Relationship between bone marrow failure syndromes and the presence of glycophosphatidyl inositol-anchored proteindeficient

Transcription

1 British Journal of Haematology, 2001, 115, 1015±1022 Relationship between bone marrow failure syndromes and the presence of glycophosphatidyl inositol-anchored proteindeficient clones Jaroslaw P. Maciejewski, 1 Candido Rivera, 2 Hoon Kook, 1 Daniel Dunn 1 and Neal S. Young 1 1 Hematology Branch, National Heart Lung and Blood Institute and 2 Department of Laboratory Medicine, Clinical Center, National Institutes of Health, Bethesda, MD, USA Received 30 January 2001; accepted 20 July 2001 Summary. Because of the insensitivity of the Ham test, paroxysmal nocturnal haemoglobinuria (PNH) has been inaccurately viewed as a late clonal complication of aplastic anaemia (AA). To clarify the relationship between PNH and marrow failure, we tested for the presence of glycosylphosphatidyl-anchored protein-deficient (GPI-AP) granulocytes in large cohorts of patients with AA, myelodysplasia (MDS), and pure haemolytic PNH. A PNH clone was detected in 32% of new AA patients and 18% of MDS patients. In serial studies, this proportion did not change up to 15 years after diagnosis, suggesting that expansion of aberrant cells is an early event (i.e. prior to initial presentation). For all patients with a PNH clone, on average 14% of PNH granulocytes were found on presentation and 37% at 10 years. Patients with PNH but without cytopenia showed higher percentages of GPI-AP-deficient cells than did those with the AA/PNH syndrome. After immunosuppression, there was no change in the contribution of PNH clone to blood production, arguing against the `immune-escape' theory in PNH. Clinically, a high proportion of GPI-AP-deficient cells correlated with marrow hypercellularity. GPI-AP-deficient cells were similarly present in patients with and without karyotypic abnormalities. Our results indicate that the GPI- AP-deficient clones show quantitative and kinetic differences between classic haemolytic PNH and PNH with marrow failure, in which the evolution rate is low later in the course of the disease. Keywords: myelodysplasia, haemolysis, pancytopenia, haematopoiesis, clonality, bone marrow. The relationship between aplastic anaemia (AA) and paroxysmal nocturnal haemoglobinuria (PNH) has been known for years (Dacie & Lewis, 1961; Dameshek, 1967; Dacie & Lewis, 1972; Young, 1992; Luzzatto et al, 1997; Young & Maciejewski, 1997; Hillmen & Richards, 2000; Young & Maciejewski, 2000) but because of the insensitivity of the Ham test, traditionally used for the diagnosis of PNH, it has been believed that this rare disease is a late clonal complication of AA or even related to the use of immunosuppression (Socie et al, 2000). Flow cytometric diagnosis of PNH has altered the view on the association between these two diseases (Schrezenmeier et al, 1995,2000; Dunn et al, 1999). The presence of PNH clones has been found more often in AA patients than was estimated from the results of the Ham test (Dunn et al, 1999; Richards et al, 2000) but the availability of sensitive tests has raised even more questions as to the nature of the pathophysiological relationship Correspondence: Jaroslaw P. Maciejewski, MD, PhD, Experimental Hematology and Hematopoiesis, Taussig Cancer Center/R40, 9500 Euclid Avenue, Cleveland, OH 44195, USA. maciejj@cc.ccf.org between these two diseases (Luzzatto et al, 1997; Hillmen & Richards, 2000; Young & Maciejewski, 2000). Although its diagnosis is invariably linked to detection of deficiency in membrane expression of glycophosphoinositolanchored proteins (GPI-AP), PNH is a clinically heterogenous disease and can overlap with AA in the PNH±AA syndrome and with myelodysplasia (MDS; Iwanaga et al, 1998; Dunn et al, 1999), as well as occurring in a `pure' haemolytic form without clinical signs of deficient blood cell production or disturbed marrow morphology. Various permutations of the clinical course of PNH are possible including development of pancytopenia after the diagnosis of haemolytic disease or, conversely, evolution of PNH from a typical AA. Once a PIG-A mutation arises in an haematopoietic stem cell (Takeda et al, 1993; Bessler et al, 1994), expansion of a PNH clone has been hypothesized to be a form of immunologically mediated escape (Rotoli & Luzzatto, 1989; Young, 1992; Luzzatto et al, 1997; Young & Maciejewski, 1997). Alternatively, an `apparently abnormal' PNH clone might serve to trigger an aberrant immune reaction leading to immune-mediated bone marrow failure q 2001 US Government 1015

