British loumaf ojHaematology. 1990, 75, 585-590

Estimation of PI-bound proteins on blood cells from PNH patients by quantitative flow cytometry T. PLESNER, N . E. HANSENA N D K. CARLSENDepartment of Medicine and Haematology C, Gentofte Hospital, Hellerup, Denmark

Received 8 January 1990; accepted for publication 11 April 1990

Summary. The phosphatidylinositol (PI) bound proteins (acetylcholin-esterase (ACE). decay accelerating factor (DAF), leucocyte function antigen type 3 (LFA-3) and Fcreceptor type I11 (FcRIII))were estimated by flow cytometry on blood cells from four patients with paroxysmal nocturnal haemoglobinuria (PNH), nine patients with ‘non-PNH’haemolytic anaemia, four patients with aplastic anaemia and a reference group of 15 healthy individuals to assess the applicability of flow cytometric measurements in the clinical mapping of the PNH defect. Estimation of DAF on granulocytes or monocytes offered the highest diagnostic sensitivity and specificity and may constitute an easy screening method for the PNH defect. One PNH patient had a negative Ham’s

test at the time of study and normal or near normal levels of PI-bound proteins on erythrocytes, but reduced expression of DAF and FcRIIIon granulocytes and DAF on monocytes. The analytical and biological coefficient of variation for flow cytometric estimation of PI-bound proteins was in the range of 48-13% and 12-24%, respectively. Blood samples should be analysed without delay, since storage produced spuriously high results. The results were expressed as molecules per cell after calibration with commercially available standards and validated by comparison with previously reported results obtained by other methods. It is proposed that this way of reporting flow cytometric results should be generally adopted to facilitate comparison of results between laboratories.

Paroxysmal nocturnal haemoglobinuria (PNH) is an acquired stem cell disorder (Rotoli & Luzzatto, 1989). Recently it has been demonstrated that a number of cell membrane proteins are absent or expressed in reduced amounts by PNH blood cells (Dockter & Morrison, 1986; Kinoshita et al, 1985; Medofet al, 1987; Nicholson-Weller et al, 1983, 1985: Zalman et al, 1987). Some of these protect the cell membrane against the action of complement. The common underlying defect seems to be a failure to form a phosphatidylinositol (PI) anchor for binding of these proteins to the cell membrane (Davitzet al, 1986; Selvaraj et al, 1987, 1988). At present there is a need to establish a test system which as broadly and sensitively as possible can detect the PNH defect in the clinical laboratory. The aim of the present study was to explore the role of flow cytometric measurements of PI-bound proteins in the diagnosis of PNH, and to establish a reference material for such measurements. In addition to the mapping of the defect such measurements may more sensitively expose the PNH defect, if, for example, increased haemolysis has destroyed the most sensitive erythrocytes,

making the conventional tests (Ham’s test and the sucrose haemolysis test) negative. MATERIALS AND METHODS

Blood samples. Blood was collected after informed consent by venous puncture during the day from four patients with a well-established diagnosis of PNH (one male and three females, age range 29-64 years). Nine patients with ‘nonPNH’ haemolytic anaemias, four patients with aplastic anaemia and 1 5 healthy staff members (seven males and eight females, age range 29-52 years) served as a reference population. The PNH patients were studied twice with a 6month interval. The ‘non-PNH’haemolytic anaemias were hereditary spherocytosis (three), sickle cell anaemia (one), Coombs positive haemolytic anaemia (three) and acquired Coombs negative haemolytic anaemia with reduced survival of transfused erythrocytes (two). Blood anticoagulated with EDTA was processed for cell marker analysis within 1 h of sampling. Bone marrow was obtained by aspiration and biopsy from the iliac crest. Immunofluorescence. Erythrocytes were labelled in 50 pl diluted blood (1:1000 with isotonic phosphate buffered saline (PBS) containing sodium azide 10 mmol/l and bovine serum albumin 1 g/l) by addition of 50 pl optimally dilutf,d

Correspondence: Dr Torben Plesner. Department of Medicine and Haematology C. Gentofte Hospital, 65 Niels Andersensvej. DK-2900 Hellerup. Denmark.

585

586

T. Plesner, N. E. Hansen and K. Carlsen Table I. List of antibodies

IgG

Ig

(CD No.)

