Immune lysis of normal human and paroxysmal nocturnal hemoglobinuria (PNH) red blood cells. 3. The membrane defects caused by complement lysis.

1. The defects produced on the membrane of the human red blood cell by the action of complement and antibody have been studied by the use of the electron microscope. These are round to slightly ovoid holes and are surrounded by an irregular ring, about 20 A thick. The mean diameter of the holes is about 103 A if human complement is used (regardless of the antibody used for sensitization) and about 88 A if guinea pig complement is used. 2. The holes in normal and PNH red cells appear to be identical, under the same conditions. The membrane defects produced by lysis of PNH cells with acidified normal serum (the Ham's test) are identical to those produced by complement lysis with specific antibody, indicating that complement is undoubtedly the cause of such lysis. 3. Evidence is presented that when human complement acts on human red cells sensitized with anti-I antibody, each complete activation of complement leads to the production of a cluster of holes. This contrasts to the action of guinea pig complement, on sheep cells, each activation of which leads to a single hole. 4. The maximum number of anti-I antibody molecules which can attach to a human red cell (i.e. the minimum number of antigen sites) is about 500,000 for both normal and PNH cells. 5. The number of holes produced during lysis of the PNH cell is the same as that of the normal cell. When all cells are lysed by am excess of C', a mean of about 90,000 holes are present on each membrane. When complement is limited, a larger proportion of PNH cells are lysed due to their peculiar sensitivity to C' but the number of holes on each lysed cell is the same as for normal cells lysed by the same concentration of C'.

Identification of human erythrocyte blood group antigens on decay-accelerating factor (DAF) and an erythrocyte phenotype negative for DAF.

Decay accelerating factor (DAF) is a glycoprotein present on the surfaces of many types ofcells in contact with plasma, including erythrocytes, leukocytes, and platelets (reviewed in reference 1). A small amount of DAF is also present in serum. Numerous investigators have demonstrated that DAF inhibits the action of C3 convertases on cell surfaces, and its absence has been shown to be at least partially responsible for the abnormal sensitivity to lysis by complement exhibited by erythrocytes of patients with the acquired stem cell disorder paroxysmal nocturnal hemoglobinuria (PNH) (2). Hereditary absence of DAF has not been previously described. Tc(a) and Cr(a) are high-frequency human erythrocyte antigens . These antigens are part of a family of blood group antigens, designated Cromer related, which are all absent from the null phenotype cell IFC(-) , or Inab (3). Recently, Spring and colleagues (4) have identified two monoclonal antibodies which bound to high frequency red cell antigens absent from the Inab phenotype. They also demonstrated that these antibodies, as well as several human antisera to Cromer-related antigens, bound to a 70-kD glycoprotein when used to stain immunoblots of human erythrocyte membrane proteins . Because the wide tissue distribution of mAb reactivity, along with some of the biochemical characterization and immunoblotting data, was similar to that of DAF, we investigated whether the Cromer-related antigens Cr(a) and Tc(a) resided on the DAF molecule.

Fixation of the first component of complement (C'la) by human antibodies.

The fixation of the first component of complement (C'1a) by human antibodies and human cells has been studied by the use of the C'1a fixation and transfer test (C'1a FT test) of Borsos and Rapp.Cold agglutinin antibodies appear to require no more than one antibody molecule to fix one molecule of C'1a. Most warm agglutinin antibodies are IgG in immunoglobulin type and require at least two molecules of antibody to fix a molecule of C'1a. Donath-Landsteiner antibody has the same requirements for C'1a fixation. A single example of a warm agglutinin antibody which appears to require one molecule of antibody for the fixation of C'1a was found. Antibodies of the Rh system do not fix significant amounts of C'1a in the absence of anti-antibody when antiserum of a single Rh specificity was used. However, when three antisera at different specificity are present, C'1a may be fixed. Under these conditions cells from a patient with paroxysmal nocturnal hemoglobinuria may be lysed when fresh serum is added to provide the other components of complement. The presence of IgG antibodies could be detected by the use of anti-IgG(Hu) antiserum and a one-to-one relationship between the concentration of antiserum in the reaction and the amount of C'1a fixed could be established. The effect of temperature, ionic strength, papainization of the red cells, and repeated washing of the red cell-antibody aggregates on the amount of C'1a fixed was investigated. Conditions of maximal C'1a fixation were established for each class of antibodies. Globulins present in normal isologous or autologous serum are absorbed in small amounts to normal red cells in a manner analogous to warm agglutinin antibody. Their presence is detectable by the C'1a fixation and transfer test only with antiglobulin antiserum. Within certain limits, the C'1a fixation and transfer test provides a quantitative measure of the reaction of human red cells and antibodies to antigens on their surface.

