- Email: [email protected]
Evaluation of an automated method of percent reactive antibody determination
Evaluation of an Automated Method of Percent Reactive Antibody Determination Matthew J. Heller, Patrick W. Adams, and Charles G. Orosz
ABSTRACT: A fluorescence-based automated method of percent reactive antibody (PRA) analysis is described. This method utilizes the conventional antibody-mediated, C'-dependent lymphocyte microcytotoxicity assay to detect alloantibodies, but replaces the eosin-based method for detection of cell death with a fluorescence-based method. To identify viable cells, lymphocytes were pretreated with carboxy fluorescein diacetate (CFDA), which fluoresces green, to identify viable cells. To identify dead cells after the reaction with antibody and C', they were treated with propidium iodide (PI), which fluoresces red. Pretreatment of lymphocytes with CFDA did not affect their ability or interact with alloantibodies in the microcytotoxicity assays. When visually analyzed, detection of cell death by fluorescence was as sensitive as detection by eosin exclusion. However electronic detection of fluores-
cence was slightly more sensitive than visual detection. Automation of the fluorescent method required a calculation that converts electronic data to an ASHI score for cell death. One such method is described and evaluated. Both the automated and the conventional methods of analysis were used to obtain PRA values for various sera. There was good correlation between the PRA values obtained with the automated method versus the conventional method. Further, there was good correlation for PRAderived alloantibody specificities obtained with the automated method versus the conventional method. These data demonstrate that automated fluorescence-based PRA analysis is an effective and practical alternative to conventional PRA analysis. Human Immunology 35, 179-187 (19921
ABBREVIATIONS
CFDA PI
5-carboxy fluoroscein diacetate (isomer free) propidium iodide
ALG PHS PMT
antilymphocyte globulin pooled human sera photomultiplier tube
INTRODUCTION Candidates awaiting transplantation often become allosensitized as a result of pregnancy, transfusion, or previous transplantation. These candidates develop circulating alloantibodies that could initiate the hyperacute rejection of a subsequent organ allograft [1]. To avoid this, transplant candidates are regularly tested for the presence of serum alloantibodies. Clinically, serum alloantibodies are detected by their ability to mediate C'dependent lymphocytolysis. Because of the large numFrom the Department of Surgery, Division of Transplantation (MJ.H., P.W.A.), and the Departments of Pathology and Pharmacology (C.G.G.), Ohio State University of College of Medicine, Columbus, Ohio, USA. Address reprint requeststo Dr. M. J. Heller, Clinical Histocompatibility Laboratory, Room N935 Doan Hall, The Ohio State University Hospitals, 410 West Tenth Avenue, Columbus, OH 43210, USA. Received May I 1, 1992; acceptedJuly 22, 1992. Human Immunology 35, 179-187 (1992) © American Society for Histocompatibility and Immunogenetics, 1992
ber of serologically distinct human leukocyte antigen (HLA) alleles, this must be done with a panel o f l y m p h o cytes obtained from a large n u m b e r of different individuals. The panel of lymphocyte donors is selected to represent the distribution and frequency of H L A specificities that are present in the local population. For any given serum, the percent of positive reactions against this panel of lymphocytes is called percent reactive antibody (PRA), and is considered to be an index of allosensitization. If the H L A specificities displayed by each panel m e m b e r are known, the pattern of reactivity in this panel can be used to identify the specific H L A antigens recognized by the alloantibodies in the serum sample. The conventional method of PRA analysis involves the microscopic evaluation of antibody-mediated C'-dependent cell death that is visualized by vital dye (eosin 179 0198-8859/92/$5.00
180
or trypan blue) exclusion [2]. This method is labor intensive and somewhat subjective. Further, each transplant candidate is tested monthly for PRA. Hence, PRA analyses can consume large amounts of time and resources, especially in centers with long transplant candidate lists. An alternative to the conventional method is offered by fluorescence-based technology. This technology, which is currently used for HLA typing, lends itself well to automation [3-6]. An automated PRA methodology would be less subjective than conventional methods and should reduce the tedium of visual PRA evaluation for large numbers of serum samples. Consequently, we have adapted existing fluorescence techniques to the automated determination of PRA. In this communication, we describe an automated method of PRA analysis and of antibody specificity determination. Further, we compare this method with the traditional visual methods for ability to yield PRA values and to define HLA-reactive antibody specificities.
M.J. Heller et al.
Alternatively, 60-member, frozen, CFDA-labeled cell panels were obtained from PelFreez Clinical Systems (Brown Deer, WI) and used according to the manufacturer's instructions. The panel members were chosen to reflect the frequency of HLA epitopes found in the midwestern United States. C'-Mediated Microcytotoxicity Assay
Eosin-based visual analysis. The Amos-modified micro-
METHODS
cytotoxicity assay was employed [2, 9]. Briefly, 1 ~zl of T cells (2-4 x 106/ml) was mixed with 1/zl of test serum in Terasaki wells, and incubated for 30 minutes at room temperature. Five microliters of medium was added to each well and the cells were permitted to settle for 10 minutes. The plates were flicked to remove the serum, 5-/zl/well rabbit complement (PelFreez Clinical Systems) was added, and the plates were further incubated for 120 minutes. Each well was stained with eosin, fixed with 10% formaldehyde in phosphate-buffered saline, and scored for percent cell death with an inverted phasecontrast microscope.
Cell Isolation
CFDA/PI-based fluorescence analysis. To label lympho-
Peripheral blood mononuclear cells (PBMC). Fresh heparinized blood was collected by venipuncture from normal individuals. Leukocytes were isolated following centrifugation and were separated on Ficoll-hypaque density gradients using the method of Boyum [7]. Cells recovered from the PBMC interface were washed three times in Hank's balanced salt solution (HBSS) and resuspended in McCoy's 5A medium containing 20% pooled human serum (PHS).
Purified T cells. Human T-lymphocytes were isolated from PBMCs by negative selection with a commercially available cocktail of monoclonal antibodies and complement (T Lympho-Kwik, One Lambda, Los Angeles, CA), using methods based on those of Clouse et al. [8]. PBMCs (5-10 x 107 cells) were washed by centrifugation (400 g, 10 minutes) and resuspended in 0.8 ml T Lympho-kwik in a Fisher centrifuge tube. The cells were incubated 60 minutes at 37°C, with gentle mixing at 15 minute intervals, overlaid with 0.25 ml of culture medium, and centrifuged for 2 minutes at 700 g. The supernatant containing cell debris was discarded, and the pelleted T cells were washed three times in HBSS and resuspended in McCoy's medium plus 20% PHS. Preparation of Panels For locally prepared cell panels, isolated T cells were labeled with CFDA (see later) and suspended in culture medium plus 8% DMSO at a concentration of 2-4 x 106 cells/ml. These cells were added to Terasaki plates (1 /A/well), which were frozen and stored in liquid nitrogen until used.
cytes with CFDA, purified T-lymphocytes were washed three times in serum-free balanced salt solution. The cell pellet (containing - 2 - 5 x 106 cells) was resuspended in 1.0 ml CFDA (20 ~g/ml; Molecular Probes, Eugene, OR) for 15 minutes at 37°C in the dark. The cells were washed three times in McCoy's 5A medium containing 20% PHS and resuspended at a concentration of 1-2 x 106/ml. These cells were used as target cells in the C'-mediated cytotoxicity assay using identical conditions as described earlier. Following incubation with C', the cells were stained with 5 ~l/well of PI/quench solution (63 tzg/ml PI, 5/A/ml Colanyl Black PR-A, and 5% NaEDTA [10]) and visually or electronically measured for cell lysis by fluorescence microscopy.
PRA determination. PRA for each serum sample was determined by dividing the number of positive reactions by the total number of cells in the panel (usually 60). In the visual analysis, an ASHI score of 6 or greater was considered a positive reaction. In the automated, fluorescent analysis, a positive reaction was determined as described in Results. In general, wells were scanned for fluorescence, which was detected as millivolt (mV) values, converted to percent cell lysis, and translated into an ASHI score.
HLA antibody specificity analysis. ASHI reaction scores of individual PRA tests were transferred into a commercially available computer program ("Screener," Jehn Enterprises, Lawrenceville, NJ) that analyzes PRA data, identifies antibody specificities, and provides a statistical evaluation of specificity assignment.
