Article Text

This article has a correction. Please see:

Download PDFPDF

Paroxysmal nocturnal haemoglobinuria: Nature's gene therapy?
  1. R J Johnson1,
  2. P Hillmen2
  1. 1Department of Haematology, Birmingham Heartlands Hospital, Bordesley Green East, Birmingham B9 5SS, UK
  2. 2Haematological Malignancy Diagnostic Service, The General Infirmary at Leeds, Great George Street, Leeds LS1 3EX, UK
  1. Correspondence to:
 Dr R J Johnson, Birmingham Heartlands Hospital, Bordesley Green East, Birmingham B9 5SS, UK;


The development of paroxysmal nocturnal haemoglobinuria (PNH) requires two coincident factors: somatic mutation of the PIG-A gene in one or more haemopoietic stem cells and an abnormal, hypoplastic bone marrow environment. When both of these conditions are met, the fledgling PNH clone may flourish. This review will discuss the pathophysiology of this disease, which has recently been elucidated in some detail.

  • paroxysmal nocturnal haemoglobinuria
  • PIG-A gene
  • aplastic anaemia
  • AA, aplastic anaemia
  • ER, endoplasmic reticulum
  • GPI, glycosyl phosphatidylinositol
  • GlcNAc, N-acetylglucosamine
  • PEA, phosphoethanolamine
  • PI, phosphatidylinositol
  • PNH, paroxysmal nocturnal haemoglobinuria
  • VSG, variant surface glycoprotein

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

The development of paroxysmal nocturnal haemoglobinuria (PNH) requires two coincident factors: somatic mutation of the PIG-A gene in one or more haemopoietic stem cells and an abnormal, hypoplastic bone marrow environment. When both of these conditions are met, the PNH clone may flourish. Recently, the pathophysiology of this disease has been elucidated in some detail and we now have rational theories concerning the clinical manifestations of PNH, its association with other haematological disorders (such as aplasia and myelodysplasia), and the sequence of events that leads to overt disease. Spontaneous somatic mutations are common in the PIG-A gene, leading to failure of synthesis of the glycosyl phosphatidylinositol (GPI) anchor. Without this structure, many proteins are unable to attach to cell surfaces. Red blood cells lose complement defence proteins, which explains the classic feature of intravascular complement mediated haemolysis. There is indirect evidence that platelet activation with consequent thrombosis is caused by a similar mechanism. The relative growth advantage of PNH cells in a hypoplastic marrow is also, presumably, a direct or indirect result of these alterations in surface antigen composition, although the precise pathophysiological mechanisms remain to be described. It is probable that the association with aplasia is explained by this relative growth advantage and that clonal evolution to a leukaemic state is a consequence of the primary insult causing the aplasia. In this way, PNH can be seen as an attempt to restore a form of useful, if abnormal, haemopoiesis in a damaged bone marrow: nature's gene therapy.


PNH as a clinical entity has puzzled physicians and scientists for 200 years. Perhaps the first description was “An account of a singular periodic discharge of blood from the urethra” written in 1794 by a Scottish surgeon, Charles Stewart. However, the first detailed description is credited to Paul Strübing in 1882.1 His patient had a six year history of passing dark urine intermittently in the mornings, always clearing by noon. His conclusions were detailed and astonishingly perceptive, suggesting as he did that there was intravascular haemolysis and that some of the patient's symptoms were the result of thrombosis. He even concluded (correctly) that a red blood cell defect was to blame. Despite this insight, the work was largely ignored and when Marchiafava reported a case in Italy 29 years later it was regarded as a new entity.2 Michelli published further observations in 1931 and used the weighty term “splenomegalic haemolytic anaemia with haemoglobinuria and haemosiderinuria, Marchiafava–Michelli type” from which the disease retains the eponym.3 The modern term PNH was first coined shortly before Michelli's paper in a description of a case from the Netherlands.4

Attempts to explain the haemolysis began in Rotterdam in 1911, when Hÿmans van den Bergh noted that PNH erythrocytes were sensitive to lysis in vitro when exposed to carbon dioxide.5 By the 1930s it had been shown that the lysis was complement mediated and pH dependent: being greatest in acidified conditions.6 This formed the basis for Ham's test,7 which was the diagnostic gold standard until its replacement by flow cytometry in recent years.

