• Peter Hillmen,

    1. Haematological Malignancy Diagnostic Service, Algernon Firth Building, Leeds General Infirmary, Great George Street, Leeds LS1 3EX, UK
    Search for more papers by this author
  • Stephen J. Richards

    1. Haematological Malignancy Diagnostic Service, Algernon Firth Building, Leeds General Infirmary, Great George Street, Leeds LS1 3EX, UK
    Search for more papers by this author

Dr Peter Hillmen Haematological Malignancy Diagnostic Service, Algernon Firth Building, Leeds General Infirmary, Great George Street, Leeds LS 1 3EX, UK.

Paroxysmal nocturnal haemoglobinuria (PNH) has been recognized as a discrete disease entity since 1882. Progress in understanding the mechanism underlying the clinical features of PNH (i.e. intravascular haemolysis, thrombosis, etc.) was initially slow but has accelerated rapidly over the last decade or so. We are now in a position to explain most of the clinical features of the disorder, but our new insights into PNH have uncovered far more intriguing questions regarding the pathophysiology of the disease itself. It appears that a full understanding of the reason why PNH occurs at all will provide unique insights into the pathophysiology of aplastic anaemia and perhaps other bone marrow failure syndromes, such as myelodysplastic syndrome, as well as revealing much about normal haematopoiesis, normal lymphopoiesis and the disturbance of the immune system in ‘autoimmune conditions’. In this review, we will summarize the important advances in the understanding of PNH but will concentrate mainly on the impact these discoveries are having on the diagnosis and management of patients with PNH, as well as attempting to outline the current position on the answers to the questions: ‘Why does PNH occur and why does it matter?’


PNH was first described as a discrete clinical entity in 1882 ( Crosby, 1951), which was a full 50 years before the recognition that the red cells in PNH had an increased sensitivity to lysis in acidified serum. This in turn led to the first reliable diagnostic test for PNH — the acidified-serum lysis test or Ham test ( Ham & Dingle, 1939). The Ham test has remained the gold standard for establishing the diagnosis of PNH until the advent of flow cytometry in the last few years. It soon became clear that the reason for the increased lysis of PNH red cells was that they have a markedly increased sensitivity to activated complement. In the 1950s, it was recognized for the first time that the neutrophils were also abnormal in PNH ( Beck & Valentine, 1951; Lewis & Dacie, 1965) and, subsequently, that the other haematopoietic cell lineages were also affected. Dacie (1963) first suggested that PNH was an acquired clonal disorder resulting from a somatic mutation in a haematopoietic stem cell. The demonstration in two G6PD heterozygote women with PNH that only one G6PD variant enzyme was present in the PNH red cells, whereas both variants were present in the residual normal red cells, provided conclusive evidence of the clonal nature of PNH ( Oni et al, 1970 ).

Over the last two decades, there has been an ever-increasing list of cell-surface antigens that have been reported as deficient from PNH cells. To date, over 20 different antigens have been described that are missing from PNH cells ( Table I). In 1980, it was first suggested that some antigens may be held onto the cell membrane via a glycolipid structure ( Low & Zilversmit, 1980), and the structure of this glycosylphosphatidylinositol (GPI) anchor has subsequently been defined ( Ferguson, 1992) (Fig 1). Proteins that are destined to be GPI linked have the anchor covalently attached as a post-translational processing step. It soon became apparent that all the proteins that were absent from PNH cells were normally GPI linked and that failure in the biosynthesis of the GPI structure was the most likely cause of the PNH abnormality.

Table 1. Table I. GPI-anchored proteins on haematopoietic cells. Bold, useful for flow cytometry in the diagnosis and monitoring of patients with PNH.* Both transmembrane and GPI-linked forms expressed on this cell type.† Expressed on B-cells and activated T-cells only.‡ Expressed on a proportion of T-cells and B-cells.§ Expressed on activated T-cells and natural killer cells.¶ Expressed on a proportion of stem cells only. Thumbnail image of
Figure 1.

and R2 represent different fatty acid chains that are inserted into the cell membrane. The protein is attached, via its carboxyl terminal, to the preformed GPI core in the rough endoplasmic reticulum.