Source

Clone

subclass

or CS

Conversion factor (P/F)

Reciprocal dilution

Anti-Leu11 (CD16) Anti-DAF (CD55) Anti-DAF (CD55) Anti-ACE (-) Anti-LFA-3 (CD58)

Becton Dickinson Kinoshita Anstee Novo-Nordisk Anstee

NKP 1 5 IA 10 BRIC 110

IgGl

Ig

5.5

20

W 2 a

Ig CS

5.1

5000

8.3 5.8 9.3

1 1

Antibody

F40 BRIC 5

IgGl IgGLb IgGza

cs CS

100

CD = cluster of differentiation. Ig = isolated immunoglobulin; CS = culture supernatant. Conversion factor = FITC to protein conversion factor.

monoclonal antibody (anti-acetylcholinesterase (ACE), antiDAF (clone IA lo), and anti-LFA-3 (for details see Table I)). After incubation in the dark for 30 min at 0-4OC the cells were washed three times with PBS at 4°C by centrifugation at 200 g for 5 min and subsequently labelled with fluorescein isothiocyanate (FITC) conjugated rabbit anti-mouse Ig (lot No. 048, dil. 1 : 100, Dakopatts). Leucocytes were labelled with anti-DAF (see above) and FITC conjugated anti-FcRII, (cf. Table I) by adding 50 p1 antibody to 50 p1 undiluted EDTA anticoagulated blood. After incubation in the dark for 1 5 min at room temperature erythrocytes were lysed by addition of 2 ml ‘Lysing Solution’TM) as suggested by the manufacturer (BectonDickinson). For indirect immunofluorescence with anti-DAF, FITC rabbit anti-mouse Ig was added after two washings of the lysed blood. Negative controls were labelled with PBS and FITC rabbit anti-mouse Ig or monoclonal FITC IgG control (Becton Dickinson). After three final washings the samples were analysed by flow cytometry. Flow cytometry. A FACScanTMflow cytometer was calibrated with CaliBriteTMbeads using the AutoCompTMcomputer program (Becton Dickinson). Erythrocytes were analysed by counting 5000 cells without gating and measuring the FITC fluorescence (median) with logarithmic amplification. Leucocytes were analysed by collecting 5000 cells with a forward scatter larger than channel No. 200 (linear amplification). Granulocytes, monocytes and lymphocytes were identified in the forward versus side scatter display (lin., lin. mode), gated and analysed separately for FITC fluorescence (median, log amplification). The number of FITC molecules bound per cell was calculated from the median fluorescence intensity using a standard calibration diagram created from flow cytometric measurements of standard beads expressing 7.0 x lo3to 1.8 x lohFITC molecules per bead (control No. 040189, Flow Cytometry Standards CorporationTM)(Fig 1)(Oonishi & Uyesaka. 1985). The FITC to protein ratio was determined for each of the monoclonal antibodies using Simply CellularTMbeads with a capacity to bind 1.4 x lo5mouse Ig molecules of IgG,. IgGz, and IgGLb isotype per bead (control No. 070189, Flow Cytometry Standards CorporationTM).The mean number of protein molecules per cell was calculated by multiplying the difference in FITC molecules per cell of sample and corresponding negative control with the estimated protein to FITC ratio assuming that each surface exposed protein molecule binds

Channel No. Median 1000

-

750 -

500 -

I

103

I

lo4

I

105

1

106

107

FITC molecules per cell Fig 1.Standard calibration diagram for estimation of number of FITC

molecules per cell. one monoclonal antibody molecule. When two peaks of cells differing in fluorescence intensity could be identified in the FACS histogram the median fluorescence intensity and number of molecules per cell was determined separately for each peak. The relative number of cells in each peak was determined from the respective peak areas. Analytical and biological Variation. Blood samples from 10 healthy individuals were divided and processed separately for double determinations. The analytical coefficient of variation (CV) was calculated from CV=JIC(Xl -XJ2/2N] x 100/X mean %. The biological day-to-day variation was determined in two healthy individuals on 10 consecutive days and CV calculated from CV=J[C(X,-X mean)’/N- 11 x 100/X mean % (Youden, 1951). Other experiments. Blood samples were stored for 3 and 6 h at room temperature and then processed for immunofluorescence labelling and flow cytometry as described. Leucocytes were isolated by sedimentation of erythrocytes with Dextran T500 (Pharmacia) (Borregaard et al, 1983). labelled with antibodies and analysed by flow cytometry as described. In some experiments the membranes of unseparated blood cells

