- Email: [email protected]
The development of a double-antibody radioimmunoassay for detecting ovarian tumor-associated antigen fraction OCA in plasma
The development of a double-antibody radioimmunoassay for detecting ovarian tumor-associated antigen fraction OCA in plasma SUZANNE G. I.
KNAUF
URBACH,
M.D.
Toronto, Ontario, Canada Ovarian tumor-associated antigen isolated from human tumor tissue was shown to have a different mobility from that of carcinoembryonic antigen (CEA) in both acrylarnide gel electrophoresis and immunoelectrophoresis in agarose. The ovarian tumor antigen is composed of six species with different electrophoretic mobility in acrylamide gel electrophoresis. Three of these species were detected in Sephadex G-100 ovarian fraction OCA (from the void volume peak) and the other three species of lower apparent molecular weight were detected in fraction OCD (from the second peak). Fractions OCA and OCD did not share common antigenic determinants as determined by immunodiffusion. CEA was shown to share antigenic determinants with both OCA and OCD. A double-antibody radioimmunoassay capable of detecting nanogram quantities of plasma OCA was developed. In a preliminary study of ovarian cancer patients, OCA appeared to be a more sensitive marker for ovarian cancer than CEA. There was virtually no correlation (r2 = 0.1) between OCA and CEA levels in these patients, as determined by radioimmunoassay. (AM. J. OBSTET. GYNECOL. 131: 780, 1978.)
TUMOR-ASSOCIATED ANTIGEN purified in our laboratory was contaminated by normal antigen.’ Therefore, the physiochemical properties of the ovarian tumor antigen preparation were examined for the purpose of discovering a method for the removal of normal antigen so that a radioimmunoassay for circulating ovarian tumor antigen couId be developed. Since carcinoembryonic antigen (CEA) is currently the tumor-associated antiqen monitored in ovarian cancer patients, ovarian tumor antigen was compared with CEA. This paper reports the results of immunologic studies of the cross-reactivity between ovarian tumor antigen and CEA, the development of a double-antibody radioimmunoassay for the ovarian tumor antigen fraction OCA, and a comparison of
OVARIAN
From the Department of Obstekcs and Gynaecology, University of Toronto, The Wellesley Hospital. Supported Foundation
by Onkwio Cancer Project No. 324
Received for publication Revised
November
Accepted Jama?
Treatment
Sepkmber
and Research
19, 1977.
29, 1977. 20, 1978.
Reprint requests: Dr. G. I. Urbach, 600 Sherboumz Suite 805, Toronto, Ontario, Canada M4X 1 W4.
780
St.,
the OCA and CEA radioimmunoassays for the assessment of circulating tumor antigen in patients with ovarian cancer. Material
and methods
Immundogic reagents. Ovarian tumor antigen was prepared as previously described.’ Briefly, human ovarian tumor tissue was homogenized in saline, extracted with an equal volume of 1.2M perchloric acid, and subjected to affinity chromatography on concanavalin A-Sepharose.* Bound protein was eluted with 10 and 25 per cent n-mannose. The 10 per cent mannose fraction was subsequently purified by passage through affinity columns that contained anti-normal female serum, anti-normal ovarian tissue, and antimannose fractions from 2, 10, and 25 per cent concanavalin A-Sepharose fractionation of normal female serum coupled to Sepharose 4B. The eluate was concentrated in an ultrafiltration unit? with a UM-10 filter. Antibodies were prepared in female New Zealand white rabbits, as previously described.’ Gamma globulin was prepared by two precipitations with 33 per cent saturated ammonium sulfate.” *Pharmacia Laboratories, tAmicon Corp., Lexington,
000%9378/78/07131-0780$00.80/O
Inc., Piscataway, Massachusetts.
@ 1978
The
New
Jersey.
C. V. Mosby
Co
Volume Number
Double-antibody
13 1 7
CEA and absorbed anti-CEA for Ouchterlony and immunoelectrophoretic analyses were obtained from Dr. Jacques P. Vandevoorde, Hoffmann-La Roche Inc., Nutley, New Jersey. Ilz5-labeled CEA was purchased from Hoffmann-La Roche Inc. Donkey anti-rabbit globulin (RD17) was purchased from Wellcome Reagents Limited. Radiolabeling of protein. Approximately 3 to 5 pg of protein was labeled with IIt5 by the chloramine T method.6 Unbound iodine was separated from labeled antigen by chromatography on bovine serum albumen (BSA)-coated Sephadex G-75 (1.6 by 30 cm.). Then 2 ml. fractions were collected. Two peaks of radioactivity were routinely obtained; the first contained labeled antigen and the second contained unbound Ilt5. Immunologic methods. Ouchterlony double diffusion was carried out on lantern slides covered with 1 per cent agarose in O.OlM Tris, O.OOlM ethylenediaminetetra-acetic acid (EDTA) (pH 8.6) containing 0. IM sodium chloride. Immunoelectrophoresis was carried out on microscope slides covered with the diffusion agarose. The electrophoresis buffer was 0.05M barbital (pH 8.6). Electrophoresis was at 110 v. for three hours. Cross-reactivity with normal serum components was determined as follows: Anti-normal human serum* was coupled to CNBr-activated Sepharose 4B according to the manufacturer’s instructions. The protein to be investigated was labeled with IIXs. A known amount of the iodinated protein (less than 10 ng.) was added to 2 ml. of a 10 per cent suspension of anti-normal human serum-Sepharose in 0.2M borate buffer (pH 8) containing 1.6M sodium chloride. A blank with uncoupled Sepharose was run for each sample. After incubation for one hour at room temperature, the samples were centrifuged at 1,500 x g for 10 minutes. Supernatants were aspirated and the pellets were washed once with 2 ml. of borate buffer. The pellets were counted for one minute in a gamma counter. Results are expressed as counts per minute (c.p.m.) in the pellet minus counts per minute of the blank divided by the total counts per minute X 100. Acrylamide gel electrophoresis. Sodium dodecyl Sdfitt (SDS) polyacrylamide gel electrophoresis (PAGE) was carried out in a slab gel apparatus? with 5.6 per cent gels.” Tris glycine gel electrophoresis was performed with 5 per cent gels in a discontinuous buffer system, as described by Ornstein4 and Davis.5 The *No. 01 I-121. Mesa, California.
Hyland
Div.,
tSE 500, Hoefer Scientific California.