Biochemical defect

The identification that the PNH defect was likely to be a complete or partial deficiency of GPI-linked proteins led to frenzied activity among several groups throughout the world in an attempt to identify the underlying abnormality. The development of cloned cell lines with the PNH abnormality (both B-cell and T-cell lines) facilitated the rapid elucidation of the defect ( Schubert et al, 1990 ; Hillmen et al, 1993a ; Nakakuma et al, 1994 ). The GPI biosynthetic pathway in PNH cells was examined and found to be abnormal (Fig 2). The same step of the pathway was found to be defective in all patients studied — namely, the transfer of N-acetylglucosamine to phosphatidylinositol (the first step in the pathway; Armstrong et al, 1992 ; Hillmen et al, 1993b ; Takahashi et al, 1993 ). In addition, the same biosynthetic abnormality was identified in both PNH cells that had a complete deficiency of GPI-linked proteins (PNH type III red cells) and those that were partially deficient(PNH type II red cells; Hillmen et al, 1993b ).

Figure 2.

genes are used in the pathway (for example, PIG-A, PIG-C, PIG-H and GPI1 are all required for the first step), but all cases of PNH thus far described result from the disruption of the PIG-A gene (see text).

The PIG-A gene

It appears that the products of four different genes are required for the first step in the GPI biosynthetic pathway. Miyata et al (1993 ) isolated one of these genes by complementation (correction) of the GPI defect in a cell line that had been selected in vitro to have a GPI-deficient phenotype and in which the defect had previously been shown to be at the same step in GPI biosynthesis as the PNH defect. This gene, named PIG-A (phosphatidylinositol-glycan complementation class A; see Table II), was transfected into GPI-deficient cell lines derived from PNH patients and completely restored the expression of all GPI-linked proteins in all cases tested ( Takeda et al, 1993 ; Bessler et al, 1994 a). Takeda et al (1993 ) were the first to describe acquired PIG-A mutations in patients with PNH and, since then, mutations of the PIG-A gene have been identified in large numbers of patients from all over the world ( Bessler et al, 1994 a; Miyata et al, 1994 ; Ware et al, 1994 ; Nagarajan et al, 1995 ; Pramoonjago et al, 1995 ; Yamada et al, 1995 ; Savoia et al, 1996 ; Lin et al, 1997 ; Pavlu et al, 1997 ). Abnormalities of the other genes, the products of which are involved in the GPI biosynthetic pathway, have not been described in PNH. Why should mutations of the PIG-A gene be responsible for all cases of PNH? The answer would appear to lie in the chromosomal localization of the various genes. The only gene that appears to situated on the X chromosome is PIG-A, with the remainder being autosomal ( Kinoshita et al, 1997 ). Thus, there is only one functional PIG-A gene in any diploid cell (including females because of X inactivation). In contrast, all the other GPI biosynthetic genes appear to be located on autosomes and, therefore, have two active copies. Thus, a single mutation of the active PIG-A would lead to a GPI-deficient phenotype, whereas for the other genes, mutations of both alleles would be necessary to produce the GPI-deficient phenotype.

Table 2. Table II. Characteristics of PIG-A. Thumbnail image of

PIG-A mutations in PNH

The somatic mutations of the PIG-A gene that result in the PNH phenotype are varied and are largely distributed throughout the entire coding region of the gene. The mutations can be deletions, insertions or point mutations, which may be missense or nonsense mutations. There appears to be no difference in the pattern of mutations between different clinical subtypes of PNH, i.e. primarily haemolytic vs. primarily hypoplastic PNH. The majority of the mutations, approximately two-thirds, are insertions or deletions. Most of these are small (1 or 2 bp), result in a frameshift in the coding region and therefore a shortened non-functional product. These frameshift mutations are typically seen in patients whose red cells are completely deficient in GPI-linked proteins (type III red cells). A minority of the mutations are point mutations, most of which are missense. Point mutations are found in either completely deficient PNH cells (type III) or partially deficient (type II) cells. Presumably in type II cells, the product of the mutated PIG-A gene has some residual activity, allowing a small quantity of GPI anchor to be produced.

