies using pediatric populations may reveal more precise basic biologic insights.

Genetic Analysis of HU-Induced HbF

Analyses of HbF regulation are crucial to understanding the spectrum of SCD severity, variability of HU response, and design of novel therapies. In addition to the established observations of ethnic variability of HbF levels, several key observations drive the rationale for identifying genetic components of HU induction of HbF in US populations of SCD:

  1. 1

    Baseline HbF levels in SCD have high heritability (55,56);

  2. 2

    HbF induction from HU therapy is also a heritable trait (55);

  3. 3

    Genome-wide SNP studies in normal nonanemic adults identified a few major loci associated with variation of low HbF levels. These regions are both cis and trans of the β-globin gene locus (56,57);

  4. 4

    These same loci are associated with baseline HbF in people with SCD in the United States (56,58,59,60,61,62). Additional loci have been identified but are not yet replicated;

  5. 5

    A modest correlation in children exists between levels of HbF at baseline and on HU (14,33,51).

Taken together, these findings lead to the prediction that genetic regulation of HbF expression at baseline overlaps with the control of HU-induced HbF The three major loci related to HbF expression in normal and SCD populations are the following: a SNP upstream of the Îł-globin gene within the globin locus on chromosome 11, previously identified by restriction enzyme analysis as the XmnI site (37,60,61); BCL11A, the gene encoding a transcription factor now recognized as a major silencer of HbF expression (51,58,59,61,63); and the intergenic interval between HBS1L and MYB (56,58,61). Additional loci have been identified and await replication (59,64,65), in addition to evaluation of known and probable epigenetic effects ( Figure 3 ) (66).

Figure 3
figure 3

Phenotypic variability in hydroxyurea (HU) response. A diagram synthesizes the varying clinical and genetic effects of HU in sickle cell disease. The β-globin locus is shown below. ARG1 and 2, arginase 1 and 2; Hb, hemoglobin; HbF, fetal Hb; LCR, locus control region; MYB, myeloblastosis oncogene.

PowerPoint slide

Full size image

Only a few published studies report on the genetics of HbF response to HU in SCD (33,51,54,64). Compared with genomic studies of more common disorders, sample sizes of studies on HU effects in SCD are inevitably modest. Using the retrospective cohort from the Multicenter Study of Hydroxyurea adult trial and assessing more than two dozen candidate genes, Ma et al. (54) reported significant associations between SNPs and HbF response to HU in loci of genes involved in the metabolism of arginine to NO and in a transcription factor that induces DNA bending. This report predated the identification of BCL11A as a central regulator of HbF expression. Most of the Multicenter Study of Hydroxyurea patients exhibited a small HbF response to HU (54), with less than 5% change in HbF from baseline. This blunted response is not universal in US adults with SCD and may be influenced by both patient characteristics and adherence to HU regimen.

Whether HbF induction by HU occurs through the direct influence of BCL11A is a concept awaiting direct testing. Effects of BCL11A on HbF are probably mediated through its protein partners, upstream or downstream effectors, chromatin structure, and/or telomerase function (recently reviewed in ref. (67)). Other reports include associations between HU response in SCD and polymorphisms in the guanosine triphosphate–binding protein gene sar1a (64), underscoring the complexity of the genetic pathways regulating the HbF response to HU ( Figure 3 ).

Two pediatric pharmacogenetic analyses using candidate SNP markers suggested that just a few genes are associated with baseline HbF, including several SNPs within BCL11A ( Table 2 ). SNP associations with induced HbF are generally not independent of baseline HbF levels (33,51). In contrast with the induced HbF level, the treatment-associated increment appears to be a less relevant marker. Both of these observations probably reflect the association between baseline and induced levels.

Table 2 SNPs associated with HbF
Full size table

In our own smaller multisite analysis, baseline levels of the candidate genes were significantly associated with SNPs within the BCL11A and the β- and ε- globin loci (HBB and HBE, respectively), with an additive attributable variance from these loci of 23% ( Table 2 ) (51). Consistent with studies by Ware et al.(14,33), we reported that baseline HbF levels explained 33% of the variance in induced levels. The variant in HBE accounted for an additional 13% of the variance in induced levels, whereas variants in the HBB and BCL11A loci did not contribute beyond baseline levels. Thus, our data suggest that the combined effects of baseline HbF and one SNP marker contributed an estimated 46% of the variance in HbF (51).