Travenol
Instruments,
Labs.,
Inc.,
Costa
San Francisco,
RIA for detecting OCA
781
gel slabs were dried on Whatman No. 1 paper’ in cacuo and autoradiographed on medical x-ray fitrn.:‘: Sephadex G-100 chromatography. Approximately 3 to 4 mg. of protein in 1 ml. of 0.2M borate buffer (pH 8) containing 1.6M sodium chloride wab applied to a 1.6 by 60 cm. Sephadex G- 100 column equilibrated at 4” C. with the same buffer. The column c~tfiurnt was monitored at an optical density of 280 nm. and 1 ml. fractions were collected. Protein determinations were carried out by the fluorescamine method 01‘ Bohlen and associates.’ Radioimmunoassay for OCA. A solution of 0.01M phosphate buffer (pH 7.8), containing U. 15M sodium chloride and 0.1 per cent sodium azidr (PBS), was prepared. BSA buffer (pH 7.5) was prepared by adding 0.2 per cent BSA to PBS. EDTA buffer (+-I 7.5) was prepared by adding O.OlM EDTA to BSA buffer. The radioimmunoassay for OCA (C)CA-KIA) was performed in the following manner: Thy sample 01 standard was added to a 10 by 75 mm. glass tube. First 100 ~1 of BSA buffer and then 20 ~1 (IF anti-OC::2 gamma globulin (diluted 114,000 in BS.4 buffer) was added to each tube. The tubes were placed in a 37” C. water bath for 45 minutes. Then 20 ~1 of I”“-OC.4 (a 1120 dilution of the Sephadex G-75 peak f’raction in BSA buffer containing approximately 4 ii IO” c.p.m. per milliliter) was added to each tube and rhe tubes were incubated 16 to 20 hours at 37” (:. ?o rach t&e 100 ~1 of precipitating antibody (a 1125 dilution of donkey anti-rabbit globulin in EDTA buffer containing 0.5 per cent normai rabbit serum) was added. The tubes were incubated at 37” C. for one haul- and placed in an ice bath. Then 700 ~1 of ice-cold ED’l’i\ buffer was added to each tube. The tubes we]-c incubated for 20 minutes in the ice bath and then centrifuged at 1,500 x g for 30 minutes. The supernatant was aspirated and the pellet was counted for one minute in a gamma counter. All standards and plasma samples wc1.e assayed in triplicate. ‘The plasma sample volume W.IS 100 ~1. A standard curve was performed with eaclr assay. The primary OC.4 standard contained I34 ~g of protein per milliliter as determined by fluorometric assay.? .4 10m4 dilution of the standard was used for thr standard curve. The avarage value of counts per minute was plotted versus nanograms of OCA. and the (X.4 content of each sample was determined from 1his standard curve. Normal pooled A+ plasma, a standard ovarian antigen solution, and three plasma poc~ls of known OCA content were internal controls irl each assav. *RP York.
RowI
M-Omat.
Eastman
Kodak
(A>.. Rocltester.
NW
782
Knauf and Urbach
I
Fig. 1. A, Autoradiogram of 5 per cent Tris glycine slab gel (5.6 per cent) of labeled CEA and ovarian antigen (OC). B, Autoradiogram of 5.6 per cent SDS slab gel of I ‘2s-labeled CEA and ovarian antigen (OC]. Radioimmunoassay for CEA. The radioimmunoassay for CEA was performed by the zirconyl phosphate gel method with the CEA-Roche test kit. CEA determinations reported in this paper were carried out in the laboratory of Dr. A. Malkin, Department of Clinical Biochemistry, Sunnybrook Medical Centre, Toronto, Ontario, Canada. Collection of plasma samples. Blood was obtained by venipuncture and collected in a lavender stoppered Vacutainer tube* containing 0.07 ml. of 15 per cent EDTA and 0.014 mg. of potassium sorbate. After collection the sample was mixed and stored at room temperature (maximum eight hours) before being centrifuged at 1,000 x g for 10 minutes. The plasma was removed and stored at 4” C. untii assayed. If the sample was not to be assayed within two weeks, the sample was stored frozen at -20” C. The sample was thawed and centrifuged at 1,000 x g for five minutes before being assayed. *Becton, Dickinson & Co., Cockeysville, Maryland.
Control normal plasma was obtained from pooled outdated blood bank plasma from A-positive donors. Individual plasma samples were mixed together, allowed to stand at room tem,perature for one hour, and then centrifuged at 1,000 x g for 10 minutes. The clarified plasma was aliquoted and stored at -70” C. until needed, at which time the plasma was thawed and filtered through Whatman No. 1 paper. Results P-labeled ovarian tumor antigen and CEA with the use of Tris glycine and SDS slab gels. The Tris glycine autoradiograph is shown in Fig. 1, A and the SDS autoradiograph is shown in Fig. 1. B. Both ovarian tumor antigen and CEA contain more than one species. There appears to be a difference in the electrophoretic mobilities of ovarian antigen and CEA. On Tris glycine gels, both CEA and ovarian tumor antigen contain species that migrate more slowly than BSA; ovarian antigen has one very diffuse band and CEA has two bands that migrate further down the gel than
Double-antibody
RIA for detecting OCA
783
.4
.3
E
.2
6 z
.l
I
OCA;
OCB ;
:
occ
;
OCD
:
:
\
0 20
30
40 FrdCtiOfl
50
bo
Number
Fig. 2. Sephadex G-100 protein elution antigen (see “Material and methods”).
