Clinical evaluation of an aerolysin-based screening test for paroxysmal nocturnal haemoglobinuria




Recently, a toxin produced by Aeromonas hydrophila was demonstrated to bind directly to the glycosyl-phosphatidyl-inositol (GPI) anchor. After coupling it to a fluorescent dye and applying it in fluorescence-activated cell scanning (FACS), this property was exploited to detect GPI-negative cells in the diagnosis of paroxysmal nocturnal haemoglobinuria (PNH).


We used this reagent according to a very simple staining protocol followed by single-colour FACS and compared the results in patients with PNH and normal controls with those obtained with antibody-mediated detection of cells lacking GPI-anchored proteins.


We observed very good concordance between the two methods, with correlation coefficients (R2) of quantified GPI-deficient cell populations ranging from 0.952 to 0.969. The lower limit of detection was determined at 0.50% GPI-negative cells, which was in the range obtained with double-colour staining with antibodies (0.20–1.00%, depending on the antibody). A significant correlation was observed between the fraction of GPI-negative granulocytes and laboratory parameters of haemolysis, with the erythrocyte creatine having the best correlation (R2 = 0.671, P < 0.0001).


Using this protocol, we were able to reliably diagnose PNH with a high sensitivity. The test allows the identification of GPI-negative granulocyte populations as small as 0.5%. © 2005 Wiley-Liss, Inc.

For many years the diagnosis of paroxysmal nocturnal haemoglobinuria (PNH) was made by demonstrating the susceptibility to complement-mediated lysis (1, 2). Subsequently, demonstration of the absence of glycosyl-phosphatidyl-inositol (GPI)–anchored proteins from the cell surface using antibodies and fluorescence-activated cell scanning (FACS) largely served to replace this approach (3–5). Typically, only a fraction of the cells shows the PNH immunophenotype, and it is the demonstration of two cell populations, one positive and one negative in GPI-anchored proteins, that is diagnostic for the disease. In different cell populations, various antigens are investigated.

Aerolysin, the toxin of the bacterium Aeromonas hydrophila, binds directly to the GPI anchor. It is secreted as an inactive protoxin, proaerolysin, that is converted to the active form, by various mammalian proteases, through proteolytic removal of a C-terminal peptide (6, 7). Aerolysin has been used in diagnostic procedures to demonstrate the resistance of PNH erythrocytes to aerolysin and to enrich for GPI-negative PNH cells (8, 9). Also, two point mutations were introduced to obtain a protein that still binds GPI upon activation but lacks lytic activity. By coupling this mutant proaerolysin to a fluorescent marker (Alexa Fluor 488), a reagent (FLAER) was produced that stains cells containing GPI proteins but not PNH cells lacking GPI (10). Because this reagent detects the GPI anchor itself, it can be used to investigate all cell types, irrespective of the proteins normally expressed, except erythrocytes, which do not express the necessary activating proteases. In contrast to immunophenotyping, where several antibodies must be employed to investigate different cell populations, FLAER can provide the same information in a single reaction.

We used FLAER in whole blood according to a very simple and fast staining protocol as a screening test for the diagnosis of PNH. We determined the lower limit of detection with this simple procedure and compared the results with those from conventional immunophenotyping performed in parallel. In addition, results obtained with FLAER were correlated with known parameters of haemolysis.


Detection of GPI Anchor by FLAER Assay

FLAER was obtained from Protox Biotech (Victoria, Canada). It was stored at 10−6 M in phosphate buffered saline (PBS) at −20°C. To 25 μl of whole blood, anticoagulated with ethylenediaminetetraacetic acid or heparin, 5 μl of the FLAER solution was added, mixed gently, and incubated for 10 to 15 min in the dark at room temperature. After addition of 500 μl of erythrocyte lysing reagent (0.829% NH4Cl, 0.1% K2HCO3, and 0.0037% ethylenediaminetetraacetic acid in distilled water), samples were measured in parallel with a normal control by single-colour FACS analysis in channel 1, corresponding to green fluorescence, on a FACScalibur (Becton Dickinson, Basel, Switzerland). Leukocyte subpopulations were gated on a forward versus side scatter plot.

A normal control was always stained in parallel. The threshold fluorescence for negative versus positive cells was set where the main peak of this normal control intersected the x-axis.


