Glucose-6-phosphate dehydrogenase deficiency

Among inborn errors of metabolism, the phrase by which inherited disorders were formerly known, glucose-6-phosphate dehydrogenase (G6PD) deficiency was the first red cell enzymopathy to be identified (Box 1), and it proved unique in at least 2 ways. First, G6PD deficiency is a polymorphic genetic trait with worldwide distribution (a current estimate is that over 500 million people are affected), but with great variation in prevalence, from 0 in the original Amerindian populations to 20% or more in parts of Africa and of Asia (Figure 1). Second, G6PD deficiency is a paradigmatic example of pharmacogenetics²³ : nearly all people with this trait have no disease, indeed no signs or symptoms, unless and until they are exposed to an exogenous agent that triggers acute hemolytic anemia (AHA), which may be severe and even life-threatening.

G6PD is an enzyme very ancient in evolution, found in all organisms except for Archaea, which are mostly anaerobic, and a few obligate surface parasites (for example, Mycoplasma genitalium) and intracellular parasites (for example, Rickettsia prowazekii).²⁴  G6PD is an oxidoreductase that catalyzes the oxidation of glucose-6-phosphate to 6-phosphoglucono-lactone coupled to the reduction of NAD phosphate (NADP) to reduced NADP (NADPH) (Figure 2). In all animals, G6PD is ubiquitously expressed, suggesting that it has an essential housekeeping function: indeed, knockout of the G6PD gene is lethal early in embryonic life.¹⁶  G6PD is often referred to as the first enzyme of the pentose phosphate pathway,²⁸  underscoring its role in the production of pentose sugars required for nucleic acid synthesis (but pentose can also be produced through the alternative transketolase-transaldolase pathway). NADPH, produced by G6PD and called its coenzyme, is the electron donor in reactions required for the biosynthesis of deoxyribonucleotides, fatty acids, and steroids; it is also the coenzyme of cytochrome P450, central to the metabolism of many drugs and other xenobiotics. The reducing power of NADPH is required in what is commonly referred to as defense against oxidative challenges.²⁸ 

There is no nucleus and therefore no DNA or RNA synthesis, no fatty acid synthesis, no endoplasmic reticulum, and therefore no cytochrome P450²⁹  in mature red cells: therefore, in their physiology, all of these factors are irrelevant except for the last function, which is paramount. Indeed, red cells are constantly exposed to an endogenous oxidative challenge because the concentration of hemoglobin (Hb) in red cells is no less than 5 mM. The reversible binding of O₂ to the 4 heme residues within Hb is a masterpiece of physical chemistry, but no physical device can be perfect. Spontaneous conformational fluctuations in the heme pocket of HbO₂ occasionally allow water or a small anion to enter, resulting in transfer of an electron from the iron to oxygen to produce methemoglobin and superoxide radicals: this auto-oxidation process affects 2% to 3% of total Hb each day (Low et al³⁰ ). Methemoglobin (Fe3+) is rapidly reduced back to Hb (Fe2+) by NADH cytochrome b5 reductase 3 (also known as methemoglobin reductase); whereas superoxide radicals, via superoxide dismutase, produce hydrogen peroxide (H₂O₂), a powerful oxidizing agent. It is here that the provision of NADPH is critical: indeed, 3 enzymatic processes in the red cell can remove H₂O₂, and NADPH is involved in all 3 (Figure 2). In red cells, G6PD is the 1 key to facing oxidative challenge.

Because mature red cells have no protein synthesis, G6PD activity physiologically decreases as red cells age.³¹  Reticulocytes have a high level of G6PD activity, whereas only approximately one-tenth of that remains in the oldest red cells; however, in G6PD normal red cells, this amount is still sufficient for their needs.

The commonest trigger of hemolysis in G6PD-deficient persons is a meal of fava beans: therefore, we briefly describe here an attack of favism (Figure 3A), a topic recently reviewed elsewhere.³²  The patient is usually a boy in the 2- to 10-year age group (but girls and adults are not exempt as discussed in “Genetics and molecular genetics”) who appears pale, jaundiced, and quite ill with abdominal pain and sometimes fever. On examination, the spleen may be enlarged, and the urine is usually dark (“like red wine” or “like Coca-Cola,” depending on cultural preferences). Blood examination shows moderate to very severe anemia, and a rather spectacular blood smear (Figure 3B). Supravital staining with methyl violet reveals Heinz bodies; there is often a neutrophil leukocytosis with shift to the left; the platelets are usually normal. The unconjugated bilirubin and lactate dehydrogrenase are elevated; haptoglobin is low or undetectable.