profile of ovarian
the diffuse ovarian tumor antigen band. In addition, there are two bands with molecular weights less than that of BSA in the ovarian antigen preparation. In the denaturing SDS gel system, CEA has two highmolecular weight bands and a third diffuse band while the ovarian tumor antigen preparation contains two high-molecular weight bands that have approximately the same mobility as the major two bands of CEA plus two other bands of lower apparent molecular weight than BSA. Since the ovarian tumor antigen is composed of at least three different species, it was fractionated on Sephadex G- 100 in an attempt to separate these species and to ascertain which of these species were normal and which were tumor-associated. The protein elution profile is presented in Fig. 2. The two peaks obtained were divided into two fractions each-fractions OCA and OCB from the first peak and fractions OCC and OCD from the second peak. Each fraction was labeled with ILzs and examined by autoradiography after SDS PAGE (Fig. 3). There are six different bands which can be detected-three in OCA, three in OCB, five in OCC, and three in OCD. It should be noted that, although bands III and IV have very similar relative mobilities, these two bands are different with respect to
Fig. 3. Autoradiogram of SDS PAGE: in 5.6 per cent gels* of ovarian antigen Sephadex G-100 fractions OCA. C)CR, OCX:, and OCD. ‘This is a composite from onp 4ah gel. shape as well as mobility. Only OCA contams band I11 material; all other fractions contain band Ii:. Unfractionated ovarian tumor antigen and fractions OCA and OCD were examined for cros>-reactivity i%.ith antibody against normal serum components. Lrnfractionated antigen contained 40 per cent. ()<:A contained less than .5 per cent, and OCD contained 30 per cent normal serum components, as estimated by binding of I’“5-labeled protein to anti-normal WI um Sepharose 4B. Antibodies were prepared against tractions O(:A (acrylamide bands I, II, III, and low-normal serum contamination) and OCD (bands IV. V. ‘L’I. and rclatively high-normal serum contamination). ‘l‘hese antibodies were subsequently used in immunologic studies of the cross-reactivity of CXA. O( :I), and CEA. In Ouchterlony diffusion, there was no c.ross-reactivity
784
Knauf and Urbach
Fig. 4. Ouchterlony
Fig. 5. Immunoelectrophoresis OCA
patterns
of OCA
ofOCAand
and
CEA
CEAagainst
with
anti-OCD(A),
anti-OCA
(I?), and anti-CEA(C).
anti-
and anti-CEA.
between OCA and OCD. Lines of identity were seen in diffusion of OCA and CEA against anti-OCA, of OCD and CEA against anti-OCD, and of OCA and CEA or of OCD and CEA against anti-CEA. When CEA was tested against anti-OCA, anti-OCD, and anti-CEA (Fig. 4), it became evident that CEA contained at least two antigenic groups, one that was detected by anti-OCA and one that was detected by anti-OCD. Both antigenic groups were detected by anti-CEA, even though there appeared to be only one precipitation line. In contrast, there was only a single line of identity with OCA against anti-CEA and anti-OCA and no precipitation arc against anti-OCD. An additional difference between OCA and CEA was noted in immunoelectrophoresis. CEA appeared to be more highly charged than OCA (Fig. 5). A double-antibody radioimmunoassay for OCA was developed for the purpose of measuring plasma OCA. The standard curve data from 28 separate assays performed over a period of three months, which included four preparations of P-OCA, three different dilutions of OCA, and two different preparations of anti-
1.0
U
nancgrams
Fig. 6. Standard
curve
3.0
2.0
data
from
OCA
28 separate
OCA
assays.
OCA, are presented in Fig. 6. To account for radioactive decay, the results are expressed as the counts per minute bound by the standard divided by the counts bound in the zero standard x 100 (B/B, X 100). Fig. 7 shows a nomogram for the assay, with both inter- and intra-assay statistics given for values up to 0.9 ng. of OCA activity. Although the radioimmunoassay appears to be very reproducible, preliminary studies indicated that it was critical to be able to distinguish between values in the
Volume Number
Double-antibody
131 7
RIA for detecting
OCA
785
6
5
.
4 l
Range
. %
3
.
.
.
l
. .
.
c
l
.
P
.. . .
l
.
2
.* l :
.
.
.
.
.
l
.
. .
.
‘=:: . .
l
.
. l
:
. ”
.
I
.
.
.. . :..
. .
l
r’ .
4 I
2
3
4
5
6
ng CEA 1
.2
.3
.4
.5
nanqrams
.6
.7
.8
.9
1.0
OCA
Fig, 7. A nomogram showing the intra- and inter-assay statistics for the OCA-RIA. range 0.5 to 0.7 ng. of OCA per 100 ~1 of plasma if we hoped to differentiate between “normal” and “abnormal” plasma OCA levels. Despite the fact that less than 5 per cent of labeled OCA could be precipitated by anti-normal human serum,’ the A+ plasma pool contained 0.35 ? 0.04 (S.D.) ng. of OCA per 100 ~1 (17 assays). It was determined that, if the A+ plasma pool was assayed twice in every assay (once at the beginning and once at the end) and the mean A+ plasma pool value was then subtracted from each sample value, the interassay precision was increased. The results presented here are calculated in this manner. A summary of a preliminary study of ovarian cancer patients, patients with noncancerous gynecologic disease, pregnant patients, and apparently disease-free control subjects is presented in Table I. Of known ovarian cancer patients who were currently receiving treatment for disease, 52 per cent had OCA levels greater than 1.9 ng. per milliliter. No pregnant patients, three of 29 patients with noncancerous gynecologic disease, and five of 61 normal subjects (8 per cent) had OCA levels greater than 1.9 ng. per milliliter. Accordingly, an OCA value less than 1.8 ng. per milliliter is considered normal and a value greater than 1.9 is considered abnormal. A value of 1.8 or 1.9 is considered questionable. The OCA-RIA was also used to evaluate further the relationship of the ovarian antigen to CEA. Unfrac-
Fig. 8. A scatter plot of data obtained by simultaneous evaluation of OCA and CEA levels in plasma samples from ovarian cancer patients. The levels are in nanograms per milliliter of plasma. rz = 0. I. tionated ovarian antigen had 3.8 ng. of total protein, 1.82 ng. of OCA activity (48 per cent), and 0.26 ng. of CEA activity (7 per cent) per milliliter. The CEA standard supplied with the CEA-Roche test kit had 125 ng. of CEA activity per milliliter and 57 ng. of OC.4 activity per milliliter. Simultaneous evaluation of OCA and CEA levels in plasma from ovarian cancer patients attending the Out-Patient Clinic at The Princess Margaret Hospital (PMH) in Toronto was carried out to determine the specificity of OCA and CEA for ovarian carcinoma. The results from 77 plasma samples obtained from these ovarian cancer patients are presented in Table II. The CEA level considered to be normal varies from study to study. The level of 2.5 ng. per milliliter is the cutoff point recommended by Hoffmann-I.a Roche Inc., but the level of 4.0 ng. per milliliter was chosen by Malkin and co-workers* for a study of ov:irian cancer patients attending the PMH clinic. Therefore, the data are divided into three groups-less than 2..i, 2.6 to 3.9, and 4.0 or more ng. of CEA activity per milliliter. It can be seen that elevated OCA levels are detected in 60 per cent while elevated CEA levels (~4.0 ng. per milliliter) were present in only 16 per cent of the patients tested. Even if the normal level of CEA is considered to be less than 2.5 ng. per milliliter, only 28 per cent of the pa*Personal communication.