Evaluation of an Automated Method of PRA Determination
Identification of HLA Antigens Commercially available antiserum trays (One Lambda) were used to test peripheral blood lymphocytes for HLA-A and B using a standard National Institutes of Health (NIH) lymphocytotoxicity test, and for HLAD R using the prolonged lymphocytotoxicity test [2, 9, 11]. C o m p u t e r P r o g r a m s and E q u i p m e n t Utilized Equipment used for the automated reading of the fluorescent plates was a Leitz Diaphot fluorescence microscope with epifluorescent capabilities provided by a ploemo-pak fitted with an N-2 filter, for providing the exitation wavelength for PI (488 nm) and an I-2 filter for providing the excitation wavelength for CFDA (485 nm). The filters allowed the appropriate emission wavelengths (PI = 570 nm; CFDA = 535 nm) to be viewed visually or detected by a photomultiplier tube. The microscope was fitted with a Leitz mechanical stage, controlled by P A T I M E D software through the MVP-MT2 electronic control unit. The software was run on an IBM PS/2 50Z computer with a 30-MB hard drive. This provided control of mechanical stage movements, PMT shutter and sensitivity, lamp intensity, data flow to computer files, and various data-reduction functions. This system was designed for HLA tissue typing with internal plate positive and negative controls. The PATIMED software directs the scanning of the plate, collects the red fluorescence data from each well, and then changes the filter and the ploemo-pak and rescans the plate to collect the green fluorescence data. The data collected as millivolt values are then stored in a Leitz proprietary file system. The spreadsheet program used to further manipulate the data was Quattro Pro (versions 2.0 and 3.0). Data were extracted from the Leitz data files and imported into a Quattro Pro template via ASCII format. The data were then calculated as the percent of cell lysis. The raw data were corrected for background fluorescence by subtracting the average value obtained in replicate control wells ( P H S / C ' / P I but no cells), and normalized by a natural logarithmic transformation to minimize high- or low-value effects [ 12]. The data were then converted to a percent cell lysis. % cell lysis = (mV red fluorescence/ (mV red + mV green fluorescence)) × 100 These data were then converted to ASHI scores [9] as follows: Percent lysis (%) ASHI scoring 0-10 1 11-20 2 21-50 4 51-80 6 81-100 8
181
TABLE 1
Specificity of cell lysis by alloantiserum and C' in microcytotoxicity assays Reactivity of PBMC
Specificity o f alloantisera PRAmethod PBMC 1 PBMC2 PBMC 3
Anti-A2
Anti-B7
A
B
C
+
+
+
A + +
B + +
C + +
Anti-B22 A
B
C
PHS A
B
C
PBMC, peripheral blood mononuclear cell. PBMCs from three individuals were tested for reactivity with various HLAreactive alloantisera or with pooled human serum alloantisera in C'-dependent microcytotoxicity assays. Cell lysis was determined by three visual methods. Method A used CFDA-iabeled PBMCs, and cell death was visualized with eosin. Method B used nonlabeled PBMCs, and eosin was used to visualize cell death. Method C utilized CFDA-labeled PBMCs, and cell death was visualized with PI. All positive wells had >90c~ lysis and all negative wells had < 1 0 ~ cell lysis. HLA antigens displayed by PBMCs: PBMC 1: A1, A32, B7, B53; PBMC 2: AI, A2, BT, B13; and PBMC 3: A1, A24, B51, B8
Results The automated method of PRA analysis required differential fluorescent labeling of viable and nonviable lymphocytes. Viable lymphocytes were visualized with CFDA, which enters live cells and is biochemically converted into a green fluorescent product that cannot escape the cell unless membrane integrity is lost [13]. Nonviable lymphocytes were visualized with PI, which enters only dead ceils and fluoresces red. We first determined whether labeling lymphocytes with CFDA interferes with the specificity of interaction between alloantibodies and lymphocyte surface determinants. PBMCs with defined M H C determinants were labeled with CFDA or were left unlabeled. These cells were then treated with various alloantisera with appropriate or inappropriate M H C specificities in a C'-dependent microcytotoxicity assay. Lymphocytotoxicity was detected three ways: (a) non-CFDA-labeled cells were stained with eosin and scored for reactivity by visual microscopy, (b) CFDA-labeled cells were stained with eosin and scored for reactivity by visual microscopy, and (c) CFDA-labeled cells were stained with PI and scored for reactivity by visual fluorescence microscopy. As shown in Table 1, PBMC 1, which was H L A - B 7 + , A 2 - , and B 2 2 - , reacted with the anti-B7 serum, but not with anti-A2 or an anti-B22 sera. PBMC 2, which was HLA-B7 + , A2 + , and B22 - , reacted with an antiB7 and an anti-A2 sera, but not with an anti-B22 serum. PBMC 3, which was H L A - A 2 - , B 7 - and B 2 2 - , reacted with none of the antisera. Importantly, no differences were observed between each of the three scoring methods, indicating that labeling lymphocytes with CFDA does not affect their performance in the microcytotoxicity assay.
182
M.J. Heller et al.
ASHI Scoring
% Red Fluor. 100 90
80
6
70
4 60 2
~'Tm" 1
50 I 2
I 4
I 8
I 16
I 32
I 64
I 128
I 256
I 512
I 1024
2048
Reciprocal Dilutions
FIGURE 1 Sensitivity of visual versus electronic detection of cell death in microcytotoxicity assays. Purified T-lymphocytes were treated with CFDA or left untreated. Both T-cell populations were cultured with various dilutions of a high PRA serum plus C' and tested for cell lysis in the following way: (a) non-CFDA-treated T cells were stained with eosin and evaluated visually for cell death (V), (b) CFDA-treated T cells were stained with eosin and evaluated visually for cell death (O), (c) CFDA-treated T cells were stained with PI and evaluated visually for cell death (W), and (d) CFDA-treated T cells were stained with PI and evaluated electronically for cell death (0). Results for groups 1-3 are plotted as the ASHI scores designated after visual analyses. Results for group 4 are plotted as the percent red fluorescence determined electronically: % red fluorescnce = (mV red fluorescence/mV red fluorescence + mV green fluorescence) x 100. Controls for the fluorescence method determined by electronic analysis consisted of replicate reaction wells of ALG (~) and PHS (+).
We next determined whether detection of cell lysis by fluorescent methods is comparable to detection by the conventional, eosin-based method. PBMCs were labeled with C F D A or left untreated. These cells were then mixed with various dilutions of serum from a highly sensitized patient and tested for lysis in a C ' - d e p e n d e n t microcytotoxicity assay. Lymphocytotoxicity was detected with the three methods listed earlier. In addition, lymphocytotoxicity was measured electronically as a
function of red and green fluorescence. As shown in Fig. 1, all three visual analyses were equally able to detect antibody at titers greater than 1 : 128. This indicates that CFDA-labeled and nonlabeled PBMCs are comparably sensitive to alloantibody-mediated lysis. Furthermore, electronic analysis detected antibody at titers greater than 1:256. This demonstrates that electronic analysis is at least as sensitive as visual analysis for this purpose. W e utilized this system of electronic analysis to automate determination of PRA with 6 0 - m e m b e r cell panels. Cell panels were prepared by pretreating purified T cells from individual donors with C F D A and plating them in 60-well Terasaki plates. Sixteen replicate plates were reacted with P H S (negative control) or ALG (positive control) plus C', and stained with PI. To define the positive and negative reactivity of lymphocytes from a given individual, corresponding wells of these replicate plates were measured electronically and plotted on a graph of red versus green fluorescence. As shown in Fig. 2, these values clustered in defined areas of the graph. The mean of the negative responses (point A) and the mean of the positive responses (point B) were determined. A line drawn between these points defines the range of possible experimental responses. T o determine the cytotoxic reactivity of a given test serum for each cell in the panel, the test serum was added to a single plate of CFDA-labeled cells, followed
Evaluation of an Automated Method of PRA Determination
183
Red Fluor: Ln(mV) 7 6
,
mm
i
..........................
C
. . . . . . . . . . . . . . . . . . . !
I
A
:
Positive-c0ntr0i8
.......