2 1016 J. P. Maciejewski et al (Young & Maciejewski, 2000). Resolving the question as to whether the expansion of a PNH clone precedes bone marrow failure or is its consequence would have important implications for the research in this field. Here, we used a flow cytometry-based test to detect the presence of PNH clones. Owing to several factors (transfusions, destruction of GPI-AP-deficient erythrocytes and transfer of GPI-AP such as CD55 and CD59 to negative erythrocytes), detection of a PNH clone in granulocytes is more accurate then erythrocyte-based assays (Dunn et al, 1999; Piedras & Lopez- Karpovitch, 2000; Richards et al, 2000). Using this precise method, we designed an analytic approach to PNH in which, in addition to categorizing patients according to their primary pathological diagnosis, we studied a cohort of patients selected solely by a singular parameter, the presence on GPI-AP-deficient clone. Using this criterion, we attempted to establish on clinical grounds the pathophysiological relationship between a PNH clone its size and manifestations of marrow failure. In contrast to many other studies that provide a `snap shot' analysis of the frequency of PNH clone in AA, we serially studied a large number of patients with AA for the presence of GPI-AP-deficient clone. MATERIAL AND METHODS Patients. For peripheral blood (PB) collection, informed consent was obtained according to a protocol approved by the Institutional Review Board of the National Heart, Lung and Blood Institute. We analysed PB samples from patients with AA, MDS, PNH and normal healthy volunteers. The diagnosis of AA was established by BM biopsy and PB count criteria according to the International Study of Aplastic Anaemia and Agranulocytosis; severity was classified by the criteria of Camitta et al (1976). For the diagnosis of severe AA (saa), in addition to a hypocellular bone marrow without evidence for karyotypic abnormalities or morphological dysplasia, patients had to fulfil two out of three PB criteria: absolute neutrophil count (ANC), /l; absolute reticulocyte count (ARC) of, /l; and platelet count of, /l of blood. Patients with hypocellular marrow without evidence of dysplasia and with normal karyotype, whose pancytopenia did not fulfil above mentioned severity criteria but with two our three parameters lower than ANC, /l, platelet count, /l and haemoglobin of, 8 gm/dl, were classified as having moderate AA (maa). `Pure' or haemolytic PNH was diagnosed by the presence of GPI-AP-deficient clone, a hypercellular marrow with normal karyotype and lacking dysplasia, and when the platelet count was /l, ANC /l, and the reticulocyte count /l. Furthermore, these patients did not have a past medical history of AA. The diagnosis of MDS was established based on characteristic morphological abnormalities and/or the presence of cytogenetic abnormalities and cytopenia (Bennett et al, 1986). In this study, patients with the presence of PNH and otherwise fulfilling criteria of saa or maa were classified as having AA/PNH syndrome and those with MDS and a PNH clone as MDS/PNH syndrome. At presentation, we analysed a total of 115 AA samples (many of these patients had at least one but, in most cases, two or more measurements performed over a period of up to 4 years after initial evaluation) and 120 samples from patients with MDS. We also have obtained samples from 171 AA patients during their subsequent visits on at least one occasion. In addition, we studied 21 patients fulfilling the diagnosis of `pure' haemolytic PNH as defined above. The control group comprised 73 healthy individuals. The response to the therapy was determined 6 months after treatment and defined as transfusion-independence and improvement in PB counts such that patients no longer fulfilled severity criteria (Rosenfeld et al, 1995). Flow cytometric analysis. Flow cytometry of GPI-APdeficient granulocytes and monocytes has been shown to be superior to the phenotyping of erythrocytes for the detection and quantification of PNH-clone (Piedras & Lopez- Karpovitch, 2000; Richards et al, 2000). Strict criteria to distinguish cells lacking GPI-AP expression were used: granulocytes were identified on the basis of light-scatter properties and the expression of a constitutively expressed granulocyte marker. We used the simultaneous absence of two GPI-AP (CD66b and CD16) that are constitutively expressed at high levels on normal granulocytes to determine PNH phenotype (Dunn et al, 1999). Erythrocytes were lysed by using Q-Prep apparatus (Coulter, Hielah, FL, USA) and were then fixed with paraformaldehyde. We used CD66b-fluorescein isothiocyante (FITC) (Immunotech, Marseille, France) and CD16-phycoerythrin/Cy5 (PECy5) (Caltag, Burlingame, CA, USA), with CD15-PE (Immunotech) as a non-gpi-ap to positively identify granulocytes. Isotypic