To 100 μl of whole blood, antibodies (Becton Dickinson) were added at the dilution suggested by the manufacturer and incubated for 15 to 30 min at room temperature in the dark. Erythrocytes were lysed with 2 ml of lysing reagent and leukocytes washed once with 3 ml of standard PBS and resuspended with 500 μl of PBS. Leukocyte subpopulations were identified on a forward versus side scatter plot, which allowed gating for granulocytes and monocytes. CD55-negative cells were investigated in granulocytes and monocytes in this manner. This test therefore was a single-colour fluorescence experiment. For dual-colour fluorescence analysis of CD16, CD24, and CD66b, a monoclonal antibody with a different fluorescent label (CD13/phycoerythrin) was added to the assay to improve the identification of myelomonocytic cells. Thus, the investigated cell populations, granulocytes and monocytes, were identified by virtue of their positivity for CD13 and their light scatter properties. CD59 staining was performed with CD3/phycoerythrin, which allowed elimination of T cells that could have contaminated the gates for granulocytes and monocytes based on forward versus side scatter characteristics. CD14 was investigated with dual-colour fluorescence analysis on monocytes. A CD64/phycoerythrin-labelled monoclonal antibody served to discern monocytes from lymphocytes contaminating the monocyte gate in forward versus side scatters. Lymphocytes were not analysed. We investigated CD16, CD66b, CD24, CD59, and CD55 on granulocytes and CD14, CD59, and CD55 on monocytes.

A normal control was always stained in parallel. The threshold fluorescence for negative versus positive cells was set between the two, readily visible populations in samples with an obvious PNH component. In cases without two clear populations, the threshold was set below the dot cloud in the normal control in such a manner that only events clearly separated from this dot cloud were counted as nonfluorescent.

Statistical Analysis

Linear correlation and calculation of the correlation coefficient (R2) were performed with StatView software.


Demonstration of GPI-Negative Cells

GPI-deficient cells were sought by using antibodies as indicated in Materials and Methods and FLAER in seven measurements in five patients and seven normal controls. Figure 1 shows the FACS results of patient 5 and the respective normal control. Diagnoses were PNH in two, myelodysplastic syndrome (MDS) with a PNH clone in one, MDS without a PNH clone in one, and aplastic anaemia with a PNH clone in 1 patient (Table 1). Relevant GPI-negative cell populations were observed in four of five patients and in none of seven controls (controls not shown). These results, expressed as dichotomous values of PNH clone present or PNH clone absent, were completely concordant between the two methods (Table 1, last column).

Figure 1.

Demonstration of GPI-negative cells by antibodies and FLAER. A representative example is shown with FACS results for a patient and for a normal control. Parameters measured are as indicated. Gating was performed in a forward versus side scatter plot (Gc, granulocytes; M, monocytes). Antibodies were coupled with fluorescein isothiocyanate (FITC) or phycoerythrin (PE). Antibodies used are indicated on the axes.

Table 1. Percentage of GPI-Negative Cells in Granulocytes and Monocytes With FLAER and Various Antibodies*
Patientno.Diagnosis%Negative granulocytes%Negative monocytesPNH clone
  • *

    Values were rounded to the nearest integer. The two values missing are due to technical errors. Patient 5 was measured three times on different occasions during treatment with antilymphocyte globulin and ciclosporin.

  • AA, aplastic anemia; MDS, myelodysplastic syndrome.

5MDS/PNH51 6753616741633759Yes
  516869535465415047 Yes

Correlation Between FACS Mediated by Antibody Versus FLAER

Fourteen measurements of antibody-mediated FACS were performed in parallel with FLAER analysis. The nonfluorescent fraction was determined for each investigation and the correlation was determined by linear regression analysis. Correlation coefficients (R2) and regression plots are presented in Table 2 and Figure 2, respectively. Excellent correlations between PNH populations detected by the two methods were observed, demonstrating a comparable accuracy of the two methods for the quantification of PNH cells.

Figure 2.

Regression plots of nonfluorescent fractions from FACS mediated by FLAER versus fluorescent monoclonal antibodies. Representative examples of regression plots are shown. Values on the axes represent percentages of nonfluorescent cells. FLAER is always on the vertical axis, and monoclonal antibodies are as indicated on the horizontal axis. Correlation coefficients (R2) are given in the plots. Gc, granulocytes; Mono, monocytes.

Table 2. Correlation Coefficients Between FACS Mediated by Fluorescent Antibodies or FLAER in Granulocytes or Monocytes*
  • *

    Excellent correlation was observed, demonstrating comparable accuracy of the two methods in quantifying PNH cells. Gc, granulocytes; Mono, monocytes.