When a drug is the trigger (Table 1), the clinical picture is similar. The largest set of patients on whom there is full documentation is that of children treated with an antimalarial combination that contained dapsone (Figure 3C). Like in favism, hemolysis was brisk, with the hemoglobin nadir on day 3³³ ; but, unlike in favism, all children were already anemic (from malaria and/or other causes) when they were exposed to dapsone: this illustrates the importance of comorbidity in the clinical assessment of AHA. Drug-induced AHA is, by definition, iatrogenic: therefore, in preventing it, we bear an extra responsibility, which has become particularly poignant with the introduction of rasburicase and its long-acting derivative pegloticase.³⁴  These drugs are used electively in patients at risk for tumor lysis syndrome or kidney injury: they are contraindicated in patients with G6PD deficiency, meaning that there is no justification for using them before testing for G6PD deficiency (Box 2). There is a tendency to limit testing to patients from ethnic groups in which G6PD deficiency is common. Given that ethnic group is a vague concept and that ancestry may be mixed or only partly known, we recommend that, before giving rasburicase, G6PD testing should always be done.

Infection is the third potential trigger of AHA, but it is more erratic. Broadly speaking, bacterial infection must be severe to cause AHA: it has been reported with pneumonia,⁵⁰  brucellosis,⁵¹  rickettsiosis,⁵²  maxillary abscesses caused by Streptococcus,⁵³  or Staphylococcus,⁵⁴  and even with Clostridium difficile⁵⁵  infection. AHA also occurs in the course of hepatitis,⁵⁶  including hepatitis A, B,⁵⁷  and E.^58,59  With viral hepatitis in G6PD-deficient patients, hyperbilirubinemia is sometimes excessive even when there is little or no evidence of AHA,^56-60  and acute renal failure can be a further complication.⁶⁰  AHA has also been reported occasionally in the course of cytomegalovirus infection⁶¹  and in 1 instance of dengue fever.⁶²  AHA triggered by infection can be quite severe: if it subsides when the infection is controlled, then we can be confident that infection was the trigger. In some cases, AHA has been mistakenly blamed, rather than on the infection, on an antibiotic used to treat it; but the reverse could also happen.

With both fava beans and drugs, G6PD deficiency-related hemolysis is characteristically dose-dependent: the cases that come to the emergency room are the “tip of the iceberg.” When the anemia is mild, the patient may not be seen in hospital and may often be undiagnosed; in many cases, there may not even be anemia, but only compensated hemolysis.

If AHA is severe, blood transfusion may be urgent⁶³ ; otherwise, only fluid support and analgesics may be needed. In most cases, the patient had been previously asymptomatic: AHA will prompt testing for G6PD deficiency, which should be done by a quantitative test (Box 2). In most cases, the result will be clear cut, that is, below the normal range of 7 to 11 IU/g Hb.⁶⁴  Reference ranges vary, and unfortunately those stated by some laboratories are too wide. A result below 80% of the lower limit of normal must be regarded as G6PD deficient. Sometimes, during or immediately after a hemolytic attack, G6PD activity from a G6PD-deficient patient may be within the “normal range” (as a result of both destruction of the oldest cells and reticulocytosis). In such cases, the test must be repeated after a few weeks, or a known mutation can be revealed by DNA testing.

In areas with high prevalence, donor blood may be G6PD deficient: in general, this can be used safely⁶⁵  and universal screening of donors is not practical. Ideally, it would of course be preferable to use G6PD normal blood when needed for AHA in a G6PD-deficient patient, when doing an exchange transfusion for neonatal jaundice (NNJ; see “NNJ” later in text), or in patients requiring regular blood transfusion.

Unlike with autoimmune AHA, which often poses considerable therapeutic challenges, with G6PD-related AHA, rapid recovery is the rule. A most striking feature is the speed of the reticulocyte response (Figure 3A). It is hard to imagine that a hypoxia-erythropoietin–mediated response can take place within 48 hours: there must be a rush order (mechanism not yet known) that prompts the exiting from the marrow of existing reticulocytes.