controls consisted of mouse IgG-FITC (Becton Dickinson, San Diego, CA, USA) and mouse IgG1-PECy5 (Caltag). Nonspecific Fc receptor-mediated binding of conjugated antibodies was blocked by mouse IgG (Caltag). Interpretation of flow cytometry data, diagnostic criteria and statistical analysis. A group of 73 normal individuals was used to establish the mean and range of double-negative cells. However, for the diagnosis of PNH clone in patients, a clearly distinguishable population of cells with a PNH phenotype was required: such a distinct population was seen at levels > 1%, although this value was slightly higher than the mean ^ 2 standard deviation (SD) of the control value. The higher cut-off value of 1% assured a higher specificity at the possible expense of sensitivity. An unpaired (or paired where appropriate) Student t-test was applied to determine whether the means in the percentages of GPI-AP-deficient cells are different between the groups. For not normally distributed samples, an unpaired Wilcoxon test was used. Proportions of patients with and without PNH clone were compared using the x 2 test. Correlations between the variables was determined using linear regression analysis. RESULTS Criteria for the diagnosis of PNH clonal expansion and percentages of GPI-AP-deficient granulocytes in patients For the diagnosis of the PNH phenotype, we selected stringent criteria: percentages above the maximum range observed in normal control subjects (0±0 98%; mean

3 1 2SD ˆ 0 45%; n ˆ 73) were chosen as threshold because only at levels. 1% was a clear distinction of a doublenegative population possible (Fig 1). When the distribution of values, 1% in patients considered negative for PNH was compared with the distribution of control values in the same range, patients showed a more shallow slope of the distribution curve (Fig 1C), indicating that numerically patients categorized as `negative' for PNH showed higher than normal proportions of GPI-AP-deficient granulocytes. Clinical features of patients with bone marrow failure and GPI- AP-deficient clone We identified 136 patients with. 1% of GPI-AP-deficient cells (Fig 1). They were analysed with regard to clinical features and associated diagnoses (Table I): 68% and 16% had AA (AA/PNH), MDS (MDS/PNH), respectively, whereas 15% were primary haemolytic PNH. Among patients with a PNH clone, those with 1±5% of PNH granulocytes were most numerous. However, there was a significant number of patients in whom most granulocytes were derived from the PNH stem cells (Fig 2A±C). Although less frequent, PNH was found in children. Young adults were affected at the highest rate (Fig 1A). There was a correlation between serum lactate Association Between Bone Marrow Failure, PNH and AA 1017 dehydrogenase (LDH) levels and the percentage of PNH cells (r ˆ 0 75, not shown). As expected, the highest proportions of GPI-AP-deficient cells were found in primary PNH. Much lower average percentages were present in AA/ PNH and MDS/PNH (Table I). When patients with a PNH clone were subclassified based on the presence or history of haemolysis (defined as low haptoglobin values, LDH elevation, anaemia and haemoglobinuria), those with haemolysis showed significantly higher percentages of GPI-AP-deficient granulocytes and had less pronounced cytopenias (data not shown). In a total of 136 patients with an expanded PNH clone, 11% had at some point in their disease developed a chromosomal abnormality. When patients with the pure haemolytic form of PNH (n ˆ 21, all by definition had a normal karyotype) were subtracted from total cases, the overall frequency of cytogenetic abnormalities was 13% of PNH-clone bearing patients. In comparison, karyotypic abnormalities developed in 7% of AA cases studied (9% of AA/PNH and 6% of patients with AA without PNH clone) and were found in 44% of patients with MDS (31% in MDS/PNH and 44% of patients with MDS without PNH). Overall, the prevalence of PNH detection in patients with normal chromosomes was not significantly different Fig 1. Percentage of GPI-AP-deficient cells in normal control subjects, patients with AA, MDS and pure haemolytic PNH. (A) Percentages of PNH granulocytes were grouped according to the primary diagnosis. One value per patient is shown. PNH clonal expansion was significantly more frequent in the setting of history of AA compared with MDS (x 2 test, P, ). (B) Frequency of detection of PNH clone and age for patients with. 1% of GPI-AP-deficient cells. Patients with PNH clone were ranked by the age compared with the total number of cases studied (solid line). (C) PNH clone size distribution plots of normal controls and AA patients with, 1% of GPI-AP-deficient granulocytes.

