Transfer of glycosylphosphatidylinositol-anchored proteins to deficient cells after erythrocyte transfusion in paroxysmal nocturnal hemoglobinuria

Paroxysmal nocturnal hemoglobinuria (PNH) is a clonal bone marrow disorder resulting from an acquired, somatic mutation of the X chromosome PIGA gene in a hematopoietic stem cell.^1-3  Absence of PIG-A function in a cell prevents synthesis of the glycosylphosphatidylinositol (GPI) moiety, which anchors many different types of proteins to the cell membrane.^4,5  Intravascular red blood cell (RBC) destruction, the hallmark of the disorder, is caused by susceptibility of the abnormal erythrocyte to complement-mediated lysis; this sensitivity is due to lack of CD59, a potent inhibitor of the late components of complement and reactive lysis.^6-8 

Transfer of GPI-anchored proteins to deficient cells has been demonstrated in tissue culture and in a few animal models. Cell-to-cell transfer of CD59 from erythrocytes to endothelial cells occurred in mice made transgenic for this GPI-anchored protein,⁹  as well as of Thy-1 in chimeric murine embryoid bodies composed of normal and PIG-A “knock-out” cells.¹⁰  Others have demonstrated transfer of the variant surface glycoprotein (VSG) from trypanosomal membranes to the RBCs of parasitemic patients,¹¹  transfer of CD59 and CD55 (a second GPI-anchored protein that also functions in complement resistance) from seminal fluid to prostasomes,¹²  and transfer of CD55 from high-density lipoproteins to deficient cells.¹³ 

Our laboratory has shown that RBC microvesicles, derived from outdated blood units during storage, can be a source of CD59 and CD55 and alter the susceptibility of RBCs to lysis.¹⁴  We hypothesized that transfusion of normal RBCs to patients with PNH might effect transfer of CD55 and CD59 to GPI-anchored protein-deficient cells in these patients. Should such transfer occur, transfusion could potentially increase cell surface CD55 and CD59 on patients'cells, decreasing their sensitivity to complement and slowing the rate of intravascular hemolysis.

Many GPI-anchored proteins can be easily detected on the cell surface by flow cytometry using fluorescent-labeled monoclonal antibodies. To test our idea, we required a further distinguishing assay for patient and donor erythrocytes, which was provided by Dolichos biflorus, a lectin specific for cells of group A₁ blood that is routinely used in a fluorescein-isothiocyanate (FITC)–conjugated form to resolve some clinical problems in transfusion medicine.¹⁵  In this study, patients with PNH with blood group A₁ were given transfusions with compatible RBCs of blood group O. Pretransfusion and posttransfusion samples were analyzed by flow cytometry for CD55 and CD59 expression on recipient cells identified by lectin labeling.

Samples of venous blood were obtained and analyzed from 6 healthy volunteers, 6 patients with PNH who had not had transfusions for the previous 4 months, and 4 patients having many transfusions (non-PNH) receiving RBCs. Informed consent was obtained according to protocols approved by the Institutional Review Board of the National Heart, Lung, and Blood Institute. The diagnosis of PNH was based on clinical signs of intravascular hemolysis and flow cytometric evidence for deficient granulocyte expression of CD16 and CD66 by flow cytometry. All 6 patients with PNH were transfusion-dependent.

Blood was drawn from group O volunteer donors who met the standards of the American Association of Blood Banks (AABB) for donation. RBC units were stored in Adsol buffer and washed to remove plasma-containing complement. Six PNH patients with blood group A₁ blood who required transfusion as judged by their treating physician were given 3 U of washed, compatible group O blood.

