Journal of Immunological Methods, 127 (1990) 171-178
171
Elsevier JIM 05473
Selection of monoclonal antibodies for use in an immunometric assay for carcinoembryonic antigen Ole P. Bt~rmer a n d Kjell N u s t a d Central Laboratory, The Norwegian Radium Hospital Oslo, Norway
(Received 13 April 1989, revised received29 June 1989, accepted 26 October 1989)
Six high-affinity mouse monoclonal antibodies against carcinoembryonic antigen produced in our laboratory were evaluated for use in a solid-phase immunoradiometric assay. The antibodies all recognized single peptide epitopes, one being highly conformation dependent. Cross-inhibition studies demonstrated that three antibodies bound to different epitopes, while the remaining three bound to the same or to very closely related epitodes. All antibodies strongly stained colorectal cancer tissue. The three antibodies binding to the same epitode also stained normal granulocytes, indicating a reactivity with the non-specific cross-reacting antigen. On the basis of the results of the characterization, two antibodies were selected for use in a sandwich immunometric assay. One of these was coupled to 2.8 ~ m magnetizable polymer particles, while the other one was radiolabelled. The assay required 2 h incubation to reach equilibrium, having a working range up to 135 # g / l and a sensitivity (Zero + 2 SD) of 0.1 ~ g / l . A comparison of the new assay with two commercial CEA kits that also use monoclonal antibodies was carried out on a small panel of serum samples from colorectal cancer patients and revealed satisfactory correlations but also discrepancies that must be attributed to differences in epitope specificity. Key words: Monoclonal antibody; Antigenic determinant; Carcinoembryonicantigen; Immunoassay
Introduction Carcinoembryonic antigen (CEA) in serum is an important parameter for the staging and follow-up of patients with some of the most common forms of cancer. First described by Gold and Freedman (1965), CEA was initially regarded as an oncofetal antigen specific for fetal gut and
Correspondence to: O.P. B6rmer, Central Laboratory, The Norwegian Radium Hospital, N-0310 Oslo, Norway. Abbreviations: BSA, bovine serum albumin; CEA, carcinoembryonic antigen; IRMA, immunoradiometric assay; ISOBM, International Society for Oncodevelopmental Biology and Medicine; NCA, non-specificcross-reacting antigen; PBS, phosphate-buffered saline.
carcinomas of the gastrointestinal tract. Later studies, however, have shown that CEA is also present in small amounts in normal adult tissues (Nap et al., 1988). CEA has recently been shown to be a member of the immunoglobulin gene superfamily, along with several related substances that may cross-react in immunological detection methods (for review, see Thompson and Zimmermann, 1988). Furthermore, CEA is heavily and heterogeneously glycosylated, implying that the molecule may show considerable variations between patients and even heterogeneity within a single patient. These factors, amplified by the use of different standards, explain most of the serious discrepancies between the various commercial assays which have complicated the routine use of CEA assays in cancer patients (Felder et al., 1987).
0022-1759/90/$03.50 (O 1990 Elsevier Science Publishers B.V. (Biomedical Division)
172 The use of monoclonal antibodies in the immunoassays does not in itself solve the problems, but may make it easier to understand the factors that determine the performance of an assay. Under the auspices of the International Society for O n c o d e v e i o p m e n t a l Biology and Medicine (ISOBM), an international workshop is studying the epitope specificity and relationship of a large number of monoclonai anti-CEA antibodies produced by the participants. This has led to increased knowledge about the antigenic structure of CEA (HammarstriSm et al., 1989). Here we report a rapid and sensitive immunometric CEA assay, using monoclonal antibodies that have been well characterized, in part through our participation in the ISOBM workshop.
Materials and methods
Preparation of CEA CEA was purified from liver metastases from colorectal carcinomas by perchloric acid extraction followed by ion exchange and gel filtration chromatography, as described (BiSrmer, 1982).
Production of monoclonal anti-CEA antibodies Mouse hybridomas producing anti-CEA antibodies were made as described (BtSrmer, 1989), using purified CEA as the antigen. The hybridomas were expanded as ascites tumors in mice, and antibodies were purified from the ascitic fluid by Protein A-Sepharose chromatography (Pharmacia, Uppsala, Sweden) at 4 ° C.
Characterization of antibody class and affinity The antibodies selected as promising after a preliminary estimation of affinity and specificity were denoted 12-140-1, -2, -4, -5, -7, and -10. Their immunoglobulin classes, and their affinities for radiolabelled CEA, were determined as described elsewhere (BiSrmer, 1989) and are summarized in Table I.
Radiolabelling of antibodies t2SI-labelling was performed by the Iodo-Gen method (cat. no. 28600, Pierce Chemical Co., Rockford, IL) (Fracker and Speck, 1978), as de-
TABLE I CHEMICAL CHARACTERIZATION OF ANTIBODIES AND EPITOPES Antibody 12-140
IgG Kd x 10H class mol/I
Percentloss a of antibody bindingafter treatment of CEA with Periodate Reduction+ alkylation
-1 -2
1 2b
-4
1
-5 -7 -10
2a 2b 1
3.1 20 30 3.4 6.0 18
48 16 40 10 17 14
12 23 14 67 89 34
a Mean of three experiments.
scribed by Paus et al. (1982). Equimolar amounts of protein and radioiodine were used.
Chemical modifications of CEA For the cross-inhibition and chemical modification experiments, CEA was adsorbed onto the wells of polyvinyl chloride 96-well microtiter plates (cat. no. 001-010-2801, Dynatech Laboratories, Alexandria, VA). The adsorption was performed overnight at room temperature with 1 /~g CEA preparation in 0.1 ml of a 0.2 m o l / I sodium carbonate buffer, pH 9.3, with 0.1% sodium azide. Periodate oxidation of adsorbed CEA, as well as reduction and alkylation of CEA followed by adsorption to microtiter plates, were performed as described by Price (1988). After adsorption and modification of the CEA preparations, the wells were saturated with 1% ( w / v ) bovine serum albumin (BSA) (cat. no. A4503, Sigma Chemical Co., St. Louis, MO) in 0.2 ml phosphate-buffered saline (PBS) (sodium-phosphate 0.01 m o l / l , NaCI 0.15 m o l / l , sodium-azide 0.05% (w/v), p H 7.5) for 2 h at room temperature.