786
Knauf and Urbach
Table I. OCA plasma levels in ovarian cancer and control samples from patients with noncancerous gynecologic disease, pregnant patients, and apparently disease-free control subjects OCA value -Cl .8 (ng. lml.)
>I .9 (ng.lml.)
1.8-1.9(ng.lml.)
Diagnosis
Total No.
NO.
%
No.
Y@
No.
%
Ovarian cancer Noncancerous gynecologic disease Pregnant Apparently disease-free
126 29 19 61
54 25 17 55
43 86 89 90
6
5 3
66 3
52
11 2
0 5
Table II. A simultaneous
evaluation
1
2 1
of OCA and CEA levels in plasma from
77 ovarian
10 i
cancer patients
OCA
1.8-l .9 ng. iml.
>I .9 ng. iml.
Total
CEA
No.
%
No.
%
No.
5%
No.
70
<2.5 ng./ml. 2.5-3.9 ng./ml. >3.9 ng./ml. Total
27 1 1 29
35 1 1 38
2 0 0 2
3 0 0 3
27 8 11 46
35
56
72
10
9
12
12
16
tients tested had elevated CEA levels. CEA levels were elevated (~2.5 ng. per milliliter) in only two patients who did not have elevated OCA titers, while OCA titers were elevated in 27 patients who did not have elevated CEA levels. That the OCA assay measures different antigenic material as compared to the CEA assay is confirmed by the data presented in Fig. 8. In this scatter plot of OCA and CEA levels in plasma from ovarian cancer patients, it can be seen that there is virtually no correlation between OCA and CFA plasma levels in these patients (r2 = 0.1).
Comment Ovarian tumor antigen and CEA differ with respect to electrophoretic mobility in acrylamide gel electrophoresis and immunoelectrophoresis. From immunodiffusion studies of Sephadex G- 100 ovarian tumor antigen fractions OCA and OCD, CEA, and their respective antibodies, it was evident that both OCA and OCD contained antigenic determinants that cross reacted with or were identical to those of CEA. However, OCA did not share antigenic determinants with OCD. OCA appears to be a subspecies of CEA, with fewer antigenic determinants and a “carrier” molecule that has a different charge from that of CEA. With the use of a double-antibody radioimmunoassay for plasma OCA, it has been determined that a plasma OCA level below 1.8 ng. per milliliter is “nor-
mal” while a level above 1.9 ng. per milliliter is an indication of the malignant state. Values of 1.8 to 1.9 ng. per milliliter are considered questionable, and a second plasma sample should be obtained. Fifty-six of 61 normal subjects and all 19 pregnant patients tested had plasma OCA levels less than 2.0. Three patients with noncancerous gynecologic disease had elevated OCA levels. One of these patients was being treated for mammary hypertrophy and menorrhagia and the other two were being treated for ovarian cysts and had a history of previous surgical removal of ovarian cysts. It is interesting that three patients with benign ovarian cysts and no previous history of cysts had normal OCA levels. Comparison of OCA and CEA levels in plasma samples from 77 ovarian cancer patients indicated that OCA is more specific for ovarian cancer in comparison to CEA. Only 21 (28 per cent) ovarian cancer patients had CEA titers of 2.5 ng. of CEA per milliliter or higher, while 46 (60 per cent) had elevated OCA titers (2.0 ng. of OCA per milliliter or higher). Two (3 per cent) patients had elevated CEA and normal OCA levels while 27 (35 per cent) patients had elevated OCA and normal CEA levels. Of 126 ovarian cancer patients being seen at the PMH Out-Patient Clinic, 66 patients (52 per cent) had elevated OCA levels and six patients (5 per cent) had OCA levels of 1.8 or 1.9 ng. per milliliter. Included in the 126 were patients with both active and inactive disease as well as patients who had operable ovarian
Volume
13 I
Yumher
7
Double-antibody
cancer and who were clinically free of further disease at the time of testing. Until more data concerning the stage and activity of disease are available, it is not possible to evaluate OCA as an indicator of malignant activity. Large-scale testing is necessary to evaluate the selectivity and specificity of the OCA assay for ovarian cancer.
RIA for detecting OCA
787
We thank Dr. R. S. Bush, The Princess Margaret Hospital, Toronto, Ontario, Canada, who submitted patient material and coordinated activity to obtain blood specimens from ovarian cancer patients. We are also indebted to Dr. M. LeTarte for helpful discussion of the data and manuscript.
REFERENCES
1. Knauf, S., and Urbach, G. I.: Purification of human ovarian tumor-associated antigen and demonstration of circulating tumor antigen in patients with advanced ovarian malignancy, AM. J. OBSTET. GYNECOL. lP7: 705, 1977. 2. Campbell, D. H., Garvey, J. S., Cremer, N. E., and Sussdorf, D. H.: Methods in immunology, New York, 1963, W. A. Benjamin, Inc., p. 118. 3. Fairbanks, G., Steck, T. L., and Wallach, D. F. H.: Electrophoretic analysis of the major polypeptides of the human erythrocyte membrane, Biochemistry IO: 2606, 1971.
4. Ornstein, L.: Disc electrophoresis. I. Background and theory, Ann. N. Y. Acad. Sci. 141: 321, 1964. 5. Davis, B.: Disc Electrophoresis. II. Method and application to human serum proteins, Ann. N. Y. Acad. Sci. 191: 404, 1964. 6. Greenwood, F. C., Hunter, W. M., and Glover, J. S.: The preparation of I isi-labeled human growth hormone of high specific radioactivity, Biochem. J. 69: 114, 1963. 7. Bohlen, P., Stein, S., Dairman, W., and Udenfriend, S.: Fluorometric assay of proteins in the nanogram range, Arch. B&hem. Biophys. 155: 213, 1973.
Inform&Ion for authors Most of the provisions of the Copyright Act of 1976 became effective on January 1, 1978. Therefore, all manuscripts must be accompanied by the following statement, signed by each author: “The undersigned author(s) transfers all copyright ownership of the manuscript entitled (title of article) to The C. V. Mosby Company in the event the work is published. The author(s) warrants that the article is original, is not under consideration by another journal, and has not been previously published.” Authors will be consulted, when possible, regarding republication of their material.