PI-bound Proteins on Blood Cells in P N H

587

Table 11. Blood and bone marrow findings in PNH patients

Patient ~~

Haemoglobin (g/dl) Leucocytes (10~11) Platelets (lO’/I) Reticulocytes (109/1) D H (U/U Bilirubin (pmolll)

Ham’s test Sucrose test Bone marrow cellularity Bone marrow erythroblasts (%) Bone marrow iron

L2

Si

AB

EJ

Reference

12.4/13.0* 5.0/4.4 138/163 262/NE 2830/335(’ 19/34

9.9/11.9 2.613.3 164/200 65/99 1016/NE 16/NE

12’6/11.9 3.713.4 146/190 156/192 239011990 18/17

12.7/13.1 2.8/2.9 196/203 42/52 761/707 8/10

11.3-16’1 3‘0-9.0 125-350 7.4-110 150-450 4-1 7

+

( W E ( +W E

(+)/(+)

I 50 D

I 30 D

(+)I-

+ )lNE

( + ) A +) (+I/-

I 48 D

NE NE NE

-1-

(

I = increased: D = decreased: NE = not examined. ( + ) =weakly positive: - =negative. * Values at first and second time of study respectively: bone marrow only studied at first occasion.

and isolated leucocytes were made permeable by brief (2 s) exposure at room temperature to buffered formol acetone (BFA)(Slaper-Cortenbach et al, 1988)to allow simultaneous labelling of surface exposed and intracellular proteins. RESIJLTS Haemoglobin, blood cell count, plasma lactate dehydrogenase and bilirubin levels from four patients with PNH are shown in Table I1 together with the results of Ham’s and sucrose haemolysis tests and bone marrow findings. All patients showed signs of active haemolysis at the times of study. Ham’s test was negative in one patient (AB), who previously was found to have a positive test. The sucrose test was positive in all cases. All patients had chronic haemolytic anaemia (duration of disease 9-20 years). The median number of PI-bound protein molecules ( x 10-j) on blood cells from PNH patients ‘non-PNH’ haemolytic anaemias, aplastic anaemias and healthy individuals is shown in Table 111. together with previously published findings in normal blood cells and the corresponding references. When the fluorescence histograms demonstrated two subpopulations differing with regard to the severity of the PNH abnormality (Fig 2), the relative number of cells (%) and result is shown for each peak (Table 111). Most often one of the peaks was within the normal range. The frequency of abnormal findings in PNH and other patients is shown in Table IV. Optimal diagnostic sensitivity and specificity was obtained by estimation of DAF on granulocytes or monocytes. The analytical variation obtained by flow cytometric double determinations of PI-bound proteins on blood cells from 1 0 healthy individuals is shown in Table V. The day-today variation of 10 estimates on two healthy individuals is shown in Table VI. The influence of storage of blood samples prior to analysis was studied by estimation of granulocyte-FcRIIIat 0, 3 and 6 h after collection of blood samples. A 40% increase was found

after 3 h, and this increased further up to 60% after 6 h. When leucocytes were isolated as buffy coat the granulocytes respectively, expressed 30% and 120% more DAF and FcRI~~ monocytes expressed 70% more DAF, while lymphocyte-DAF remained constant compared to unseparated blood cells. Permeability of cell membranes due to buffered formolacetone resulted in a 40% increase of both granulocyte-DAF and -FcRlllra 90% increase of monocyte-DAF and constant levels of lymphocyte-DAFin unseparated blood. Permeability of buffy coat cells resulted in a further increase of monocyteDAF to 120% that of unseparated, untreated monocytes, while granulocyte-DAF and -FcRiiI and lymphocyte-DAF remained constant. Treatment of buffy coat cells with ‘Lysis Buffer’,which is used to remove erythrocytes after labelling of leucocytes in unseparated blood, resulted in a 60% and 70% reduction of DAF on granulocytes and monocytes respectively, while lymphocyte-DAF and granulocyte-FcRIII remained constant. DISCUSSION We have measured a number of PI-bound proteins on erythrocytes and leucocytes from PNH patients and a reference population by flow cytometry and expressed the results as number of protein molecules per cell using commercially available calibration standards. Estimation of DAF on granulocytes or monocytes offered the highest diagnostic sensitivity and specificity. Of particular interest was the finding of normal or near normal expression of PIbound proteins on erythrocytes in one PNH patient with a negative Ham’s test. The diagnosis was well established in this case since a positive Ham’s test had been found on previous occasions and the sucrose test remained positive. Our finding may be explained by a selective loss of the most severely affected erythrocytes during periods of active haemolysis. It illustrates the importance of also studying the leucocyte populations in suspected cases of PNH. Heterogeneity of expression of PI-bound proteins was a frequent