In 1944, Sir John Dacie first noted the association of PNH with aplasia in a case of Fanconi's anaemia.8 In the 1950s, the defect was shown to be present in other haemopoietic lineages, with the observation that the neutrophil alkaline phosphatase was reduced or absent in PNH.9 This led Dacie to suggest that PNH was an acquired clonal disorder arising in a haemopoietic stem cell.10 This important and perceptive idea was later confirmed by an elegant report in two patients with PNH who were also heterozygous for the enzyme glucose-6-phosphate dehydrogenase.11 In this study, it was shown that the patients' red cells with a PNH phenotype contained only one isotype of glucose-6-phosphate dehydrogenase whereas their residual, non-PNH cells contained both.

“Ham's test was the diagnostic gold standard until its replacement by flow cytometry in recent years”

From the 1960s onwards, an increasing number of proteins were shown to be missing from the cell surface in PNH. These included molecules known to be involved in the regulation of complement at cell surfaces. It was hypothesised that their absence caused a complement mediated intravascular haemolysis. Decay accelerating factor (DAF/CD55), which has a role in the inactivation of complement at an early stage of the cascade, was thought to be important but individuals with the Inab red cell phenotype, who have an inherited DAF deficiency, had no clinical illness or in vitro red cell complement sensitivity.12 CD59, which inhibits the formation of the membrane attack complex (the final step in the complement cascade) was the next candidate. Clinical evidence for the importance of this molecule came in 1992 when a 22 year old man was described who had a homozygous deficiency of CD59 on all his cells and suffered PNH-like symptoms, with haemolysis and cerebral thrombosis.13 It is probable that the haemolytic and thrombotic features of PNH are mediated by complement sensitivity and that CD59 deficiency is an important cause of this.

The biochemical explanation for the absence of these proteins became clear when the GPI anchor was described as a novel mechanism of attachment of antigens to cells in 1980.14 It was subsequently shown that all the proteins absent in PNH were GPI linked and that all GPI linked antigens are missing from PNH cells. In the past decade, the structure and biochemistry of the GPI anchor have been described and there is a consistent biosynthetic abnormality in all patients with PNH described to date.15

A gene whose cDNA is able to correct this defect in all transfected human cell lines has now been cloned.16 It is situated on the short arm of the X chromosome and has been named PIG-A, a term derived from its ability to restore the GPI synthetic defect in class A murine cell lines.17 Since the gene has been cloned, mutations have been found in all patient samples reported.18

This sequence of historical milestones has taken PNH from its most obvious clinical manifestation right back to a single gene defect in a haemopoietic stem cell. The story is in one sense complete but there are still many intriguing questions to be answered.



PNH is a rare disease. One of the largest epidemiological studies looked back at data from French centres from 1946 to 1995 and found only 220 reported cases.19 The annual incidence is about 4/million and the overall frequency is probably similar to that of aplastic anaemia (AA) with which it has a close association. It is probable that greater awareness and improved diagnostic methods will increase the number of diagnosed cases. The UK PNH registry in Leeds has been collecting new and existing cases since 1990 and currently has over 140 recorded patients (P Hillmen, personal observation, 2001).

Clinical features

Patients with PNH may have a long term chronic illness but the disease does shorten life. The median survival from diagnosis was 10 to 15 years in two large historical studies.19, 20 Patients most commonly die of thrombosis or progressive cytopenias. Leukaemic transformation is uncommon (< 5%). Many patients will continue to have intermittent paroxysms of haemolysis but some eventually achieve a spontaneous remission.20 The identification of patients destined to remit is clearly an important requirement to prevent the use of toxic treatments in patients with a good prognosis.

Haemolysis is the cardinal feature. It is classically paroxysmal and most apparent in the first urine passed on waking—hence the name of the disease. In practice, patients often have a chronic haemolysis with exacerbations. This results in a variable transfusion requirement with iron deficiency often contributing to the anaemia. All patients should receive daily folic acid, because a low degree of haemolysis is usual between paroxysms. Heavily transfused patients can, paradoxically, become iron overloaded and this should be monitored to avoid compounding the problem with iron supplements.