4 1018 J. P. Maciejewski et al Table I. Nosological classification of patients with flow cytometric detection of PNH clone. All patients with PNH clone AA/PNH MDS/PNH Pure haemolytic PNH N (%) 136 (100%) 93 (68%) 22 (16%) 21 (15%) GPI-AP-deficient granulocytes in percentage (median; 95% C.I.) 11 (5 3±18) 7 9 (2 9±14) 4 8 (2±17) 89 (73±94) Age in years (mean; 95% C.I.) 35 (32±39) 54 (48±61) 43 (35±51) ANC 10 9 /L (median; 95% C.I.) 1 (0 76±1 46) 1 6 (0 8±2 7) 1 5 (1 3±2 3) Reticulocytes in percentage (median; 95% C.I.) 1 7 (1 3±2 2) 2 1 (1 4±2 3) 6 9 (6 2±9 8) Platelets 10 9 /L (median; 95% C.I.) 27 (24±40) 37 (37±119) 80 (44±107) Statistical analysis: t-test for comparisons of the percentages of GPI-AP-deficient granulocytes between the groups: AA/PNH versus MDS/ PNH, ns; AA/PNH versus PNH, P, 0 01; MDS/PNH versus PNH, P Fig 2. Characteristics of patients with an expanded PNH clone. All patients with an expanded PNH clone were subdivided according to the diagnosis and analysed in (A) (AA/PNH), (B) (MDS/PNH) and (C) (Primary PNH). Left panels: frequency distribution diagram of AA/PNH patients ranked by the size of PNH clone. Right panels: frequency distribution diagram of AA/PNH patients ranked according to age.

5 (P ˆ 0 07) from those found in patients with abnormal karyotype: a PNH clone was present in 22% of all patients with karyotypic abnormalities (n ˆ 68) and in 34% of patients with a normal karyotype. Classification of patients with GPI-AP-deficient clonal expansion in relation to their primary diagnosis In the group of patients with a PNH clone, the proportions of patients with history of AA or MDS might be skewed by the patients spectrum seen at our institution. Therefore, we determined the prevalence of a GPI-AP-deficient clone in AA and MDS. In AA and MDS, a PNH clone was diagnosed in 42% and 18% of patients respectively (Fig 1). Previous medical history suggestive of secondary MDS after initial PNH or AA was found in nine out of 22 patients with the MDS/PNH syndrome, thus decreasing the prevalence of PNH within true primary MDS to 11%. Although most patients with AA/PNH and MDS/PNH showed low percentages of GPI-AP-deficient granulocytes (1±2%), there were AA patients with high proportions of PNH neutrophils. As expected, such high percentages were found in all patients with primary PNH. For patient age, the prevalence of the PNH detection generally followed the age distribution curve of a typical AA and MDS (Fig 2). Both, AA/PNH and pure haemolytic PNH were found in children. PNH clones and AA Among patients in whom an expanded GPI-AP-deficient clone was found, AA was the most frequently associated diagnosis (Table I). Therefore, we studied in more detail the clinical features of the AA±PNH syndrome. At presentation, an expanded GPI-AP-deficient clone was detected in 34% of AA patients. This proportion did not significantly differ in saa and maa (34% versus 27%). At presentation, except for age (28 versus 35 years; P ˆ 0 04) there was no difference in the ANC, reticulocyte and platelets counts between patients with a PNH clone of < 5% affected granulocytes and those with. 5% (data not shown). Among the patients studied later in the course of the disease, those with a PNH clone. 5% showed a trend towards higher counts and more frequently had a hypercellular marrow (P ˆ 0 02) than those with < 5% of affected cells (data not shown). Association Between Bone Marrow Failure, PNH and AA 1019 To determine whether the incidence of PNH increased with the duration of disease, we subdivided all patients based on time elapsed after initial presentation. In this `snap shot' approach, there was a trend to a higher incidence of PNH detection, which did not reach significance (Table II, upper portion). We also serially studied 119 AA patients over a mean follow-up period of 1 28 years (range: 3±50 months); flow analysis was performed at 3-monthly to yearly intervals (Fig 3A and B). While the prevalence of a PNH clone at presentation was high, the subsequent risk of conversion to the AA/ PNH appeared to be low. We also found a low subsequent risk of conversion in patients whose flow cytometric follow-up started at a later time point (6 months- 10 years) after initial diagnosis. In some patients with AA/PNH, the GPI-AP-deficient