KNAUF
URBACH,
M.D.
Toronto, Ontario, Canada Ovarian tumor-associated antigen isolated from human tumor tissue was shown to have a different mobility from that of carcinoembryonic antigen (CEA) in both acrylarnide gel electrophoresis and immunoelectrophoresis in agarose. The ovarian tumor antigen is composed of six species with different electrophoretic mobility in acrylamide gel electrophoresis. Three of these species were detected in Sephadex G-100 ovarian fraction OCA (from the void volume peak) and the other three species of lower apparent molecular weight were detected in fraction OCD (from the second peak). Fractions OCA and OCD did not share common antigenic determinants as determined by immunodiffusion. CEA was shown to share antigenic determinants with both OCA and OCD. A double-antibody radioimmunoassay capable of detecting nanogram quantities of plasma OCA was developed. In a preliminary study of ovarian cancer patients, OCA appeared to be a more sensitive marker for ovarian cancer than CEA. There was virtually no correlation (r2 = 0.1) between OCA and CEA levels in these patients, as determined by radioimmunoassay. (AM. J. OBSTET. GYNECOL. 131: 780, 1978.)
TUMOR-ASSOCIATED ANTIGEN purified in our laboratory was contaminated by normal antigen.’ Therefore, the physiochemical properties of the ovarian tumor antigen preparation were examined for the purpose of discovering a method for the removal of normal antigen so that a radioimmunoassay for circulating ovarian tumor antigen couId be developed. Since carcinoembryonic antigen (CEA) is currently the tumor-associated antiqen monitored in ovarian cancer patients, ovarian tumor antigen was compared with CEA. This paper reports the results of immunologic studies of the cross-reactivity between ovarian tumor antigen and CEA, the development of a double-antibody radioimmunoassay for the ovarian tumor antigen fraction OCA, and a comparison of
OVARIAN
From the Department of Obstekcs and Gynaecology, University of Toronto, The Wellesley Hospital. Supported Foundation
by Onkwio Cancer Project No. 324
Received for publication Revised
November
Accepted Jama?
Treatment
Sepkmber
and Research
19, 1977.
29, 1977. 20, 1978.
Reprint requests: Dr. G. I. Urbach, 600 Sherboumz Suite 805, Toronto, Ontario, Canada M4X 1 W4.
780
St.,
the OCA and CEA radioimmunoassays for the assessment of circulating tumor antigen in patients with ovarian cancer. Material
and methods
Immundogic reagents. Ovarian tumor antigen was prepared as previously described.’ Briefly, human ovarian tumor tissue was homogenized in saline, extracted with an equal volume of 1.2M perchloric acid, and subjected to affinity chromatography on concanavalin A-Sepharose.* Bound protein was eluted with 10 and 25 per cent n-mannose. The 10 per cent mannose fraction was subsequently purified by passage through affinity columns that contained anti-normal female serum, anti-normal ovarian tissue, and antimannose fractions from 2, 10, and 25 per cent concanavalin A-Sepharose fractionation of normal female serum coupled to Sepharose 4B. The eluate was concentrated in an ultrafiltration unit? with a UM-10 filter. Antibodies were prepared in female New Zealand white rabbits, as previously described.’ Gamma globulin was prepared by two precipitations with 33 per cent saturated ammonium sulfate.” *Pharmacia Laboratories, tAmicon Corp., Lexington,
000%9378/78/07131-0780$00.80/O
Inc., Piscataway, Massachusetts.
@ 1978
The
New
Jersey.
C. V. Mosby
Co
Volume Number
Double-antibody
13 1 7
CEA and absorbed anti-CEA for Ouchterlony and immunoelectrophoretic analyses were obtained from Dr. Jacques P. Vandevoorde, Hoffmann-La Roche Inc., Nutley, New Jersey. Ilz5-labeled CEA was purchased from Hoffmann-La Roche Inc. Donkey anti-rabbit globulin (RD17) was purchased from Wellcome Reagents Limited. Radiolabeling of protein. Approximately 3 to 5 pg of protein was labeled with IIt5 by the chloramine T method.6 Unbound iodine was separated from labeled antigen by chromatography on bovine serum albumen (BSA)-coated Sephadex G-75 (1.6 by 30 cm.). Then 2 ml. fractions were collected. Two peaks of radioactivity were routinely obtained; the first contained labeled antigen and the second contained unbound Ilt5. Immunologic methods. Ouchterlony double diffusion was carried out on lantern slides covered with 1 per cent agarose in O.OlM Tris, O.OOlM ethylenediaminetetra-acetic acid (EDTA) (pH 8.6) containing 0. IM sodium chloride. Immunoelectrophoresis was carried out on microscope slides covered with the diffusion agarose. The electrophoresis buffer was 0.05M barbital (pH 8.6). Electrophoresis was at 110 v. for three hours. Cross-reactivity with normal serum components was determined as follows: Anti-normal human serum* was coupled to CNBr-activated Sepharose 4B according to the manufacturer’s instructions. The protein to be investigated was labeled with IIXs. A known amount of the iodinated protein (less than 10 ng.) was added to 2 ml. of a 10 per cent suspension of anti-normal human serum-Sepharose in 0.2M borate buffer (pH 8) containing 1.6M sodium chloride. A blank with uncoupled Sepharose was run for each sample. After incubation for one hour at room temperature, the samples were centrifuged at 1,500 x g for 10 minutes. Supernatants were aspirated and the pellets were washed once with 2 ml. of borate buffer. The pellets were counted for one minute in a gamma counter. Results are expressed as counts per minute (c.p.m.) in the pellet minus counts per minute of the blank divided by the total counts per minute X 100. Acrylamide gel electrophoresis. Sodium dodecyl Sdfitt (SDS) polyacrylamide gel electrophoresis (PAGE) was carried out in a slab gel apparatus? with 5.6 per cent gels.” Tris glycine gel electrophoresis was performed with 5 per cent gels in a discontinuous buffer system, as described by Ornstein4 and Davis.5 The *No. 01 I-121. Mesa, California.
Hyland
Div.,
tSE 500, Hoefer Scientific California.