This type of AHA seems almost like experimental hematology (reminiscent of a phenylhydrazine-treated rabbit).⁶⁶  The zero time is exposure to the trigger; from then on, we witness what happens when, in a person with a previously normal blood count, up to two-thirds of red cells are suddenly destroyed by a process referred to as oxidative damage. The culprit is a chemical (primaquine, or any of the other drugs listed in Table 1, or divicine and isouramil from fava beans) that can generate reactive oxygen radicals (ROS) and H₂O₂ (Figure 2): these can also be produced directly in the course of bacterial infection. As noted in the previous section, ROS are generated all the time through auto-oxidation of Hb; however, the rate at which they are thus produced is low, so that even G6PD-deficient red cells can provide enough NADPH to cope with them. When the levels of ROS are instead higher, red cells with normal G6PD respond by recycling NADPH at an increased rate, thereby preventing damage. G6PD-deficient red cells are unable to do so, which results in formation of ferryl hemoglobin and of hemichromes (partially denatured hemoglobin); at the same time, lipids and thiol groups in cytoplasmic and membrane proteins are oxidized. Binding of hemichromes to the membrane cytoskeleton leads to formation of Heinz bodies; aggregation of membrane proteins leads to formation of cross-bonded rigid hemighosts.^32,67  Glutathione (GSH) plays a special role in defense against oxidative damage because, in addition to being the substrate of GSH peroxidase, it can also regenerate thiol groups in proteins that had been oxidized to disulphides.⁶⁸  The more severely damaged red cells will undergo intravascular hemolysis, with consequent hemoglobinuria. In other red cells, clusters of oxidized band 3 in the membrane will bind immunoglobulin G and complement factor C3c⁶⁹ ; thus, opsonized, they will undergo erythrophagocytosis, that is, extravascular hemolysis. Intravascular hemolysis accounts, of course, for hemoglobinuria, and also for abdominal pain, because plasma Hb will bind nitric oxide causing smooth muscle dystonia; extravascular hemolysis accounts for jaundice and for splenomegaly. In G6PD-deficient red cells, like in normal red cells, there is a downward gradient in G6PD activity as they age⁷⁰ : therefore, the oldest red cells will be the first to suffer damage; whereas the youngest red cells, including reticulocytes, are relatively resistant to hemolysis.

This rather cumbersome heading originates from history. Hereditary spherocytosis (HS) has been, for over a century, a prototype of congenital hemolytic anemias (other than hemoglobinopathies), but in some patients with this kind of condition, spherocytes may not be prominent on a blood smear: after G6PD deficiency was discovered, some of them were found to be G6PD deficient.

In contrast to the high prevalence of G6PD deficiency that entails the risk of AHA, chronic nonspherocytic hemolytic anemia (CNSHA) is a rare disease (estimated frequency, <10 per million). The clinical picture is generally similar to that of HS, including jaundice and gallstones, with wide-ranging severity (see examples in Table 2): from mild (for example, diagnosed incidentally in an adult) to severe enough (in a minority of patients) to require recurrent blood transfusion. Unlike in HS, a history of NNJ (see next section), often severe, is the rule. Needless to say, all agents capable of causing AHA in any G6PD-deficient person will cause acute on chronic hemolysis in patients with CNSHA. In some transfusion-dependent patients, splenectomy has proven highly beneficial (Figure 4).

The clinical picture of CNSHA is so different from asymptomatic G6PD deficiency that it prompted early on⁹  a classification of underlying variants (Table 3). It is now clear that different mutations underlie these 2 phenotypes, as discussed later in text.

In newborns, G6PD deficiency entails the risk of NNJ (Box 1). The vast pertinent literature (see Luzzatto and Poggi⁶³  and Kaplan and Hammerman⁸⁸ ) cannot be fully reviewed here, but some facts must be noted. First, NNJ in G6PD-deficient babies can be severe enough to cause kernicterus, therefore, prompt management, which may require exchange blood transfusion, is imperative if permanent neurologic damage is to be avoided. Second, unlike with severe NNJ caused, for example, by Rhesus incompatibility, in most cases there is little if any evidence of hemolysis: indeed, we have to admit that the mechanism is not yet clear.⁸⁹  Third, although not all G6PD-deficient babies develop NNJ, in numerous studies from different parts of the world (for example, in Africa), the frequency of NNJ has been always higher in G6PD-deficient babies compared with G6PD normal controls.^90,91  The coexistence of a UGT1A1 promoter mutation further increases this frequency. Fourth, the peak incidence of hyperbilirubinemia is on day 2 to day 3 after birth: this is of great practical importance because mother and baby may have been discharged already, and therefore there may be a delay or even a failure of a diagnosis that requires urgent management by phototherapy or exchange blood transfusion.

