Pure red cell aplasia.

Abstract

Pure red cell aplasia (PRCA), a disorder first described in 1922 (Kaznelson, 1922), can be characterized as an anaemia with the almost complete absence of red-cell precursors in the bone marrow, but essentially normal granulopoiesis and megakaryopoiesis. Typical bone marrow findings in PRCA are shown in Fig 1. The reticulocyte count is low while the platelet count, leucocyte count and leucocyte differential are normal. Clinically, the patients present with symptoms of severe anaemia in the absence of haemorrhagic phenomena. Depending on the cause, the course can be acute and self-limiting or chronic with rare spontaneous remissions. Bone marrow histology of a patient with PRCA associated with thymoma. In the aplastic phase (A and B) there is a complete absence of red-cell precursors, as documented by immunohistochemistry using an antibody against haemoglobin with a positive reaction in some partly squashed erythrocytes (A). Most myeloid cells, except for the eosinophils (marked by arrows) stain positive for naphthol-AS-d-chloroacetate esterase (B). Upon recovery of erythropoiesis following removal of the thymoma (C and D), the anti-haemoglobin immunohistochemistry reaction clearly visualizes erythroblasts (C). These cells are also detectable as the naphthol-AS-D-chloroacetate esterase-negative cell clusters (D). (Fig 1A–D, magnification ×370). The pathophysiology of PRCA is heterogeneous, as summarized in Table I. There is a ‘congenital’ form (Diamond–Blackfan anaemia) where most cases appear to be a result of several genetic defects predominantly affecting the erythropoietic lineage. Acquired PRCA induced by parvovirus B19 infection typically produces an acute self-limiting disease, called ‘transient aplastic crisis’ (TAC). In immunosuppressed individuals, B19 infection may result in a more chronic type of bone marrow failure, clinically apparent as PRCA. However, most cases of PRCA are autoimmune-mediated. Various autoimmune mechanisms of PRCA have been described (Table I), such as antibodies against erythroblasts or erythropoietin. T cells or natural killer (NK) cells have been proposed to secrete factors selectively inhibiting erythroid colonies in the bone marrow or directly lysing erythroblasts. Lysis of erythroblasts by T cells could result from ‘classic’ T-cell receptor (TCR)-mediated antigen recognition. In addition, PRCA can be mediated by major histocompatibility complex (MHC) unrestricted effector-target cell recognition, as erythroid progenitors progressively lose expression of MHC class I and, thus, become susceptible to destruction by NK-type cells. This mechanism is similar to NK-mediated lysis of tumour cells following loss of HLA class I by the tumour cells. Autoimmune PRCA can occur as a primary ‘idiopathic’ form or be associated with (i) infections, (ii) autoimmune disease, and (iii) neoplasias such as thymoma, lymphoma or carcinoma. Specific situations under which immune-mediated PRCA has been described are pregnancy and post-allogeneic bone marrow or stem cell transplantation, in which recipient antibody against incompatible donor ABO blood group antigens may inhibit red-cell regeneration. Moreover, most cases of transient erythroblastopenia of childhood (TEC), an acute self-limiting form of PRCA, are probably caused by humoral immune mechanisms, aetiologically induced following infection with an unknown virus that is distinct from B19 parvovirus. Rarely, PRCA may represent the initial manifestation of a preleukaemic syndrome. Finally, PRCA has been associated with several drugs and toxins, documented by the demonstration that the anaemia remits shortly after removal of the causative agent. PRCA has been also studied in cats, either as a rare form occurring spontaneously that resolves following immunosuppressive therapy (Stokol & Blue, 1999) and another form induced by subgroup C feline leukaemia retrovirus that is thought to be mediated by a direct cytopathic effect of the virus (Dean et al, 1992), rather than by antibody and T cells (Abkowitz et al, 1987). This article will review the present knowledge on the aetiology of the diverse types of PRCA with particular attention to the cases associated with expansions of large granular lymphocytes (LGLs) (Lacy