Quantitative immunology of immune hemolytic anemia: II. The relationship of cell-bound antibody to hemolysis and the effect of treatment.

The concentration of cell-bound and serum antibody was determined in a series of patients with warm antibody immune hemolytic anemia by determining the amount of C[unk]1 fixed to the cells by anti-IgG. This was compared to the rate of hemolysis as determined by hemoglobin concentration and reticulocyte count, or the endogenous production of carbon monoxide. The rate of hemolysis was, in general, proportional to the concentration of cell-bound antibody. In splenectomized patients, the rate of hemolysis was very much less than in unsplenectomized patients for a given concentration of cell-bound antibody. When prednisone was given, three effects were noted: (a) at high doses of drug, the concentration of cell-bound antibody decreased rapidly and the concentration of serum antibody increased, suggesting that the affinity of antibody for antigen had been altered; (b) in patients achieving remission, the concentration of serum antibody fell to low levels but rose again if the dose of prednisone was insufficient; (c) in one patient, prednisone appeared to inhibit sequestration of highly sensitized cells.

Quantitative immunology of immune hemolytic anemia. I. The fixation of C1 by autoimmune antibody and heterologous anti-IgG antibody.

The fixation of the first component of complement (C[unk]1) by rabbit and goat anti-IgG antibodies reacting with auto- or isoimmune antibodies attached to red cells has been investigated. Two molecules of the rabbit IgG anti-IgG were required to fix a single molecule of C[unk]1, whereas only one molecule of goat IgM anti-IgG was required. The relationship between the number of auto- or isoimmune antibody molecules attached to the red cells and the amount of C[unk]1 fixed by anti-IgG was determined by the concentration of anti-IgG. A concentration of anti-IgG was found such that the number of molecules of C[unk]1 fixed was directly proportional to the concentration of auto- or isoimmune antibody. By this method a sensitive, reproducible minimum estimate of the amount of cell-bound and serum antibody could be made.

Mechanism of complement-mediated activation of human blood platelets in vitro: comparison of normal and paroxysmal nocturnal hemoglobinuria platelets.

The paroxysmal nocturnal hemoglobinuria (PNH) platelet differs from the normal human platelet in its interaction with activated complement components: (a) when complement is activated by the alternative pathway, greater amounts of C3 are fixed to the PNH platelet than to the normal platelet; (b) the platelet-release reaction, as measured by serotonin release, occurs after C3 fixation to the PNH platelet. This reaction does not occur with normal platelets; (c) although serotonin release mediated by antibody alone was the same for normal and PNH platelets, antibody-initiated complement activation resulted in the fixation of greater amounts of C3 to PNH platelets and greater consequent serotonin release; and (d) nearly maximal serotonin release; and (d) nearly maximal serotonin release from PNH platelets occurs after the fixation of C3 (or perhaps C5) to the membrane without completion of the terminal sequence. In contrast, completion of the terminal complement sequence beyond C5 is required for maximal serotonin release from normal platelets. These abnormalities of interaction of complement components and PNH platelets may explain the occurrence of thromboses in this disease.

Interactions of the platelets in paroxysmal nocturnal hemoglobinuria with complement. Relationship to defects in the regulation of complement and to platelet survival in vivo.

The blood cells of patients with paroxysmal nocturnal hemoglobinuria (PNH) have abnormal interactions with complement. The activity of the alternative pathway C3 convertase on the platelets of 9 out of 19 patients with PNH was elevated. 10 patients had C3 convertase activity within the normal range even though 80-95% of their platelets lacked the complement regulatory protein decay accelerating factor (DAF) that is absent from the affected blood cells in PNH. PNH and normal platelets released factor H when C3 was bound to their surfaces. This may account for the apparent regulation of C3 convertase activity on platelets that lack DAF. The abnormal uptake of the membrane attack complex of complement by PNH III erythrocytes was not seen in PNH platelets. 111Indium-labeled platelet survival times were normal in five of eight patients, which suggests that the lack of the membrane attack complex defect results in normal platelet survival in PNH.

Quantitative influence of antibody and complement coating of red cells on monocyte-mediated cell lysis.