Clinical features

There have been two recent papers reporting the natural history of groups of patients with PNH over prolonged follow-up ( Hillmen et al, 1995 ; Socie et al, 1996 ). These studies have clarified the frequency of some of the complications of PNH and of the long-term outcomes. Clearly, this information is of great importance when electing therapeutic options for individual patients, especially relatively high-risk therapies such as bone marrow transplantation. The median survival in PNH is between 10 and 15 years from the time of diagnosis.

The most frequent and sinister complication of PNH is venous thrombosis, which occurs in up to half of patients with haemolytic disease and is the cause of death in a third. Patients with PNH/aplastic anaemia (PNH/AA) appear to have a much lower risk of thrombosis, reported to be in the region of 5% ( Fujioka & Asai, 1989; Dunn et al, 1991 ). The most obvious difference between haemolytic PNH and PNH/AA is that, in the latter, the proportion of PNH cells is smaller (usually less than 20% of PNH neutrophils as opposed to more than 50% in most patients with haemolytic disease). In view of the high risk of thrombosis, which may occur without warning, a policy of routine prophylactic anticoagulation in patients with large PNH populations and with no contraindication to anticoagulation (such as severe thrombocytopenia) has the potential to prevent potentially life-threatening thromboses and is worthy of serious consideration.

Progressive pancytopenia occurs in a proportion of patients with PNH, and it is unclear whether this is more of a concern in PNH/AA rather than haemolytic PNH. It appears that in the region of 10% of patients will die from aplastic anaemia associated with PNH. The reversal of pancytopenia in PNH in response to immunosuppression (antilymphocyte globulin and/or cyclosporin A) appears to be at least as good as the responses observed in uncomplicated aplastic anaemia ( Ebenbichler et al, 1996 ; Stoppa et al, 1996 ; Paquette et al, 1997 ; Schubert et al, 1997 ).

The risk of leukaemia in PNH has been somewhat overstated in the literature. The incidence of acute myeloid leukaemia (AML) in PNH appears to be similar to the risk of AML in aplastic anaemia, in the region of 5%. It appears likely that aplastic anaemia predisposes to clonal haematopoietic disorders, such as AML, PNH and myelodysplastic syndrome, but that the development of a PNH clone does not increase the risk of AML/MDS. Thus, the GPI-deficient phenotype is not preleukaemic.

Another important finding from these clinical studies is that a significant proportion of patients survive for prolonged periods (≈ 25% surviving over 25 years) and that in the region of 15% will experience a spontaneous recovery from their PNH with no sequelae attributable to their disease ( Hillmen et al, 1995 ).


Ham test

The mainstay for the diagnosis of PNH since the 1930s has been tests that rely on the lysis of PNH red cells by activated complement. Numerous different ways of activating complement through both classical and alternative pathways have been devised, but the gold standard has remained the acidification of serum (the Ham test). The Ham test, when performed scrupulously, remains a specific and relatively sensitive test for haemolytic PNH. There are, however, major problems inherent in this type of test. They are at best semi-quantitative and will not reliably detect small populations of red cells (< 5%). These relatively simple complement-based tests cannot differentiate PNH type III red cells (extremely sensitive to complement) from PNH type II red cells (intermediate sensitivity to complement). The complement lysis sensitivity test ( Rosse, 1973) can differentiate type III from type II red cells but is extremely laborious and is not suitable for routine diagnostic use. The proportion of PNH neutrophils gives a much more accurate estimation of the true size of the PNH clone, because it is not dependent on the degree of haemolysis or recent transfusions, but the Ham test cannot provide information on cell lineages other than the red cells.