et al, 1996; Handgretinger et al, 1999). Diamond–Blackfan anaemia Following the first report on red cell aplasia in infancy (Josephs, 1936), four more cases were presented (Diamond & Blackfan, 1938) and several names proposed, including congenital hypoplastic anaemia, chronic congenital aregenerative anaemia or erythrogenesis imperfecta. The diagnostic criteria for Diamond–Blackfan anaemia (DBA) include normochromic, at times macrocytic, anaemia developing early in childhood, a normocellular bone marrow with a selective deficiency of red-cell precursors beyond the level of proerythroblasts, reticulocytopenia, normal or slightly decreased leucocyte counts and normal, increased or reduced platelet counts. The incidence of DBA has been reported to be five per million live births with a manifestation in the neonatal period or in early infancy (Ball et al, 1996). Approximately three-quarters of the cases are sporadic, but dominant or recessive inheritance of DBA has been reported in different families (Diamond et al, 1976). Some patients have chromosomal abnormalities (Tartaglia et al, 1966). Physical abnormalities are present in about 30% of the patients (Diamond et al, 1976). They include short stature, atrial or ventricular septal defects, urogenital abnormalities, microcephaly, cleft palate, micrognathia, macroglossia and a deformed thumb (Alter, 1978; Diamond, 1978). Further haematological findings are increased levels of HbF (Diamond, 1978), erythrocyte adenosine deaminase (an enzyme of the purine salvage pathway) in most patients (Glader et al, 1983) and erythropoietin. Earlier reports, suggesting an immune cell-mediated pathogenesis of DBA (Hoffman et al, 1976), have not been confirmed by others (Freedman & Saunders, 1978; Nathan et al, 1978a). Current evidence suggests more than one pathogenic mechanism for the erythroid failure in DBA, suggesting that DBA is composed of more than one disorder (McGuckin et al, 1995; Giri et al, 2000; Willig et al, 2000). It appears that the erythroid progenitor compartment is intrinsically defective in DBA as marrow cultures have shown reduced or absent erythroid blast-forming units (BFU-E) or erythroid colony-forming units (CFU-E) in most, although not in all, DBA patients (Freedman et al, 1976; Nathan et al, 1978a; Tsai et al, 1989; Santucci et al, 1999). Erythroid progenitors in some DBA patients appear to display a reduced sensitivity to erythropoietin that could be corrected by the addition of glucocorticoids in vitro (Chan et al, 1982a). This may provide an explanation for the empirical response of 60–70% of the DBA patients to steroids (Diamond et al, 1976; Chan et al, 1982b; Janov et al, 1996). However, ‘spontaneous remissions’ are known to occur in 20–30% of DBA cases (Diamond et al, 1976; Janov et al, 1996), for the most part in patients with a family history of DBA that were treated before the era of steroid therapy. Although such patients with DBA in remission no longer require transfusion therapy, they are not cured as their red cells maintain fetal characteristics (Alter & Nathan, 1979). Thus, such patients may have a more silent clinical manifestation of the underlying defects. Different in vitro assays of DBA bone marrow, as well as molecular studies, have failed to demonstrate a direct role for different growth factors or their receptors in the pathogenesis of DBA (Tsai et al, 1989; Bagnara et al, 1991; McGuckin et al, 1995; Dianzani et al, 1996). Nevertheless, clinical trials with interleukin (IL)-3 observed an improvement of erythropoiesis in some DBA patients (Dunbar et al, 1991), while erythropoietin seemed to be ineffective in vivo (Niemeyer et al, 1991). Stem cell factor has not been tested in DBA thus far. However, as there is a significantly increased risk of DBA patients developing haematological malignancies (as high as 23% by the end of the fourth decade) (Janov et al, 1996), therapeutic approaches of DBA with growth factors are not without risks. In patients with ‘spontaneous’ or steroid induced remissions there remains an increased risk of developing leukaemia (Janov et al, 1996). Normal haematopoiesis in DBA patients can be restored by bone marrow allografting (Greinix et al, 1993). Genetic mapping studies localized a major DBA locus to chromosome 