Monocyte-mediated lysis in vitro of human red cells coated with measured amounts of immunoglobulin G (IgG) or complement were studied. 1,000-1,500 molecules of IgG anti-D are necessary to effect measurable lysis, and lysis increases linearly with increasing levels of antibody sensitization. 100 microgram/ml of IgG1 abolished lysis even at maximal levels of anti-D sensitization (15,000 molecules/cell). Two isoimmune IgG anti-A or anti-B antisera were 5 to 10-fold less efficient in promoting phagocytosis or lysis per molecule of IgG bound; however, because of the greater antigen density of A or B, more than 100,000 molecules IgG/cell could be bound, producing equivalent lysis to anti-D-coated cells. Although inhibition by IgG1 was similar at equivalent levels of sensitization with anti-A, anti-B, or anti-D at high levels of coating with anti-A or anti-B (not attainable with anti-D), lysis was not inhibited by IgG1. Cells coated with human complement components alone were not lysed by monocytes; however, complement coating augmented IgG-mediated lysis and reduced the quantity of anti-D necessary to produce lysis to less than 1,000 molecules/cell. After thorough degradation of C3b by serum to C3d, complement augmentation persisted.

Reactions of immunoglobulin G-binding ligands with platelets and platelet-associated immunoglobulin G.

Immunoglobulin G (IgG) bound to platelets is usually detected by one of two general methods: binding of labeled anti-IgG or consumption of anti-IgG. The latter method gives, in general, values 5-10-fold greater than the former under the same conditions. To investigate these discrepancies, we have compared the detection of platelet-bound IgG by a labeled anti-IgG binding assay and by a quantitative antiglobulin consumption test using the same antibodies. The interaction of 125I-labeled monoclonal anti-IgG or polyclonal anti-IgG with washed and IgG-coated platelets was studied. The binding of these ligands to washed normal platelets was largely (50-80%) nonspecific; the binding was not saturable and was only partially inhibitable by excess unlabeled anti-IgG. The binding of anti-IgG to platelets coated with anti-PIA1, a platelet-specific IgG antibody, appeared to be saturable and inhibitable; the dissociation constant (KD) of this IgG-anti-IgG reaction was 4.9 X 10(-9) for monoclonal and 1.4 X 10(-7) for polyclonal anti-IgG. The ratio of sites present on the membrane (determined by 131I-labeled anti-PIA1) to the number of binding sites for anti-IgG determined by Scatchard analysis was 0.53 for monoclonal anti-IgG and 1.3 for polyclonal anti-IgG. The binding of monoclonal anti-IgG to platelet-bound immune complexes or IgG aggregates appeared to be complex. 131I-Labeled IgG was affixed to platelets and was detected by three tests: direct binding of radiolabeled monoclonal anti-IgG and quantitative antiglobulin consumption (QAC) tests, which were quantitated either by measuring directly the amount of radiolabeled anti-IgG consumed from fluid phase (direct QAC), or indirectly by reference to a calibration curve relating the consumption of anti-IgG by known amounts of fluid-phase, non-immune IgG (indirect QAC). The amount of platelet-bound IgG detected by the direct binding of 125I-labeled monoclonal anti-IgG and by the direct QAC approximated that known to be bound to the platelet. The results of the indirect QAC test were 10-fold greater. The discrepancy appears to be due to the fact that there is a difference between the IgG-anti-IgG interaction when IgG is bound to a platelet and when it is in solution or bound to plastic nonspecifically or specifically. This difference results in a falsely high value for platelet-bound IgG when fluid-phase or plastic-bound IgG is used to calibrate the antiglobulin consumption test.

Abnormality of glycophorin-alpha on paroxysmal nocturnal hemoglobinuria erythrocytes.