Flow cytometry — diagnosis

The use of flow cytometry coupled with specific combinations of monoclonal antibodies has made a significant contribution to the understanding of the biology of PNH. Multiparameter flow cytometry has enhanced our ability to detect small PNH clones within multiple haematopoietic cell lineages, confirming phenotypically the stem cell nature of this disorder. In addition, accurate quantification of clone size is possible, thus providing the opportunity for serial monitoring of clone size and evaluation of correlations with clinical course. The sensitivity of detection, particularly for optimally designed multicolour combinations, allows GPI-deficient clones of around 0.1% to be reliably identified. This increased level of sensitivity in detection may well be reflected in the identification of an increased number of new cases of PNH, particularly among patients with aplastic anaemia who were previously negative by the Ham test. As a consequence, this has helped to clarify further the classification of PNH into two types: (1) haemolytic PNH, characterized by overt episodes of intravascular haemolysis and typically with large PNH clones; and (2) hypoplastic PNH with a clinical picture dominated by cytopenias (neutropenia, thrombocytopenia and/or anaemia) with no overt haemolysis and usually small PNH clones or predominantly PNH type II red cells.

Peripheral blood samples are recommended for screening procedures. For red cell screening, it is important to use antibodies to two different GPI-linked antigens. CD55 and CD59 are the best characterized and provide the clearest definition of type I (normal), type II (partially deficient) and type III (completely deficient) populations (Fig 3). For granulocyte screening, combinations of antibodies against two GPI-linked antigens (e.g. CD55/CD16 or CD59/CD16) plus an additional antibody directed against a non-GPI-linked granulocyte antigen (CD15 or CD33) provide the clearest results (Fig 4). The presence of a residual population of normal granulocytes in the majority of PNH patients provides an internal control. The delineation of type III and type II populations of granulocytes is far less clear than is seen by flow cytometry of red cells.

Figure 3.

Fig 3. Expression of the GPI-linked antigen CD59 on normal red cells (a) and on the red cells from five PNH patients (b–f) by flow cytometry. The histograms show: (a) a normal histogram profile of CD59 expression. All cells are positive and are classified as type I; (b) a bimodal distribution with PNH type III red cells (complete deficiency) and type I red cells (normal); (c) major populations of PNH type II red cells (partial deficiency) and residual normal (type I) red cells with a minor type III population; (d–f) show various distributions of PNH populations. In some patients, the differentiation between the type I, II and III red cells is well demarcated (d), whereas in others, the distinction is less clear (f).

Figure 4.

Fig 4. Identification of GPI-deficient granulocytes by three-colour flow cytometry. Plot (a) shows the gating strategy used to identify neutrophils reliably using CD15 positivity (non-GPI-linked antigen expressed on neutrophils) and high light-side scatter (indicating granular cells). Plot (b) shows the expression of two GPI-linked antigens (CD16 and CD55) on the cells gated as neutrophils from plot (a). The PNH clone comprises 33% and the normal granulocytes 63% of the total.

Flow cytometry — lymphocyte studies

In PNH, the presence of GPI-deficient clones within the platelet, monocyte and granulocyte components of the myeloid lineage is well documented. A small number of investigators have also unequivocally described PNH clones within peripheral blood lymphocyte populations, including B-cells, T-cells and NK cells. Indeed, persistence of GPI-deficient T-lymphocyte populations has been described in patients in long-term clinical remission of PNH with no detectable granulocyte or red cell clones ( Hillmen et al, 1995 ). The absence of GPI-linked antigens from PNH lymphocytes provides a unique model with which to study the kinetics of de novo lymphopoiesis in adults and also to monitor phenotypic changes in residual normal T-cell and B-cell populations ( Richards et al, 1998 )

Flow cytometry — serial monitoring

The clinical relevance of monitoring the size of the PNH clone by flow cytometry is unclear at present. We are currently co-ordinating a PNH registry, which contains over 100 patients, and we are performing repeated flow cytometric analyses on individual PNH patients. It is clear from a preliminary analysis of these data that the size of individual patients' PNH clones is not constant with time. The information obtained from such a prospective study may provide important insights to allow the prediction of an individual's clinical course. The ability to predict which patients will progress to MDS/AML, follow a stable clinical course or undergo spontaneous remission would greatly improve patient management, particularly when considering the decision to perform bone marrow transplantation or deciding which patients should be prophylactically anticoagulated.