19q13.2 on the basis of (i) the identification of a balanced reciprocal translocation X;19 in a singular DBA patient (Gustavsson et al, 1997a), (ii) linkage analysis in different DBA families that documented the involvement of chromosome 19q13 in the majority of familial cases (Gustavsson et al, 1997b), and (iii) microdeletions on chromosome 19q13, associated with a few of the sporadic cases of the disease (Gustavsson et al, 1998). The gene encoding the ribosomal protein S19 (RSP19) was identified at the breakpoint of the reciprocal X;19 chromosome translocation in the singular DBA case cited above and mutations of RSP19 were documented in 10 out of 40 unrelated DBA patients, including four out of 19 of the sporadic cases analysed (Draptchinskaia et al, 1999). There is a possibility that haploinsufficiency for the RSP19 protein results in a protein synthesis defect in some tissues with high proliferative activity and this may also explain some of the variable clinical outcomes and hereditary features of DBA. However, in some of the familial cases the DBA candidate region on 19q13 was excluded by the segregation of marker alleles (Gustavsson et al, 1998) and the majority of sporadic cases do not seem to have RSP19 mutations (Draptchinskaia et al, 1999). Thus, mutations in RSP19 cause DBA in a subset of patients, but other causes of DBA remain to be identified. The current knowledge on DBA was the subject of a recent detailed review (Willig et al, 2000). Transient erythroblastopenia of childhood Transient erythroblastopenia of childhood (TEC) is an acquired anaemia in previously healthy children. More than 80% of the patients are 1 year of age or older at diagnosis. The children have a temporary reticulocytopenia and, typically, the bone marrow shows erythroblastopenia with normal white blood and platelet counts (Ware & Kinney, 1991; Freedman, 1993). However, significant neutropenia and hypocellular marrows have also been observed in many patients with TEC that may be as a result of a common pathogenic mechanism to that producing the anaemia (Rogers et al, 1989; Skeppner & Wranne, 1993; Cherrick et al, 1994). In addition, increased numbers of lymphoid cells with a common pre-B-acute lymphoblastic leukaemia (ALL) phenotype have been noticed in the bone marrow of some patients with TEC, thus this should not be misdiagnosed as acute leukaemia (Foot et al, 1990). Transient central neurological changes have been described in a few TEC cases (Green et al, 1986). In some of the patients, a preceding viral illness had been observed, but a common viral aetiology, including parvovirus B19 association, could not be found. TEC can occur in siblings simultaneously (Labotka et al, 1981). It appears that most cases of TEC are as a result of antibodies against red-cell progenitors by a mechanism similar to autoimmune thrombocytopenic purpura or autoimmune haemolytic anaemia in childhood (Freedman, 1993). TEC can be differentiated from DBA by the typically older age of onset, the lack of associated anomalies, the normocytic type of anaemia in TEC vs. the mostly macrocytic type in DBA, the normal levels of HbF and red-cell adenosine deaminase activity and the spontaneous recovery in TEC, usually within 4–8 weeks without recurrence (Freedman, 1993). However, in children presenting with TEC in the first year of life this distinction may be difficult at times (Ware & Kinney, 1991). TEC is generally a disease with a good prognosis and most children recover fully with a blood transfusion alone (Skeppner & Wranne, 1993). Interestingly, a recent study from Sweden suggested that TEC may also involve hereditary factors (Skeppner et al, 1998). The first report on this condition was probably that of a 42-year-old man suffering from recurrent jaundice, slowly increasing splenomegaly, enhanced urobilinuria and pigmented cholelithiasis (Minkowski, 1900). Other members of his family in three generations were affected by the same syndrome. This patient died following a short episode of influenza-like symptoms, severe malaise and lobar pneumonia. This early description was followed by repeated cause-related reports on patients suffering from congenital haemolytic diseases (such as familial spherocytosis, sickle cell anaemia, acquired immune