Antibody binding to solid-phase native and modified CEA The effects of chemical modification of CEA on antibody binding were studied by incubating the wells, previously coated with the various CEA preparations, with 0.75 ~g monoclonal anti-CEA in 75 ~1 1% BSA-PBS for 1 h. After washing with PBS, the amount of monoclonal anti-CEA bound
173 TABLE I1 EPITOPE RELATIONSHIP AND IMMUNOHISTOCHEMICAL SPECIFICITY OF ANTI-CEA ANTIBODIES Antibody 12-140
Epitope group a
Percent inhibition b of the binding to CEA of the labelled antibodies
Histochemical staining c of Colon Normal tissues carcinomas Liver Lung
-1
-2
-4
-5
-7
-10
5 2d
99 0
0 96
0 0
0 0
0 0
0 0
++ ++
-
-4
1
88
+ +
-
-5
4 4 4
+++ + + 4+++
+ -
-1 -2
-7 -10
0 0 0
0
30 32 32
98
0 0 0
0
0
0
95 90 75
97 95 84
94 92 92
Granulocytes
+++ +++ ++
a 'Gold' epitope group (Hammarstr~m et al., 1989). b By an excess of the various unlabelled antibodies (see text). CScale_ to + + + . d This antibody was not included in the ISOBM Workshop, but classified later (see text).
was q u a n t i t a t e d b y i n c u b a t i n g the wells for 2 h with a p p r o x . 4 ng (50,000 c p m ) l:5I-labelled imm u n o p u r i f i e d sheep a n t i - m o u s e i m m u n o g l o b u l i n a n t i b o d i e s in 0.1 ml 1% BSA-PBS with 2% n o r m a l sheep serum. A f t e r washing with PBS, the b o u n d r a d i o a c t i v i t y in the s e p a r a t e d m i c r o t i t e r wells was c o u n t e d , a n d a n t i b o d y b i n d i n g was calculated as a p e r c e n t a g e o f the b i n d i n g to u n t r e a t e d C E A . T h e e p i t o p e relationships of the m o n o c l o n a l a n t i b o d i e s were m a p p e d by cross-inhibition studies, essentially as d e s c r i b e d by W a g e n e r et al. (1983b). Briefly, d u p l i c a t e wells with a d s o r b e d C E A were i n c u b a t e d for 2 h with approx. 2.5 ng (30,000 c p m ) labelled a n t i - C E A to be tested, in 50 #1 1% BSA-PBS, together with d o u b l i n g dilutions of various u n l a b e l l e d c o m p e t i n g antibodies, plus a ' b l a n k ' with no u n l a b e l l e d a n t i b o d y a d d e d . A f t e r w a s h i n g with PBS, the b o u n d r a d i o a c t i v i t y in the s e p a r a t e d wells was counted, a n d the results c a l c u l a t e d as the percent inhibition of the b i n d i n g of labelled a n t i b o d y . Except for a n t i b o d y 12-140-2, this e x p e r i m e n t was p e r f o r m e d with all a n t i b o d i e s received t h r o u g h the I S O B M w o r k s h o p as c o m p e titors, as r e p o r t e d b y Hammarstrt~m et ai. (1989). T h e results o b t a i n e d with o u r own a n t i b o d i e s at 2 /Lg c o m p e t i n g a n t i b o d y p e r well are shown in T a b l e II.
Immunohistochemical studies Previously we have e v a l u a t e d the a n t i b o d i e s for use in i m m u n o h i s t o c h e m i s t r y (Dasovi6-Kne~.evi6
et al., 1990). T h e results o b t a i n e d with p a r a f f i n sections of colon c a r c i n o m a s a n d several n o r m a l tissues are s u m m a r i z e d in T a b l e II.
Preparation of solid-phase antibodies for immunoassay use A n t i b o d i e s were c o v a l e n t l y c o u p l e d to tosyla t e d m a g n e t i z a b l e 2.8 ~ m m o n o d i s p e r s e p o l y m e r particles ( D y n a b e a d s M-280, D y n a l , Oslo, N o r way) ( N u s t a d et al., 1988), using 20 m g a n t i b o d y / g p o l y m e r particles. A f t e r a n t i b o d y coupling, the covalent b i n d i n g c a p a c i t y of the i m m u n o b e a d s was s a t u r a t e d with BSA.
The immunoradiometric assay (IRMA) T h e b u f f e r used c o n t a i n e d s o d i u m p h o s p h a t e 0.05 m o l / l , N a C ! 0.1 m o l / l , p H 6.5, with 1% ( w / v ) BSA, 0.1% ( w / v ) m e r t h i o l a t e , a n d 0.1% ( v / v ) T w e e n 20. N o r m a l a d u l t bovine, rat, a n d m o u s e sera were a d d e d to 5, 0.5 a n d 0.1% ( v / v ) , respectively. O u r purified C E A was used as a s s a y s t a n d a r d , in a m a t r i x o f h e a t - i n a c t i v a t e d b l o o d b a n k sera, selected for low C E A content. F o r the zero stand a r d , the s e r u m p o o l was a b s o r b e d with a n t i - C E A 12-140-5 c o u p l e d to c y a n o g e n b r o m i d e a c t i v a t e d S e p h a r o s e 4B. S t a n d a r d s were c h o s e n to cover the range 0 - 1 3 5 # g / l . D u p l i c a t e s o f 0.1 ml s t a n d a r d s , c o n t r o l s or p a t i e n t sera were i n c u b a t e d for 1 h with t25Ilabelled a n t i b o d y 12-140-1 ( a p p r o x . 150,000
174
c p m / 1 0 ng) in 0.1 ml assay buffer. Immunobeads (0.5 mg) with antibody 12-140-10 were added in 0.1 ml assay buffer, and the incubation was continued for 1 h with shaking. The tubes were then washed with 3 x 0.5 ml PBS on magnetic racks (Amerlex-M Separator, Amersham International, Amersham, U.K.), and counted. All radioactivity counting was performed with an LKB/Wallac 1277 Gammamaster (Wallac OY, Turku, Finland). In the immunometric assay the standard curve, the results for unknowns and controls, and the precision profile were calculated by the L K B / W a l l a c 1224 Riacaic program.
of 12-140-4, while the converse was not the case. This possibly reflects steric hindrance not involving the epitope itself. Also shown in Table II is the ISOBM Workshop epitope group classification ('Gold' groups (HammarstriSm et al., 1989)) of the antibodies. Antibody 12-140-2, not included in the Workshop, has been tested by us in cross inhibition experiments with the antibodies Mab 35 (Gold group 2) and CE27 (Gold group 3), kindly provided by Drs. J.P. Mach and F. Buchegger. These experiments demonstrated that 12-140-2 was inhibited up to 80% by Mab 35, while CE27 had no effect. Thus, 12-140-2 should probably be assigned to the Gold group 2.