Why does PNH occur?

There is now conclusive evidence that PNH only develops in individuals who have a predisposition to the development or, more likely, the expansion of GPI-deficient haematopoietic clones. Thus, the development of PNH requires at least two events: (1) a somatic mutation in a haematopoietic stem cell, affecting the PIG-A gene in all cases thus far described, which results in a GPI-deficient clone; and (2) selection in favour of the PNH clone to allow it to proliferate preferentially compared with the residual haematopoiesis. This dual pathogenesis theory for the development of PNH was first suggested by Dacie (1980) and clearly defined by Rotoli & Luzzatto (1989).

What is the evidence to support the dual pathogenesis of PNH hypothesis?

Both PNH and aplastic anaemia are uncommon disorders with a prevalence of only a few cases per million individuals. Therefore, the fact that PNH clones are reported in the region of a quarter to a half of patients with aplastic anaemia cannot be mere coincidence ( Tichelli et al, 1988 ; Schrezenmeier et al, 1995 ). In addition, there is evidence that, even in patients with PNH who do not have a history of antecedent aplastic anaemia, there is an underlying bone marrow failure syndrome. Thrombocytopenia and/or neutropenia is present in over 80% of cases of PNH, suggesting an underlying hypoplasia (unpublished observation). When haematopoietic cell cultures are performed in patients with PNH [either short-term (BFU-E or CFU-GM) or long-term], the number of colonies produced is an order of magnitude less than normal ( Rotoli & Luzzatto, 1982; Maciejewski et al, 1997 ). In addition, if the normal cells are separated from the PNH cells before culture, both components are equally poor at producing colonies. Thus, bone marrow hypoplasia, in the form of aplastic anaemia, appears to be permissive for GPI-deficient or PNH clones.

The analysis of GPI-anchored molecules on the surface of the haematopoietic cells, particularly the red cells, in PNH reveals that ≈ 50% of patients have a trimodal distribution: a completely deficient population (PNH type III cells), a partially deficient population (PNH type II cells) and a residual normal population. Thus, even by the rather insensitive technique of flow cytometry, almost half of patients appear to have more than one PNH population. Analysis of the PIG-A mutations in PNH has uncovered many individuals with two or more independent PNH clones (different discrete PIG-A mutations), and individuals with as many as four independent GPI-deficient clones have been described ( Bessler et al, 1994b ; Yamada et al, 1995 ; Endo et al, 1996 ; Nishimura et al, 1997 ). Many of these patients with multiple clones appear to have only a single population of PNH red cells by flow cytometry. Thus, there is convincing evidence that there are multiple clones in the majority of patients with PNH. The elegant study of a patient with PNH who underwent a syngeneic bone marrow transplant (BMT) and subsequently relapsed has provided further insights into the pathophysiology of PNH ( Nafa et al, 1998 ). The patient underwent a syngeneic BMT without conditioning for PNH in 1973 and had a good response. He had a laboratory relapse of PNH in 1983 and a clinical relapse in 1987. The authors identified the PIG-A mutation at relapse and showed that this was entirely different compared with the PIG-A mutations subsequently identified from material stored before the transplant. Thus, this patient must have had a tendency to develop PNH clones. Therefore, PNH is a clonal disorder (a somatic mutation leads to the clonal proliferation of haematopoietic cells), but several clones usually co-exist — PNH is frequently not a monoclonal disorder.

Is the dual pathogenesis model of PNH supported by animal models?