,
,oeo~o.o .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
C
:
Toet Serum
~ o o ~ o o o o /
B:NegatJve
Controle
Comparison Well
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
Green Fluor: Ln(mV) FIGURE 2 Method for automated fluorescence-based determination of percent lymphocyte death. Electronic values obtained in control plates for a given member of a cell panel are plotted as a function of millivolts of red versus green fluorescence. Values for positive controls are obtained from replicate plates treated with ALG + C'. The mean of these values is calculated and plotted (point A). Values for negative controls are obtained from replicate panels treated with PHS and C'. The mean of these values is calculated and plotted (point B). The line between these two points represents the
range of detectable cell viabilities. To define the antibody activity in a test serum, a cell panel is treated with the test serum + C', and the electronic values for the specific panel member are plotted as a function of millivolts of red versus green fluorescence (point C). This point is extrapolated to a point on lineAB (point D), defined as the intersection between line AB and a line drawn from point C to the origin of the graph. The percent viability in that serum-cell combination = (distance DB/distance AB) × 100. A separate, similar analysis is performed for each member of the cell panel.
by C' and PI as described previously. T h e red and green fluorescent values were obtained for each well, and compared with the positive and negative controls for that given individual. For example, Fig. 2 illustrates one value (point C) obtained in this manner. T o convert this value to the A S H I system of microcytotoxicity scoring, a line is drawn between point C and the origin of the graph. Point D represents the intersection of this line with line AB and reflects the relative viability of the cell population in question. This relative viability is determined as a percent of the distance along line AB and is converted to an A S H I score as follows: 0 % - 1 0 % -- 1, 1 1 % - 2 0 % = 2, 2 1 % - 5 0 % = 4, 5 1 % - 8 0 % = 6, and > 8 0 % = 8 [9]. A similar calculation was p e r f o r m e d with fluorescence information for each of the 60 wells on each plate using a Leitz automated fluorescence microscope, P A T I M E D software, and a computer algorithm developed in this laboratory that analyzes and
scores each of the 60 wells in this manner. For PRA determination, each well with an A S H I score greater than or equal to 6 is considered positive, and the percent of positive wells constitutes the PRA value• We next determined whether PRA values obtained electronically correlate with those obtained by PI-based visual analysis and by eosin-based visual analysis. To do this, sera from 23 patients were tested for reactivity with a 60-cell panel of CFDA-stained T cells. To compare PRA results obtained by electronic versus visual fluorescent analysis, each plate was stained with PI, and fluorescence was measured first with the automated Leitz system and then by visual microscopic examination. The results are represented graphically in Fig. 3A. These two methods of evaluation show a high degree of correlation (r 2 = 0.965). To compare PRA results obtained by electronic fluorescence analysis versus eosin-based visual analysis, parallel plates were stained with PI or with
A
Fluorescence-Based Electronic Analysis 100
80
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
•
.
.
.
.
.
.
.
.
!
.
.
.
.
.
.
.
.
.
.
.
.
.
60
40 n=23 20 •
O
R2= 0.965
;
0
10
20
30
40
50
60
70
80
90
100
Fluorescence-Based Visual Analysis
B
Fluorescence-Based Electronic Analysis 100 m
80 . . . . . . . . . . .
'
.
.
.
.
.
.
.
.
.
.
60 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
40 . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
•. . . . . . . . . . . . . . . . . . . . . . . . .
"....
~' . . . . . . . . . . . . . . . . . . . . . . . . . . . .
n=23 20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . •
2
• 0
,
0
....
R = 0.850 I
10
. . . .
I
20
~
~
~
~
I
30
~
~
~
~1
40
~
~
~
~
I
~
~
50
~
~
I
~
60
~
~
k
I
~
70
~
~
~
I
80
~
~
~
~
J
90
r
~
~
L
100
Eosin-Based Visual Analysis F I G U R E 3 Comparison of PRA values obtained by electronic analyses versus visual analyses. The method described in Fig. 2 was used to evaluate the reactivity of a 60-member cell panel with various test sera. Panel members with > 5 0 % lysis were considered positive, and the percent positive wells provided the PRA value. In A, the same plate was evaluated
elecronically and visually for fluorescence-based detection of cell lysis. In B, two parallel plates were evaluated for electronic, fluorescence-based detection versus visual, eosin-based detection of cell lysis. For both studies, the PRA values obtained by the first method are plotted in comparison to the PRA values obtained by the second method.
Evaluation of an Automated Method of PRA Determination
TABLE 2
Serum sample
Comparison of automated and conventional methods for analysis of PRA and alloantibody specificity
185
23
A B
17 3
B13/B61 B13/B61
0.745/0.291 0.557/0.701
24
A B
32 38
A2/A28 A2/A28
0.805/0.184 0.933/0.340
25
A B
33 30
A24/B8/A23 A24/B8/A23
0.697/0.541/0.541 0.757/0.690/0.690
Analytic method
PRA
Specificity
r value
I
A B
45 43
A2 A2/B64/A28
O.934 0.966/0.697/0.697
26
A B
38 42
A2 A2
0.933 1.000
2
A B
50 50
A2/B13/A28 A2/B13/A28
0.840/0.749/1.00 0.845/0.750/1.00
27
A B
45 43
A1/A3 A1/A3
0.499/0.465 0.362/0.350
3
A B
35 35
A24/A23/B64 A24/A23
0.643/0.627/0.562 0.669/0.673
28
A B
17 35
B58 B7/B58
0.415 0.284/0.253
4
A B
75 81
A24/A23/A2 A24/A23/A2
0.462/0.535/0.488 0.326/0.276/0.388
5
A B
35 40
A2 A2
0.900 0.966
6
A B
20 17
A24/B73 A24/B73
0.808/1.000 0.705/1.000
7
A B
55 60
B44/A29 B44/A29
0.522/0.522 0.482/0.482
8
A B
38 10
A1 A1
0.601 O.56O
9
A B
70 67
A1/B49 A1/B49
0.310/0.275 0.335/0.310
10
A B
58 73
A2/A24 A2/A24
0.714/0.525 0.510/0.498
11
A B
60 52
A2 A2
0.621 O.479
12
A B
0 15
None A26
O.380
A B
17 15
None B38/B50
0.442/0.332
A B
0 15
None A26
0.549
15
A B
47 43
B35 B35
0.449 0.480
16
A B
32 22
A24/A23 A24
0.569/0.569 0.858
17
A B
12 3
B57/C6/B47 C6
0.735/0.313/.700 0.353
18
A B
42 87
B60 B60
0.540 0.131
19
A B
13 35
A2/B56/B61 A2/B27/B61
0.265/0.332/0.332 0.443/0.419/0.384
20
A B
23 35
A1/A23 A1/A24/B52
0.757/0.546 0.063/0.505/0.564
21
A B
38 47
A1/B65 A1/B65
0.424/0.672 0.507/0.512
22
A B
20 60
A26/A25 None
0.603/0.523
13 14
PRA, percent reactive antibody; method A, eosin-based visual reading; and method B, fluorescence-based electronic reading.
eosin. The PI-stained plate was evaluated with the automated Leitz system, and the eosin-stained plate was evaluated by visual microscopic examination. As shown in Fig. 3B, these two methods demonstrate good correlation (r2 -- 0.850), despite the fact that they represent results obtained by two completely different detection methods on two separate plates. One use of PRA information is to determine the HLA specificities for antibodies that are present in a given serum sample. We next determined whether PRAs determined electronically could be used to define antibody specificities and, if so, how these specificities compared to those defined by PRAs determined by eosin-based visual analysis. To do this, various sera were tested for PRA by these two methods, and the reaction values were computer analyzed with an antibody specificity program provided by Dan Doran, PhD. Of the 69 sera tested, 43 sera had PRAs > 1 0 . Of these, 28 sera had PRAs that resulted in detectable HLA specificities with r values >0.5. Shown in Table 2 are the HLA specificities determined via electronic versus visual PRA analysis for these sera. Both methods detected the same antibody specificities in 23 of the 28 sera, despite some differences in actual PRA values. Further, similar r values were often obtained for these antibody specificities. Five sera (nos. 12, 13, 14, 17, and 22) displayed antibody specificities that were detectable with only one method. Three of these (nos. 12, 13, and 14) were nonreactive by visual scoring, but were mildly reactive (15% PRA) when scored electronically. This resulted in an HLA specificity that was definable after electronic, but not visual analysis. The opposite was true for serum 17. The remaining serum (no. 22) was reactive using both methods, but the PRA obtained electronically was much larger than the PRA obtained visually. This obscured the HLA specificity determined by visual analysis. Despite these few exceptions, the overall correlation of antibody specificities
186
determined by visual versus electronic methods is impressive. DISCUSSION We have examined the use of automated fluorescence technology for the determination of PRA levels in the sera of transplant candidates. This determination utilized the same basic technique, antibody-mediated, C'-dependent lymphocyte microcytotoxicity, that is used conventionally for this analysis. Fluorescence technology was used only to determine the degree of cell death in this system. T o do this, lymphocytes were prelabeled with CFDA, causing them to fluoresce green as long as they remained viable. After the reaction with antibody and C ~, the lymphocytes were treated with PI, a red fluor that can only enter dead cells. Hence, live cells fluoresce green and dead cells fluoresce red. This can be evaluated visually by fluorescence microscopy or electronically by automated fluorescence microscopy, if the appropriate equipment is available. Our studies employed the Leitz P A T I M E D system for this purpose. Initially, we demonstrated that pretreatment of lymphocytes with CFDA did not alter the HLA epitopes recognized by alloantisera in the lymphocytotoxicity assay (Table 1). We further demonstrated that CFDAlabeled lymphocytes are as sensitive to alloantiserummediated lysis as nonlabeled lymphocytes (Fig. 1). We next compared visual fluorescence detection of cell lysis with the conventional, eosin-based visual method of detection and found that fluorescence detection was as sensitive as eosin (Fig. 1). Because fluorescence can easily be measured electronically (with a photomultiplier tube), this detection system was amenable to automation. Indeed, a comparison of electronic detection versus visual detection indicated that electronic detection may be slightly more sensitive (Fig. 1). Many laboratories are using automated fluorescence technology to identify M H C class I and class II on human lymphocytes [ 3 - 6 ] . To our knowledge, no one has described an automated method for the determination of PRA and serum antibody M H C specificities, despite the large amount of time and effort that these analyses require. The automation o f PRA analysis required a conversion of electronic data (photometric values) to conventional N I H scores as an index of cell lysis. This was accomplished in the following generalized way: A photometric value was obtained when a test serum was reacted in the assay with lymphocytes from a specific individual. This value was mathematically related to positive and negative control values, which were obtained by reacting the same cells in parallel assays with ALG or PHS, respectively. This computation provided an index of cell
M.J. Heller et at.