Results
Characterization of antibodies and their epitopes The dissociation constants of the antibodies (Table I) suggest that even low concentrations of antibody should be able to bind CEA in the ~tg/l range, which is relevant for serum measurements. All antibodies retained more than 50% of their binding after periodate treatment of CEA (Table I) indicating that the epitopes are probably not carbohydrate structures (Price, 1988). After reduction and alkylation of CEA, the binding of antibodies 12-140-5 and -7 was greatly reduced, suggesting that the corresponding epitopes were strongly dependent on protein conformation (Price, 1988). The epitopes for the other antibodies were considerably less affected by this treatment. The relationships between the epitopes for the various antibodies were first evaluated in sandwich immunometric assays, which demonstrated that all antibodies bound to single epitopes on the CEA molecule since no binding of labelled antibody occurred when any one of them was used as both solid-phase and tracer antibody (data not shown). The results of the cross-inhibition experiments (Table II) revealed that antibodies 12-140-5 and -7 bound to the same epitope. Antibody 12-140-10 apparently bound to a closely related but not identical structure. The remaining three antibodies recognized separate epitopes, although some interference occurred between antibodies 12-140-4 and -1 : binding of 12-140-1 was inhibited by an excess
Immunohistochernical studies The results of the immunohistochemical studies, also shown in Table II, demonstrate that all antibodies bound strongly to colon carcinoma cells. Antibodies 12-140-5, -7 and -10 (Gold group 4) also bound to normal granulocytes, indicating that these antibodies recognized an epitope common to CEA and non-specific cross-reacting antigen, NCA. Antibody 12-140-5 also bound slightly to normal hepatocytes, possibly reflecting its higher affinity (Table I). Consequently this antibody was able to recognize smaller amounts of NCA. The remaining three antibodies, 12-140-1, -2, and -4, appeared to be CEA specific in these studies with paraffin sections. Selection of antibodies for the assay Complement factors in normal serum were found to inhibit severely the binding of CEA to solid-phase IgG2a and IgG2b antibodies (B~Srmer, 1989). This complement interference excluded antibodies 12-140-2, -5, and -7 from use on the solid phase. The routine radiolabelling of tracer antibodies made it necessary for these to have a high affinity, since only limited amounts could be used due to radiation safety considerations. In practice, only antibodies 12-140-1 and -5 yielded a fully satisfactory tracer antibody sensitivity. Of these, antibody 12-140-5 was non-specific, reacting also with NCA, as pointed out above. Because a rapid immunometric assay required the use of comparatively large amounts of solid-phase antibody, even a
175
'specific' solid-phase antibody might bind crossreacting substances with lower affinity. Using a non-specific tracer antibody, the overall assay specificity would then be lost. Consequently the CEA-specific antibody 12-140-1 was chosen for radiolabelling. Due to complement interference, only the two remaining IgGl antibodies could be considered for use on the solid phase. Of these, 12-140-4 exhibited some interference with the binding of labelled 12-140-1 to CEA (Table II). Consequently the other lgG1 antibody, 12-140-10, was chosen for coupling to the solid phase. Even if this antibody showed some cross reaction with NCA, sufficient assay specificity was secured by using comparatively low concentrations of the more specific tracer antibody 12-140-1. The immunoradiometric assay The optimal conditions for coupling antibody to the magnetizable polymer particles were determined in separate experiments (data not shown). The composition of the assay buffer was chosen to minimize non-specific binding of tracer antibody to the immunobeads. This was achieved at pH 6.7, adding BSA to 1% ( w / v ) and Tween 20 to 0.1% (v/v). In order to eliminate interference from heterophilic antibodies and other mouse immunoglobulin-binding substances in the human sera, which would give falsely elevated CEA results, normal adult bovine, rat and mouse sera were added (Boscato and Stuart, 1986). Assay buffer with adult bovine serum added to 50% was used initially as the matrix for CEA standards. However, with this matrix the non-specific binding of ageing tracer antibody to the immunobeads increased slightly more in the standards than in the samples, giving a systematic assay drift of approx. 1 # g / I during the 6-week life of radiolabelled antibody. We therefore chose to use a pool of heat-inactivated blood blank sera, selected for low CEA content, as standard matrix. Attempts to remove all CEA from the serum pool with the aid of antibody 12-140-5 coupled to Sepharose were also unsuccessful, since a slight leakage of antibody into the serum caused the standard values to drift by 1-2 #g/1 during a few days in the refrigerator. We therefore used the absorbed serum pool only as a zero standard, and carefully
8 ~ / 0 - -
0
0
8
g Z,, ko
O
~ e / O
~2 o/ o" 01
60
120 180 MINUTESINCUBATION
240
Fig. 1. Binding kinetics of the assay components. The volumes and concentrations of the reagents were as in the immunometric assay (see text). The CEA sample concentration was 5 # g / I . After washing with 3×0.5 ml PBS on magnetic racks, the radioactivity bound to the immunobeads was counted, and calculated as the percentage of radiolabelled antibody added. ( o ) Binding of complexes of CEA and radiolabelled antibody 12-140-1, preformed by incubation overnight, to immunobeads with antibody 12-140-10; ( I ) binding to immunobeads of CEA and radiolabelled antibody, added simultaneously; (r,) binding of radiolabelled antibody to CEA, followed by a short (15 min) incubation with immunobeads.
calibrated the CEA content of the untreated serum pool before preparing the other standards. The experiments performed to determine the necessary incubation times in the new assay (Fig. 1) yielded results of practical importance. When all components (sample, tracer antibody, and immunobeads) were added simultaneously, more than 4 h were needed to reach equilibrium. On the other hand, the binding of sample CEA to tracer antibody reached equilibrium within 1 h when they were incubated alone, and the subsequent binding of the complexes so formed to the im10
1o0
%
"~ 80 123 z om 60
8
2~
0
4,,
0 2 i i i 15 45 135 CEA CONCENTRATION, /~g/I Fig. 2. Typical standard curve, and precision profile. The precision was calculated from the differences between duplicates of 830 samples during six consecutive runs, and is expressed as the coefficient of variation (CV). Assay sensitivity (Zero + 2 SD) was better than 0.1 # g / I in all runs.