The G6PD gene (which maps to Xq28) has 13 exons (exon 1 being noncoding: Figure 5), and it encodes a polypeptide chain of 514 amino acids, the dimer and the tetramer of which (Figure 6) are the enzymatically active forms of G6PD (the monomer has no enzyme activity).

The consequences of X-linkage are very important. First, because a male has only 1 G6PD allele, he can only be either hemizygous G6PD normal or hemizygous G6PD deficient; a female instead, having 2 G6PD alleles, can be homozygous G6PD normal, or homozygous G6PD deficient, or heterozygous for G6PD deficiency. Second, in keeping with the laws of population genetics, in any given population, heterozygotes are more frequent than hemizygous G6PD-deficient males, whereas homozygous G6PD-deficient females are much more rare (see examples in Luzzatto et al³⁸ ). Third, because G6PD, like most X-linked genes, is subject to the phenomenon of X-chromosome inactivation (Box 1), heterozygous females are epigenetic mosaics: in their blood, there is a mixture of G6PD normal and G6PD-deficient red cells. The ratio between the 2 types has a modal value of 1 (ie, 50% of each type), but this ratio is highly variable from 1 person to another: therefore, the phenotype of heterozygous females ranges from G6PD normal to G6PD, just as deficient as that of a hemizygous male. Some of these facts are sometimes misunderstood. At the population level, many studies report the frequency of G6PD deficiency in a sex-pooled population sample. This is unfortunate because only a subset of G6PD-deficient heterozygous females will have been classified as G6PD deficient, and therefore, in that population, we will not know the true frequency of the G6PD-deficient allele(s): instead, this could have been exactly ascertained from the frequency of G6PD-deficient males alone, which is identical to the frequency of the respective allele. It is also often stated that G6PD deficiency “is more frequent in males,” or “more expressed in males”: both statements are not correct. It is true, instead, that whereas G6PD deficiency is regularly expressed in G6PD-deficient hemizygous males and homozygous females, in heterozygous females, the expression, and therefore the potential severity of hemolysis, is highly variable. Surprisingly, G6PD deficiency is often qualified, even in textbooks, as “X-linked recessive”: of course this is wrong because G6PD deficiency is frequently expressed in heterozygotes, both biochemically and clinically.

There are now 230 G6PD variants with known mutations (Table 3; Figure 5; supplemental Table 2). All but 2 of them are either missense mutations (each 1 causing a single amino acid replacement) or small in-frame deletions (causing loss of 1 or a few amino acids): these will produce a qualitatively abnormal G6PD protein, which still has some residual enzyme activity. Frame-shift mutations, which would result in zero enzyme activity, are conspicuously absent; and the only nonsense mutation (G6PD Georgia; supplemental Table 2), which would do the same, has been found in a heterozygote. These data support the notion that complete loss of G6PD activity is lethal. Some variants have 2 mutations in cis: the best known is G6PD A^-, in which a mutation in codon 126 (N126D, not causing G6PD deficiency), coexists with 1 of 3 other mutations (M68V, R227L, L323P). It is the combination of the 2 mutations that causes G6PD deficiency.⁹⁴  Note: the term variant is currently often used to designate proteins resulting from mutations that may or may not have pathological implications. Algorithms predicting pathogenicity of variants have become popular, but, in the case of G6PD, the finding of enzyme deficiency is the best test because every enzyme-deficient variant has the potential to cause clinical manifestation.

Given the large number of point mutations identified within the G6PD gene, we should try to understand why and how they give a particular phenotype. First, we must consider what will cause G6PD enzyme activity to be deficient. A low rate of G6PD protein synthesis is unlikely, as no regulatory mutations have been found. In principle, we might be dealing therefore with decreased stability or decreased catalytic activity (k_cat). Either or both occur in different cases (see examples in Table 2), but decreased stability (that may imply impaired folding or impaired stability of dimer or tetramer) is more prominent in more cases; this is not surprising, in view of the long lifespan of red cells, whereby the physiological decrease in G6PD activity as red cells age is greatly accelerated. If instability is moderate, there will be enough G6PD for the steady-state requirements of red cells, and these will fail (ie, undergo hemolysis) only when an exogenous oxidative stress is applied. If instability is severe, G6PD activity will be insufficient to enable normal red cell survival even in the absence of oxidative stress: this will give the CNSHA (class I) phenotype.