Gc CD160.959 
Gc CD240.950 
Gc CD550.9680.956
Gc CD590.9690.901
Gc CD66b0.957 
CD14 0.918

Lower Limit of Detection

For analysis of the lower limit of detection, only granulocytes were investigated because they represent the largest fraction and were more easily gated in the forward versus side scatter plot. Fifty normal control samples were assayed by FLAER and seven by immunophenotyping. Values between 0% and 0.48% FLAER-negative events were counted in the granulocyte gate. The values observed with fluorescent antibodies were in the same range in double-colour FACS results, as presented in Table 3 and by a percentile plot in Figure 3A. However, we performed the CD55 staining as a single-colour fluorescence assay, and a higher background was observed (0.07–3.62%; Fig. 3B). In addition, in CD16 staining, higher values were measured in normal controls, but this was due to the absence of CD16 from normal eosinophil granulocytes. In 50 normal controls, the highest value of fluorescence-negative events with FLAER was 0.48%. Therefore, a lower limit of detection can be set at 0.5%.

Figure 3.

Percentile plot of fluorescence-negative events in FLAER versus fluorescent monoclonal antibody-mediated FACS in normal controls. Values on the vertical axis are percentages of fluorescence-negative events. Percentiles are indicated on the horizontal axis. Only the granulocyte (Gc) gate was investigated. A: FLAER versus antibodies used in a dual-colour FACS experiment. B: FLAER versus CD55 staining that was performed as a single-colour FACS experiment.

Table 3. Fluorescence-Negative Events in Normal Controls by FLAER Analysis and Antibody-Mediated FACS*
  • *

    Values are percentages of nonfluorescent granulocytes gated by forward versus side scatter.


FLAER Correlates With Haemolytic Activity

The fraction of FLAER-negative cells as a percentage correlated only very weakly with the parameters of haemolysis LDH, reticulocyte count, and bilirubin. A weak but definitely better correlation was found with the haemolysis parameter of erythrocyte creatine (R2 = 0.671, P < 0.0001; Fig. 4) (11).

Figure 4.

Correlation between the number of FLAER-negative cells and parameters for haemolysis. Linear correlation plots compare percentages of FLAER-negative cells with known parameters of haemolysis. The best correlation was observed with erythrocyte creatine (ec crea;P < 0.0001) (11), although all correlations were rather weak. Gc, granulocytes; LDH, lactate dehydrogenase.


PNH enters the differential diagnosis in every case of Coombs-negative haemolysis. In addition, it has recently been shown that patients with myelodysplastic syndrome carrying a PNH clone, even if it is very small, respond more readily to immunosuppressive treatment (12). Therefore, although PNH is a rare disease, it is quite frequently investigated, and a simple, reliable screening test with sufficient sensitivity for small PNH clones would be of benefit.

Haemolysis in PNH arises from a lack of complement inhibitors from the surface of erythrocytes. This is due to the inability of these cells to synthesize the GPI anchor. Until a few years ago, diagnosis had to rely on indirect methods that demonstrated increased susceptibility to complement-mediated lysis (1, 2) or the lack of GPI-anchored proteins by immunological methods (3–5). With aerolysin, the toxin of A. hydrophila, which specifically binds to the GPI anchor, a more straightforward diagnostic approach has become possible (10).

We have shown that the diagnosis of PNH can be made reliably using this reagent with a very simple and rapid staining protocol and single-colour fluorescence detection. Nonfluorescent events in the FACS analysis of normal controls were always less than 0.5% of peripheral granulocytes. Therefore, the test has a lower limit of detection for PNH clones as small as 0.5%. A comparable performance was obtained with monoclonal antibodies only with dual-colour staining.

Probably by counterstaining the cell population of interest with an antibody labelled with a second fluorescent dye in a FLAER experiment, the lower limit of detection for GPI-negative cells could be decreased. However, this was not the goal of the present method, which was deliberately designed as a rapid and simple screening test for routine analysis.

Lymphocytes could not be investigated with the present method because the gating strategy did not permit separating them reliably from remaining non-lysed erythrocytes and probably erythrocyte ghosts. Staining with an antibody carrying a second fluorescent label would likely solve this problem in a setting where the investigation of lymphocytes was important. Moreover, erythrocytes are not amenable to analysis by FLAER because they lack the proteases necessary for proteolytic activation of proaerolysin (10).

Because the percentage of FLAER-negative cells correlated to a certain degree with haemolytic activity, it can be used as a follow-up parameter to estimate disease severity and response to treatment.

In conclusion, by using this very rapid FLAER-based protocol for FACS, we were able to reliably detect GPI-negative cell populations in patients carrying a PNH clone. PNH clones as small as 0.5% were readily detected. Because of its simplicity and correlation to haemolytic activity, this test also lends itself to serial follow-up of the course of PNH.


We thank Mrs. Ruth Kalberer, Mrs. Evelyne Giabbani, Mrs. Michelle Handschin, Mrs. Francesca Bianda, and Mrs. Leta Fuchs for excellent technical assistance in FACS.