Thrombosis is the most feared complication. There is a well established predilection for the hepatic veins but a diversity of predominantly venous sites has been described. Patients with AA and only laboratory evidence of PNH are much less likely to suffer a thrombosis than those with active haemolysis and a large proportion of PNH cells in their blood. In this last group, thrombosis may occur in up to 50% and be the cause of death in one third.19, 20

The intimate connection between AA and PNH is underlined by the clinical course of the illness in individual patients. Some degree of cytopenia is a consistent finding, even in haemolytic PNH. This may range from a mild reduction in one cell lineage to life threatening bone marrow failure. Even when blood counts are normal, bone marrow examination and progenitor culture assays reveal impaired haemopoiesis.

Malignancy, such as myelodysplasia or acute myeloid leukaemia supervenes in around 5% of cases, but is probably a result of the process leading to AA rather than a specific risk related to the PNH clone itself, which is not considered preleukaemic.21, 22


The demonstration of non-immune haemolysis with haemosiderinuria should lead to an investigation for PNH. Alternatively, its presence may be sought because of AA or a venous thrombosis at an unusual anatomical site. The diagnosis is definitively established by the demonstration of GPI linked protein deficiencies on red blood cell and neutrophil surfaces by multiparameter flow cytometry.23 The Ham test has been largely abandoned where flow cytometry is available, because it is relatively insensitive and labour intensive and only gives information on red blood cells. The solid phase gel techniques used for antibody detection in blood transfusion provide a rapid screen but again are limited to red blood cell antigens. The proportion of affected red blood cells often gives a falsely low assessment of the true clone size because of the effects of the selective haemolysis of PNH red cells compared with their normal counterparts and because of the effect of transfusion. The neutrophil series is not affected by these variables and therefore allows an accurate measurement of the clone size. It is possible to detect PNH clones that comprise < 1% of neutrophils using multiparameter flow cytometry.24 It is important to include a transmembrane antigen as a lineage marker (for example, CD15 for neutrophils) and at least two GPI linked antigens (for example, CD55 and CD59) to exclude the rare inherited deficiencies of single antigens such as the Inab phenotype (CD55 deficiency).

“It is probable that the haemolytic and thrombotic features of paroxysmal nocturnal haemoglobinuria are mediated by complement sensitivity and that CD59 deficiency is an important cause of this”

Using flow cytometry, it is possible to demonstrate patterns of complete or partial GPI linked protein deficiency on the red blood cell series. Normal cells are designated type I, partially deficient type II, and completely deficient type III (fig 1). The clinical severity of the disease is directly related to the proportion of type III red blood cells.

Figure 1

An example of peripheral blood phenotyping in paroxysmal nocturnal haemoglobinuria (PNH). Plots from a flow cytometer showing the clear discrimination between populations of normal and PNH cells in a patient's peripheral blood sample. (A,B) Plots showing the three types of red blood cells that can be detected in these patients. Type 1 cells are normal, type II are partially deficient in glycosyl phosphatidylinositol (GPI) linked proteins, and type III are wholly deficient. CD55 and CD59 are GPI linked antigens found on red blood cells. (C) Plot showing granulocytes that have been dual stained with two GPI linked surface antigens (CD16 and CD66) in a patient with PNH. This clearly delineates the normal and the PNH cells. (D) Plot using the same double staining method but with antigens relevant to monocytes (CD64 and CD14): the same clear demarcation is shown. Courtesy of Dr S Richards, Haematological Malignancy Diagnostic Service, The General Infirmary, Leeds, UK.

If these techniques are applied to other haemopoietic cell lineages, GPI deficiency can be documented on platelets, monocytes, and lymphocytes, confirming the stem cell nature of the disorder25 (fig 1).


This is another area much influenced by the interplay of PNH and AA. Those with AA or progressive pancytopenia may be candidates for intensive disease modifying treatment, including immunosuppression or bone marrow transplantation. These approaches are usually not appropriate for classic PNH without bone marrow failure. Interesting exceptions to this rule are disease occurring in patients with a syngeneic twin. In this circumstance, there is little risk from transplant, although it appears that at least moderate doses of conditioning chemotherapy before stem cell infusion are necessary to prevent re-expansion of the PNH clone.26