176 munobeads also approached equilibrium in 1 h. Consequently, by first incubating sample and tracer antibody for 1 h, then adding immunobeads and continuing incubation for another hour, the necessary incubation time was more than halved• Evaluation of assay performance The performance of the assay was characterized in several ways. The precision and sensitivity proved to be excellent, as demonstrated by the data in Fig. 2. The overall coefficient of variation (CV) for control samples at 5 and 50 /~g/l for every 23rd patient sample in routine assays over long periods of time was 5% at both levels. The reference limit was determined by analyzing sera from 305 blood donors and patients with non-carcinomatous diseases, showing a median CEA value of 0.9 p,g/l. 5 # g / l defined the 98.4th percentile and was conservatively chosen as the practical reference limit. Combined dilution and recovery experiments were performed with sera from ten colorectal cancer patients having moderately elevated CEA values (20-71 # g / l ) . Aliquots were diluted with equal volumes of sera having a low CEA content or with assay buffer. Mean recovery was 100% (range 96-105%) in serum, 104% (range 98-109%) in assay buffer. Assay performance at extremely elevated CEA values was studied by analyzing samples with purified CEA added up to 50,000 # g / l . As expected a ' h o o k effect' occurred, so that the 50,000 ~g/1 sample was measured as 15/xg/l. The contribution of cross-reactions with N C A in the assay was evaluated by adding purified N C A (a kind gift from Dr. P. Burtin, Villejuif,
TABLE |I1 CROSS-REACTIONS WITH PURIFIED NCA a NCA added
CEA measured (/~g/I) by
(/~g/l)
IRMA
Abbott b
Roche c
750 2250 6750
0.6 0.9 1.9
0.2 0.5 1.3
0.4 0.9 3.2
a A gift from Dr. P. Burtin. h Abbott CEA-RIA monoclonai. c CEA EIA test 'Roche'.
r-.98
o o
oO
o .41•
•" o
..'"Sm,~-I
..*"
o •
o,ik. *
o O . . ."
o..+.,,~00 ,, °+°' 40
1Q
c a laor~-./i 40 r-.M
'~,3o o
o
LJ •
.-"o
~"
o
.¢," o
•
10
c o , troop,~ / 0
r,-.g3 • ..'
~'20 10 n
..'" ..'o
.'"
# ~1~ ° °° , , 10
o
•
•
o
o.~'"
•
8
~3
40
c ~ Ag8o~. ~g/i
Fig. 3. Correlation of CEA values obtained by three different assays in sera from 35 patients treated for colorectai cancer. Top panel.. The IRMA vs. Abbott CEA-RIA monocional. Middle: The IRMA vs. CEA EIA test 'Roche'. Bottom: Abbott CEA-RIA monoclonal vs. CEA-EIA test" Roche'.
France) to the assay buffer. Similar experiments were performed also with the commercial assays CEA EIA test 'Roche' (F. Hoffmann-LaRoche, Basel, Switzerland) and Abbott C E A - R I A monoclonal (Abbott Laboratories, Chicago, IL), dissolving NCA in their respective sample diluents. The results in Table III show that all assays gave clinically insignificant cross-reactions. Finally, the correlation of the new I R M A with the two commercial assays was investigated by analyzing serum samples from 35 colorectai cancer patients having CEA values in the new assay ranging from 1.1 to 30 # g / l . The results showed satisfactory correlations (Fig. 3) but also demonstrated the analytical problems of CEA assays.
177 Discussion Our objective was to develop a rapid, precise and sensitive immunometric assay for CEA as a replacement of our former radioimmunoassay (B/Srmer, 1982). The IRMA described, requiring 2 h incubation and yielding highly satisfactory sensitivity and precision, successfully meets the required objectives. Current knowledge indicates that CEA has a well-defined protein backbone. The large antigenic heterogeneity observed is due to post-translational modifications, mainly extensive glycosylations, which obviously may also influence the presentation of protein epitopes. Such variations in epitope presentation may affect antibody affinity and consequently the results of immunoassays, since these assume that standard and sample epitopes bind with the same affinity. The emerging insight into the epitope structure of CEA may lead to a better standardization of C E A assays, and hopefully also to the identification of the epitopes that convey the most relevant clinical information. In solid-phase immunometric assays the dependency on epitope presentation can be expected to be most pronounced for the binding of the tracer antibody, which is used at a comparatively low concentration. In contrast, the large excess of solid-phase antibody usually employed renders this assay step less susceptible to such variations. Although our choice of antibodies for the IRMA was in effect determined by other factors, the antibody 12-140-1 used as radiolabelled tracer turned out to be the least susceptible to conformational changes following reduction and alkylation of CEA, as is demonstrated in Table II. Notably the antibody 12-140-10, used on the solid phase, seemed to be considerably less dependent on CEA conformation than did antibodies 12-140-5 and -7, even though these recognized epitopes in the same (Gold 4) group (Hammarstri3m et al., 1990). It is reasonable to assume that the use of antibodies with a low dependency on conformation, as tested in our experiments, will yield an assay that is comparatively resistant to variations in CEA presentation in samples from patients. Due to the presence of cross-reacting substances, particularly NCA, in several normal cells and tissues, immunohistochemical evaluation of
specificity is an important step in the characterization of anti-CEA antibodies. It should be noted that an antibody cross-reacting with NCA may have an affinity for NCA that is several orders of magnitude lower than its affinity for CEA (Wagener et al., 1983a). Our studies of antibody binding to radiolabelled and solid-phase NCA confirmed this to be true also for the non-specific antibodies 12-140-5, -7 and -10 (data not shown). This means that even an assay using cross-reacting antibodies will usually have a low sensitivity for NCA. However, the concentration of N C A in normal sera may be up to 500 #g/1 (Von Kleist et al., 1977), increasing up to several thousand # g / I in patients with chronic myeloid leukemia (Wahren et al., 1980). Consequently the CEA assays should be tested for cross-reaction at these high NCA levels. The results in Table III show cross-reactions judged to be insignificant both in the IRMA and in the two commercial assays, especially as a