From the evolutionary point of view, G6PD-deficient variants are enriched in amino acid residues of intermediate conservation (supplemental Figure 1 and Notaro et al²⁴ ). We presume that mutations in more highly conserved residues are more likely to be lethal, and mutations in nonconserved residues are less likely to cause G6PD deficiency. The next question is why, once a mutation gives G6PD deficiency, the phenotype is mild (class II or III; Table 3) or severe (class I; CNSHA). The most obvious factor must be the level of residual G6PD activity: in this respect, there is a striking density of class I mutations in exon 10 (Figure 5), which comprises many of the amino acids involved in the dimer interface (failure of dimerization entails the maximum of instability); amino acid replacements near the structural NADP site (Figure 6A) or the dimer-dimer interface⁹⁵  may also affect stability. Another factor is substrate affinity, because the intraerythrocytic concentrations of G6P and NADP are much below saturation⁹⁶ . Therefore, a high K_m^G6P will be generally an adverse feature (Table 2), whereas a low K_m^G6P will be an advantage in the steady state (less so under oxidative stress: it might be the reason why G6PD Mediterranean, for instance, is in class II rather than in class I). The relative roles of k_cat and of enzyme instability in G6PD variants have been recently explored⁷⁸  and subjected to rigorous principal component analysis.⁸¹ 

The regulation of G6PD transcription has been investigated in some detail. The G6PD core promoter region includes 7 guanine-cytosine (GC) boxes, 2 of which, through binding of the Sp1 and AP-2 transcription factors, are essential.⁹⁷  G6PD expression requires a specific chromatin conformation, as histone deacetylase inhibitors, through enhanced recruitment of Sp1 and of RNA polymerase II, enhance G6PD transcription.⁹⁸ 

Animal models of G6PD deficiency have been reviewed elsewhere.³⁸  Although models exist in mouse and in zebrafish, they have not yet been adopted for routine safety testing of new drugs.

By transduction of hematopoietic stem cells with a retroviral vector harboring the human G6PD complementary DNA (cDNA), stable expression of human G6PD was obtained in primary and secondary recipient syngeneic mice.⁹⁹  A similar vector was also competent for human G6PD expression in Macaque monkeys.¹⁰⁰  In principle, severe CNSHA due to G6PD deficiency could be treated by allogeneic bone marrow transplantation or by gene therapy, but this has not yet been attempted.

In view of the fact that histone deacetylase inhibitors selectively increase the synthesis of G6PD (but not of 16 other red cell enzymes that have been tested⁹⁸ ), butyrate or valproate might correct G6PD deficiency, but chronic administration would be required (or, in theory, one might contemplate using 1 of these agents to prevent or curb hemolysis under special circumstances). Very recently, high-throughput screening has been used to identify a small molecule, AG1 (2,2′-disulfanedylbis-N-(2-(1H-indol-3-yl)ethyl)ethan-1-amine), which is able to activate G6PD¹⁰¹ . Although the activation is less than twofold, AG1 might be a lead compound for finding more effective activators.

A close geographic correlation between the frequency of G6PD deficiency and malaria endemicity, reminiscent of what is well known for Hb S, has been observed in very many studies (Box 3 and Figure 1); as a result, the coexistence of the 2 abnormalities is not infrequent, and it was natural to wonder whether G6PD deficiency makes sickle cell anemia worse. Overall, the effect is small.^115,116  When G6PD-deficient red cells are infected by Plasmodium falciparum, they are sensed by macrophages as abnormal at an early stage; they are removed,¹⁰⁹  which seems a highly plausible protective mechanism. The original notion that protection is a prerogative of heterozygous females,¹⁰⁷  as in classic balanced polymorphisms, has been confirmed by several recent studies.^117,118  Perhaps G6PD deficiency protects against cerebral malaria but not against malaria with severe anemia,¹¹⁹  although this finding has been challenged as possibly due to “collider bias.”¹²⁰  It remains counterintuitive that having only a portion of G6PD-deficient red cells (as in heterozygous females) is more protective than having all G6PD-deficient red cells (as in hemizygous males). A convincing mechanistic explanation is still needed.