588

T. Plesner, N. E. Hansen and K. Carlsen Table 111. Expression of PI-bound proteins on blood cells from PNH patients non-PNH haemolytic anaemias (HA), aplastic anaemia (AA) and healthy individuals (median number protein molecules x per cell)

PNH patients

Cell type

Protein LP

Granulocyte FcR111 I* I1

Granulocyte DAF

SJ

Lymphocyte DAF

Monocyte

DAF

Erythrocyte

ACE

I I1

Erythrocyte

LFA-3 I I1

Erythrocyte

DAF

I

I1

72-176

7.7

6.0

77%=2.2 40%= 2.8 AA=43-127 23%= 34 60%= 57

(135; Selvaraj et al. 1988)

1.5

5.6

80%=2.6 46%=O.O 20%= 39 54%=35

HA= 25-71

22-107

5.1

0.0

74%= 0.0 26%=28

AA=49-138

(85: Kinoshita et al, 1985)

HA=10-66

5.6-62

17

53%=1.0 47%=41

43

12

2.0

7.1

7.7

2.0

0-0

14 4.6 C

15 1.0

I

I1

HA= 30-138

81%=2.8 48%= 2.2 19%=33 52%=55

I I1

EJ

Own reference interval and (previously published values; ref.)

1.7 21

I I1

AB

Non-PNH haemolytic anaemia and aplastic anaemia (range)

5 6 X ~ 1 . 5 AA=8.7-112 44%=55

(33; Kinoshita et al, 1985)

56%=1.5 HA=44-204 44%=87

39-143

75%=0.0 42%=0.0 AA=42-179 25%=51 58%=49

1.2 48%=0.0 52%=22

9.3

14

HA = 5.2-23

8.7

4.1

15

AA= 10-24

2.4

11

46%=0.0 54%= 56

35

30%=0.9 HA=20-73 70% = 48

(68; Kinoshita et nl, 1985) 9.9-20

34-71

7.8 52%=1.9 27 48%=38

42

AA= 34-73

2.6 50%=0.0 50%=31

15

26%=0.0 HA=13-41 74%=31

16-40

5.1 43%=1.5 57%=28

23

23%=0.0 AA=24-43 77%=32

(3.3: Kinoshitaet a!. 1985)

Two values indicate that two subpopulations were seen in the fluorescence histogram. The relative number of cells in each peak is given in per cent (%). * I =first, I1 =second time of study.

finding in PNH, but peaks of cells with low and high expression could easily be identified in the fluorescence histograms for separate determination of their median fluorescence intensity. The relative number of cells with low and high expression could be estimated from the areas of the respective peaks in the FACS histograms. Our results corresponded well with previously published values from normal blood cells obtained with different methods except for erythrocyte-DAF,where we found significantly higher amounts (mean value 30 versus 3 . 3 x lo3 molecules per cell, Table 111). The reason for this discrepancy is presently not clear. Estimation of erythrocyte-DAF with two different monoclonal antibodies gave similar results in our experiments. The analytical precision of our flow cyto-

metric measurements was adequate with a coefficient of variation of 4-8-13%, and similar to previous findings by flow cytometry on cell lines (Poncelet & Cayaron, 1985).The biological variation was 12-24% in healthy individuals and may be larger in some diseases. In our study we found examples of both reduced and elevated expression of PIbound protein in 'non-PNH haemolytic anaemias and aplastic anaemias. Of particular interest in this clinical setting was the finding of reduced expression of granulocyteFcRIIIin some cases of haemolytic anaemia (one sickle cell anaemia and one spherocytosis) and in two of four cases of aplastic anaemia and also the low expression of one or more PI-bound erythrocyte proteins in three of four cases of extraerythrocytic haemolytic anaemias. Flow cytometric studies of

PI-bound Proteins on Blood Cells in PNH 10

10

I I

I

.