Most patients without pronounced cytopenias simply require supportive management. Blood transfusion is the mainstay for those with symptomatic anaemia. Folate supplementation is mandatory and iron status should be monitored because patients can become iron deficient through urinary loss or iron overloaded from blood. Because thrombosis is a leading cause of mortality in this group, all those with haemolytic disease or a large percentage of PNH cells in their blood (perhaps > 50% PNH neutrophils) should be considered for primary prophylaxis with warfarin if there are no contraindications.20


A failure of GPI anchor synthesis is a constant and key feature in all cases of PNH. The genetic basis of this abnormality is now well described, as is the detail of the biochemical consequences.16, 18, 27 GPI deficiency causes a loss of many proteins from the cell surface. The resulting cell phenotype explains the clinical features and suggests a mechanism for expansion of the PNH clone. These assumptions, while reasonable, await further experimental proof and the correlation of all clinical sequelae with specific protein loss has yet to be achieved. Before speculating on this, it is worth describing what we know about the GPI anchor, the missing proteins, and the underlying genetic defect.

The GPI anchor

Most cell surface proteins are attached via a sequence of hydrophobic amino acids that spans the lipid portion of the bilayer. This transmembrane domain achieves a stable interaction, which is not easily disrupted without destroying the membrane itself. In the 1980s, another method of cellular attachment was described whereby proteins were linked to a GPI molecule, which was itself inserted into the cell membrane.28 The structure and the biosynthesis of this GPI anchor were determined from work in trypanosomes, whose major surface glycoprotein is attached by this mechanism.29 The backbone of the GPI structure is highly conserved between species.

There are essentially three parts to the GPI anchor (fig 2). The membrane attachment is achieved through the insertion of the lipid moiety of phosphatidylinositol (PI) into the outer leaflet of the membrane. There is then a glycan core consisting of a molecule of N-acetylglucosamine (GlcNAc) linked to three mannose residues and then to an ethanolamine. The protein attachment site is to the phosphoethanolamine (PEA) molecule linked to the terminal mannose. The C-terminus of the relevant protein is linked to the amino group of the PEA molecule by an amide bond.18, 30, 31

Figure 2

The glycosyl phosphatidylinositol (GPI) anchor. This is a simplified diagram of the GPI structure. The C-terminus of the anchored protein links to an ethanolamine residue on the GPI anchor. The anchor itself consists of this ethanolamine moiety attached to a glycan core. The GPI structure attaches to the cell membrane via phosphatidylinositol. The glycan core consists of a molecule of GlcNAc linked to three mannose residues. The first step in GPI synthesis is the linkage of the GlcNAc to PI. It is this reaction that fails in paroxysmal nocturnal haemoglobinuria because the genetic lesion disrupts the production of a necessary enzyme complex.

“Glycosyl phosphatidylinositol deficiency causes a loss of many proteins from the cell surface”

Biosynthesis of the GPI moiety occurs in the rough endoplasmic reticulum (ER). The precise location at which each reaction takes place is still a matter of some doubt. Some of the steps take place on the cytoplasmic surface of the ER and some within the cisternal space.32 The first step in the process is the addition of a molecule of GlcNAc to a PI residue. This does take place on the cytoplasmic surface of the ER and, at some point, the developing molecule is flipped to the cisterna. A series of three mannosylations follows using dolichol-phosphate mannose as a donor. Ethanolamine is added to each of these sugars. Transamidation of the newly synthesised protein leads to its attachment to the PEA molecule on the terminal mannose and finally the protein–GPI complex is transported to the external surface of the cell.

The importance of GPI linkage

The high degree of interspecies conservation of the GPI structure and its wide distribution argues for an important biological role for this mechanism of protein attachment. Most available information comes from work on trypanosomes and animal studies and the relevance to humans is speculative.

Enzymes have been characterised that cleave GPI anchors, thus releasing the tethered proteins.33 These phospholipases are present in other mammals and trypanosomes and their existence suggests that some proteins may be anchored through GPI to allow their selective removal. An example of this is the deliberate cleavage of the major surface protein (variant surface glycoprotein; VSG) of the trypanosome and its replacement with an immunologically discrete variant VSG from its repertoire. This allows the organism to evade the immune response in an infected host.34

Proteins attached through GPI are less tightly bound than their transmembrane counterparts, which allows a degree of transfer from one species (or cell) to another, as has been described in the parasitic infection caused by Schistosoma mansoni. In this case, the parasite seems to acquire host CD55, which in turn helps it to avoid complement mediated immune attack.35

In addition to allowing “loss or gain” of proteins, the biochemical membrane associations and mobility are different for GPI anchors and transmembrane domains. It is possible that the anchor's characteristics and localisation are integral to the normal function of the associated protein.