death that could be converted to an ASHI score and compared with ASHI scores obtained with conventional, eosin-based testing. The mathematical basis for the conversion of electronic data to percent cell lysis was described by van Lambalgen and Bradley [12] and is illustrated in Fig. 2. Although this evaluation can be performed manually, it can be programmed and performed by computer. We used IBM-based computer hardware and Quattro Pro spreadsheet software for this purpose, and interested individuals with compatible hardware can obtain a copy of our analysis software on request. This is separate from the Leitz PATIMED software used to control the automated plate reader, which can be purchased from Leitz. Our software further provides a conversion of the percent cell lysis to an ASHI score and organizes the data for analysis of antibody specificity. When we compared PRA values obtained by fluorescence and evaluated electronically versus visually, we found a very high degree of correlation (Fig. 3A). When we compared PRA values obtained by fluorescence and evaluated electronically versus those obtained with eosin and evaluated visually (i.e., the conventional testing method), we found slightly less correlation (Fig. 3B). This probably reflects a number of factors, including (a) the tests were run on two parallel plates and (b) fluorescence detection seems to be slightly more sensitive and less subjective than visual analysis. Nevertheless, when PRA values were used to determine antibody specificities, there was strong agreement between specificities identified by visual versus electronic analysis (Table 2), despite differences in PRA values. A few of the samples tested for antibody specificity gave discrepant results using visual analysis versus automated analysis; however, in our experience, this discrepancy was not unexpected. Indeed, when commercial trays from different vendors are compared using the eosin-visual techniques, we routinely note discrepancies in both PRA value and specificities. Ideally, specificities based on reactivity with small numbers of cells should be confirmed by testing against additional cells bearing the specificity in question. There are some technical considerations regarding automated analysis that need to be addressed: (a) EosinY solutions fluoresce, so the same plate cannot be stained simultaneously with PI for electronic analysis and eosin for visual analysis. However, CFDA-treated cells can be stained with eosin and visually scored if necessary, without influencing the PRA value. (b) Data obtained from wells with too many or too few cells are unreliable, so efforts should be made to ensure the deposition of 1 0 0 0 - 2 0 0 0 cells per Terasaki plate well. Unavoidably, an occasional well will contain an aberrant number of cells. If unexpected PRA values are obtained, visual
Evaluation of an Automated Method of PRA Determination
187
examination of the suspect wells is recommended. It may also be informative to review raw photometric data for those experimental wells. For example, if the sum of the red plus green fluorescence in a given well falls below the mean + 3 SD of the fluorescence observed in replicate wells containing all reagents except lymphocytes, we exclude that well from consideration for PRA determination. (c) Plastic binds a small amount of PI, so controls must be included to correct for this background fluorescence. Cell debris, which can be a problem for visual analysis, does not seem to interfere with automated analysis. In general, there seem to be few technical problems encountered by the transition from conventional to automated PRA analysis. T h e r e are also some general considerations regarding PRA and antibody specificity determination that should be addressed. It is well known that the composition of the panel can dramatically influence the PRA value for a given serum. If different fluoresceinated cell panels are used for automated PRA analysis, a significant variation in PRA values, when compared with previously obtained values, is possible. T h e PRA results reported herein were obtained with a commercially available, frozen CFDA-treated cell panel. Thawing these panels results in background cell lysis, which is usually minimized by washing with serum-supplemented culture medium. To account for this background, we have arbitrarily defined a positive reaction for PRA as an A S H I score of 6 or greater. Fresh cell panels do not exhibit significant background cell lysis. Hence, a lower A S H I score could be used as a cutoff for positive reactions, thereby increasing the sensitivity of the analysis. This is important for the rare sera with low titer alloantibodies. Currently, our program performs approximately 200 PRAs/month. Using this method, we have noted a significant reduction in the time required to perform these analyses. Automation has also reduced the tedium associated with visual analysis of this large number of tests. Indeed, the ease of analysis leads to a tendency to rely blindly on the results obtained by automated analysis. We r e c o m m e n d that sera containing well-characterized allospecificities be routinely included to insure reliability of the results. In general, the automated method of PRA analysis and antibody specificity determination is a practical and efficient alternative to conventional analytic methods.
Mr. Irving Katz from Leitz for his technical support. This is paper no. 62 from the Therapeutic Immunology Laboratories at the Ohio State University. This publication was supported, in part, by N I H grant AI24676 (CO), and PHS grant P30CA1605814.
ACKNOWLEDGMENTS
The authors thank Ms. Teri Bailey for her administrative assistance, Ms. Marsha Stalker for her secretarial assistance, and
REFERENCES 1. Patel R, Terasaki P: Significance of the positive crossmatch test in kidney transplantation. N Engl J Med 280:735, 1969. 2. NIAID Manual of Tissue Typing Techniques. N I H Lymphocyte Microcytotoxicity, 1979. Bethesda, MD, National Institute of Allergy and Infectious Diseases, Transplantation in Immunology Branch, 1979. 3. Naipal AM, D'Amaro J, BruningJW, van Leeuwen A, van Rood JJ: Automated reading of P.I. lymphocytotoxicity tests of HLA-DR,MB,MT typing. Tissue Antigens 24:302, 1984. 4. BruningJW, Claas FHJ, Kardol MJ, Lansbergen Q, Naipal AM, Tanke HJ: Automated reading of HLA-A,B,C typing and screening (PI method). Hum Immunol 5:225, 1982. 5. Bruning JW: Cell preparation and staining techniques for a photometric reading of the microlymphocytotoxicity assay: CFDA and PI method. Leitz Photomicroscopy Manual 64:1, 1982. 6. Rubinstein P, Walker ME, Swope E, Rothman W, Calahan T, Siegel J, Genzer M: Serology for automated cytotoxicity testing. II. Routine reading of HLA-typing using the contrast fluorescence test. Tissue Antigens 27:209, 1986. 7. Boyum A: Separation of leukocytes from blood and bone marrow. Scand J Clin Lab Invest 21:1, 1968. 8. Clouse K, Adams P, Sheridan J, Orosz C: Rapid method for purification of human T lymphocytes for further functional studies. J Immunol Methods 105:253, 1987. 9. Hopkins K: Basic microlymphocytotoxicity test. In Zachary A, Teresi G (eds): ASHI Laboratory Manual. Lenexa, KS 1990. 10. Bruning JW: Double stain fluorescence test for HLAA,B,C typing. Leitz Communication, 1988. 11. Sullivan KA, Amos DB: The HLA system and its detection. In Rose N, Friedman H, Fabey J (eds): Washington, DC, American Society for Microbiology, 1986. 12. van Lambalgen R, Bradley BA: Investigations into the principles underlying the automatic reading of tissue typing plates by double fluorochromasia. Tissue Antigens 26:87, 1985. 13. Bethell Dr, Dawson M, LaFor MJ: Characterization of monoclonal antibodies to cell surface antigens by particle concentration fluorescence immunoassay (PCFIA). BioTechniques 3:466, 1985.