Travenol
Instruments,
Labs.,
Inc.,
Costa
San Francisco,
RIA for detecting OCA
781
gel slabs were dried on Whatman No. 1 paper’ in cacuo and autoradiographed on medical x-ray fitrn.:‘: Sephadex G-100 chromatography. Approximately 3 to 4 mg. of protein in 1 ml. of 0.2M borate buffer (pH 8) containing 1.6M sodium chloride wab applied to a 1.6 by 60 cm. Sephadex G- 100 column equilibrated at 4” C. with the same buffer. The column c~tfiurnt was monitored at an optical density of 280 nm. and 1 ml. fractions were collected. Protein determinations were carried out by the fluorescamine method 01‘ Bohlen and associates.’ Radioimmunoassay for OCA. A solution of 0.01M phosphate buffer (pH 7.8), containing U. 15M sodium chloride and 0.1 per cent sodium azidr (PBS), was prepared. BSA buffer (pH 7.5) was prepared by adding 0.2 per cent BSA to PBS. EDTA buffer (+-I 7.5) was prepared by adding O.OlM EDTA to BSA buffer. The radioimmunoassay for OCA (C)CA-KIA) was performed in the following manner: Thy sample 01 standard was added to a 10 by 75 mm. glass tube. First 100 ~1 of BSA buffer and then 20 ~1 (IF anti-OC::2 gamma globulin (diluted 114,000 in BS.4 buffer) was added to each tube. The tubes were placed in a 37” C. water bath for 45 minutes. Then 20 ~1 of I”“-OC.4 (a 1120 dilution of the Sephadex G-75 peak f’raction in BSA buffer containing approximately 4 ii IO” c.p.m. per milliliter) was added to each tube and rhe tubes were incubated 16 to 20 hours at 37” (:. ?o rach t&e 100 ~1 of precipitating antibody (a 1125 dilution of donkey anti-rabbit globulin in EDTA buffer containing 0.5 per cent normai rabbit serum) was added. The tubes were incubated at 37” C. for one haul- and placed in an ice bath. Then 700 ~1 of ice-cold ED’l’i\ buffer was added to each tube. The tubes we]-c incubated for 20 minutes in the ice bath and then centrifuged at 1,500 x g for 30 minutes. The supernatant was aspirated and the pellet was counted for one minute in a gamma counter. All standards and plasma samples wc1.e assayed in triplicate. ‘The plasma sample volume W.IS 100 ~1. A standard curve was performed with eaclr assay. The primary OC.4 standard contained I34 ~g of protein per milliliter as determined by fluorometric assay.? .4 10m4 dilution of the standard was used for thr standard curve. The avarage value of counts per minute was plotted versus nanograms of OCA. and the (X.4 content of each sample was determined from 1his standard curve. Normal pooled A+ plasma, a standard ovarian antigen solution, and three plasma poc~ls of known OCA content were internal controls irl each assav. *RP York.
RowI
M-Omat.
Eastman
Kodak
(A>.. Rocltester.
NW
782
Knauf and Urbach
I
Fig. 1. A, Autoradiogram of 5 per cent Tris glycine slab gel (5.6 per cent) of labeled CEA and ovarian antigen (OC). B, Autoradiogram of 5.6 per cent SDS slab gel of I ‘2s-labeled CEA and ovarian antigen (OC]. Radioimmunoassay for CEA. The radioimmunoassay for CEA was performed by the zirconyl phosphate gel method with the CEA-Roche test kit. CEA determinations reported in this paper were carried out in the laboratory of Dr. A. Malkin, Department of Clinical Biochemistry, Sunnybrook Medical Centre, Toronto, Ontario, Canada. Collection of plasma samples. Blood was obtained by venipuncture and collected in a lavender stoppered Vacutainer tube* containing 0.07 ml. of 15 per cent EDTA and 0.014 mg. of potassium sorbate. After collection the sample was mixed and stored at room temperature (maximum eight hours) before being centrifuged at 1,000 x g for 10 minutes. The plasma was removed and stored at 4” C. untii assayed. If the sample was not to be assayed within two weeks, the sample was stored frozen at -20” C. The sample was thawed and centrifuged at 1,000 x g for five minutes before being assayed. *Becton, Dickinson & Co., Cockeysville, Maryland.
Control normal plasma was obtained from pooled outdated blood bank plasma from A-positive donors. Individual plasma samples were mixed together, allowed to stand at room tem,perature for one hour, and then centrifuged at 1,000 x g for 10 minutes. The clarified plasma was aliquoted and stored at -70” C. until needed, at which time the plasma was thawed and filtered through Whatman No. 1 paper. Results P-labeled ovarian tumor antigen and CEA with the use of Tris glycine and SDS slab gels. The Tris glycine autoradiograph is shown in Fig. 1, A and the SDS autoradiograph is shown in Fig. 1. B. Both ovarian tumor antigen and CEA contain more than one species. There appears to be a difference in the electrophoretic mobilities of ovarian antigen and CEA. On Tris glycine gels, both CEA and ovarian tumor antigen contain species that migrate more slowly than BSA; ovarian antigen has one very diffuse band and CEA has two bands that migrate further down the gel than
Double-antibody
RIA for detecting OCA
783
.4
.3
E
.2
6 z
.l
I
OCA;
OCB ;
:
occ
;
OCD
:
:
\
0 20
30
40 FrdCtiOfl
50
bo
Number
Fig. 2. Sephadex G-100 protein elution antigen (see “Material and methods”).