haemolytic anaemia, hereditary elliptocytosis, thalassaemia, glucose-6-phosphate dehydrogenase and pyruvate kinase deficiencies, etc.) who developed severe acute anaemia combined with fever, anorexia, nausea, vomiting, headache, abdominal pain and chills (Horne et al, 1945; Dameshek & Bloom, 1948; Owren, 1948; Heilmeyer, 1950). The haemoglobin concentrations fell as low as 5 g/dl, sometimes requiring transfusions, with most patients recovering from their severe illness within 10 d. A meticulous analysis of six cases within families with congenital haemolytic anaemias deduced that these ‘haemolytic crises’ were as a result of a sudden cessation of red blood cell production (Owren, 1948). This disproved the initial idea that the crises were because of enhanced haemolysis. Subsequently, it was reported (Gasser, 1950) that a febrile infection triggered these ‘transient aplastic crises’ in patients with chronic haemolytic disorders and that this type of infection even provoked mostly clinically inapparent acute erythroblastic aplasias in normal subjects. The causative agent, parvovirus B19, was identified in 1981 as the infective cause of outbreaks of aplastic crises in sickle cell anaemia (Pattison et al, 1981; Serjeant et al, 1981). B19 infection studies in normal volunteers confirmed that a 5–10 d erythropoietic aplasia does not produce significant anaemia in healthy subjects whose red cells have a life span of about 120 d (Potter et al, 1987). In contrast, in patients with shortened erythrocyte survival (e.g. 15 d in familial spherocytosis), B19 infection produces a transient decompensation of the erythropoietic system with its clinical manifestation as severe anaemia. B19-induced aplastic crisis may also occur in patients with non-inherited forms of haemolysis, such as iron deficiency anaemia (Kudoh et al, 1994). Parvovirus B19 was found to be a rather small (15–28 nm) non-enveloped single-stranded DNA virus related to several animal parvoviruses (Cossart et al, 1975). It is transmitted via the respiratory tract or, rarely, by blood products. The main target of B19 infection is the erythroid progenitor cell and the basis of this erythroid tropism is the B19 cellular receptor globoside, known as the blood group P antigens (P, P1 and Pk antigens; Brown & Young, 1995, 1997). Very rare individuals with erythrocytes lacking the P antigens (p phenotype) cannot be infected with parvovirus B19. Most infections with parvovirus B19 are clinically inapparent, with up to 15% of children below 5 years, 50–60% of young adults and up to 90% of elderly individuals being seropositive. The clinical manifestations of B19 infection include erythema infectiosum (‘fifth disease’), an acute polyarthropathy syndrome, hydrops fetalis, the ‘transient aplastic crisis’ (TAC) described above and a more chronic type of bone marrow failure manifesting as a chronic pure red cell aplasia (Kurtzman et al, 1989; Frickhofen et al, 1994). Only the latter two will be further discussed in this review. B19-mediated TAC may be rarely associated with changes in the other blood lineages, varying degrees of neutropenia and thrombocytopenia. As a characteristic finding, the bone marrow contains abnormally large proerythroblasts, ‘gigantoproerythroblasts’, that appear towards the end of the aplastic phase (Fig 2) (Gasser, 1950; Schaefer, 1992). These cells contain large vesicular nuclei with loosely distributed chromatin and prominent, inclusion body-like nucleoli, surrounded by a basophilic cytoplasm. Some of these blasts may somehow resemble Hodgkin cells (Fig 2). As the anaemia resolves, usually between the fifth and tenth day after onset, the gigantoproerythroblasts completely disappear from the bone marrow and are followed by regenerating normoblasts. This is mostly as a result of a humoral protective immune response to the virus, virus-specific immunoglobulin (Ig)M and IgG antibodies to the viral capsid proteins (Brown & Young, 1995). It appears that only antibodies against the 83 kd minor capsid species VP1 are virus-neutralizing antibodies, while the early antibody response against the 58 kd major capsid protein VP2 is insufficient to clear the infection. A protective antibody response generally results in life-long immunity against reinfection. Because