ABSTRACT: A fluorescence-based automated method of percent reactive antibody (PRA) analysis is described. This method utilizes the conventional antibody-mediated, C'-dependent lymphocyte microcytotoxicity assay to detect alloantibodies, but replaces the eosin-based method for detection of cell death with a fluorescence-based method. To identify viable cells, lymphocytes were pretreated with carboxy fluorescein diacetate (CFDA), which fluoresces green, to identify viable cells. To identify dead cells after the reaction with antibody and C', they were treated with propidium iodide (PI), which fluoresces red. Pretreatment of lymphocytes with CFDA did not affect their ability or interact with alloantibodies in the microcytotoxicity assays. When visually analyzed, detection of cell death by fluorescence was as sensitive as detection by eosin exclusion. However electronic detection of fluores-
cence was slightly more sensitive than visual detection. Automation of the fluorescent method required a calculation that converts electronic data to an ASHI score for cell death. One such method is described and evaluated. Both the automated and the conventional methods of analysis were used to obtain PRA values for various sera. There was good correlation between the PRA values obtained with the automated method versus the conventional method. Further, there was good correlation for PRAderived alloantibody specificities obtained with the automated method versus the conventional method. These data demonstrate that automated fluorescence-based PRA analysis is an effective and practical alternative to conventional PRA analysis. Human Immunology 35, 179-187 (19921
ABBREVIATIONS
CFDA PI
5-carboxy fluoroscein diacetate (isomer free) propidium iodide
ALG PHS PMT
antilymphocyte globulin pooled human sera photomultiplier tube
INTRODUCTION Candidates awaiting transplantation often become allosensitized as a result of pregnancy, transfusion, or previous transplantation. These candidates develop circulating alloantibodies that could initiate the hyperacute rejection of a subsequent organ allograft [1]. To avoid this, transplant candidates are regularly tested for the presence of serum alloantibodies. Clinically, serum alloantibodies are detected by their ability to mediate C'dependent lymphocytolysis. Because of the large numFrom the Department of Surgery, Division of Transplantation (MJ.H., P.W.A.), and the Departments of Pathology and Pharmacology (C.G.G.), Ohio State University of College of Medicine, Columbus, Ohio, USA. Address reprint requeststo Dr. M. J. Heller, Clinical Histocompatibility Laboratory, Room N935 Doan Hall, The Ohio State University Hospitals, 410 West Tenth Avenue, Columbus, OH 43210, USA. Received May I 1, 1992; acceptedJuly 22, 1992. Human Immunology 35, 179-187 (1992) © American Society for Histocompatibility and Immunogenetics, 1992
ber of serologically distinct human leukocyte antigen (HLA) alleles, this must be done with a panel o f l y m p h o cytes obtained from a large n u m b e r of different individuals. The panel of lymphocyte donors is selected to represent the distribution and frequency of H L A specificities that are present in the local population. For any given serum, the percent of positive reactions against this panel of lymphocytes is called percent reactive antibody (PRA), and is considered to be an index of allosensitization. If the H L A specificities displayed by each panel m e m b e r are known, the pattern of reactivity in this panel can be used to identify the specific H L A antigens recognized by the alloantibodies in the serum sample. The conventional method of PRA analysis involves the microscopic evaluation of antibody-mediated C'-dependent cell death that is visualized by vital dye (eosin 179 0198-8859/92/$5.00
180
or trypan blue) exclusion [2]. This method is labor intensive and somewhat subjective. Further, each transplant candidate is tested monthly for PRA. Hence, PRA analyses can consume large amounts of time and resources, especially in centers with long transplant candidate lists. An alternative to the conventional method is offered by fluorescence-based technology. This technology, which is currently used for HLA typing, lends itself well to automation [3-6]. An automated PRA methodology would be less subjective than conventional methods and should reduce the tedium of visual PRA evaluation for large numbers of serum samples. Consequently, we have adapted existing fluorescence techniques to the automated determination of PRA. In this communication, we describe an automated method of PRA analysis and of antibody specificity determination. Further, we compare this method with the traditional visual methods for ability to yield PRA values and to define HLA-reactive antibody specificities.
M.J. Heller et al.
Alternatively, 60-member, frozen, CFDA-labeled cell panels were obtained from PelFreez Clinical Systems (Brown Deer, WI) and used according to the manufacturer's instructions. The panel members were chosen to reflect the frequency of HLA epitopes found in the midwestern United States. C'-Mediated Microcytotoxicity Assay
Eosin-based visual analysis. The Amos-modified micro-
METHODS
cytotoxicity assay was employed [2, 9]. Briefly, 1 ~zl of T cells (2-4 x 106/ml) was mixed with 1/zl of test serum in Terasaki wells, and incubated for 30 minutes at room temperature. Five microliters of medium was added to each well and the cells were permitted to settle for 10 minutes. The plates were flicked to remove the serum, 5-/zl/well rabbit complement (PelFreez Clinical Systems) was added, and the plates were further incubated for 120 minutes. Each well was stained with eosin, fixed with 10% formaldehyde in phosphate-buffered saline, and scored for percent cell death with an inverted phasecontrast microscope.
Cell Isolation
CFDA/PI-based fluorescence analysis. To label lympho-
Peripheral blood mononuclear cells (PBMC). Fresh heparinized blood was collected by venipuncture from normal individuals. Leukocytes were isolated following centrifugation and were separated on Ficoll-hypaque density gradients using the method of Boyum [7]. Cells recovered from the PBMC interface were washed three times in Hank's balanced salt solution (HBSS) and resuspended in McCoy's 5A medium containing 20% pooled human serum (PHS).
Purified T cells. Human T-lymphocytes were isolated from PBMCs by negative selection with a commercially available cocktail of monoclonal antibodies and complement (T Lympho-Kwik, One Lambda, Los Angeles, CA), using methods based on those of Clouse et al. [8]. PBMCs (5-10 x 107 cells) were washed by centrifugation (400 g, 10 minutes) and resuspended in 0.8 ml T Lympho-kwik in a Fisher centrifuge tube. The cells were incubated 60 minutes at 37°C, with gentle mixing at 15 minute intervals, overlaid with 0.25 ml of culture medium, and centrifuged for 2 minutes at 700 g. The supernatant containing cell debris was discarded, and the pelleted T cells were washed three times in HBSS and resuspended in McCoy's medium plus 20% PHS. Preparation of Panels For locally prepared cell panels, isolated T cells were labeled with CFDA (see later) and suspended in culture medium plus 8% DMSO at a concentration of 2-4 x 106 cells/ml. These cells were added to Terasaki plates (1 /A/well), which were frozen and stored in liquid nitrogen until used.
cytes with CFDA, purified T-lymphocytes were washed three times in serum-free balanced salt solution. The cell pellet (containing - 2 - 5 x 106 cells) was resuspended in 1.0 ml CFDA (20 ~g/ml; Molecular Probes, Eugene, OR) for 15 minutes at 37°C in the dark. The cells were washed three times in McCoy's 5A medium containing 20% PHS and resuspended at a concentration of 1-2 x 106/ml. These cells were used as target cells in the C'-mediated cytotoxicity assay using identical conditions as described earlier. Following incubation with C', the cells were stained with 5 ~l/well of PI/quench solution (63 tzg/ml PI, 5/A/ml Colanyl Black PR-A, and 5% NaEDTA [10]) and visually or electronically measured for cell lysis by fluorescence microscopy.
PRA determination. PRA for each serum sample was determined by dividing the number of positive reactions by the total number of cells in the panel (usually 60). In the visual analysis, an ASHI score of 6 or greater was considered a positive reaction. In the automated, fluorescent analysis, a positive reaction was determined as described in Results. In general, wells were scanned for fluorescence, which was detected as millivolt (mV) values, converted to percent cell lysis, and translated into an ASHI score.
HLA antibody specificity analysis. ASHI reaction scores of individual PRA tests were transferred into a commercially available computer program ("Screener," Jehn Enterprises, Lawrenceville, NJ) that analyzes PRA data, identifies antibody specificities, and provides a statistical evaluation of specificity assignment.