To investigate the greater enzymatic activity of the alternative pathway convertase (and the subsequent greater fixation of C3b) on paroxysmal nocturnal hemoglobinuria (PNH) erythrocytes, we have examined the topography of binding of C3b to PNH and normal erythrocytes. Using sodium dodecyl sulfate-polyacrylamide gel electrophoresis and autoradiography, the alpha-chain of C3b was found to bind via predominantly ester bonds to free hydroxyl groups on glycophorin-alpha, the major erythrocyte sialoglycoprotein. The pattern of binding of nascent C3b was the same for normal and PNH erythrocytes. Thus, although C3b binding to a different membrane constituent did not appear to account for the greater enzymatic activity of the alternative pathway convertase when affixed to PNH erythrocytes, it seemed possible that the glycoproteins to which C3b bound might be qualitatively abnormal on the PNH cells, and that structural differences in these molecules might impose modifications in the enzyme-substrate interactions of the alternative pathway convertase. Using methods for radiolabeling both protein and carbohydrate residues, we therefore compared the electrophoretic pattern of the cell-surface glycoproteins on PNH and normal erythrocytes. The glycophorin-alpha dimer was found to be qualitatively abnormal on the PNH cells as evidenced by its greater susceptibility to trypsin-mediated proteolysis. In addition, the abnormal erythrocytes from patients with PNH had fewer periodate oxidizable constituents than did normal erythrocytes, indicating a relative deficiency of cell-surface sialic acid. These investigations suggest that abnormalities in membrane glycoproteins may underlie the aberrant interactions of complement with the hematopoietic elements of PNH.

Increased enzymatic activity of the alternative pathway convertase when bound to the erythrocytes of paroxysmal nocturnal hemoglobinuria.

To investigate the greater fixation of C3 to the erythrocytes of patients with paroxysmal nocturnal hemoglobinuria (PNH) upon activation of complement, we have examined the formation and the reaction of the C3 nephritic factor-stabilized alternative pathway convertase made with purified components on normal and PNH erythrocytes. Each convertase complex converts four to five times more fluid-phase C3 to C3b when affixed to a PNH cell than when affixed to a normal cell. The greater activity of the convertase on PNH cells is not due to differences in the intrinsic or extrinsic stability of the convertase complex. The excessive binding of C3 to PNH cell si due to this increased conversion of fluid-phase C3, because the efficiency of binding of nascent C3b was identical for the two cell types. This is the first instance in which the enzyme activity of a complement complex has been shown to be increased by being affixed to an abnormal surface.

Potassium permeability in -thalassemia minor red blood cells.

The intracellular content of K(+) in thalassemia minor red blood cells is markedly reduced after incubation in autologous serum for 24 hr at 37 degrees C. There is no compensatory increase in intracellular Na(+) concentration of the cell thus reduced. This change is due to an acquired increase in selective permeability of the membrane to K(+). This phenomenon follows the depletion of energy sources in the thalassemia minor cells but does not follow comparable depletion in normal cells. The loss of osmotically active intracellular contents probably accounts for the increased resistance of incubated thalassemia minor red blood cells to osmotic lysis.

Mechanisms of immune lysis of red blood cells in vitro. I. Paroxysmal nocturnal hemoglobinuria cells.

The effect of five different reactions which activate complement (antibody activation, reduction in ionic strength, acidification, cobra venom factor (CoF) activation, and inulin activation) upon normal and PNH cells was investigated, using normal serum and serum devoid of the fourth component of complement (C4) activity from patients with hereditary angioneurotic edema (HANE) as a source of complement. Both normal and HANE serum lysed paroxysmal nocturnal hemoglobinuria (PNH) cells when complement was activated by acidification, CoF and inulin, indicating that the terminal steps of complement were activated in the absence of C4, presumably by the alternate or properdin pathway. Normal but not HANE serum lysed cells coated with anti-I, indicating that complement was activated by the C1-dependent classic pathway. HANE serum partially supported lysis by serum at reduced ionic strength, indicating that the activation of terminal complement components had occurred through both of the pathways of activation. The amount of the third component of human complement (C3) which was bound to the membrane of lysed and unlysed cells by these procedures was determined by anti-C3 absorption and was found to differ for each method of complement activation. In general, more C3 was bound to lysed cells than to unlysed cells. For given conditions, more was bound to PNH cells than to normal cells. However, very much less bound C3 was required for lysis of the PNH cells than for lysis of normal cells. These two phenomena, especially the latter, account for the marked lysis of PNH cells when complement is activated. Normal cells treated with AET (aminoethylisothiouronium bromide) did not bind more C3 than untreated cells and the lysed cells had less bound C3 than lysed PNH cells.

Measurement of the third component of complement bound to red blood cells in patients with the cold agglutinin syndrome.

The amount of the third component of complement (C3) bound to red cells of patients with the cold agglutinin syndrome was determined by a quantitative assay, measuring the fixation of the first component of complement by anti-C3. Abrupt reduction in the serum concentration of cold agglutinin by plasmapheresis markedly decreased the hemolytic rate, but the amount of C3 bound to circulating cells did not change appreciably. When this patient was transfused with normal cells. C3 accumulated on the transfused cells within 48 h to the level present on his own cells, but selective destruction of the transfused cells did not occur. When patients were subjected to acute cold stress, cell-bound C3 rose abruptly and intravascular hemolysis occurred. These studies suggest that most of the C3 detected on the circulating red cells of cold agglutinin patients is in an inactive form, and that the rate of attachment of C3 to the membrane is important in determining hemolysis.