clone subsequently disappeared (Fig 3B), but this `off-phenomenon' was observed only in patients whose percentage of GPI-AP-deficient leucocytes was, 5%. Some PNH patients present with almost completely clonal haematopoiesis at diagnosis (all blood cells of PNH phenotype). When all patients with AA/PNH were grouped based on time after initial diagnosis, the contribution of the PNH clone significantly increased when newly diagnosed patients were compared with those at later time points after initial diagnosis of AA (Table II, lower portion). In serially followed-up patients with a PNH clone, there was a trend towards increase in the clonal size but statistical significance levels were not reached as the changes were not consistent in all patients (38 ^ 6%, 42 ^ 6, 49 ^ 6 and 49 ^ 4 at 0, 1, 2 and 3 years respectively). Similarly, expansion of the GPI-AP-deficient clone over time was also seen when the numbers of patients with PNH clone. 10,. 40 and. 80% of GPI-AP-deficient granulocytes were compared (Table III). Although only two (out of 43) patients at presentation showed. 80% PNH granulocytes, 19 out of 83 patients studied later in their disease had. 80% of PNH-cells (P ˆ 0 017). When a lower percentage of GPI-AP-deficient granulocytes was used for such a comparison, there was no statistical difference between the cohorts of patients studied at presentation and later in their disease (for. 40% of GPI-AP-deficient cells: four out of 43 and 21 out of 83, P ˆ 0 09; for. 10% of Table II. Frequency of GPI-AP-deficient granulocytes in patients with aplastic anaemia patients at different times after presentation. Duration of disease Initial presentation 1±4 years. 4 years Total patients Patients with PNH phenotype (%) 43 (34%) 45 (35%) 26 (50%) In patients with PNH clone, % of GPI-AP-deficient granulocytes* 5 8 (2 6±9 7) 21 (7±69) 12 (2±25) Statistical analysis: there was no difference in the proportion of patients with PNH clone at presentation, 1±4 and. 4 years after diagnosis (x 3 test, P ˆ 0 61, 0 25). *Median (95% confidence interval) of percentage GPI-AP-deficient granulocytes, unpaired Wilcoxon test: comparisons between initial presentation, 1±4 years (P ˆ ) and. 4 years after presentation (P ˆ 0 008).

6 1020 J. P. Maciejewski et al time of initial presentation and immunosuppressive therapy over a period of up to 4 years after immunosuppression, there was no consistent pattern of increase or decrease in the percentages of GPI-AP-deficient cells. Neither the initial PB values nor the age of the patients allowed prediction of subsequent clonal expansion. Overall, the presence of PNH was predictive of responsiveness to immunosuppression applied for the treatment of bone marrow failure (71% responders among patients with PNH clone versus 55% response rate among those without; P ˆ 0 004). However, in patients with PNH clone who received immunosuppressive therapy, the outcome of immunosuppressive therapy did not determine whether the GPI-AP-deficient clones decreased or remained stable at 3±6 months and at 1±2 years, nor was the initial size of the PNH clone predictive of further expansion (Table IV). DISCUSSION Fig 3. Probability of evolution and disappearance of GPI-APdeficient clones in patients with AA. Kaplan±Meier curves depicting event probability calculated for patients followed up for a period of up 50 months. (A) For evolution (appearance of an expanded PNH clone) and (B) for disappearance. All patients were followed up sequentially from the time of initial diagnosis (A; n ˆ 119) and (B; n ˆ 22). GPI-AP-deficient cells: 12 out of 43 and 38 out of 83 patients, P ˆ 0 1). PNH clonal expansion and immunosuppression In patients with AA/PNH who were studied serially from the Table III. Relative size of PNH clone in patients with bone marrow failure. Presence of PNH clone (percentages of PNH cells) At presentation After treatment, 80% % 2 (5%) 18 (20%) P ˆ 0 017, 40% % 4 (9%) 21 (23%) P ˆ 0 09, 10% % 12 (27%) 38 (41%) P ˆ 0 1 Numbers of patients with AA/PNH (studied at presentation and during the course of their disease) have been subdivided based on the percentages of GPI-AP-deficient granulocytes as indicated in the table. Proportions of patients fulfilling the criteria have been determined. Statistical analysis