By trend analysis, children with an allele associated with higher HbF (“favorable” allele) in one of the BCL11A and/or either globin marker had significantly higher average values of baseline HbF than those who lacked a favorable allele (51). Effects on baseline HbF from a SNP in each these two genes were additive and were associated with two-fold higher HbF for patients with favorable alleles in both loci. Similarly, having at least one favorable allele in either globin locus and in BCL11A was associated with a higher level of induced HbF. Statistical significance did not withstand adjustment for baseline HbF, probably reflecting the interrelatedness of HbF regulation under both physiological conditions. Genetic studies examining larger pediatric populations on HU, unusual responders, and the influence of specific sequence variants are needed to evaluate the contribution of these and other genetic loci responsible for HbF response.

Other Physiological Effects of HU

Cellular Biology

The effects of HU largely depend on its effects on nucleic acid synthesis in dividing RBC progenitors. HU affects the S-phase by inhibiting ribonucleotide reductase, an enzyme important for DNA synthesis. Depletion of DNA precursors by HU causes arrest of the replication fork, leading to cell death. A cell-based ex vivo assay for HbF induction, burst-forming unit erythroid colonies grown in methylcellulose from blood of children with SCD, demonstrated that HU decreases the number of burst-forming unit erythroid colonies. HU and other ribonucleotide reductase inhibitors increase HbF production in that system. Interestingly, other cytotoxic agents that are not of that drug category, such as cytarabine and alkylating agents, decreased burst-forming unit erythroid counts but did not induce HbF (41,68).

HU’s lethal effect of on ribonucleotide reductase and cell survival are also seen in laboratory bacteria such as Escherichia coli (69). Whether the bactericidal effects influence investigation using animal models or even in patients has not previously been studied. Direct bacterial effects on the HU-dampened expression of adhesion molecules should be addressed in a murine model of bacterial infection.

HbF response to temporary marrow toxicity is probably attributable to transcriptional and epigenetic effects on the progenitor developmental program (66,70). HU signaling appears to involve cGMP (cyclic guanine monophosphate), cAMP (cyclic adenosine monophosphate), p38MAPK (mitogen-activated protein kinase), and other pathways. Activation of cGMP may induce HbF via enhancing production of NO (47,68,71,72). NO may also support HbF production (47). HU induces a small guanosine triphosphate–binding protein, the secretion-associated and Ras-related protein (SAR) (73). SAR may be involved in the activation of transcription factors and signal transduction pathways in erythroleukemia K562 cells and in human bone marrow–derived progenitor cells. HU may also function through kinase and signal transduction pathways, such as globin transcription factor-1, to enhance γ- and β-globin synthesis in erythroid cells (47).

Gene Expression

Comparing whole mRNA at the pre- and post-initiation stages of therapy revealed that HU affects expression of a number of genes involved in transcription, translation, ribosome assembly, and chromosomal organization (66,70). Results may vary with age, dosing, or other clinical conditions. Variation in cell source, whether from bone marrow or purified early reticulocytes, would be expected to affect detection of expressed genes. HU may also affect expression of genes that link HU and HbF to BCL11A (66,70). Epigenetic analysis of the Îł-globin promoter did not reveal much impact from HU (74). Interestingly, HU appears to upregulate specific microRNAs (74). These results require further investigation but underscore the view that HU is involved with complex pathways of gene regulation.

Testing for Oncogenicity

The primary effect is damaging DNA replication by inhibiting ribonucleotide reductase. This effect raises concerns about an oncogenic potential, especially after prolonged use. These fears have been amplified by its original use as chemotherapeutic agent for chronic myeloid leukemia, the latent phase of acute leukemia. Although links with acute leukemia outside of chronic myeloid leukemia have been disproven (75), concerns for the safety of long-term use in children persist. Several studies have tested DNA and cellular toxicity from pediatric HU users. No genotoxicity was detected using several different assays in vitro, including karyotype, illegitimate VDJ recombination of the variable regions of rearranged T-cell receptor genes, and chromatid breaks (9). Increased reticulocyte micronuclei were observed, but this effect was highly variable among patients and did not increase with time (76). In all, oncogenicity of HU is probably quite low or nonexistent. A few cases of acute leukemia were reported in patients after many years of HU treatment, but these do not appear to be more frequent than in the untreated population (75).