profile of ovarian
the diffuse ovarian tumor antigen band. In addition, there are two bands with molecular weights less than that of BSA in the ovarian antigen preparation. In the denaturing SDS gel system, CEA has two highmolecular weight bands and a third diffuse band while the ovarian tumor antigen preparation contains two high-molecular weight bands that have approximately the same mobility as the major two bands of CEA plus two other bands of lower apparent molecular weight than BSA. Since the ovarian tumor antigen is composed of at least three different species, it was fractionated on Sephadex G- 100 in an attempt to separate these species and to ascertain which of these species were normal and which were tumor-associated. The protein elution profile is presented in Fig. 2. The two peaks obtained were divided into two fractions each-fractions OCA and OCB from the first peak and fractions OCC and OCD from the second peak. Each fraction was labeled with ILzs and examined by autoradiography after SDS PAGE (Fig. 3). There are six different bands which can be detected-three in OCA, three in OCB, five in OCC, and three in OCD. It should be noted that, although bands III and IV have very similar relative mobilities, these two bands are different with respect to
Fig. 3. Autoradiogram of SDS PAGE: in 5.6 per cent gels* of ovarian antigen Sephadex G-100 fractions OCA. C)CR, OCX:, and OCD. ‘This is a composite from onp 4ah gel. shape as well as mobility. Only OCA contams band I11 material; all other fractions contain band Ii:. Unfractionated ovarian tumor antigen and fractions OCA and OCD were examined for cros>-reactivity i%.ith antibody against normal serum components. Lrnfractionated antigen contained 40 per cent. ()<:A contained less than .5 per cent, and OCD contained 30 per cent normal serum components, as estimated by binding of I’“5-labeled protein to anti-normal WI um Sepharose 4B. Antibodies were prepared against tractions O(:A (acrylamide bands I, II, III, and low-normal serum contamination) and OCD (bands IV. V. ‘L’I. and rclatively high-normal serum contamination). ‘l‘hese antibodies were subsequently used in immunologic studies of the cross-reactivity of CXA. O( :I), and CEA. In Ouchterlony diffusion, there was no c.ross-reactivity
784
Knauf and Urbach
Fig. 4. Ouchterlony
Fig. 5. Immunoelectrophoresis OCA
patterns
of OCA
ofOCAand
and
CEA
CEAagainst
with
anti-OCD(A),
anti-OCA
(I?), and anti-CEA(C).
anti-
and anti-CEA.
between OCA and OCD. Lines of identity were seen in diffusion of OCA and CEA against anti-OCA, of OCD and CEA against anti-OCD, and of OCA and CEA or of OCD and CEA against anti-CEA. When CEA was tested against anti-OCA, anti-OCD, and anti-CEA (Fig. 4), it became evident that CEA contained at least two antigenic groups, one that was detected by anti-OCA and one that was detected by anti-OCD. Both antigenic groups were detected by anti-CEA, even though there appeared to be only one precipitation line. In contrast, there was only a single line of identity with OCA against anti-CEA and anti-OCA and no precipitation arc against anti-OCD. An additional difference between OCA and CEA was noted in immunoelectrophoresis. CEA appeared to be more highly charged than OCA (Fig. 5). A double-antibody radioimmunoassay for OCA was developed for the purpose of measuring plasma OCA. The standard curve data from 28 separate assays performed over a period of three months, which included four preparations of P-OCA, three different dilutions of OCA, and two different preparations of anti-
1.0
U
nancgrams
Fig. 6. Standard
curve
3.0
2.0
data
from
OCA
28 separate
OCA
assays.
OCA, are presented in Fig. 6. To account for radioactive decay, the results are expressed as the counts per minute bound by the standard divided by the counts bound in the zero standard x 100 (B/B, X 100). Fig. 7 shows a nomogram for the assay, with both inter- and intra-assay statistics given for values up to 0.9 ng. of OCA activity. Although the radioimmunoassay appears to be very reproducible, preliminary studies indicated that it was critical to be able to distinguish between values in the
Volume Number
Double-antibody
131 7
RIA for detecting
OCA
785
6
5
.
4 l
Range
. %
3
.
.
.
l
. .
.
c
l
.
P
.. . .
l
.
2
.* l :
.
.
.
.
.
l
.
. .
.
‘=:: . .
l
.
. l
:
. ”
.
I
.
.
.. . :..
. .
l
r’ .
4 I
2
3
4
5
6
ng CEA 1
.2
.3
.4
.5
nanqrams
.6
.7
.8
.9
1.0
OCA
Fig, 7. A nomogram showing the intra- and inter-assay statistics for the OCA-RIA. range 0.5 to 0.7 ng. of OCA per 100 ~1 of plasma if we hoped to differentiate between “normal” and “abnormal” plasma OCA levels. Despite the fact that less than 5 per cent of labeled OCA could be precipitated by anti-normal human serum,’ the A+ plasma pool contained 0.35 ? 0.04 (S.D.) ng. of OCA per 100 ~1 (17 assays). It was determined that, if the A+ plasma pool was assayed twice in every assay (once at the beginning and once at the end) and the mean A+ plasma pool value was then subtracted from each sample value, the interassay precision was increased. The results presented here are calculated in this manner. A summary of a preliminary study of ovarian cancer patients, patients with noncancerous gynecologic disease, pregnant patients, and apparently disease-free control subjects is presented in Table I. Of known ovarian cancer patients who were currently receiving treatment for disease, 52 per cent had OCA levels greater than 1.9 ng. per milliliter. No pregnant patients, three of 29 patients with noncancerous gynecologic disease, and five of 61 normal subjects (8 per cent) had OCA levels greater than 1.9 ng. per milliliter. Accordingly, an OCA value less than 1.8 ng. per milliliter is considered normal and a value greater than 1.9 is considered abnormal. A value of 1.8 or 1.9 is considered questionable. The OCA-RIA was also used to evaluate further the relationship of the ovarian antigen to CEA. Unfrac-
Fig. 8. A scatter plot of data obtained by simultaneous evaluation of OCA and CEA levels in plasma samples from ovarian cancer patients. The levels are in nanograms per milliliter of plasma. rz = 0. I. tionated ovarian antigen had 3.8 ng. of total protein, 1.82 ng. of OCA activity (48 per cent), and 0.26 ng. of CEA activity (7 per cent) per milliliter. The CEA standard supplied with the CEA-Roche test kit had 125 ng. of CEA activity per milliliter and 57 ng. of OC.4 activity per milliliter. Simultaneous evaluation of OCA and CEA levels in plasma from ovarian cancer patients attending the Out-Patient Clinic at The Princess Margaret Hospital (PMH) in Toronto was carried out to determine the specificity of OCA and CEA for ovarian carcinoma. The results from 77 plasma samples obtained from these ovarian cancer patients are presented in Table II. The CEA level considered to be normal varies from study to study. The level of 2.5 ng. per milliliter is the cutoff point recommended by Hoffmann-I.a Roche Inc., but the level of 4.0 ng. per milliliter was chosen by Malkin and co-workers* for a study of ov:irian cancer patients attending the PMH clinic. Therefore, the data are divided into three groups-less than 2..i, 2.6 to 3.9, and 4.0 or more ng. of CEA activity per milliliter. It can be seen that elevated OCA levels are detected in 60 per cent while elevated CEA levels (~4.0 ng. per milliliter) were present in only 16 per cent of the patients tested. Even if the normal level of CEA is considered to be less than 2.5 ng. per milliliter, only 28 per cent of the pa*Personal communication.