by x 2 test. Clarification of the relationship between AA, an immunemediated bone marrow failure, and PNH, the result of expansion of a somatically mutated haematopoietic stem cell, has eluded investigators for several decades (Dacie & Lewis, 1961; Dameshek, 1967; Dacie & Lewis, 1972; Young, 1992). In this study, we attempted to characterize the clinical significance of PNH clonal expansion in a large population of patients suffering from a variety of haematological diseases. In contrast to many early studies, which were based on the insensitive Ham test, we used a flow cytometric assay that is both specific and sensitive. Using the presence of GPI-AP-deficient granulocytes as a basis, we studied two parameters: the relative clone size in individual patients and the frequency of detection of an expanded PNH clone in various patient populations. In choosing a threshold level of. 1% granulocytes lacking two GPI-AP, we recognized the discordance between this value and the mean ^ 2SD (95% confidence interval) in our large number of controls. The more conservative, higher figure was chosen to avoid false positives because we did not observe a distinct population of double-negative cells below the 1% level. Nevertheless, we cannot exclude that such cells in controls and in patients represent minimal expansion of a minor PNH population and not a technical artefact. AA patients, who were considered negative for the presence of PNH clone, showed higher values within the population of GPI-AP-deficient granulocytes, suggesting that low levels of PNH cells may be present even more frequently in AA than we have determined. The presence of PIA-G mutation in small numbers of cells has been also demonstrated in normal individuals (Araten et al, 1999). Other investigators have used very small numbers of GPI- AP-deficient erythrocytes and granulocytes to conclude that very high proportions of AA patients have concurrent PNH (Brodsky et al, 2000) but, clearly, selection of the indicator proteins may affect results. We used two GPI-AP that are constitutively expressed on granulocytes; others have chosen CD55 or CD59 (Piedras & Lopez-Karpovitch, 2000; Richards et al, 2000; Schrezenmeier et al, 2000). Our selection was based on direct experimental comparisons, as

7 Table IV. Evolution of the PNH clone after immunosuppressive therapy. (A) Association Between Bone Marrow Failure, PNH and AA 1021 Time after therapy 3±6 months 1±2 years Decrease Increase or no change () Decrease Increase or no change () Responders 13 8 (1) 3 13 (5) Non-responders 4 7 (3) 3 4 (2) All 17 (53%) 15 (47%) 6 (26%) 17 (74%) Change in the percentage of GPI-AP-deficient granulocytes in relation to the responsiveness to immunosuppression. Statistical analysis: x 2 test: 3±6 months: responders versus non-responders P ˆ 0 31; 1±2 years: responders versus non-responders P ˆ 0 48; for all patients comparison of proportions 3±6 months P ˆ 0,82; 1±2 years P ˆ (B) Time after therapy 3±6 months 1±2 years Decrease Increase or no change () Decrease Increase or no change (), 5% 10 7 (3) 4 6 (0). 5% 7 8 (1) 2 11 (3) Change in the percentage of GPI-AP-deficient granulocytes in relation to the size of the PNH clone at presentation. Statistical analysis: x 2 test 3±6 months:, 5% versus. 5% P ˆ 0 78; 1±2 years: responders versus non-responders P ˆ well as the fact that CD16 and CD66b are constitutively expressed on normal granulocytes but not on erythocytes or platelets. It is also possible that some abundant GPI-AP, such as CD55 and CD59, are transferred in vivo from positive to negative cells (Sloand et al, 1998), which would lead to a decrease in the sensitivity of PNH detection. In the patient populations tested, small size PNH clones were most common. The size of abnormally expanded clones did not fit a normal distribution, suggesting that the initial appearance of a PNH clone was not a single, stochastically determined event. Greatly expanded clones correlated well with the presence of clinical haemolysis. However, despite measurement on granulocytes, there was little correlation between the detection of GPI-AP-deficient cells and the degree of cytopenia. Furthermore, significant expansion of PNH clones was observed despite persistent or recurrent cytopenias, from which we