Patient EJ

Patient EJ -Granulocyte

- control

-- -Granulocyte

- DAF

.*

.

5

5

C

0

10

.’.*

O

.

a

- control Granulocyte - FcRll,

-Granulocyte

...

* . .

-.-..... .... lb’

rb”

lb’

1L4

L

Patient AB

Patient AB

-Granulocyte - control

-Granulocyte - control

- - - Granulocyte - DAF

Granulocyte - FcR,,,

Fig 2 . Histograms obtained by flow cytometry of decay accelerating factor (DAF) and Fc-receptor type 111 (FCRIII)on granulocytes from PNH patients AB and EJ.

Table IV. Frequency of reduced expression of PIbound proteins on blood cells from PNH-patients, non-PNH haemolytic anaemias (HA) and aplastic anaemias (AA) PNH

Granulocyte-FcRIIl Granulocyte-DAF Lymphocyte-DAF Monocyte-DAF Erythrocyte-ACE Erythrocyte-LFA-3 Erythrocyte-DAF

I

I1

414

414 414 114

414 314 414 314

414 314

314

314

414

314

HA

AA

219

214

019 019 019

014

319 319 119

014 014

Table V. Analytical variation (protein molecules x per cell)

Cell type

Protein

Mean

SD

(37%

Granulocyte

F~RIII DAF

98 45

4.7 5.1

4.8 11

Lymphocyte

DAF

35

3.9

13

Monocyte

DAF

61

5.5

Erythrocyte

ACE LFA-3 DAF

17 56 30

1.7 2.8

2.9

9.0 10

5.1 9.6

014 014 014

I and I1 = First and second time of study respectively.

larger patient numbers including cases with the aplastic variant of PNH will be needed to validate estimation of DAF on granulocytes or monocytes, which seemed most promising in our study, as a reliable screening method for PNH. We have used Simply CellularTM beads for the estimation of protein to FITC ratio. Compared to spectrophotometric estimation of this ratio (The 6; Feltkamp, 1970) the beads offer the possibility of using unconjugated monoclonal

antibodies for estimation of membrane proteins. Also problems with spectrophotometric absorbance from contaminating proteins (e.g. ‘carrier proteins’) are circumvented. We have expressed our results as number of molecules per cell, but acknowledge the fact that no reference method is presently available to give the true value. Our results depend on the antibodies and calibration standards used. However, we prefer this way of reporting our results, because with an increasing interest in standardized procedures, interlaboratory comparison of results will be facilitated. It was found that storage and isolation of leucocytes altered the results stressing the importance of standardizing and reporting all rele-

590

T. Plesner, N. E. Hansen and K. Carlsen Table VI. Biological (‘day-to-day’) variation of PI-bound proteins in two healthy individuals (median number of protein molecules x per cell) Donor I