Evaluation of an Automated Method of PRA Determination
Identification of HLA Antigens Commercially available antiserum trays (One Lambda) were used to test peripheral blood lymphocytes for HLA-A and B using a standard National Institutes of Health (NIH) lymphocytotoxicity test, and for HLAD R using the prolonged lymphocytotoxicity test [2, 9, 11]. C o m p u t e r P r o g r a m s and E q u i p m e n t Utilized Equipment used for the automated reading of the fluorescent plates was a Leitz Diaphot fluorescence microscope with epifluorescent capabilities provided by a ploemo-pak fitted with an N-2 filter, for providing the exitation wavelength for PI (488 nm) and an I-2 filter for providing the excitation wavelength for CFDA (485 nm). The filters allowed the appropriate emission wavelengths (PI = 570 nm; CFDA = 535 nm) to be viewed visually or detected by a photomultiplier tube. The microscope was fitted with a Leitz mechanical stage, controlled by P A T I M E D software through the MVP-MT2 electronic control unit. The software was run on an IBM PS/2 50Z computer with a 30-MB hard drive. This provided control of mechanical stage movements, PMT shutter and sensitivity, lamp intensity, data flow to computer files, and various data-reduction functions. This system was designed for HLA tissue typing with internal plate positive and negative controls. The PATIMED software directs the scanning of the plate, collects the red fluorescence data from each well, and then changes the filter and the ploemo-pak and rescans the plate to collect the green fluorescence data. The data collected as millivolt values are then stored in a Leitz proprietary file system. The spreadsheet program used to further manipulate the data was Quattro Pro (versions 2.0 and 3.0). Data were extracted from the Leitz data files and imported into a Quattro Pro template via ASCII format. The data were then calculated as the percent of cell lysis. The raw data were corrected for background fluorescence by subtracting the average value obtained in replicate control wells ( P H S / C ' / P I but no cells), and normalized by a natural logarithmic transformation to minimize high- or low-value effects [ 12]. The data were then converted to a percent cell lysis. % cell lysis = (mV red fluorescence/ (mV red + mV green fluorescence)) × 100 These data were then converted to ASHI scores [9] as follows: Percent lysis (%) ASHI scoring 0-10 1 11-20 2 21-50 4 51-80 6 81-100 8
181
TABLE 1
Specificity of cell lysis by alloantiserum and C' in microcytotoxicity assays Reactivity of PBMC
Specificity o f alloantisera PRAmethod PBMC 1 PBMC2 PBMC 3
Anti-A2
Anti-B7
A
B
C
+
+
+
A + +
B + +
C + +
Anti-B22 A
B
C
PHS A
B
C
PBMC, peripheral blood mononuclear cell. PBMCs from three individuals were tested for reactivity with various HLAreactive alloantisera or with pooled human serum alloantisera in C'-dependent microcytotoxicity assays. Cell lysis was determined by three visual methods. Method A used CFDA-iabeled PBMCs, and cell death was visualized with eosin. Method B used nonlabeled PBMCs, and eosin was used to visualize cell death. Method C utilized CFDA-labeled PBMCs, and cell death was visualized with PI. All positive wells had >90c~ lysis and all negative wells had < 1 0 ~ cell lysis. HLA antigens displayed by PBMCs: PBMC 1: A1, A32, B7, B53; PBMC 2: AI, A2, BT, B13; and PBMC 3: A1, A24, B51, B8
Results The automated method of PRA analysis required differential fluorescent labeling of viable and nonviable lymphocytes. Viable lymphocytes were visualized with CFDA, which enters live cells and is biochemically converted into a green fluorescent product that cannot escape the cell unless membrane integrity is lost [13]. Nonviable lymphocytes were visualized with PI, which enters only dead ceils and fluoresces red. We first determined whether labeling lymphocytes with CFDA interferes with the specificity of interaction between alloantibodies and lymphocyte surface determinants. PBMCs with defined M H C determinants were labeled with CFDA or were left unlabeled. These cells were then treated with various alloantisera with appropriate or inappropriate M H C specificities in a C'-dependent microcytotoxicity assay. Lymphocytotoxicity was detected three ways: (a) non-CFDA-labeled cells were stained with eosin and scored for reactivity by visual microscopy, (b) CFDA-labeled cells were stained with eosin and scored for reactivity by visual microscopy, and (c) CFDA-labeled cells were stained with PI and scored for reactivity by visual fluorescence microscopy. As shown in Table 1, PBMC 1, which was H L A - B 7 + , A 2 - , and B 2 2 - , reacted with the anti-B7 serum, but not with anti-A2 or an anti-B22 sera. PBMC 2, which was HLA-B7 + , A2 + , and B22 - , reacted with an antiB7 and an anti-A2 sera, but not with an anti-B22 serum. PBMC 3, which was H L A - A 2 - , B 7 - and B 2 2 - , reacted with none of the antisera. Importantly, no differences were observed between each of the three scoring methods, indicating that labeling lymphocytes with CFDA does not affect their performance in the microcytotoxicity assay.
182
M.J. Heller et al.
ASHI Scoring
% Red Fluor. 100 90
80
6
70
4 60 2
~'Tm" 1
50 I 2
I 4
I 8
I 16
I 32
I 64
I 128
I 256
I 512
I 1024
2048
Reciprocal Dilutions
FIGURE 1 Sensitivity of visual versus electronic detection of cell death in microcytotoxicity assays. Purified T-lymphocytes were treated with CFDA or left untreated. Both T-cell populations were cultured with various dilutions of a high PRA serum plus C' and tested for cell lysis in the following way: (a) non-CFDA-treated T cells were stained with eosin and evaluated visually for cell death (V), (b) CFDA-treated T cells were stained with eosin and evaluated visually for cell death (O), (c) CFDA-treated T cells were stained with PI and evaluated visually for cell death (W), and (d) CFDA-treated T cells were stained with PI and evaluated electronically for cell death (0). Results for groups 1-3 are plotted as the ASHI scores designated after visual analyses. Results for group 4 are plotted as the percent red fluorescence determined electronically: % red fluorescnce = (mV red fluorescence/mV red fluorescence + mV green fluorescence) x 100. Controls for the fluorescence method determined by electronic analysis consisted of replicate reaction wells of ALG (~) and PHS (+).
We next determined whether detection of cell lysis by fluorescent methods is comparable to detection by the conventional, eosin-based method. PBMCs were labeled with C F D A or left untreated. These cells were then mixed with various dilutions of serum from a highly sensitized patient and tested for lysis in a C ' - d e p e n d e n t microcytotoxicity assay. Lymphocytotoxicity was detected with the three methods listed earlier. In addition, lymphocytotoxicity was measured electronically as a
function of red and green fluorescence. As shown in Fig. 1, all three visual analyses were equally able to detect antibody at titers greater than 1 : 128. This indicates that CFDA-labeled and nonlabeled PBMCs are comparably sensitive to alloantibody-mediated lysis. Furthermore, electronic analysis detected antibody at titers greater than 1:256. This demonstrates that electronic analysis is at least as sensitive as visual analysis for this purpose. W e utilized this system of electronic analysis to automate determination of PRA with 6 0 - m e m b e r cell panels. Cell panels were prepared by pretreating purified T cells from individual donors with C F D A and plating them in 60-well Terasaki plates. Sixteen replicate plates were reacted with P H S (negative control) or ALG (positive control) plus C', and stained with PI. To define the positive and negative reactivity of lymphocytes from a given individual, corresponding wells of these replicate plates were measured electronically and plotted on a graph of red versus green fluorescence. As shown in Fig. 2, these values clustered in defined areas of the graph. The mean of the negative responses (point A) and the mean of the positive responses (point B) were determined. A line drawn between these points defines the range of possible experimental responses. T o determine the cytotoxic reactivity of a given test serum for each cell in the panel, the test serum was added to a single plate of CFDA-labeled cells, followed
Evaluation of an Automated Method of PRA Determination
183
Red Fluor: Ln(mV) 7 6
,
mm
i
..........................
C
. . . . . . . . . . . . . . . . . . . !
I
A
:
Positive-c0ntr0i8
.......