slight CEA contamination in the NCA preparation cannot be excluded. Some practical aspects of the new IRMA require comment. Due to the heterogeneity of CEA mentioned above, nonlinearity of the responses, obtained with dilution series of some samples from patients is a common experience with most or all CEA assays. We therefore consider the dilution and recovery results obtained with the new assays to be very satisfactory. The 'hook effect' seen at extremely elevated CEA levels can be completely avoided by running the assay with two separate incubation steps, firstly incubating standards and samples with immunobeads, next washing away unbound CEA, and then incubating with labelled antibody. This procedure requires an extra washing step. Since such extremely elevated CEA values never occur in patients who are not suspected of having advanced disease, we have chosen the more simple incubation procedure for the routine assay. The reference limit was conservatively set at 5 # g / l ; only 1.6% (5/305) of the reference population had higher serum CEA levels. However, with the excellent sensitivity and reproducibility of the assay, changes within the reference range may be highly significant during the follow-up of a cancer patient. The correlation between the results obtained
178
with the new I R M A and those obtained with the two commercial assays on serum samples from colorectal cancer patients is shown in Fig. 3. The results demonstrate a good correlation between our I R M A and the Abbott C E A - R I A monoclonal, whereas the correlation with the CEA-EIA ' Roche' and between the two commercial assays was poorer. Due to the wide scatter of the comparatively few results no regression analysis was performed. The lines drawn have a slope = unity, and the distribution of results around these indicates that our assay and the Abbott kit were similarly calibrated, while the Roche assay yielded slightly lower CEA values. With the current knowledge of the CEA epitope structure it is reasonable to believe that the scatter of CEA values found with the various assays is due to differences in the epitope specificity of the antibodies used. The results in Fig. 3 suggest that the Abbott assay employs antibodies with a specificity similar to ours, while the antibodies used in the Roche assay appear to be different. Studies are in progress to classify the specificity of several commercial CEA assays within the ISOBM Workshop epitope groups, correlating the results obtained with the assays in various patient categories.
Acknowledgements We are grateful to Ms. Tone Varaas and Ms. Kari Thrane-Steen for skilled technical assistance, to Elisabeth Paus, Ph.D., for valuable discussions, and to Prof. Alexander Pihl for advice in the preparation of the manuscript. The work was supported by The Norwegian Cancer Society.
References Bt~rmer, O.P. (1982) A direct assay for carcinoembryonic antigen in serum and its diagnostic value in metastatic breast cancer. Clin. Biochem. 15, 128. Btirmer, O.P. (1989) Interference of complement with the binding of carcinoembryonic antigen to solid-phase monoclonal antibodies. J. Immunol. Methods 121, 85. Boscato, L.M. and Stuart, M.C. (1986) Incidence and specificity of interference in two-site immunoassays. Clin. Chem. 32, 1491.
Dasovir-Kne~.evir, M., IKrrmer, O.P., Holm, R., Hoie, J., Sobrinho-SimSes, M. and Nesland, J. (1990) Carcinoembryonic antigen in medullary thyroid carcinoma. An immunohistochemical study applying six novel monoclonal antibodies. Mod. Pathol., in press. Felder, R.A., MacMillan III, R.H. and Bruns, D.E. (1987) Two monoclonal-based assays for carcinoembryonic antigen compared. Clin. Chem. 33, 700. Fracker, P.J. and Speck, J.C. (1978) Protein and cell membrane iodination with the sparingly soluble chloramide 1.3.4.6-tetrachloro-3a,6a-diphenyl glycoluril. Biochem. Biophys. Res. Commun. 80, 849. Gold. P. and Freedman, S.O. (1965) Demonstration of tumorspecific antigens in human colonic carcinomata by immunological tolerance and absorption techniques. J. Exp. Med. 121,439. Hammarstr~Sm, S., Shively, J.. Paxton, R.J., Beatty, B.G., Larsson, A., Ghosh, R., B/3rmer, O., Buchegger, F., Mach, J.-P., Burtin, P., Fuks, A., Kalthoff, H., Von Kleist, S., Grunert. F., Matsuoka, Y., Kuroki, M., Wagener, C., Weber, T., Yachi, A. and Imai, K. (1989) Antigenic sites in carcinoembryonic antigen. Cancer Res. 49, 4852. Nap, M., Mollgard, K., Burtin, P. and Fleuren. G.J. (1988) Immunohistochemistry of carcino-embryonic antigen in the embryo, fetus and adult. Tumor Biol. 9. 145. Nustad, K., Danielsen, H., Reith, A., Funderud, S., Lea, T., Vartdal, F. and Ugelstad, J. (1988) Monodisperse polymer particles in immunoassays and cell separation. In: A. Rembaum and Z.A. T~Skrs (Eds.), Microspheres: Medical and Biological Applications. CRC Press, Boca Raton, FL, p. 53. Paus, E., BiSrmer, O. and Nustad, K. (1982) Radioiodination of proteins with the lodogen method. In: Radioimmunoassay and Related Procedures in Medicine 1982. International Atomic Energy Agency, Vienna, p. 161. Price, M.R. (1988) Epitopes of carcinoembryonic antigen (CEA) defined by monoclonal antibodies. Br. J. Cancer 57, 165. Thompson, J. and Zimmermann. W. (1988) The carcinoembryonic antigen gene family: Structure, expression and evolution. Tumor Biol. 9, 63. Von Kleist, S., Troupel, S., King, M. and Burtin, P. (1977) A clinical comparison between non-specific cross-reacting antigen and CEA in patients' sera. Br. J. Cancer 35, 875. Wagener, C., Clark, B.R., Rickard, K.J. and Shively, J.E. (1983a) Monoclonal antibodies for carcinoembryonic antigen and related antigens as a model system: Determination of affinities and specificities of monoclonal antibodies by using biotin-labeled antibodies and avidin as precipitating agent in a solution phase immunoassay. J. lmmunol. 130, 2302. Wagener, C., Yang, Y.H.J., Crawford, F.G. and Shively, J.E. (1983b) Monoclonal antibodies for carcinoembryonic antigen and related antigens as a model system: A systematic approach for the determination of epitope specificities of monoclonal antibodies. J. lmmunol. 130, 2308. Wahren, B., Gahrton, G. and Hammarstrbm, S. (1980) Nonspecific cross-reacting antigen in normal and leukemic myeloid cells and serum of leukemic patients. Cancer Res. 40, 2039.