Acute myeloblastic leukemia in paroxysmal nocturnal hemoglobinuria. Evidence of evolution from the abnormal paroxysmal nocturnal hemoglobinuria clone.

Paroxysmal nocturnal hemoglobinuria (PNH) is an acquired hematopoietic stem cell disorder in which the blood cells demonstrate aberrant interactions with serum complement. In part, this is due to the absence of the complement regulatory protein, decay accelerating factor (DAF). A small number of patients with PNH have gone on to develop acute nonlymphocytic leukemia, which is thought to arise from the injured marrow as a second hematopoietic disorder. We have studied a patient with PNH who developed acute myeloblastic leukemia (AML); the blasts from this patient were found to lack DAF as measured by polyclonal antibody binding and fluorescence flow cytometry as well as by immunoblotting. The blasts from 11 other patients with AML bound anti-DAF antibody in amounts similar to normal mononuclear cells from healthy donors. Cells of the human leukemia cell lines HL-60, K562, U937, and HEL also bound anti-DAF antibody. In addition to DAF deficiency, blasts from the PNH patient had undetectable alkaline phosphatase activity, in contrast to human leukemia cell lines. These data suggest that the leukemic cells of the PNH patient arose out of the PNH clone and that AML in the setting of PNH is not a separate disorder.

Mechanisms of immune lysis of the red cells in hereditary erythroblastic multinuclearity with a positive acidified serum test and paroxysmal nocturnal hemoglobinuria.

The red cells of patients with hereditary erythroblastic multinuclearity with a positive acidified serum test (HEMPAS), a form of congenital dyserythropoietic anemia, and the cells of patients with paroxysmal nocturnal hemoglobinuria (PNH) are lysed more readily than normal cells by certain antibodies, notably cold agglutinins (anti-I) and complement. With some but not other examples of anti-I, HEMPAS and PNH cells adsorbed more antibody than normal cells. Equal quantities of adsorbed antibody bound equal quantities of the first component of complement (C1) to normal, PNH, and HEMPAS cells. However, for a given quantity of bound antibody and C1, much more of the fourth component of complement (C4) was bound to HEMPAS cells than to normal cells. This resulted in the binding of proportionately larger quantities of the third component of complement (C3) to these cells. The same amount of bound C3 was found on the membranes of normal and HEMPAS cells for a given degree of lysis. Hence, the marked increase in lysis of HEMPAS cells is due to the increased adsorption of antibody and/or increased binding of C4.PNH cells bound the same amount of C4 per bound C1 as normal cells but bound more C3 than normal cells. However, the mean concentration of C3 on the membrane of PNH cells was one-third to one-fifth that on normal cells for a given degree of lysis. Hence, the increased lysis of PNH cells is due to the increased binding of C3 and increased hemolytic effectiveness of the bound C3.

Characterization of the complement sensitivity of paroxysmal nocturnal hemoglobinuria erythrocytes.

The affected erythrocytes of paroxysmal nocturnal hemoglobinuria (PNH II and PNH III cells) are abnormally sensitive to complement-mediated lysis. Normal human erythrocytes chemically modified by treatment with 2-amino-ethylisothiouronium bromide (AET) have been used as models for PNH cells inasmuch as they also exhibit an enhanced susceptibility to complement. To investigate the bases for the greater sensitivity of these abnormal cells to complement-mediated lysis, we compared binding of C3 and constituents of the membrane attack complex to normal, PNH II, PNH III, and AET-treated cells after classical pathway activation by antibody and fluid-phase activation by cobra venom factor complexes. When whole serum complement was activated by antibody, there was increased binding of C3 and C9 to PNH II, PNH III, and AET-treated cells, although the binding of these complement components to PNH II and PNH III cells was considerably greater than their binding to the AET-treated cells. In addition, all of the abnormal cell types showed a greater degree of lysis per C9 bound than did the normal erythrocytes. PNH III and AET-treated cells were readily lysed by fluid-phase activation of complement, whereas normal and PNH II erythrocytes were not susceptible to bystander lysis. The greater hemolysis of PNH III and AET-treated cells in this reactive lysis system was due to a quantitative increase in binding of constituents of the membrane attack complex. This more efficient binding of the terminal components after fluid-phase activation of whole serum complement was not mediated by cell-bound C3 fragments. These investigations demonstrate that the molecular events that characterize the enhanced susceptibility of PNH II, PNH III, and AET-treated erythrocytes to complement-mediated lysis are heterogeneous.