of their bizarre aspect, gigantoproerythroblasts are to some extent diagnostic for the parvovirus B19 infection. However, immunostaining for the B19 antigen fails to demonstrate the presence of virus in these cells. Positive results are seen in smaller regressive erythroblasts with a dense ring of chromatin that surrounds an inclusion-like central part of the nucleus harbouring the immunoreactive B19 antigen (Fig 2A). These decaying erythroblastic virocytes become visible for only a very short time at the early phase of the infection. At the time of bone marrow assessment, the erythroblastic virocytes have just disappeared in many cases with gigantoproerythroblasts being the only hallmark of the previous cytolytic viral infection. When B19 infection in pregnancy leads to fatal hydrops fetalis, erythroblastic virocytes are easily detectable by immunostaining in various organs. We propose that these gigantoproerythroblasts in B19 infection represent very early red-cell precursors that appear after clearance of the B19 virus as evidence of regenerating erythropoiesis. Typical ‘gigantoproerythroblasts’ in parvovirus B19-associated PRCA. (A) Bone marrow histology with several gigantoproerythroblasts (marked by asterisks) that resemble Hodgkin cells. A small regressive erythroblast (marked with an arrow) typically contains the immunoreactive B19-antigen. (B) Bone marrow cytology of another patient with B19 infection showing two gigantoproerythroblasts (marked by asterisks). (Fig 2A, magnification ×750; Fig 2B, magnification ×400). Persisting B19 infection can occur in a wide variety of conditions of immunosuppression [congenital immunodeficiency, human immunodeficiency virus (HIV) infection, lymphoproliferative disorders, post organ transplant, etc.], even in cases of only minimal or clinically previously unrecognized immune dysfunction (Kurtzman et al, 1989; Frickhofen et al, 1994). In such patients, the presentation may be as a chronic PRCA or, more rarely, as pancytopenia (Brown & Young, 1995). The bone marrow examination in chronic B19-induced PRCA also shows gigantoproerythroblasts, although in some cases they cannot be found. B19-specific antibody is absent or low and directed against the major capsid protein VP2. Typically, BFU-E and CFU-E are decreased and the patients' sera inhibit normal BFU-E, CFU-E, but not granulocyte-macrophage CFU (CFU-GM), colony formation because of viraemia. The diagnosis is established by the demonstration of viral DNA in the serum by nucleic acid hybridization assays. Although polymerase chain reaction (PCR) is much more sensitive it carries a high risk of contamination and false-positive results (Brown & Young, 1997). In chronic B19 infections, antibody studies often do not produce clear results, although high titres of B19 IgG make a diagnosis of a persistent infection improbable. As indicated above, fetal infection with parvovirus can lead to a miscarriage, hydrops fetalis and to infants born with chronic anaemia (Brown et al, 1994, 1995). Thus, a maternally transmitted B19 infection should be suspected in infants with congenital red cell aplasia. Although the patients may be seriously ill, parvovirus B19-induced ‘aplastic crisis’ in immunocompetent individuals is generally a self-limiting problem, usually requiring only supportive therapy including blood transfusions. Chronic B19-associated PRCA has been successfully treated by the repeated administration of immunoglobulin until the virus is no longer detectable by PCR (Kurtzman et al, 1989; Brown & Young, 1995). Patients generally respond within 2 weeks of treatment, but should be monitored for recurrence of viraemia. PRCA has been associated with autoimmune diseases such as rheumatoid arthritis (Dessypris et al, 1984; Rodrigues et al, 1988) and systemic lupus erythematosus (Cassileth & Myers, 1973; Dainiak et al, 1980). In addition, red cell aplasia occurred as a paraneoplastic syndrome (Field et al, 1968), complicating thymoma (Jacobs et al, 1959; Roland, 1964), large granular lymphocyte expansions (Loughran & Starkebaum, 1987), chronic lymphocytic leukaemia (Mangan et al, 1982; Chikkappa et al, 1986), Hodgkin's disease (Morgan et al, 1978) and diverse carcinomas (Mitchell et al, 1971). Although early observations