171
Elsevier JIM 05473
Selection of monoclonal antibodies for use in an immunometric assay for carcinoembryonic antigen Ole P. Bt~rmer a n d Kjell N u s t a d Central Laboratory, The Norwegian Radium Hospital Oslo, Norway
(Received 13 April 1989, revised received29 June 1989, accepted 26 October 1989)
Six high-affinity mouse monoclonal antibodies against carcinoembryonic antigen produced in our laboratory were evaluated for use in a solid-phase immunoradiometric assay. The antibodies all recognized single peptide epitopes, one being highly conformation dependent. Cross-inhibition studies demonstrated that three antibodies bound to different epitopes, while the remaining three bound to the same or to very closely related epitodes. All antibodies strongly stained colorectal cancer tissue. The three antibodies binding to the same epitode also stained normal granulocytes, indicating a reactivity with the non-specific cross-reacting antigen. On the basis of the results of the characterization, two antibodies were selected for use in a sandwich immunometric assay. One of these was coupled to 2.8 ~ m magnetizable polymer particles, while the other one was radiolabelled. The assay required 2 h incubation to reach equilibrium, having a working range up to 135 # g / l and a sensitivity (Zero + 2 SD) of 0.1 ~ g / l . A comparison of the new assay with two commercial CEA kits that also use monoclonal antibodies was carried out on a small panel of serum samples from colorectal cancer patients and revealed satisfactory correlations but also discrepancies that must be attributed to differences in epitope specificity. Key words: Monoclonal antibody; Antigenic determinant; Carcinoembryonicantigen; Immunoassay
Introduction Carcinoembryonic antigen (CEA) in serum is an important parameter for the staging and follow-up of patients with some of the most common forms of cancer. First described by Gold and Freedman (1965), CEA was initially regarded as an oncofetal antigen specific for fetal gut and
Correspondence to: O.P. B6rmer, Central Laboratory, The Norwegian Radium Hospital, N-0310 Oslo, Norway. Abbreviations: BSA, bovine serum albumin; CEA, carcinoembryonic antigen; IRMA, immunoradiometric assay; ISOBM, International Society for Oncodevelopmental Biology and Medicine; NCA, non-specificcross-reacting antigen; PBS, phosphate-buffered saline.
carcinomas of the gastrointestinal tract. Later studies, however, have shown that CEA is also present in small amounts in normal adult tissues (Nap et al., 1988). CEA has recently been shown to be a member of the immunoglobulin gene superfamily, along with several related substances that may cross-react in immunological detection methods (for review, see Thompson and Zimmermann, 1988). Furthermore, CEA is heavily and heterogeneously glycosylated, implying that the molecule may show considerable variations between patients and even heterogeneity within a single patient. These factors, amplified by the use of different standards, explain most of the serious discrepancies between the various commercial assays which have complicated the routine use of CEA assays in cancer patients (Felder et al., 1987).
0022-1759/90/$03.50 (O 1990 Elsevier Science Publishers B.V. (Biomedical Division)
172 The use of monoclonal antibodies in the immunoassays does not in itself solve the problems, but may make it easier to understand the factors that determine the performance of an assay. Under the auspices of the International Society for O n c o d e v e i o p m e n t a l Biology and Medicine (ISOBM), an international workshop is studying the epitope specificity and relationship of a large number of monoclonai anti-CEA antibodies produced by the participants. This has led to increased knowledge about the antigenic structure of CEA (HammarstriSm et al., 1989). Here we report a rapid and sensitive immunometric CEA assay, using monoclonal antibodies that have been well characterized, in part through our participation in the ISOBM workshop.
Materials and methods
Preparation of CEA CEA was purified from liver metastases from colorectal carcinomas by perchloric acid extraction followed by ion exchange and gel filtration chromatography, as described (BiSrmer, 1982).
Production of monoclonal anti-CEA antibodies Mouse hybridomas producing anti-CEA antibodies were made as described (BtSrmer, 1989), using purified CEA as the antigen. The hybridomas were expanded as ascites tumors in mice, and antibodies were purified from the ascitic fluid by Protein A-Sepharose chromatography (Pharmacia, Uppsala, Sweden) at 4 ° C.
Characterization of antibody class and affinity The antibodies selected as promising after a preliminary estimation of affinity and specificity were denoted 12-140-1, -2, -4, -5, -7, and -10. Their immunoglobulin classes, and their affinities for radiolabelled CEA, were determined as described elsewhere (BiSrmer, 1989) and are summarized in Table I.
Radiolabelling of antibodies t2SI-labelling was performed by the Iodo-Gen method (cat. no. 28600, Pierce Chemical Co., Rockford, IL) (Fracker and Speck, 1978), as de-
TABLE I CHEMICAL CHARACTERIZATION OF ANTIBODIES AND EPITOPES Antibody 12-140
IgG Kd x 10H class mol/I
Percentloss a of antibody bindingafter treatment of CEA with Periodate Reduction+ alkylation
-1 -2
1 2b
-4
1
-5 -7 -10
2a 2b 1
3.1 20 30 3.4 6.0 18
48 16 40 10 17 14
12 23 14 67 89 34
a Mean of three experiments.
scribed by Paus et al. (1982). Equimolar amounts of protein and radioiodine were used.