Several groups have attempted to produce mice with the PIG-A gene disrupted. No mouse has been produced without PIG-A activity, and it appears that a functional PIG-A, and presumably therefore GPI-linked proteins, are necessary for successful fertilization and/or fetal development. Mice that are chimaeric for a non-functional PIG-A gene have been produced, and the proportion of GPI-deficient cells either remains stable or decreases with time ( Kawagoe et al, 1996 ; Rosti et al, 1997 ). The technique of spatio-temporally controlled site-specific mutagenesis, which inactivates the PIG-A gene early in embryogenesis, has been used to create mice with up to 53% GPI-deficient red cells. The proportion of PNH red cells, granulocytes and lymphocytes falls over the first 4 months of the mouse's life but then remains stable with time ( Dominguez et al, 1998 ). Thus, GPI-deficient cells do not have a growth advantage over the residual normal cells in animals without bone marrow failure.

However, when bone marrow, either from normal individuals or from patients with PNH, was transplanted into sublethally irradiated SCID mice and the mice treated with human cytokines, a rather surprising observation was made ( Iwamoto et al, 1996 ). After 7 months, the mice transplanted with normal bone marrow had no detectable human cells. In contrast, human cells were detectable in the erythroid, myeloid and lymphoid compartments in the mice transplanted with PNH bone marrow, and these cells all had a GPI-deficient phenotype. At first glance, this finding may appear somewhat surprising, as we have already described that the bone marrow from patients with PNH has an impaired ability to form haematopoietic colonies in vitro. The explanation may well be that, because of the sublethal irradiation treatment given before bone marrow infusion, the mice were rendered hypoplastic and, therefore, the human cells were transferred into an environment that selects in favour of cells with a GPI-deficient phenotype — analogous to the dual pathogenesis of PNH hypothesis.

GPI-deficient cells in ‘normal’ individuals

In view of the high frequency of PNH in aplastic anaemia, it could be predicted that the PIG-A mutations are common occurrences, but that they only develop into sizeable clones and therefore clinically apparent PNH when there is selection in their favour. Therefore, it would be predicted that PIG-A mutations are present in a high proportion of the population. If this is true, why have they not been found? The main problems are that, without the selective advantage of aplastic anaemia, only a very low proportion of cells are GPI deficient and, because the PIG-A mutations are acquired somatic mutations, they are all different. Therefore, it is not trivial to design a method of detection for an unknown mutation that may be anywhere within the PIG-A gene and is present at a very low copy number. Indirect evidence that PIG-A mutations are present in many individuals has been provided by the observation that, if patients are treated with a monoclonal antibody against a GPI-linked molecule, namely CAMPATH-1H, which is directed against the GPI-linked molecule CD52, then many of these individuals develop GPI-deficient T-lymphocytes ( Hertenstein et al, 1995 ; Taylor et al, 1997 ; Rawstron et al, 1999 ). In one case, the PIG-A mutation, which was subsequently identified, was detected in the patient's mononuclear cells harvested before CAMPATH-1H therapy ( Rawstron et al, 1999 ). Thus, intense selection against a single GPI-linked antigen allows the clonal expansion of pre-existing PNH-like T-cells in the majority of individuals treated. This indicates that PIG-A mutations are indeed present in many individuals at extremely low levels. The presence of extremely low levels of GPI-deficient neutrophils in eight out of eight normal individuals has been described recently, and the PIG-A mutations have been identified in the majority of these cases ( Araten et al, 1999 ). Thus, PNH-like cells are present in the majority of normal individuals, as predicted by the dual pathogenesis theory of PNH.

Why does spontaneous recovery occur in PNH?

In almost all cases of PNH, there remains some normal haematopoiesis, even if this is at a markedly reduced level. There is evidence that the cytopenias associated with aplastic anaemia improve with prolonged follow-up, presumably as a result of a reduction in the underlying aplastic process. A possible hypothesis to explain the spontaneous recovery occasionally seen in PNH is that the aplastic process that is positively selecting for PNH clones reduces in intensity with time. At some point, the selection in favour of the PNH clone will lessen, and the advantage in haematopoiesis will swing towards the residual normal cells. Thus, the proportion of PNH cells will decrease and eventually disappear as the bone marrow function returns to normal. This is exactly what is seen in patients with PNH who undergo spontaneous recovery. Thus, it appears that the hypothesized dual pathogenesis of PNH explains many, if not all, of the clinical and laboratory features of PNH (Fig 5).