GPI linked proteins are not randomly distributed over the cell membrane. In polar cells they are frequently located at the apical pole. In all cells, they associate with each other in regions of the membrane that are rich in glycolipids (sphingolipids and cholesterol), in so called glycolipid rafts. The importance of these structures remains unclear.

The GPI abnormality in PNH

Affected cells in PNH synthesise little or no GPI anchor. This results from a failure in the first step in the synthetic process—the addition of GlcNAc to PI. This has been demonstrated using different techniques by several workers. Biochemical studies using labelled precursors show an almost complete lack of intermediates containing mannose or glucosamine, indicating that the block in the pathway is at this first stage.36 Murine cell lines incapable of synthesising GPI structures have been studied. In these experiments it could be shown that lines in which the defect occurred at different points in the pathway complemented one another (restored the synthesis when fused together). Three lines, classes A, C, and H, were individually unable to add GlcNAc to PI but when fused together they complemented one another.37, 38 This implied that there were at least three gene products controlling this step. When PNH cells from patients are fused with these lines, they always complement cells of classes C and H but never those of class A. Thus, it became clear that the defect in PNH was always the same as that found in class A cells and led to a failure in the first step in GPI synthesis.39 This led to the subsequent discovery and expression cloning of the gene involved, which was termed PIG-A (phosphatidylinositol glycan complementation class A).16 It appears that the product of PIG-A, along with at least three other gene products (PIG-C, PIG-H, and hGP1), form the enzyme complex responsible for the transfer of GlcNAc to PI (R Watanabe et al. In: Proceedings of the International Symposium on Glycosyltransferases and cellular communication, 1997, Osaka, Japan, abstract 6).

GPI linked proteins in PNH

If no GPI molecule is produced then the unlinked proteins are degraded in the ER and are absent from the cell surface.40 Some can still be expressed in an alternative transmembrane bound form (for example, FcχRIII/CD16 or LFA-3/CD58), but whether they retain the same function is not clear.41, 42 A further complicating factor in the analysis of surface phenotype in PNH is that some patients can synthesise small quantities of GPI anchor and there appears to be competition between proteins for this residue, leading to partial expression of certain molecules and complete absence of others. This is best observed in red blood cells that are divided into three types on this basis by flow cytometry.43 Type I are normal in their surface expression, type II show reduced but detectable amounts of GPI linked proteins, and type III are completely deficient. Patients with florid haemolysis usually have a large proportion of type III cells, whereas those with non-haemolytic PNH in association with overt aplasia may have either type II cells or small type III clones.24 Type II cells probably arise in patients with PIG-A mutations that allow a small amount of residual protein to be produced—for example, some missense point mutations.44

A wide range of proteins use the GPI linkage mechanism for cell surface attachment. There is no obvious similarity between them, belonging as they do to different functional groups. They include complement defence proteins, enzymes, blood group antigens, adhesion molecules, cell receptors, and others of unknown function.27, 45 If the proteins are normally expressed on haemopoietic cell lineages then they are absent in the cells of the PNH clone. Table 1 illustrates the diversity of proteins that have been described, although it is by no means exhaustive.

Table 1

The range of glycosyl phosphatidylinositol (GPI) linked proteins

“The high degree of interspecies conservation of the glycosyl phosphatidylinositol structure and its wide distribution argues for an important biological role for this mechanism of protein attachment”