Potential Pharmacological Alternatives to HU

Other HbF inducers have been assessed during the past few decades, including nucleoside analogs such as 5-azacytidine and decitabine. However, they are often poorly tolerated, potentially oncogenic, and lack proof of effectiveness comparable with HU (recently reviewed in ref. (41)). Additional HbF-inducing drugs are histone deacetylase inhibitors, erythropoietin (already high in SCD and shown not to induce HbF in SCD), valproate, thalidomide derivatives (e.g., pomalidimide), and kit ligand. In all, a variety of cellular stresses and stimuli can promote coordinated stress responses, including activation of the γ-globin gene (66,70). Based on results from SCD mouse models, inhibitors of phosphodiesterase 9 (71) or hypoxia-inducible factor-1α (HIF-1α) (77), alone or in combination with HU, may be clinically useful to stimulate cGMP and NO for HbF production and/or to enhance its antisickling impact (71).

Barriers to Use of HU

Outside of clinical trials with HU, ample documentation exists of incomplete clinical effectiveness of HU. Uneven drug adherence has been well documented (78,79). Provider non- and underutilization is well documented (75,80). Our recent multisite survey of parents of children with SCD revealed several family barriers to use of HU, such as lack of FDA approval, near-universal safety concerns, and highly varied knowledge about its benefits, including many for whom its basic property of decreasing episodes of pain was unknown (81). Use of HU was positively correlated with fundamental knowledge of parents regarding the basic positive effects of HU on disease, independent of parental demographics such as education level, language spoken, or ethnicity. Barriers in effective communication between providers and families may be exacerbated by issues arising from medical delivery systems.

The mixed uptake of HU by families may also reflect family perspectives on the long-term effects of SCD. A single-site survey of parents revealed that the majority believed that the disease effects were going to diminish over time and would not affect life goals or life span (82). These poignant perceptions will need to be addressed if families are to embrace the long-term benefits of HU against its inconveniences and largely theoretical risks.

Conclusion

HU is a remarkably effective drug for a large proportion of children with SCD. Efforts to achieve expanded understanding of the scientific underpinnings of its effects on SCD, predict individual response, and perfect the clinical applications for modifying disease effects are ongoing. Clinical trials will continue to test the uncertain benefits of HU ( Table 1 ), such as primary prevention of brain infarcts (clinicaltrials.gov/show/NCT01389024). Murine models will facilitate insight into the benefits provided by induced HbF, altered expression of adhesion molecules, reduced BCL11A levels, and other mechanisms. Genetic epidemiology will be used to identify specific variants in regulatory genes and gene pathways.

The accumulating science of HU is anticipated to lead to three direct effects for children with HbSS: (i) use at earlier ages; (ii) wider clinical indications; and (iii) delineation of children who are less likely to enjoy substantive benefit from HU. For this last group, more aggressive consideration of chronic transfusion, hematopoietic stem cell transplantation, or trial of emerging alternative agents may be warranted. To date, early clinical trials of other experimental HbF-inducing drugs have demonstrated considerable short- and long-term toxicity compared with HU. Therefore, HU is predicted to remain the mainstay of pharmacological therapy for SCD in the foreseeable future.

Progress toward the Healthy People 2020 goals will occur through increased use of HU. Nonetheless, the entire SCD population may not benefit from HU alone. Dampened impact on clinical complications and HbF induction occurs for those with certain genotypes, many patients with HbSC (compound heterozygous for HbS and HbC), some adult patients, and those with renal compromise. New, effective, and safe therapies, alone or in combination with HU, are still needed to maximize pharmacological benefit for everyone living with SCD. Constructive engagement must be made to assist families in undertaking long-term HU use to help them to balance the optimism of HU treatment and its potential toxicities with the risk of accumulating disease consequences. Although several important crisis modulators are currently under investigation, disease modifiers that prevent crises and other morbidities are arguably the primary therapeutic targets.

Statement of Financial Support

The authors acknowledge support from the Clinical Translational Science Award at Columbia University, 5UL1RR024156 (H.N. Ginsberg, principal investigator).

Disclosure

No commercial or other conflict of interest is reported by the authors.