Chemical modifications of CEA For the cross-inhibition and chemical modification experiments, CEA was adsorbed onto the wells of polyvinyl chloride 96-well microtiter plates (cat. no. 001-010-2801, Dynatech Laboratories, Alexandria, VA). The adsorption was performed overnight at room temperature with 1 /~g CEA preparation in 0.1 ml of a 0.2 m o l / I sodium carbonate buffer, pH 9.3, with 0.1% sodium azide. Periodate oxidation of adsorbed CEA, as well as reduction and alkylation of CEA followed by adsorption to microtiter plates, were performed as described by Price (1988). After adsorption and modification of the CEA preparations, the wells were saturated with 1% ( w / v ) bovine serum albumin (BSA) (cat. no. A4503, Sigma Chemical Co., St. Louis, MO) in 0.2 ml phosphate-buffered saline (PBS) (sodium-phosphate 0.01 m o l / l , NaCI 0.15 m o l / l , sodium-azide 0.05% (w/v), p H 7.5) for 2 h at room temperature.
Antibody binding to solid-phase native and modified CEA The effects of chemical modification of CEA on antibody binding were studied by incubating the wells, previously coated with the various CEA preparations, with 0.75 ~g monoclonal anti-CEA in 75 ~1 1% BSA-PBS for 1 h. After washing with PBS, the amount of monoclonal anti-CEA bound
173 TABLE I1 EPITOPE RELATIONSHIP AND IMMUNOHISTOCHEMICAL SPECIFICITY OF ANTI-CEA ANTIBODIES Antibody 12-140
Epitope group a
Percent inhibition b of the binding to CEA of the labelled antibodies
Histochemical staining c of Colon Normal tissues carcinomas Liver Lung
-1
-2
-4
-5
-7
-10
5 2d
99 0
0 96
0 0
0 0
0 0
0 0
++ ++
-
-4
1
88
+ +
-
-5
4 4 4
+++ + + 4+++
+ -
-1 -2
-7 -10
0 0 0
0
30 32 32
98
0 0 0
0
0
0
95 90 75
97 95 84
94 92 92
Granulocytes
+++ +++ ++
a 'Gold' epitope group (Hammarstr~m et al., 1989). b By an excess of the various unlabelled antibodies (see text). CScale_ to + + + . d This antibody was not included in the ISOBM Workshop, but classified later (see text).
was q u a n t i t a t e d b y i n c u b a t i n g the wells for 2 h with a p p r o x . 4 ng (50,000 c p m ) l:5I-labelled imm u n o p u r i f i e d sheep a n t i - m o u s e i m m u n o g l o b u l i n a n t i b o d i e s in 0.1 ml 1% BSA-PBS with 2% n o r m a l sheep serum. A f t e r washing with PBS, the b o u n d r a d i o a c t i v i t y in the s e p a r a t e d m i c r o t i t e r wells was c o u n t e d , a n d a n t i b o d y b i n d i n g was calculated as a p e r c e n t a g e o f the b i n d i n g to u n t r e a t e d C E A . T h e e p i t o p e relationships of the m o n o c l o n a l a n t i b o d i e s were m a p p e d by cross-inhibition studies, essentially as d e s c r i b e d by W a g e n e r et al. (1983b). Briefly, d u p l i c a t e wells with a d s o r b e d C E A were i n c u b a t e d for 2 h with approx. 2.5 ng (30,000 c p m ) labelled a n t i - C E A to be tested, in 50 #1 1% BSA-PBS, together with d o u b l i n g dilutions of various u n l a b e l l e d c o m p e t i n g antibodies, plus a ' b l a n k ' with no u n l a b e l l e d a n t i b o d y a d d e d . A f t e r w a s h i n g with PBS, the b o u n d r a d i o a c t i v i t y in the s e p a r a t e d wells was counted, a n d the results c a l c u l a t e d as the percent inhibition of the b i n d i n g of labelled a n t i b o d y . Except for a n t i b o d y 12-140-2, this e x p e r i m e n t was p e r f o r m e d with all a n t i b o d i e s received t h r o u g h the I S O B M w o r k s h o p as c o m p e titors, as r e p o r t e d b y Hammarstrt~m et ai. (1989). T h e results o b t a i n e d with o u r own a n t i b o d i e s at 2 /Lg c o m p e t i n g a n t i b o d y p e r well are shown in T a b l e II.
Immunohistochemical studies Previously we have e v a l u a t e d the a n t i b o d i e s for use in i m m u n o h i s t o c h e m i s t r y (Dasovi6-Kne~.evi6
et al., 1990). T h e results o b t a i n e d with p a r a f f i n sections of colon c a r c i n o m a s a n d several n o r m a l tissues are s u m m a r i z e d in T a b l e II.
Preparation of solid-phase antibodies for immunoassay use A n t i b o d i e s were c o v a l e n t l y c o u p l e d to tosyla t e d m a g n e t i z a b l e 2.8 ~ m m o n o d i s p e r s e p o l y m e r particles ( D y n a b e a d s M-280, D y n a l , Oslo, N o r way) ( N u s t a d et al., 1988), using 20 m g a n t i b o d y / g p o l y m e r particles. A f t e r a n t i b o d y coupling, the covalent b i n d i n g c a p a c i t y of the i m m u n o b e a d s was s a t u r a t e d with BSA.