suggested that up to 50% of patients with PRCA had thymomas (Jacobs et al, 1959; Roland, 1964; Krantz, 1973), more recent studies found thymomas in less than 10% of PRCA patients (Charles et al, 1996; Lacy et al, 1996). An autoimmune mechanism of the anaemia was suggested by remissions of the red cell aplasia following resection of the thymoma in about 25% of cases (Zeok et al, 1979) and the response of some PRCA patients to steroids (Finkel et al, 1967), cyclophosphamide (Marmont et al, 1975) and splenectomy (Eisemann & Damashek, 1954). Several studies published in recent years helped to elucidate diverse immunological mechanisms that can explain red-cell progenitor destruction in individual patients. PRCA mediated by antibody Initial demonstrations of factors in the patients' sera that inhibited erythropoiesis in vivo (Jepson & Lowenstein, 1968) and in vitro (Krantz & Kao, 1967) were followed by the demonstration that these inhibitors could be IgG antibodies directed against erythroblasts (Krantz & Kao, 1969). These antibodies either inhibited haemoglobin synthesis (Krantz & Kao, 1967) or they were complement-binding and directly cytotoxic for erythroblasts in vitro (Zaentz & Krantz, 1973). Expanding on these experiments, immunosuppressive drugs were used for the treatment of PRCA (Krantz & Kao, 1969). Later studies of another patient who did not respond to immunosuppressive therapy revealed that such antibodies could specifically block differentiation of BFU-E in vitro (Messner et al, 1981). The pathogenic role of such antibodies was strongly suggested by the demonstration that plasmapheresis was effective to induce remission in this patient (Messner et al, 1981). In other patients, the inhibitory antibodies were found to be directed against erythropoietin (Marmont et al, 1975; Peschle et al, 1975; Casadevall et al, 1996). While erythropoietin levels in other patients with PRCA are typically elevated for the degree of anaemia, they are low in PRCA patients with erythropoietin-specific antibodies. In vitro inhibition of erythroid colonies could be reversed by the addition of purified erythropoietin and the patient's serum could immunoprecipitate purified erythropoietin (Casadevall et al, 1996). These observations gave final proof for this pathogenic mechanism. In a minority of cases, antibody-mediated PRCA could spontaneously remit with supportive therapy alone, although this could take more than a year (Clark et al, 1984). Such a self-limiting course is reminiscent of transient erythroblastopenia in children that is also believed to be antibody-mediated and that may be associated with a viral infection. Thus, some cases of adult autoimmune PRCA might be induced by antibodies produced following a viral or bacterial infection that might cross-react with erythroid precursor cells or erythropoietin. Antibody-mediated PRCA has been described in association with thymoma, systemic lupus erythematosus, rheumatoid arthritis, Hodgkin's disease and other diseases. However, only few of these studies unambiguously demonstrated a serum immunoglobulin as the mechanism of inhibition of erythropoiesis (Al-Mondhiry et al, 1971; Cassileth & Myers, 1973; Morgan et al, 1978; Dessypris et al, 1984). In some of the earlier studies, parvovirus B19 viraemia itself might have been misinterpreted as a circulating cytotoxic erythropoiesis inhibitor (Zaentz et al, 1975). The possible immunological mechanism of antibody-mediated inhibition of erythropoiesis includes direct complement-mediated lysis of red-cell progenitors and the formation of immune complexes with erythropoietin that result in functional inactivation and the removal of erythropoietin from the circulation. Although this has not yet been demonstrated experimentally, anti-erythroblast antibodies that are not directly cytotoxic might impair red-cell progenitor maturation by blocking the erythropoietin receptor or another red-cell signalling pathway. It should be kept in mind that T-helper cells (Th2 cells) could play a role in the production of such autoantibodies. This might explain why inhibitory humoral factors have been detected in the serum of some patients with a disturbed T-cell function, such as patients suffering from thymoma, Hodgkin's disease and diverse autoimmune