Erythrocytes from patients were obtained before and after transfusion, washed 3 times with phosphate-buffered saline (PBS), adjusted to a concentration of 2 × 10⁴/μL, and incubated in the dark and at room temperature with phycoerythrin (PE)–conjugated CD55- and CD59-specific monoclonal antibodies (mAbs; Monosan, Caltag, San Francisco, CA). Samples were then incubated for 30 minutes at 4°C with fluorescein isothiocyanate (FITC)–conjugated lectin, Dolichos biflorus (which binds specifically to RBCs of group A₁; Sigma, St Louis, MO). For myeloid cells, unseparated whole blood was diluted and incubated with FITC-conjugated CD15 mAb (Becton Dickinson, San Jose, CA), which binds specifically to granulocytes, and then stained with either PE-conjugated CD55- or CD59-specific mAbs (Becton Dickinson). In a separate set of experiments, RBC-derived microvesicles reacting with antiglycophorin were examined and distinguished from intact erythrocytes on the basis of size and granularity in a flow cytometer and from cellular debris by use of specific fluorescent mAb binding (glycophorin). Binding of CD59 and CD-55 PE-conjugated mAb to RBCs or RBC-derived microvesicles was expressed as mean channel fluorescence (MCF) intensity or as a percentage of the particles binding mAb.

Samples were analyzed using an EPICS XL-MCL flow cytometer (Beckman Coulter, Miami, FL) equipped with a 15-MW air-cooled argon ion laser operating at 488 nm. Single-cell suspensions were first analyzed for size and granularity by forward and side scatter, using 488 laser light to establish an RBC gate. The FITC-conjugated lectin was measured by a fluorescent detector equipped with 525-BP filter while blocking the 488 laser light with a 488 blocking filter; the PE florescence was measured by the detector equipped with 575-nmBP filter. To assess binding of CD55 and CD59, the analysis was confined to lectin-positive cells (> 99%). PE- or FITC-conjugated mouse isotype-matched mAb served as controls (Becton Dickinson). Light scatter properties were used to establish a granulocyte gate when appropriate.

To determine if FITC–Dolichos biflorus distinguished group O and group A₁ blood cells, we stained individual samples of patient and donor blood, mixed these cells in defined proportions, and performed flow cytometric analysis (Figure 1).

Multiple control experiments were performed to validate the reproducibility of the experiments. The optimal amount of anti-CD59 mAb used to stain a fixed number of RBCs was determined by titration; subsequently an amount in excess was used to stain patient cells. Similar experiments were performed using labeled lectin, but it was not possible to add an excess of the reagent to the RBC mixture because it produced agglutination. Even though we were cautious to avoid agglutination, RBC gates were drawn very tightly, based on side scatter/forward scatter to exclude clumped cells. The number of group O cells staining with lectin was always less than 0.5%; however, the number of group A₁ cells that did not stain with the lectin varied from 0% to 8% (because it was not possible to add excess lectin to ensure binding to all cells and because of differential expression of A₁ antigen¹⁶ ). The small number of recipient A₁ cells that failed to bind lectin was not considered significant for our study because lectin binding was used only to positively identify group A₁ cells and to exclude group O cells.

To verify that RBC microvesicles were responsible for transfer of CD59 to deficient RBCs, we transfused human RBCs and RBC microvesicles prepared from outdated group O erythrocyte units for transfusion into nonobese/severe combined immune deficiency (NOD/SCID) mice. Microvesicles were obtained from stored RBCs by sonication for 10 seconds at 4°C and then separated from intact erythrocytes by differential centrifugation, followed by 5 cycles of washing in PBS. Varying numbers of microvesicles or RBCs were transfused into 7 NOD/SCID mice; another 2 mice received RBCs only. We used flow cytometry and a CD59-specific mAb that did not cross-react with mouse cells to assess transfer of human CD59 to mouse cells; we positively identified mouse erythrocytes by staining with TER119 and CD24 and excluded human erythrocytes by staining with mAb to human glycophorin.

Correlations between tests were measured using the Pearson coefficient of correlation or Fisher exact test where appropriate. The Fisher exact test was used to determine statistical significance.