Relationship between decay accelerating factor deficiency, diminished acetylcholinesterase activity, and defective terminal complement pathway restriction in paroxysmal nocturnal hemoglobinuria erythrocytes.

Paroxysmal nocturnal hemoglobinuria (PNH) erythrocytes exhibit abnormalities in decay accelerating factor (DAF), acetylcholinesterase, and resistance to autologous C5b-9 attack. To investigate the nature of the lesion underlying PNH cells, we examined the relationship of these abnormalities to one another. Analyses of DAF in acetylcholinesterase-negative erythrocytes revealed that these two abnormalities involve functionally independent molecules, coincide precisely in the same cell populations, and are similarly expressed in PNH II and more complement-sensitive PNH III erythrocytes. The DAF and acetylcholinesterase deficiencies contrast with the C3b/C4b receptor (CR1) deficit, which is less profound and similarly distributed in complement-insensitive cell populations. Hemolytic studies showed that defective resistance to autologous C5b-9 attack is mediated by another mechanism. Whereas reconstitution of PNH II erythrocytes with DAF completely corrected their complement sensitivity, DAF reconstitution of PNH III erythrocytes restored their ability to circumvent C3b uptake but had no effect on their heightened susceptibility to reactive lysis. Assays of complement-insensitive (PNH I) erythrocytes surviving after reactive lysis disclosed partial DAF and acetylcholinesterase deficits. These findings indicate that the PNH lesion involves multiple membrane components and that PNH I erythrocytes are also abnormal.

Relationship between the membrane inhibitor of reactive lysis and the erythrocyte phenotypes of paroxysmal nocturnal hemoglobinuria.

Susceptibility to hemolysis initiated by activated cobra venom factor (CoF) complexes is a characteristic that distinguishes the most complement-sensitive type III erythrocytes of paroxysmal nocturnal hemoglobinuria (PNH) from the intermediately sensitive type II and the normally sensitive type I cells. Recently we isolated a membrane constituent from normal erythrocytes that inhibits CoFBb-initiated hemolysis, and this protein was designated membrane inhibitor of reactive lysis (MIRL). To investigate the molecular basis of the variability in complement sensitivity among PNH erythrocytes, the surface expression of MIRL and decay accelerating factor (DAF) on the three phenotypes of PNH was quantified immunochemically. Both complement regulatory proteins were markedly deficient on the erythrocytes from a patient with predominately type III cells. The erythrocytes from patients with a majority of either type II or I cells were also significantly deficient in both MIRL and DAF. While cytofluorometric analysis confirmed the quantitative deficiencies, segregation of erythrocytes into discrete subpopulations that expressed either no MIRL or normal amounts of MIRL was not observed. The results of immunoprecipitation studies were consistent with quantitative, but not qualitative abnormalities of MIRL and DAF. Selective removal of the sensitive erythrocytes indicated that approximately 20% of the normal amount of MIRL is sufficient to protect cells from CoF-initiated lysis. These studies suggest that relatively subtle quantitative differences in membrane complement regulatory proteins underlie the variability in complement sensitivity of PNH erythrocytes.

Recurrent thrombocytopenia following idiopathic thrombocytopenic purpura. The importance of platelet-bound IgG in establishing cause.

A 20-year-old man experienced two separate thrombocytopenic illnesses. The first episode represented classic idiopathic thrombocytopenic purpura (ITP) and was associated with elevated platelet-bound IgG values. Adequate control of thrombocytopenia could not be obtained with prednisone therapy, and splenectomy produced a clinical remission. Four weeks after splenectomy, an acute febrile illness typical of cytomegalovirus (CMV) infection developed, and CMV grew grom a sample of the patient's blood. Thrombocytopenia recurred during the CMV infection but was not associated with elevated platelet-bound IgG levels. Since the second episode of thrombocytopenia was associated with normal amounts of platelet-bound IgG, it was not ascribed to relapse of the ITP, and the thrombocytopenia resolved rapidly, without specific therapy. There are various therapeutic implications of an accurate causative diagnosis of thrombocytopenia.