There is clearly a link between the clinical features of PNH and certain specific proteins lost through GPI synthetic failure. Understanding this would require a detailed understanding of the pathogenesis of the disease, which is at present incomplete. It would also require a full knowledge of the function of the proteins that are absent, which is also beyond us at present. Despite these problems, there are some associations that are quite well characterised and serve to whet our appetite. Most prominent among these is the intravascular haemolysis from which PNH derived its name. This is caused by a lack of one or more complement defence proteins from the red blood cell surface, allowing inappropriate and unopposed activation of complement. The candidate molecules include CD59 and CD55. Evidence from clinical studies, antibody blocking experiments, and gene transfer protocols favours CD59 as the most important molecule in this process, but a contribution from others cannot be ruled out.12, 13, 46 The thrombotic tendency in PNH is less well understood but may also result from CD59 deficiency.47 In this scenario, complement activation on platelet surfaces leads to increased exovesiculation, which exposes phospholipid as a site for thrombin generation. The biological consequences of the loss of the myriad of other GPI linked proteins from cells is not clear. In many instances, because of an element of redundancy in most biological pathways, there may be no sequelae, and in others the abnormalities may be subtle. The cause of the most intriguing feature of PNH clones—their relative growth advantage in a damaged bone marrow—remains obscure. It is possible that altered surface protein expression affects cellular responses or localisation within the microenvironment and the consequent alteration in the biology of the PNH cells allows them to evade the ongoing marrow insult and prosper in comparison with their normal counterparts. Our group has reported an example of abnormal localisation of haemopoietic stem cells in PNH. Most precursors in the peripheral blood of patients with PNH had the normal (non-PNH) phenotype, despite most bone marrow based progenitors being “PNH” in these individuals. We went on to show that treatment of these patients with granulocyte colony stimulating factor released largely PNH stem cells into the blood—altering the steady state.48 Elucidation of the specific (presumably GPI linked) mechanisms by which this and other cellular interactions occur is the Holy Grail of PNH research. The answers may also have much to tell us about haemopoietic stem cell biology, the pathogenesis of bone marrow failure states, and the pathophysiology of autoimmunity.

The PIG-A gene

In 1993, Miyata et al described the expression cloning of the PIG-A gene by the correction of a GPI deficient murine cell line, which had a similar GPI biosynthetic defect to that observed in PNH.16 The PIG-A gene is somatically mutated in all cases of PNH, presumably because it is the only gene of the GPI biosynthetic pathway that is found on the X chromosome at Xp22.1.17, 49 This means that a single mutation of the gene on the active X chromosome of a haemopoietic stem cell in women or the only X chromosome in men, will result in the PNH phenotype. The PIG-A gene consists of six exons spanning 17 kb of genomic DNA. It has an open reading frame of 1452 bp encoding a protein of 484 amino acids.17, 49 There is a short 5` non-coding region with the initiation of transcription in exon 2 and a relatively large 3` non-coding region. The PIG-A promoter sequences are characteristic of a housekeeping gene, which presumably reflects the widespread expression of GPI linked proteins in all cell types.

PIG-A mutations in PNH

Because the mutations are somatic in PNH, they are extremely varied, with very few reported more than once.39, 45, 50–63 Approximately two thirds are small insertions or deletions resulting in a frameshift and early termination of transcription. In this circumstance, no active PIG-A product is produced and the PNH cells are completely deficient in all GPI linked proteins. The remainder are point mutations and these may result in a complete or partial deficiency of GPI linked proteins. More than 100 PIG-A mutations have now been described and very few are repeated (fig 3).

Figure 3

The PIG-A gene. (A) The genomic structure of the PIG-A locus. The promoter sequences of the PIG-A gene are depicted in greater detail and consist of four CAAT boxes, two AP-2 sequences, and a CRE (cAMP response element) sequence. These features are consistent with the ubiquitous expression of PIG-A. (B) The coding region of PIG-A. The reported point mutations are depicted by the arrows: the open arrows indicate missense mutation, the solid arrows are nonsense mutations, and the hatched arrows are at the site of a polymorphism that does not affect the code and has been reported in normal individuals. Codons 48 and 128 are affected by several different point mutations, indicating that these are potentially crucial areas of the PIG-A protein. The solid bar demonstrates the region of homology between PIG-A and several other glycosyltransferases, suggesting that it may be the binding site for N-acetylglucosamine. One point mutation has been reported in this domain which resulted in a relatively neutral amino acid substitution (asparagine to aspartic acid) and a partial deficiency of glycosyl phosphatidylinositol linked antigens in this patient.