786
Knauf and Urbach
Table I. OCA plasma levels in ovarian cancer and control samples from patients with noncancerous gynecologic disease, pregnant patients, and apparently disease-free control subjects OCA value -Cl .8 (ng. lml.)
>I .9 (ng.lml.)
1.8-1.9(ng.lml.)
Diagnosis
Total No.
NO.
%
No.
Y@
No.
%
Ovarian cancer Noncancerous gynecologic disease Pregnant Apparently disease-free
126 29 19 61
54 25 17 55
43 86 89 90
6
5 3
66 3
52
11 2
0 5
Table II. A simultaneous
evaluation
1
2 1
of OCA and CEA levels in plasma from
77 ovarian
10 i
cancer patients
OCA
1.8-l .9 ng. iml.
>I .9 ng. iml.
Total
CEA
No.
%
No.
%
No.
5%
No.
70
<2.5 ng./ml. 2.5-3.9 ng./ml. >3.9 ng./ml. Total
27 1 1 29
35 1 1 38
2 0 0 2
3 0 0 3
27 8 11 46
35
56
72
10
9
12
12
16
tients tested had elevated CEA levels. CEA levels were elevated (~2.5 ng. per milliliter) in only two patients who did not have elevated OCA titers, while OCA titers were elevated in 27 patients who did not have elevated CEA levels. That the OCA assay measures different antigenic material as compared to the CEA assay is confirmed by the data presented in Fig. 8. In this scatter plot of OCA and CEA levels in plasma from ovarian cancer patients, it can be seen that there is virtually no correlation between OCA and CFA plasma levels in these patients (r2 = 0.1).
Comment Ovarian tumor antigen and CEA differ with respect to electrophoretic mobility in acrylamide gel electrophoresis and immunoelectrophoresis. From immunodiffusion studies of Sephadex G- 100 ovarian tumor antigen fractions OCA and OCD, CEA, and their respective antibodies, it was evident that both OCA and OCD contained antigenic determinants that cross reacted with or were identical to those of CEA. However, OCA did not share antigenic determinants with OCD. OCA appears to be a subspecies of CEA, with fewer antigenic determinants and a “carrier” molecule that has a different charge from that of CEA. With the use of a double-antibody radioimmunoassay for plasma OCA, it has been determined that a plasma OCA level below 1.8 ng. per milliliter is “nor-
mal” while a level above 1.9 ng. per milliliter is an indication of the malignant state. Values of 1.8 to 1.9 ng. per milliliter are considered questionable, and a second plasma sample should be obtained. Fifty-six of 61 normal subjects and all 19 pregnant patients tested had plasma OCA levels less than 2.0. Three patients with noncancerous gynecologic disease had elevated OCA levels. One of these patients was being treated for mammary hypertrophy and menorrhagia and the other two were being treated for ovarian cysts and had a history of previous surgical removal of ovarian cysts. It is interesting that three patients with benign ovarian cysts and no previous history of cysts had normal OCA levels. Comparison of OCA and CEA levels in plasma samples from 77 ovarian cancer patients indicated that OCA is more specific for ovarian cancer in comparison to CEA. Only 21 (28 per cent) ovarian cancer patients had CEA titers of 2.5 ng. of CEA per milliliter or higher, while 46 (60 per cent) had elevated OCA titers (2.0 ng. of OCA per milliliter or higher). Two (3 per cent) patients had elevated CEA and normal OCA levels while 27 (35 per cent) patients had elevated OCA and normal CEA levels. Of 126 ovarian cancer patients being seen at the PMH Out-Patient Clinic, 66 patients (52 per cent) had elevated OCA levels and six patients (5 per cent) had OCA levels of 1.8 or 1.9 ng. per milliliter. Included in the 126 were patients with both active and inactive disease as well as patients who had operable ovarian
Volume
13 I
Yumher
7
Double-antibody
cancer and who were clinically free of further disease at the time of testing. Until more data concerning the stage and activity of disease are available, it is not possible to evaluate OCA as an indicator of malignant activity. Large-scale testing is necessary to evaluate the selectivity and specificity of the OCA assay for ovarian cancer.
RIA for detecting OCA
787
We thank Dr. R. S. Bush, The Princess Margaret Hospital, Toronto, Ontario, Canada, who submitted patient material and coordinated activity to obtain blood specimens from ovarian cancer patients. We are also indebted to Dr. M. LeTarte for helpful discussion of the data and manuscript.
REFERENCES
1. Knauf, S., and Urbach, G. I.: Purification of human ovarian tumor-associated antigen and demonstration of circulating tumor antigen in patients with advanced ovarian malignancy, AM. J. OBSTET. GYNECOL. lP7: 705, 1977. 2. Campbell, D. H., Garvey, J. S., Cremer, N. E., and Sussdorf, D. H.: Methods in immunology, New York, 1963, W. A. Benjamin, Inc., p. 118. 3. Fairbanks, G., Steck, T. L., and Wallach, D. F. H.: Electrophoretic analysis of the major polypeptides of the human erythrocyte membrane, Biochemistry IO: 2606, 1971.
4. Ornstein, L.: Disc electrophoresis. I. Background and theory, Ann. N. Y. Acad. Sci. 141: 321, 1964. 5. Davis, B.: Disc Electrophoresis. II. Method and application to human serum proteins, Ann. N. Y. Acad. Sci. 191: 404, 1964. 6. Greenwood, F. C., Hunter, W. M., and Glover, J. S.: The preparation of I isi-labeled human growth hormone of high specific radioactivity, Biochem. J. 69: 114, 1963. 7. Bohlen, P., Stein, S., Dairman, W., and Udenfriend, S.: Fluorometric assay of proteins in the nanogram range, Arch. B&hem. Biophys. 155: 213, 1973.
Inform&Ion for authors Most of the provisions of the Copyright Act of 1976 became effective on January 1, 1978. Therefore, all manuscripts must be accompanied by the following statement, signed by each author: “The undersigned author(s) transfers all copyright ownership of the manuscript entitled (title of article) to The C. V. Mosby Company in the event the work is published. The author(s) warrants that the article is original, is not under consideration by another journal, and has not been previously published.” Authors will be consulted, when possible, regarding republication of their material.