Six patients with PNH with episodic severe hemolysis and anemia required periodic RBC transfusion, but none had received blood during the 4 months previous to infusion of group O donor erythrocytes. By flow cytometry, donor and patient blood cells could be distinguished by staining with FITC-labeled Dolichos biflorus (Figure 1). Before transfusion, all circulating RBCs in the patients stained with FITC-labeled lectin, but following transfusion, a distinct lectin-negative population was readily observed. When evaluation gates were drawn to include only the labeled groupA₁ (recipient) cells for further analysis, there were significant increases in the level of binding of mAbs specific for GPI-anchored proteins CD59 (and CD55 in some samples) as early as 1 to 2 days following transfusion (Figure 2A-B), but no significant change in staining was observed 1 hour after transfusion. The proportion of

cells and hemoglobin levels were similar at 1 hour and 7 days after transfusion, suggesting that preferential hemolysis of group A₁ GPI-anchored protein-negative cells was unlikely to explain the increased representation of CD59 on recipient cells (Table 1). (Hemolysis of 2-4 g/dL hemoglobin would be required to theoretically account for the observed increase.) In addition, transfer of CD59 to myeloid cells could be demonstrated (Figure 3). Patient 6 received corticosteroids to treat her acute hemolytic episode and showed increases in the amount of hemoglobin A₁ in the weeks following transfusion; the increase in survival of the patient's GPI-anchored protein-negative cells could explain our failure to demonstrate significant transfer in the immediate weeks following transfusion.

Although selective hemolysis seemed an unlikely explanation for the increases in CD59 expression on PNH patient RBCs, we examined myeloid cells to rigorously exclude this possibility because granulocytes are not present in appreciable numbers in leukocyte-depleted RBC units used for transfusion (< 10⁶/U blood). CD16, a GPI-anchored protein on granulocytes but absent from erythrocytes, showed no increases following transfusion, as expected (data not shown). However, granulocytes, defined flow cytometrically by staining with a mAb to CD15, showed increased CD59 following transfusion in all cases (Figure 3).

In vitro GPI-anchored protein transfer from microvesicles to deficient cells is more efficient than cell-to-cell transfer.¹⁴  Although microvesicles are relatively abundant in stored blood,¹⁷  where vesiculation is promoted by low pH, decreased adenosine triphosphate (ATP) levels, and the anticoagulant, they are infrequent in circulating blood¹⁴  where vesiculation primarily occurs extravascularly in the spleen.¹⁸  When we identified microvesicles by size and staining with antiglycophorin and anti-CD59 mAb and assessed microvesicle number in healthy nontransfused controls (n = 5), we found a mean of only 6 microvesicles/μL blood (data not shown). Patients receiving frequent transfusions (aplastic anemia patients without PNH who received an average of 2 U twice monthly) had a mean of 20-fold more microvesicles than did healthy controls and this number increased more than 2-fold at 1 hour after transfusion of unwashed blood. The patients with PNH described in this study who received washed blood showed a 10- to 100-fold increase in microvesicles the day following transfusion (n = 4; Figure 4).

We next assessed the number and GPI-anchored protein content of RBC microvesicles in the units of blood that were transfused and then examined microvesicles in stored blood following washing in unbuffered normal saline (a standard blood banking procedure). When RBC microvesicles were identified by size and antiglycophorin staining and stained with CD55- and CD59-specific mAbs, CD59 expression predominated (Figure 5A); washing blood resulted in a 2- to 4-fold increase in the number of microvesicles in the product (Figure 5B) with a 3-fold increase in CD59 staining of those microvesicles (n = 4). Figure 5C shows an electron micrograph of CD59, antiglycophorin-positive microvesicles purified by flow cytometry sorting.

Six NOD/SCID mice were given transfusions with a mixture of microvesicles derived from outdated units of human blood; another 2 mice were given transfusions of human RBCs. In this model we used flow cytometry and CD59-specific mAbs, which did not cross-react with mouse cells, to assess transfer of human CD59 to mouse cells; we positively identified mouse erythrocytes by staining with TER119 and CD24 and excluded human erythrocytes by staining with mAb to human glycophorin. Small but consistent increases in staining with CD59 mAb were seen in all 6 mice, whereas no significant change in CD59 binding was seen in mice given intact RBCs. The extent of increase in CD59 was roughly proportional to the number of microvesicles transfused. Data from representative mice are seen in Figure 6, which shows scattergrams of TER119-PE⁺ mouse cells staining with human CD59. When double-positive (TER119-PE/CD59-FITC) mouse cells were sorted and examined by fluorescent light microscopy (× 60), mouse RBCs showed homogeneous membrane staining under the FITC filter.