,
,oeo~o.o .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
C
:
Toet Serum
~ o o ~ o o o o /
B:NegatJve
Controle
Comparison Well
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
Green Fluor: Ln(mV) FIGURE 2 Method for automated fluorescence-based determination of percent lymphocyte death. Electronic values obtained in control plates for a given member of a cell panel are plotted as a function of millivolts of red versus green fluorescence. Values for positive controls are obtained from replicate plates treated with ALG + C'. The mean of these values is calculated and plotted (point A). Values for negative controls are obtained from replicate panels treated with PHS and C'. The mean of these values is calculated and plotted (point B). The line between these two points represents the
range of detectable cell viabilities. To define the antibody activity in a test serum, a cell panel is treated with the test serum + C', and the electronic values for the specific panel member are plotted as a function of millivolts of red versus green fluorescence (point C). This point is extrapolated to a point on lineAB (point D), defined as the intersection between line AB and a line drawn from point C to the origin of the graph. The percent viability in that serum-cell combination = (distance DB/distance AB) × 100. A separate, similar analysis is performed for each member of the cell panel.
by C' and PI as described previously. T h e red and green fluorescent values were obtained for each well, and compared with the positive and negative controls for that given individual. For example, Fig. 2 illustrates one value (point C) obtained in this manner. T o convert this value to the A S H I system of microcytotoxicity scoring, a line is drawn between point C and the origin of the graph. Point D represents the intersection of this line with line AB and reflects the relative viability of the cell population in question. This relative viability is determined as a percent of the distance along line AB and is converted to an A S H I score as follows: 0 % - 1 0 % -- 1, 1 1 % - 2 0 % = 2, 2 1 % - 5 0 % = 4, 5 1 % - 8 0 % = 6, and > 8 0 % = 8 [9]. A similar calculation was p e r f o r m e d with fluorescence information for each of the 60 wells on each plate using a Leitz automated fluorescence microscope, P A T I M E D software, and a computer algorithm developed in this laboratory that analyzes and
scores each of the 60 wells in this manner. For PRA determination, each well with an A S H I score greater than or equal to 6 is considered positive, and the percent of positive wells constitutes the PRA value• We next determined whether PRA values obtained electronically correlate with those obtained by PI-based visual analysis and by eosin-based visual analysis. To do this, sera from 23 patients were tested for reactivity with a 60-cell panel of CFDA-stained T cells. To compare PRA results obtained by electronic versus visual fluorescent analysis, each plate was stained with PI, and fluorescence was measured first with the automated Leitz system and then by visual microscopic examination. The results are represented graphically in Fig. 3A. These two methods of evaluation show a high degree of correlation (r 2 = 0.965). To compare PRA results obtained by electronic fluorescence analysis versus eosin-based visual analysis, parallel plates were stained with PI or with
A
Fluorescence-Based Electronic Analysis 100
80
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
•
.
.
.
.
.
.
.
.
!
.
.
.
.
.
.
.
.
.
.
.
.
.
60
40 n=23 20 •
O
R2= 0.965
;
0
10
20
30
40
50
60
70
80
90
100
Fluorescence-Based Visual Analysis
B
Fluorescence-Based Electronic Analysis 100 m
80 . . . . . . . . . . .
'
.
.
.
.
.
.
.
.
.
.
60 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
40 . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
•. . . . . . . . . . . . . . . . . . . . . . . . .
"....
~' . . . . . . . . . . . . . . . . . . . . . . . . . . . .
n=23 20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . •
2
• 0
,
0
....
R = 0.850 I
10
. . . .
I
20
~
~
~
~
I
30
~
~
~
~1
40
~
~
~
~
I
~
~
50
~
~
I
~
60
~
~
k
I
~
70
~
~
~
I
80
~
~
~
~
J
90
r
~
~
L
100
Eosin-Based Visual Analysis F I G U R E 3 Comparison of PRA values obtained by electronic analyses versus visual analyses. The method described in Fig. 2 was used to evaluate the reactivity of a 60-member cell panel with various test sera. Panel members with > 5 0 % lysis were considered positive, and the percent positive wells provided the PRA value. In A, the same plate was evaluated
elecronically and visually for fluorescence-based detection of cell lysis. In B, two parallel plates were evaluated for electronic, fluorescence-based detection versus visual, eosin-based detection of cell lysis. For both studies, the PRA values obtained by the first method are plotted in comparison to the PRA values obtained by the second method.
Evaluation of an Automated Method of PRA Determination
TABLE 2
Serum sample
Comparison of automated and conventional methods for analysis of PRA and alloantibody specificity
185
23
A B
17 3
B13/B61 B13/B61
0.745/0.291 0.557/0.701
24
A B
32 38
A2/A28 A2/A28
0.805/0.184 0.933/0.340
25
A B
33 30
A24/B8/A23 A24/B8/A23
0.697/0.541/0.541 0.757/0.690/0.690
Analytic method
PRA
Specificity
r value
I
A B
45 43
A2 A2/B64/A28
O.934 0.966/0.697/0.697
26
A B
38 42
A2 A2
0.933 1.000
2
A B
50 50
A2/B13/A28 A2/B13/A28
0.840/0.749/1.00 0.845/0.750/1.00
27
A B
45 43
A1/A3 A1/A3
0.499/0.465 0.362/0.350
3
A B
35 35
A24/A23/B64 A24/A23
0.643/0.627/0.562 0.669/0.673
28
A B
17 35
B58 B7/B58
0.415 0.284/0.253
4
A B
75 81
A24/A23/A2 A24/A23/A2
0.462/0.535/0.488 0.326/0.276/0.388
5
A B
35 40
A2 A2
0.900 0.966
6
A B
20 17
A24/B73 A24/B73
0.808/1.000 0.705/1.000
7
A B
55 60
B44/A29 B44/A29
0.522/0.522 0.482/0.482
8
A B
38 10
A1 A1
0.601 O.56O
9
A B
70 67
A1/B49 A1/B49
0.310/0.275 0.335/0.310
10
A B
58 73
A2/A24 A2/A24
0.714/0.525 0.510/0.498
11
A B
60 52
A2 A2
0.621 O.479
12
A B
0 15
None A26
O.380
A B
17 15
None B38/B50
0.442/0.332
A B
0 15
None A26
0.549
15
A B
47 43
B35 B35
0.449 0.480
16
A B
32 22
A24/A23 A24
0.569/0.569 0.858
17
A B
12 3
B57/C6/B47 C6
0.735/0.313/.700 0.353
18
A B
42 87
B60 B60
0.540 0.131
19
A B
13 35
A2/B56/B61 A2/B27/B61
0.265/0.332/0.332 0.443/0.419/0.384
20
A B
23 35
A1/A23 A1/A24/B52
0.757/0.546 0.063/0.505/0.564
21
A B
38 47
A1/B65 A1/B65
0.424/0.672 0.507/0.512
22
A B
20 60
A26/A25 None
0.603/0.523
13 14
PRA, percent reactive antibody; method A, eosin-based visual reading; and method B, fluorescence-based electronic reading.
eosin. The PI-stained plate was evaluated with the automated Leitz system, and the eosin-stained plate was evaluated by visual microscopic examination. As shown in Fig. 3B, these two methods demonstrate good correlation (r2 -- 0.850), despite the fact that they represent results obtained by two completely different detection methods on two separate plates. One use of PRA information is to determine the HLA specificities for antibodies that are present in a given serum sample. We next determined whether PRAs determined electronically could be used to define antibody specificities and, if so, how these specificities compared to those defined by PRAs determined by eosin-based visual analysis. To do this, various sera were tested for PRA by these two methods, and the reaction values were computer analyzed with an antibody specificity program provided by Dan Doran, PhD. Of the 69 sera tested, 43 sera had PRAs > 1 0 . Of these, 28 sera had PRAs that resulted in detectable HLA specificities with r values >0.5. Shown in Table 2 are the HLA specificities determined via electronic versus visual PRA analysis for these sera. Both methods detected the same antibody specificities in 23 of the 28 sera, despite some differences in actual PRA values. Further, similar r values were often obtained for these antibody specificities. Five sera (nos. 12, 13, 14, 17, and 22) displayed antibody specificities that were detectable with only one method. Three of these (nos. 12, 13, and 14) were nonreactive by visual scoring, but were mildly reactive (15% PRA) when scored electronically. This resulted in an HLA specificity that was definable after electronic, but not visual analysis. The opposite was true for serum 17. The remaining serum (no. 22) was reactive using both methods, but the PRA obtained electronically was much larger than the PRA obtained visually. This obscured the HLA specificity determined by visual analysis. Despite these few exceptions, the overall correlation of antibody specificities
186
determined by visual versus electronic methods is impressive. DISCUSSION We have examined the use of automated fluorescence technology for the determination of PRA levels in the sera of transplant candidates. This determination utilized the same basic technique, antibody-mediated, C'-dependent lymphocyte microcytotoxicity, that is used conventionally for this analysis. Fluorescence technology was used only to determine the degree of cell death in this system. T o do this, lymphocytes were prelabeled with CFDA, causing them to fluoresce green as long as they remained viable. After the reaction with antibody and C ~, the lymphocytes were treated with PI, a red fluor that can only enter dead cells. Hence, live cells fluoresce green and dead cells fluoresce red. This can be evaluated visually by fluorescence microscopy or electronically by automated fluorescence microscopy, if the appropriate equipment is available. Our studies employed the Leitz P A T I M E D system for this purpose. Initially, we demonstrated that pretreatment of lymphocytes with CFDA did not alter the HLA epitopes recognized by alloantisera in the lymphocytotoxicity assay (Table 1). We further demonstrated that CFDAlabeled lymphocytes are as sensitive to alloantiserummediated lysis as nonlabeled lymphocytes (Fig. 1). We next compared visual fluorescence detection of cell lysis with the conventional, eosin-based visual method of detection and found that fluorescence detection was as sensitive as eosin (Fig. 1). Because fluorescence can easily be measured electronically (with a photomultiplier tube), this detection system was amenable to automation. Indeed, a comparison of electronic detection versus visual detection indicated that electronic detection may be slightly more sensitive (Fig. 1). Many laboratories are using automated fluorescence technology to identify M H C class I and class II on human lymphocytes [ 3 - 6 ] . To our knowledge, no one has described an automated method for the determination of PRA and serum antibody M H C specificities, despite the large amount of time and effort that these analyses require. The automation o f PRA analysis required a conversion of electronic data (photometric values) to conventional N I H scores as an index of cell lysis. This was accomplished in the following generalized way: A photometric value was obtained when a test serum was reacted in the assay with lymphocytes from a specific individual. This value was mathematically related to positive and negative control values, which were obtained by reacting the same cells in parallel assays with ALG or PHS, respectively. This computation provided an index of cell
M.J. Heller et at.