Darwinian selection is not the only link between malaria and G6PD deficiency, as the latter was originally discovered through the study of AHA caused by exposure to primaquine (Table 1), the only drug, until recently, that effectively eliminates the dormant liver forms (hypnozoites) of Plasmodium vivax. This requires a 14 day-course of primaquine (0.75 mg/kg per day), which will regularly cause hemolysis in a G6PD-deficient person, including the majority of heterozygous females. A much lower dose (0.25 mg/kg once only) is recommended to eliminate gametocytes (after a P falciparum malaria attack has been treated with a standard course of an artemisinin combination): this is safe for G6PD-deficient persons.¹²¹  The primaquine analog tafenoquine (Table 1) also causes AHA in G6PD-deficient persons¹²² : we may surmise that damage to the parasite and damage to G6PD-deficient red cells go hand in hand because they are both mediated by the ability of these 4-aminoquinolines to produce ROS. Tafenoquine is attractive because, due to its longer in vivo half-life, a single dose is sufficient; however, this becomes a liability in a G6PD-deficient person, as the drug cannot be discontinued if and when signs and symptoms of AHA develop. In view of this, tafenoquine¹²³  has been licensed with a label that positively prescribes testing for G6PD deficiency: this in turn has been a strong stimulus to the development of simple and reliable point-of-care tests for G6PD deficiency, including heterozygous females (Box 2).

Because G6PD is ubiquitously expressed, one might expect manifestations of G6PD deficiency not to be limited to red cells, but rather to occur in other cells or tissues as well. Overall, these are not very prominent, mainly because, as stated herein, in most cases the molecular basis of G6PD deficiency is enzyme instability, and presumably nucleated cells can compensate for instability, if need be, through increased G6PD synthesis. The consequence of G6PD deficiency in granulocytes has been best characterized in some of the patients who have CNSHA (for example, G6PD Barcelona¹²⁴ ), who suffered severe bacterial infections. The likely explanation is that granulocytes with severe G6PD deficiency are unable to produce enough NADPH to provide an effective oxidative burst (so much so that, a long time ago, G6PD deficiency had been confused with X-linked chronic granulomatous disease). A possibly related finding is the increased frequency of severe sepsis in G6PD-deficient patients after major trauma¹²⁵ : interestingly, this was observed in patients with the common variant G6PD A⁻. There have been isolated case reports in G6PD-deficient persons of acute rhabdomyolysis with myoglobinuria and sometimes acute renal failure.¹²⁶  This is a very rare occurrence when compared with AHA, with which it is usually but not always associated.¹²⁷ 

The cells of the eye lens share with red cells loss of organelles, including the nucleus. There has been controversy as to whether G6PD deficiency favors cataract formation, but this does not seem to be the case.¹²⁸  There is extensive literature on G6PD and cancer, and the results have not been uniform. A recent study has found that adenomatous polyps and colon cancer are less frequent in persons with G6PD deficiency/¹²⁹  we think it is preferable to await confirmation before speculating on how G6PD deficiency might be protective. The risk of cardiovascular disease has been reported to be somewhat higher (odds ratio, 1.39; confidence interval, 1.04-1.87) in G6PD-deficient males compared with controls.¹³⁰ 

G6PD deficiency is not a disease (except for the very small minority of patients who have CNSHA): it is a widespread genetic trait that can protect heterozygotes from dying of malaria. Yet, AHA in a G6PD-deficient child or adult is a medical emergency that, if not promptly and appropriately treated, can be fatal. On a worldwide basis, the commonest trigger of AHA is the ingestion of fava beans: favism is seen in at least 35 countries, and there are probably thousands of cases every year.³²  The next common trigger, in the same and in other countries, is iatrogenic: deaths have been reported with both primaquine¹³¹  and rasburicase,¹³²  and these deaths are preventable. Favism is preventable by population screening and health education¹³³ ; possibly also by the introduction of fava bean cultivars with absent or very low levels of vicine and convicine.¹³⁴ 

As hematologists, we have learnt much from G6PD-deficient patients. At the same time, the biology of G6PD is eminently interdisciplinary, having been a model system in biochemical genetics and in understanding how the red cell responds to oxidative attack; a tool for studying X-chromosome inactivation (the most spectacular epigenetic event in human development); a tool, for years, for studying clonal populations; a pioneer in the molecular genetics of enzymopathies; and the best characterized example in humans of an X-linked genetic polymorphism balanced by Darwinian selection exerted by malaria.

The authors thank all of their collaborators in Ibadan, Naples, London, New York, Genova, Firenze, Bangkok, Ouagadougou, Dar es Salaam, and elsewhere, and particularly José Bautista for careful review of the manuscript.

This paper is dedicated to the memory of Olaniyi Babalola and Giorgio Battistuzzi, who did early studies on the biochemistry and genetics of G6PD variants, and of Graziella Persico, who cloned the human G6PD gene.

Contribution: All authors contributed to writing the paper and to assembling the large amount of supplementary materials.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Lucio Luzzatto, Hematology, Muhimbili University of Health and Allied Sciences, PO Box 65001, Dar es Salaam, United Republic of Tanzania; e-mail: [email protected].