In over half of affected patients, flow cytometric analysis of the red blood cells identifies two discrete populations of PNH cells (type III, with complete deficiency; and type II, with partial deficiency of GPI linked antigens). This indicates that there are at least two unrelated PNH clones in these patients. Several groups have now described patients with more than one PNH clone at a molecular level; in fact, as many as four separate GPI deficient clones with different PIG-A mutations have been identified in a single patient.64 In most cases, the individual clones occur at the same time, but one patient was studied before and many years after a bone marrow transplant, at the time of relapse, and the PIG-A mutations at relapse were different to those of the original disease.65 In many of the cases with more than one mutated clone, red blood cell flow cytometry only identifies a single PNH population. Thus, it appears that most patients have multiple PNH clones. This indicates that patients are permissive for the development and/or expansion of PNH clones and therefore suggests a factor extrinsic to the PNH clone(s) that favours their development.

Is a PIG-A deficient clone sufficient to result in PNH?

The PIG-A gene is essential for embryogenesis and therefore mice that have “knocked out” PIG-A genes are not viable. When PIG-A deficient embryonic stem cells are microinjected into murine blastocysts, chimaeric mice are occasionally produced but only have a small number of cells derived from the PIG-A negative embryonic stem cells.66, 67 The deficient stem cells contribute to the haemopoietic compartment of the resulting chimaeric mice but their proportion in any individual mouse remains constant with time. In addition, when mice are produced with higher proportions of GPI deficient haemopoietic cells, by using the Cre-Lox P system and/or by transplantation experiments, the proportion of GPI deficient haemopoietic cells remains constant over time.68 These findings show that PNH cells do not have a growth advantage over their normal counterparts in mice without bone marrow failure.

“The cause of the most intriguing feature of paroxysmal nocturnal haemoglobinuria clones—their relative growth advantage in a damaged bone marrow—remains obscure”

Araten and his colleagues recently reported the presence of rare GPI deficient neutrophils in normal individuals.69 These cells have a frequency of 10–51/million cells and when collected by flow sorting were shown to contain mutations of the PIG-A gene. These findings show that such mutations exist frequently among normal individuals, but this alone is not sufficient for the development of PNH.


We may now put forward a general outline of the factors leading to PNH and of its interplay with aplasia. The dual pathogenesis model appears to withstand the rigours of both time and experimental research. In this hypothesis, somatic mutations in PIG-A lead to PNH only if the affected cells are in a bone marrow under hypoplastic stress. The mutation and the abnormal bone marrow environment are both required. PNH cells possess only a relative growth advantage and will not prosper in a normal bone marrow.

Take home messages

  • It appears that the development of paroxysmal nocturnal haemoglobinuria (PNH) requires two coincident factors: somatic mutation of the PIG-A gene and a hypoplastic bone marrow environment

  • The gene mutation results in failure of synthesis of the glycosyl phosphatidylinositol (GPI) anchor, without which many proteins are unable to attach to cell surfaces

  • It is thought that these alterations in surface antigen composition give the PNH cells a relative growth advantage in a hypoplastic marrow, although the precise pathophysiological mechanisms are unclear

  • The failure of GPI production results in the loss of complement defence proteins from red blood cells, which explains the classic feature of intravascular complement mediated haemolysis

  • There is indirect evidence that platelet activation with consequent thrombosis is caused by a similar mechanism

If this is correct, then certain observations would be expected. Hypoplasia should be found invariably. This clinical association has of course been long recognised. Some degree of single or multiple cytopenia can be found in up to 80% of cases and in others there is laboratory evidence of diminished progenitor growth potential.20, 70 It appears then, that hypoplasia and PNH do indeed go hand in hand. For dual pathogenesis, one would require PIG-A mutations to be frequent in normal individuals; otherwise the coincidence of such an occurrence in the rare disease of aplasia would be extraordinary. This was in fact suspected from the presence of multiple, separate mutations in certain patients with PNH and has now been confirmed by observations from several authors.70, 71 So it appears that the concept of dual pathogenesis accounts for the observed facts.

The mechanism by which aplasia imparts a relative growth advantage to PNH cells is more speculative. Although few doubt the immune component in aplasia, it is unclear whether this is a primary alteration in the stem cell pool against which an immune response develops or an autoreactive state, where otherwise normal stem cells are targeted. In either scenario, the PNH cells may prosper by evading the immune destruction. This could be mediated directly by loss of a GPI linked “recognition molecule” or occur through altered biology or localisation of the cells through GPI linked mechanisms. The answers to these questions will not only enlighten PNH research but may greatly enhance our understanding of aplasia and stem cell behaviour.