Cell-to-cell transfer of functionally active GPI-anchored protein has been reported in a variety of in vitro systems. Transferred protein appears to be stable and biologically functional.^{6,10,12,19,20}  For example, transfer of CD55 and CD59 to RBCs confers resistance to complement-mediated lysis in the Ham test.¹⁴  For effective transfer to occur, both the GPI anchor and the protein moiety must be intact.²¹  Transferred molecules are inserted into the outer leaflet of the plasma membrane by the acyl/alkyl chains on the GPI moiety. GPI-anchored molecules are clustered in membrane microdomains that resist extraction by detergents.²²  These GPI-anchored proteins actively take part in membrane vesicle formation, resulting in vesicles rich in GPI-anchored protein.^22,23  Some laboratory evidence suggests that direct cell-to-cell transfer may neither occur nor be required; GPI-anchored proteins may be initially transferred from vesicles or liposomes originating from the donor cell.²⁴  Our previously published data were consistent with these reports because we could demonstrate little transfer between normal and abnormal cells after short incubation times in vitro, whereas significant transfer was observed when deficient cells were incubated with microvesicles obtained by sonication of normal erythrocytes.¹⁴  The absence of direct cell-to-cell transfer and the relative rarity of naturally occurring circulating microvesicles²⁵  may explain the lack of significant in vivo transfer of GPI-anchored proteins between patients' normal cells and the PNH clone and the persistence of a clear phenotypically defined population of deficient cells. RBCs stored for transfusion readily form microvesicles, which may facilitate transfer of GPI-anchored proteins following transfusion.^26-29 

In this study, we observed 2 GPI-anchored proteins, CD55 and CD59, to be increased on recipient erythrocytes and granulocytes after blood transfusion. Significant transfer could be demonstrated as early as 1 day following transfusion and persisted several days. Increased concentrations of CD59 on patient granulocytes, but no transfer of a GPI-anchored protein (CD16) absent from RBCs, suggest that these findings were unlikely to be a result of selective hemolysis or destruction of recipient cells. Furthermore, the proportion of patient group A₁ blood remained constant from 1 hour to 1 day and 1 week following transfusion, and the calculated amount of hemolysis required to explain observed increases in CD59 expression would have had to be large and likely to be obvious. The absence of similar CD55 increases in many patients also argues against hemolysis.

Extrapolation of our observations to clinical practice should be undertaken with caution. In principle, if in vivo transfer, like in vitro transfer, confers functional activity on recipient cells, a program of regular transfusion might decrease hemolysis through partial correction of the PNH cell membrane defect in complement inactivation. However, even a reduction in the frequency or volume of transfusion might be counterbalanced by transfusional iron overload if intravascular RBC destruction and iron loss were entirely avoided. Hemolysis is frequently associated with a variety of symptoms that do not correlate with the level of anemia, prominently disproportionate fatigue, and also dyspnea and impotence, which might improve with transfusion. Eculizumab,³⁰  a humanized antibody that inhibits the activation of terminal complement components, may potentially be useful in inhibiting hemolysis in these patients in the future and clinical trials will soon be underway.

Of potentially greater impact than hemolysis is the striking thrombotic proclivity of PNH.³¹  Venous thrombosis at unusual sites is a major cause of morbidity and mortality in PNH,³²  and this complication is often refractory to conventional anticoagulation therapies and an indication for allogeneic stem cell transplantation.³³  The mechanism of thrombophilia in PNH is unknown, but if protein transfer were effective in correcting GPI-anchored protein deficiencies on platelets³⁴  or if erythrocyte lytic products were themselves thrombogenic,^35,36  regular transfusion could be prophylactic. Washed cells, historically administered in PNH to remove complement,³⁷  might provide RBC microvesicles more effectively and require further study.