infer that evolution of PNH does not always coincide with haematopoietic recovery or reflect escape from immune attack. Although institutional referral bias may play a role in the interpretation of our data, patients with a history of AA were those most frequently observed with expanded PNH clones, and even symptomatic haemolysis was more often associated with AA than occurred as typical primary haemolytic PNH. Pure haemolytic PNH was less frequent as a clinical manifestation of a large PNH clone, and it may represent the extreme pole of a broad clinical spectrum of possible PNH manifestations. Although some AA patients showed greatly expanded PNH clones at first clinical presentation, in many others expansion was associated with the development of a more cellular bone marrow and haematological improvement, consistent with compensatory repopulation or escape. There was no relationship between an expanded PNH clone and the development of clonal karyotypic abnormalities. Expanded PNH clones were found in all age groups, and there was no correlation between patient age and the size of the clone. Like others, we found a high incidence of expanded PNH clones in patients with AA (Schrezenmeier et al, 1995; Azenishi et al, 1999; Dunn et al, 1999). We also systematically analysed large numbers of patients over time. Assuming a constant rate of evolution of PNH, a steady increase in the frequency of detection of the PNH clone as patients are observed would be expected. However, in our series, almost all PNH was discovered at the time of diagnosis of aplasia, suggesting that the evolution of PNH may be linked to the initiating events of AA. Expansion of a PNH clone correlated with higher neutrophil counts and increased marrow cellularity. In addition, a highly expanded PNH clone was more frequently found in patients with a long history of disease compared with the low values seen in most of the patients at presentation. Nevertheless, in most patients, PNH clone size did not change in a consistent pattern as a result of immunosuppressive therapy or relapse. If PNH evolution was as a result of escape from immune system attack, as has been proposed (Rotoli & Luzzatto, 1989; Young, 1992), response to immunosuppression

8 1022 J. P. Maciejewski et al should result in a decrease in the relative size of the GPI-APdeficient clone. However, immune pathophysiology of PNH clone was suggested by the relationship between the presence of the GPI-AP-deficient clone and response to immunosuppression. In MDS, a much lower proportion of patients showed an expanded PNH clone and in some patients with the MDS/ PNH syndrome, AA preceded the development of secondary MDS. Nevertheless, PNH in association with typical MDS (Iwanaga et al, 1998; Dunn et al, 1999) suggests a close pathophysiological relationship between MDS and AA (Barrett & Saunthararajah, 2000). The high proportion of AA patients with concurrent PNH at diagnosis demands explanation, and further laboratory investigations are required to clarify the operation of possible mechanisms including secondary immune escape of PNH clone or its primary involvement in the induction of the immunological process leading to AA. ACKNOWLEDGEMENTS The authors thank Mr Michael Riordan and Ms Natalie Murray for their skilled technical assistance. REFERENCES Araten, D.J., Nafa, K., Pakdeesuwan, K. & Luzzatto, L. (1999) Clonal populations of hematopoietic cells with paroxysmal nocturnal hemoglobinuria phenotype are present in normal individuals. Proceedings of the National Academy of Science of theunited States of America, 96, 5209±5214. Azenishi, Y., Ueda, E., Machii, T., Nishimura, J., Hirota, T., Shibano, M., Nakao, S., Kinoshita, T., Mozoguchi, H. & Kitani, T. (1999) CD59-deficient blood cells and PIG-A gene abnormalities in Japanese patients with aplastic anaemia. British Journal of Haematology, 104, 423±429. Barrett, J.A. & Saunthararajah, Y. (2000) Immune mechanisms and modulation in MDS. Hematology, 117±121. Bennett, J.M., Catovsky, D., Daniel, M.T., Flandrin, G., Galton, D.A.G., Gralnick, H.R. & Sultan, C. (1986) Proposal for the classification of the myelodysplastic syndromes. British Journal of Haematology, 51, 181±199. Bessler, M., Mason, P.J., Hillmen, P., Miyata, T., Yamada, N., Takeda, J., Luzzatto, L. & Kinoshita, T. (1994) Paroxysmal