diseases. Moreover, this could be why some patients with apparently antibody-mediated PRCA responded to anti-thymocyte globulin (Marmont et al, 1975). Occasionally, ABO-incompatibility between donor and recipient following allogeneic bone marrow or stem cell transplantation can result in antibody-mediated PRCA with elevated titres of the incompatible agglutinins (Gmür et al, 1990). Red-cell engraftment and reticulocyte recovery may occur spontaneously within weeks or several months as the agglutinin titres fall, but there are reports of successful treatment of this condition with erythropoietin, increased immunosuppressive therapy or plasmapheresis (Gmür et al, 1990; Heyll et al, 1991). PRCA mediated by T cells and NK cells In patients with PRCA in whom a plasma inhibitor of erythropoiesis cannot be demonstrated, lymphocyte-mediated inhibition of erythropoiesis is the most probable mechanism of pathogenesis. Originally, investigations in a patient with T-cell chronic lymphocytic leukaemia showed that the patient's malignant T lymphocytes suppressed erythroid colony formation by allogeneic human bone marrow cells (Hoffman et al, 1978). Pretreatment of the patient's bone marrow T cells with anti-thymocyte globulin and complement reversed this suppression and markedly augmented autologous erythropoiesis in vitro. Subsequent studies demonstrated that T cells with receptors for the Fc portion of the IgG molecule (CD16), the so-called Tγ cells from patients with B-cell chronic lymphocytic leukaemia (B-CLL), suppressed erythroid colony formation in vitro (Nagasawa et al, 1981). Initially, it was controversial whether this was an active suppressive effect of these Tγ cells or whether this phenomenon was owing to the lack of normal T-cell help in vitro because these cultures originated from patients with B-CLL (Nathan et al, 1978b). However, subsequent studies demonstrated that removal of the Tγ cells by E-rosetting markedly augmented erythroid colony growth, suggesting that the Tγ cells actively suppressed erythropoiesis (Mangan et al, 1982). Subsequently, the Tγ cells were renamed large granular lymphocytes (LGLs) and several groups found that expansions of LGLs may be the disorder most commonly associated with PRCA (Abkowitz et al, 1986; Levitt et al, 1988; Tefferi et al, 1994; Charles et al, 1996; Lacy et al, 1996). These LGLs may be of T-cell type or of natural killer (NK)-cell type (Loughran, 1993). T-LGLs express CD3 and a T-cell receptor (TCR) of αβ- (in the majority of cases) or γδ-type. In contrast, NK-LGLs are CD3− and, consequently, do not express a TCR at the cell surface. Typically, these NK-LGLs do not rearrange the TCR genes α, β, γ and δ. Non-malignant LGLs typically display MHC unrestricted cytolytic activity against HLA class I-deficient NK-sensitive tumour cells, such as the erythroleukaemia K562. In some, although not all, patients with clonal LGL proliferations of the CD3+ (T-LGL)- and the CD3− (NK-LGL)-type, the LGLs are directly cytotoxic against K562 cells in vitro when freshly isolated (Partanen et al, 1984; Kaufmann et al, 1987; Handgretinger et al, 1999). If these fresh LGLs are not cytotoxic against K562 cells, cytotoxicity can be induced by antibody cross-linking the TCR of the T-LGLs with the Fc receptor on target cells, or by antibody cross-linking the Fc receptor on the T-LGLs or NK-LGLs (the CD16 molecule) with any specific ligand for the antibody on the target cells (Loughran et al, 1987). As in vitro cytotoxicity by the clonal LGLs against K562 cells may no longer be detectable once the LGLs have been cryopreserved, some of the published studies on the function of leukaemic LGLs may not have adequately assessed the LGLs cytolytic abilities. Thus, in principle, most neoplastic LGLs may bear a functional cytotoxic machinery that is able to destroy erythroblasts in vivo. This may be mediated by similar mechanisms as fresh LGLs kill the erythroleukaemia K562 in vitro (Partanen et al, 1984; Handgretinger et al, 1999). Triggering of cytolysis against erythroblasts could occur (i) via the TCR of αβ- or of γδ-type that could recognize unknown ligands expressed by erythroid progenitors (Fig 3A and B), (ii) via antibodies against red-cell progenitors binding to CD16 on the LGLs (Tγ cells), or (iii