death that could be converted to an ASHI score and compared with ASHI scores obtained with conventional, eosin-based testing. The mathematical basis for the conversion of electronic data to percent cell lysis was described by van Lambalgen and Bradley [12] and is illustrated in Fig. 2. Although this evaluation can be performed manually, it can be programmed and performed by computer. We used IBM-based computer hardware and Quattro Pro spreadsheet software for this purpose, and interested individuals with compatible hardware can obtain a copy of our analysis software on request. This is separate from the Leitz PATIMED software used to control the automated plate reader, which can be purchased from Leitz. Our software further provides a conversion of the percent cell lysis to an ASHI score and organizes the data for analysis of antibody specificity. When we compared PRA values obtained by fluorescence and evaluated electronically versus visually, we found a very high degree of correlation (Fig. 3A). When we compared PRA values obtained by fluorescence and evaluated electronically versus those obtained with eosin and evaluated visually (i.e., the conventional testing method), we found slightly less correlation (Fig. 3B). This probably reflects a number of factors, including (a) the tests were run on two parallel plates and (b) fluorescence detection seems to be slightly more sensitive and less subjective than visual analysis. Nevertheless, when PRA values were used to determine antibody specificities, there was strong agreement between specificities identified by visual versus electronic analysis (Table 2), despite differences in PRA values. A few of the samples tested for antibody specificity gave discrepant results using visual analysis versus automated analysis; however, in our experience, this discrepancy was not unexpected. Indeed, when commercial trays from different vendors are compared using the eosin-visual techniques, we routinely note discrepancies in both PRA value and specificities. Ideally, specificities based on reactivity with small numbers of cells should be confirmed by testing against additional cells bearing the specificity in question. There are some technical considerations regarding automated analysis that need to be addressed: (a) EosinY solutions fluoresce, so the same plate cannot be stained simultaneously with PI for electronic analysis and eosin for visual analysis. However, CFDA-treated cells can be stained with eosin and visually scored if necessary, without influencing the PRA value. (b) Data obtained from wells with too many or too few cells are unreliable, so efforts should be made to ensure the deposition of 1 0 0 0 - 2 0 0 0 cells per Terasaki plate well. Unavoidably, an occasional well will contain an aberrant number of cells. If unexpected PRA values are obtained, visual
Evaluation of an Automated Method of PRA Determination
187
examination of the suspect wells is recommended. It may also be informative to review raw photometric data for those experimental wells. For example, if the sum of the red plus green fluorescence in a given well falls below the mean + 3 SD of the fluorescence observed in replicate wells containing all reagents except lymphocytes, we exclude that well from consideration for PRA determination. (c) Plastic binds a small amount of PI, so controls must be included to correct for this background fluorescence. Cell debris, which can be a problem for visual analysis, does not seem to interfere with automated analysis. In general, there seem to be few technical problems encountered by the transition from conventional to automated PRA analysis. T h e r e are also some general considerations regarding PRA and antibody specificity determination that should be addressed. It is well known that the composition of the panel can dramatically influence the PRA value for a given serum. If different fluoresceinated cell panels are used for automated PRA analysis, a significant variation in PRA values, when compared with previously obtained values, is possible. T h e PRA results reported herein were obtained with a commercially available, frozen CFDA-treated cell panel. Thawing these panels results in background cell lysis, which is usually minimized by washing with serum-supplemented culture medium. To account for this background, we have arbitrarily defined a positive reaction for PRA as an A S H I score of 6 or greater. Fresh cell panels do not exhibit significant background cell lysis. Hence, a lower A S H I score could be used as a cutoff for positive reactions, thereby increasing the sensitivity of the analysis. This is important for the rare sera with low titer alloantibodies. Currently, our program performs approximately 200 PRAs/month. Using this method, we have noted a significant reduction in the time required to perform these analyses. Automation has also reduced the tedium associated with visual analysis of this large number of tests. Indeed, the ease of analysis leads to a tendency to rely blindly on the results obtained by automated analysis. We r e c o m m e n d that sera containing well-characterized allospecificities be routinely included to insure reliability of the results. In general, the automated method of PRA analysis and antibody specificity determination is a practical and efficient alternative to conventional analytic methods.
Mr. Irving Katz from Leitz for his technical support. This is paper no. 62 from the Therapeutic Immunology Laboratories at the Ohio State University. This publication was supported, in part, by N I H grant AI24676 (CO), and PHS grant P30CA1605814.
ACKNOWLEDGMENTS
The authors thank Ms. Teri Bailey for her administrative assistance, Ms. Marsha Stalker for her secretarial assistance, and
REFERENCES 1. Patel R, Terasaki P: Significance of the positive crossmatch test in kidney transplantation. N Engl J Med 280:735, 1969. 2. NIAID Manual of Tissue Typing Techniques. N I H Lymphocyte Microcytotoxicity, 1979. Bethesda, MD, National Institute of Allergy and Infectious Diseases, Transplantation in Immunology Branch, 1979. 3. Naipal AM, D'Amaro J, BruningJW, van Leeuwen A, van Rood JJ: Automated reading of P.I. lymphocytotoxicity tests of HLA-DR,MB,MT typing. Tissue Antigens 24:302, 1984. 4. BruningJW, Claas FHJ, Kardol MJ, Lansbergen Q, Naipal AM, Tanke HJ: Automated reading of HLA-A,B,C typing and screening (PI method). Hum Immunol 5:225, 1982. 5. Bruning JW: Cell preparation and staining techniques for a photometric reading of the microlymphocytotoxicity assay: CFDA and PI method. Leitz Photomicroscopy Manual 64:1, 1982. 6. Rubinstein P, Walker ME, Swope E, Rothman W, Calahan T, Siegel J, Genzer M: Serology for automated cytotoxicity testing. II. Routine reading of HLA-typing using the contrast fluorescence test. Tissue Antigens 27:209, 1986. 7. Boyum A: Separation of leukocytes from blood and bone marrow. Scand J Clin Lab Invest 21:1, 1968. 8. Clouse K, Adams P, Sheridan J, Orosz C: Rapid method for purification of human T lymphocytes for further functional studies. J Immunol Methods 105:253, 1987. 9. Hopkins K: Basic microlymphocytotoxicity test. In Zachary A, Teresi G (eds): ASHI Laboratory Manual. Lenexa, KS 1990. 10. Bruning JW: Double stain fluorescence test for HLAA,B,C typing. Leitz Communication, 1988. 11. Sullivan KA, Amos DB: The HLA system and its detection. In Rose N, Friedman H, Fabey J (eds): Washington, DC, American Society for Microbiology, 1986. 12. van Lambalgen R, Bradley BA: Investigations into the principles underlying the automatic reading of tissue typing plates by double fluorochromasia. Tissue Antigens 26:87, 1985. 13. Bethell Dr, Dawson M, LaFor MJ: Characterization of monoclonal antibodies to cell surface antigens by particle concentration fluorescence immunoassay (PCFIA). BioTechniques 3:466, 1985.