nocturnal hemoglobinuria is caused by somatic mutations in the PIG-A gene. EMBO Journal, 13, 110±117. Brodsky, R.A., Mukhina, G.L., Li, S., Nelson, K.L., Chiurazzi, P.L., Bucley, J.T. & Borowitz, M.J. (2000) Improved detection and characterization of paroxysmal nocturnal hemoglobinuria using fluorescent aerolysin. American Journal of Clinical Pathology, 114, 465±466. Camitta, B.M., Thomas, E.D., Nathan, D.G., Santos, G., Gordon- Smith, E.C., Gale, R.P., Rappeport, J.M. & Storb, R. (1976) Severe aplastic anemia: a prospective study of the effect of early marrow transplantation on acute mortality. Blood, 48, 63±70. Dacie, J.V. & Lewis, S.M. (1961) Paroxysmal nocturnal hemoglobinuria: variation in clinical severity in association with bone marrow hypoplasia. British Journal of Haematology, 7, 442±457. Dacie, J.V. & Lewis, S.M. (1972) Paroxysmal nocturnal hemoglobinuria: clinical manifestations, hematology and nature of the disease. Series Hematologica, 5, 3±23. Dameshek, W. (1967) Riddle: what do aplastic anemia, paroxysmal nocturnal hemoglobinuria (PNH) and hypoplastic anemia have in common? Blood, 30, 251±254. Dunn, D.E., Tanawattanacharoen, P., Boccuni, P., Nagakura, S., Green, S.W., Kirby, M.R., Kumar, M.S., Rosenfeld, S. & Young, N.S. (1999) Paroxysmal nocturnal hemoglobinuria cells in patients with bone marrow failure syndromes. Annals of Internal Medicine, 131, 401±408. Hillmen, P. & Richards, S.J. (2000) Implications of recent insights into the pathophysiology of paroxysmal nocturnal haemoglobinuria. British Journal of Haematology, 108, 470±479. Iwanaga, M., Furukawa, K., Amenomori, T., Mori, H., Nakamura, H., Fuchigami, K., Kamihira, S., Nakakuma, H. & Tomonaga, M. (1998) Paroxysmal nocturnal haemoglobinuria clones in patients with myelodysplastic syndromes. British Journal of Haematology, 102, 465±474. Luzzatto, L., Bessler, M. & Rotoli, B. (1997) Somatic mutations in paroxysmal nocturnal hemoglobinuria: a blessing in disguise? Cell, 88, 1±4. Piedras, J. & Lopez-Karpovitch, X. (2000) Flow cytometric analysis of glycosylphosphatidyl-inositol-anchored proteins to assess paroxysmal nocturnal hemoglobinuria clone size. Cytometry, 42, 234±238. Richards, S.J., Rawston, A.C. & Hillmen, P. (2000) Application of flow cytometry to the diagnosis of paroxysmal nocturnal hemoglobinuria. Cytometry, 42, 223±233. Rosenfeld, S.J., Kimball, J., Vining, D. & Young, N.S. (1995) Intensive immunosuppression with antithymocyte globulin and cyclosporine as treatment for severe acquired aplastic anemia. Blood, 85, 3058±3065. Rotoli, B. & Luzzatto, L. (1989) Paroxysmal nocturnal hemoglobinuria. Seminars in Hematology, 26, 201±207. Schrezenmeier, H., Hertenstein, B., Wagner, B., Raghavachar, A. & Heimpel, H. (1995) A pathogenetic link between aplastic anemia and paroxysmal nocturnal hemoglobinuria is suggested by a high frequency of aplastic anemia patients with a deficiency of phosphatidylinositol glycan anchored proteins. Experimental Hematology, 1, 81±87. Schrezenmeier, H., Hildebrand, A., Rojewski, M., Hacker, H., Heimpel, H. & Raghavachar, A. (2000) Paroxysmal nocturnal haemoglobinuria: a replacement of haematopoietic tissue? Acta Haematologica, 103, 41±48. Sloand, E., Maciejewski, J.P., Dunn, D., Moss, J., Brewer, B., Kirby, M. & Young, N.-S. (1998) Correction of the PNH defect by GPIanchored protein transfer. Blood, 92, 4439±4445. Socie, G., Rosenfeld, S., Frickhofen, N., Gluckman, E. & Tichelli, A. (2000) Late clonal diseases of treated aplastic anemia. Seminars in Hematology, 37, 91±101. Takeda, J., Miyata, T., Kawagoe, K., Iida, Y., Endo, Y., Fujita, T., Takahashi, M., Kitani, T. & Kinoshita, T. (1993) Deficiency of the GPI anchor caused by a somatic mutation of the PIG-A gene in paroxysmal nocturnal hemoglobinuria. Cell, 73, 703±711. Young, N.S. (1992) The problem of clonality in aplastic anemia. Dr. Dameshek's riddle, restated. Blood, 79, 1385±1387. Young, N.S. & Maciejewski, J.P. (1997) The pathophysiology of acquired aplastic anemia. New England Journal of Medicine, 336, 1365±1372. Young, N.S. & Maciejewski, J.P. (2000) Genetic and environmental effects in paroxysmal nocturnal hemoglobinuria: this little PIG- AS goes `Why? Why? Why?'. Journal of Clinical Investigation, 96, 637±641.