Figure 5.

Fig 5. Diagram depicting the correlation between dual pathogenesis of PNH and the clinical features of PNH. The vertical axis portrays the proliferative activity of normal (grey line) and PNH (black line). The horizontal axis reflects the intensity of the aplastic process (the second component of PNH). The PNH clone does not have a growth advantage when the aplastic process is minimal but, as the intensity of bone marrow suppression increases, the PNH cells are relatively spared and acquire a growth advantage over the residual normal haematopoiesis. The theory explains the mechanism for the spontaneous remission of PNH as, when the aplastic process reduces the PNH, cells lose their advantage, and the residual normal cells proliferate. Occasional patients experience progressive aplasia with loss of the PNH clone, and this is explained because, when the aplastic process is maximal, both normal and PNH haematopoiesis is suppressed. The intensity of bone marrow failure and the proliferative activity of the PNH clone in an individual patient will determine whether the size of their PNH population increases, remains stable or decreases after immune suppression.

The evidence outlined above indicates that the growth advantage of GPI-deficient cells in PNH is relative rather than absolute and that it is not intrinsic to the PNH stem cell but results from escape from some extrinsic process. Thus, there is overwhelming evidence that the dual pathogenesis theory for the pathophysiology of PNH is true but ‘what is the selective mechanism?’ The answer to this intriguing question remains obscure and, in order to attempt to answer it, one must understand the mechanism for aplastic anaemia. The overwhelming evidence supports the view that aplastic anaemia is caused by an immune-mediated attack against haematopoietic stem cells. It is possible that the stem cells in aplastic anaemia are altered in some way, which results in an abnormal recognition of them by the immune system, which in turn results in selection against the stem cells. The other possibility is that there is simply an autoreactive immune attack against normal stem cells. Thus, there is clearly an immune component to aplastic anaemia, but it is unclear whether there is also a stem cell defect. There are several conceivable mechanisms as to how a GPI-deficient clone may have an advantage in an aplastic milieu.

(a) Immune-mediated aplastic process is directed through GPI-linked proteins

It is possible that the aplastic process is directed through one or more GPI-linked proteins and that, therefore, PNH stem cells are not subject to this insult. This mechanism appears to be unlikely, as PNH can occur in aplastic anaemia as a result of several different causes, such as post-hepatitic, following drug or chemical exposure, in Fanconi anaemia, etc. It appears unlikely that all these aplasias are caused by a common mechanism directed through a single GPI-linked molecule.

(b) Failure of cytotoxic T-cells to identify or kill PNH cells

It is possible that the cytotoxic T-cells that appear to be at least partly responsible for aplastic anaemia can recognize PNH stem cells but may be unable to kill them efficiently, perhaps because of a disruption of one or more of the co-stimulatory pathways that are necessary for effective T-cell function. It is possible that some of these pathways involve molecules that are GPI linked.

(c) Altered biology of the PNH stem cell

There is increasing evidence that the physiology of PNH stem cells is not normal. Analysis of patients with PNH demonstrates that, even when the vast majority of haematopoiesis is derived from the PNH clone (> 95% bone marrow stem cells are GPI deficient), the vast majority, if not all, of the most primitive (CD34+ , Thy-1+) circulating stem cells are derived from the residual normal haematopoiesis ( Prince et al, 1995 ; Johnson et al, 1998 ). When patients are treated with recombinant granulocyte colony-stimulating factor (G-CSF), the stem cells that then appear in the peripheral blood almost all have a PNH phenotype (similar to the pattern seen in the bone marrow). The reason for this abnormal behaviour is unclear but may well contribute to the reason why PNH cells escape the aplastic process. For example, the PNH stem cells may be partially protected from attack if they happen not to be in the wrong place at the wrong time!