Donor I1

Cell type

Protein

Range*

Mean

Granulocyte Granulocyte Lymphocyte Monocyte Erythrocyte Erythrocyte Erythrocyte

FcRm DAF DAF DAF ACE LFA-3 DAF

88-1 71 77-107 31-62 87-143 9’9-16 36-53 16-28

124 94’ 51 110 13 43 22

23 17 19 17 18 13 16

Ranget

Mean

CV% __

66-149 31-77 27-59 56-1 12 9‘9-20 34-61 18-36

105 63 38 88 15 47 27

21 24 18 23 23 12 18

* Ten estimations. vant details about the handling of cells prior to estimation of membrane proteins. The increased amount of FcRIII on isolated granulocytes and DAF on isolated granulocytes and monocytes compared to unseparated blood cells may possibly be explained by translocation from the intra- to the extracellular domain, since permeability of membrane of unseparated cells due to BFA gave higher values. Also steric changes on the membrane may alter the accessibility of antibodies to the target antigen and influence the results. Such changes can be induced by chemicals (e.g. BFA, ammonium chloride in the ‘Lysing Solution’), changes in temperature, etc.. and these factors should be carefully controlled to give reproducible and comparable results. ACKNOWLEDGMENTS This work was supported by the Danish Medical Research Council and the Danish Cancer Society. Monoclonal antiDAF (IA 10)was donated by Dr T. Kinoshita, anti-LFA-3 and anti-DAF (BRIC 110)by Dr D. J. Anstee and anti-ACE by Dr 0. J. Bjerrum. E. Kjaersgaard’shelpful advice during preparation of the manuscript is gratefully acknowledged. KEFERENCES Borregaard, N., Heiple. J.M., Simons, E.R. & Clark, R.A. (1983) Subcellular localization of the b-cytochrome component of the human neutrophil microbicidal oxidase: translocation during activation. Journal of Cell Biology, 97, 52-61. Davitz. M.A., Low, M.G. & Nussenzweig, V. (1986) Release of decayaccelerating factor (DAF)from the cell membrane by phosphatidylinositol-specificphospholipase C (PIPLC).Journal of Experimental Medicine. 163, 1150-1161. Dockter. M.E. & Morrison, M. (1986) Paroxysmal nocturnal hemoglobinuria erythrocytes are of two distinct types: positive or negative for acetylcholinesterase. Blood, 67, 540-543. Kinoshita. T., Medof. M.E.. Silber, R. & Nussenzweig, V. (1985) Distribution of decay-accelerating factor in the peripheral blood of normal individuals and patients with paroxysmal nocturnal hemoglobinuria. Journal of Experimental Medicine, 162, 75-92. Medof. M.E., Gottleib. A.. Kinoshita. T.. Hall, S., Silber, R., Nussenzweig. V. & Rosse, W.F. (1987) Relationship between decay accelerating factor deficiency, diminished acetylcholinesterase

activity, and defective terminal complement pathway restriction in paroxysmal nocturnal hemoglobinuria erythrocytes. Journal of Clinical Investigation, 80, 165-174. Nicholson-Weller, A., March, J.P., Rosenfeld, S.I. & Austen, K.F. (1983) Affected erythrocytes of patients with paroxysmal nocturnal hemoglobinuria are deficient in the complement regulatory protein, decay accelerating factor. Immunology. 80, 5066-5070. Nicholson-Weller, A,. Spicer, D.B. & Austen, K.F. (1985) Deficiency of the complement regulatory protein, ‘Decay-AcceleratingFacror,’ on membranes of granulocytes, monocytes and platelets in paroxysmal nocturnal hemoglobinuria. New England Journal of Medicine, 312, 1091-1097. Oonishi. T. & Uyesaka, N. (1985) A new standard fluorescence microsphere for quantitative flow cytometry. Journal o/ Immunological Methods, 84, 143-1 54. Poncelet, P. & Cayaron, P. (1985)Cytofluorometricquantification of cell-surfaceantigens by indirect immunofluorescence using monoclonal antibodies. Journal of Immunological Methods, 85, 65-74. Rotoli, B. & Luzzato. L. (1989) Paroxysmal nocturnal haemoglobinuria. Bailliere’s Clinical Haematology, Vol. 2, No. 1. Aplastic Anemia (ed. by E. C. Gordon-Smith), pp. 113-138. Bailliere Tindall, London. Selvaraj, P.. Dustin, M.L., Silber, R., Low, M.G. & Springer, T.A. (1987) Deficiency of lymphocyte function-associated antigen 3 (LFA-3) in paroxysmal nocturnal hemoglobinuria. Functional correlates and evidence for a phosphatidylinositol membrane anchor. Journal of Experimental Medicine, 166, 1011-1025. Selvaraj, P., Rosse, W.F.. Silber,R. &Springer,T.A. (1988)The major Fc receptor in blood has a phosphatidylinositol anchor and is deficient in paroxysmal nocturnal haemoglobinuria. Nature. 333, 565-567. Slaper-Cortenbach, I.C.M., Admiraal. L.G., Kerr, J.M., van Leeuwen. E.F., von dem Borne, A.E.G.K. & Tetteroo, P.A.T. (1988) Flowcytometric detection of deoxynucleotidyl transferase and other intracellular antigens in combination with membrane antigens in acute lymphatic leukemias. Blood, 72, 1639-1 644. The, T.H. & Feltkamp. T.E.W. (1970) Conjugation of fluorescein isothiocyanate to antibodies. Immunology, 18,865-873. Youden, W.J. (1951) Statistical Methodsfor Chemists, pp. 1 3 and 16. John Wiley & Sons, New York. Zalman, L.S., Wood. L.M., Frank, M.M. & Muller-Eberhard, H.J. (198 7) Deficiency of the homologous restriction factor in paroxysmal nocturnal hemoglobinuria. Journal of Experimental Medicine, 165, 572-577.