Implications for PNH

Although PNH is an uncommon disease, it is a chronic condition, which frequently affects relatively young individuals and often presents extremely difficult management problems, which may persist for many years. Most haematologists who have had a ‘difficult’ patient under their care can immediately recall the problem for the remainder of their career! As PNH is unique in many ways — it is an acquired single gene disorder — it appears to be an attractive entity to consider for gene therapy initiatives. In order to consider such novel and potentially hazardous therapeutic options, we must be able to answer two vital questions: (1) which patients are destined to do poorly — we do not want to harm a patient who may remit spontaneously; and (2) what is the mechanism for the growth advantage of PNH clones — a PNH cell that is ‘corrected’ to normal by the introduction of a normal PIG-A gene will then express all the missing GPI-linked proteins and may then be selected against by the aplastic process. It must be remembered that PNH can be considered as a ‘natural form’ of gene therapy, in that the PNH cells escape from the aplastic process by virtue of their mutated PIG-A and, therefore, correcting the genetic abnormality is nonsensical — two wrongs do not make a right!

Implications for aplastic anaemia and other bone marrow failure syndromes

An understanding of the pathogenesis of PNH will provide a unique insight into the cause of aplastic anaemia. It is unclear whether aplastic anaemia is a single entity with a common pathway of stem cell insult or whether there are several different mechanisms of stem cell suppression. The fact that up to a half of patients with aplastic anaemia develop PNH clones indicates that a large proportion of cases have a similar common final effector mechanism. Fully understanding the growth advantage of PNH clones in aplastic anaemia will not only allow a fuller understanding of the pathophysiology of aplastic anaemia but may, in turn, permit the development of strategies to ameliorate the condition. For example, if the process is mediated through a single GPI-linked protein, then blocking this antigen could prevent or diminish the immune-mediated suppression of normal stem cells and, hence, ameliorate the pancytopenia. In other words, we would take advantage of the growth advantage of PNH clones without the clinical problems associated with PNH (unless CD59 is the implicated antigen). In view of the increasing evidence to support the hypothesis that there may be a pathogenetic link between aplastic anaemia and other bone marrow failure syndromes, such as myelodysplastic syndrome, it is conceivable that any novel therapeutic strategies developed could be effective in other associated bone marrow failure conditions.

Implications for stem cell biology

The PNH stem cell appears to behave in an abnormal fashion. The most immature PNH stem cells do not appear to circulate normally and thus may have an altered adhesion/association with bone marrow stroma. This may well be important for the pathogenesis of PNH, but may also allow important insights into the function/biology of normal stem cells. If the adhesion of stem cells to bone marrow stroma is mediated, at least partially, through one or more GPI-linked proteins, then fully understanding this mechanism may provide novel strategies to facilitate the ‘mobilization’ of normal stem cells.

Implications for autoimmunity

PNH cells escape an immune-mediated attack, and one possible mechanism appears to be that they avoid T-cell cytotoxicity, perhaps because of a disturbance of one of the co-stimulatory pathways. Clearly, a full understanding of these pathways should provide insights into the function of T-cells and perhaps how the system breaks down in autoimmune disorders.


There has been an incredible increase in our understanding of PNH over the last few years, and what we have learnt has merely uncovered more intriguing questions to be answered. A fuller understanding of the pathophysiology of PNH may have far wider implications than for PNH itself. The critical analysis of the expression of GPI-linked proteins for the diagnosis and monitoring of PNH in a large group of patients may well allow the prediction of an individual patient's prognosis and therefore facilitate informed individualized clinical decision-making.


To register a patient on the PNH registry and to have flow cytometry performed, please send 5–10 ml of peripheral blood in EDTA by first class post to the Haematological Malignancy Diagnostic Service laboratory (address below).


The authors would like to thank the other members of the Haematological Malignancy Diagnostic Service laboratory for their continued laboratory support. We also thank Cymbus Bioscience Limited for their invaluable support of the PNH registry, and the clinicians from throughout the United Kingdom who have entered patients onto the registry.