The immunoradiometric assay (IRMA) T h e b u f f e r used c o n t a i n e d s o d i u m p h o s p h a t e 0.05 m o l / l , N a C ! 0.1 m o l / l , p H 6.5, with 1% ( w / v ) BSA, 0.1% ( w / v ) m e r t h i o l a t e , a n d 0.1% ( v / v ) T w e e n 20. N o r m a l a d u l t bovine, rat, a n d m o u s e sera were a d d e d to 5, 0.5 a n d 0.1% ( v / v ) , respectively. O u r purified C E A was used as a s s a y s t a n d a r d , in a m a t r i x o f h e a t - i n a c t i v a t e d b l o o d b a n k sera, selected for low C E A content. F o r the zero stand a r d , the s e r u m p o o l was a b s o r b e d with a n t i - C E A 12-140-5 c o u p l e d to c y a n o g e n b r o m i d e a c t i v a t e d S e p h a r o s e 4B. S t a n d a r d s were c h o s e n to cover the range 0 - 1 3 5 # g / l . D u p l i c a t e s o f 0.1 ml s t a n d a r d s , c o n t r o l s or p a t i e n t sera were i n c u b a t e d for 1 h with t25Ilabelled a n t i b o d y 12-140-1 ( a p p r o x . 150,000
174
c p m / 1 0 ng) in 0.1 ml assay buffer. Immunobeads (0.5 mg) with antibody 12-140-10 were added in 0.1 ml assay buffer, and the incubation was continued for 1 h with shaking. The tubes were then washed with 3 x 0.5 ml PBS on magnetic racks (Amerlex-M Separator, Amersham International, Amersham, U.K.), and counted. All radioactivity counting was performed with an LKB/Wallac 1277 Gammamaster (Wallac OY, Turku, Finland). In the immunometric assay the standard curve, the results for unknowns and controls, and the precision profile were calculated by the L K B / W a l l a c 1224 Riacaic program.
of 12-140-4, while the converse was not the case. This possibly reflects steric hindrance not involving the epitope itself. Also shown in Table II is the ISOBM Workshop epitope group classification ('Gold' groups (HammarstriSm et al., 1989)) of the antibodies. Antibody 12-140-2, not included in the Workshop, has been tested by us in cross inhibition experiments with the antibodies Mab 35 (Gold group 2) and CE27 (Gold group 3), kindly provided by Drs. J.P. Mach and F. Buchegger. These experiments demonstrated that 12-140-2 was inhibited up to 80% by Mab 35, while CE27 had no effect. Thus, 12-140-2 should probably be assigned to the Gold group 2.
Results
Characterization of antibodies and their epitopes The dissociation constants of the antibodies (Table I) suggest that even low concentrations of antibody should be able to bind CEA in the ~tg/l range, which is relevant for serum measurements. All antibodies retained more than 50% of their binding after periodate treatment of CEA (Table I) indicating that the epitopes are probably not carbohydrate structures (Price, 1988). After reduction and alkylation of CEA, the binding of antibodies 12-140-5 and -7 was greatly reduced, suggesting that the corresponding epitopes were strongly dependent on protein conformation (Price, 1988). The epitopes for the other antibodies were considerably less affected by this treatment. The relationships between the epitopes for the various antibodies were first evaluated in sandwich immunometric assays, which demonstrated that all antibodies bound to single epitopes on the CEA molecule since no binding of labelled antibody occurred when any one of them was used as both solid-phase and tracer antibody (data not shown). The results of the cross-inhibition experiments (Table II) revealed that antibodies 12-140-5 and -7 bound to the same epitope. Antibody 12-140-10 apparently bound to a closely related but not identical structure. The remaining three antibodies recognized separate epitopes, although some interference occurred between antibodies 12-140-4 and -1 : binding of 12-140-1 was inhibited by an excess
Immunohistochernical studies The results of the immunohistochemical studies, also shown in Table II, demonstrate that all antibodies bound strongly to colon carcinoma cells. Antibodies 12-140-5, -7 and -10 (Gold group 4) also bound to normal granulocytes, indicating that these antibodies recognized an epitope common to CEA and non-specific cross-reacting antigen, NCA. Antibody 12-140-5 also bound slightly to normal hepatocytes, possibly reflecting its higher affinity (Table I). Consequently this antibody was able to recognize smaller amounts of NCA. The remaining three antibodies, 12-140-1, -2, and -4, appeared to be CEA specific in these studies with paraffin sections. Selection of antibodies for the assay Complement factors in normal serum were found to inhibit severely the binding of CEA to solid-phase IgG2a and IgG2b antibodies (B~Srmer, 1989). This complement interference excluded antibodies 12-140-2, -5, and -7 from use on the solid phase. The routine radiolabelling of tracer antibodies made it necessary for these to have a high affinity, since only limited amounts could be used due to radiation safety considerations. In practice, only antibodies 12-140-1 and -5 yielded a fully satisfactory tracer antibody sensitivity. Of these, antibody 12-140-5 was non-specific, reacting also with NCA, as pointed out above. Because a rapid immunometric assay required the use of comparatively large amounts of solid-phase antibody, even a
175
'specific' solid-phase antibody might bind crossreacting substances with lower affinity. Using a non-specific tracer antibody, the overall assay specificity would then be lost. Consequently the CEA-specific antibody 12-140-1 was chosen for radiolabelling. Due to complement interference, only the two remaining IgGl antibodies could be considered for use on the solid phase. Of these, 12-140-4 exhibited some interference with the binding of labelled 12-140-1 to CEA (Table II). Consequently the other lgG1 antibody, 12-140-10, was chosen for coupling to the solid phase. Even if this antibody showed some cross reaction with NCA, sufficient assay specificity was secured by using comparatively low concentrations of the more specific tracer antibody 12-140-1. The immunoradiometric assay The optimal conditions for coupling antibody to the magnetizable polymer particles were determined in separate experiments (data not shown). The composition of the assay buffer was chosen to minimize non-specific binding of tracer antibody to the immunobeads. This was achieved at pH 6.7, adding BSA to 1% ( w / v ) and Tween 20 to 0.1% (v/v). In order to eliminate interference from heterophilic antibodies and other mouse immunoglobulin-binding substances in the human sera, which would give falsely elevated CEA results, normal adult bovine, rat and mouse sera were added (Boscato and Stuart, 1986). Assay buffer with adult bovine serum added to 50% was used initially as the matrix for CEA standards. However, with this matrix the non-specific binding of ageing tracer antibody to the immunobeads increased slightly more in the standards than in the samples, giving a systematic assay drift of approx. 1 # g / I during the 6-week life of radiolabelled antibody. We therefore chose to use a pool of heat-inactivated blood blank sera, selected for low CEA content, as standard matrix. Attempts to remove all CEA from the serum pool with the aid of antibody 12-140-5 coupled to Sepharose were also unsuccessful, since a slight leakage of antibody into the serum caused the standard values to drift by 1-2 #g/1 during a few days in the refrigerator. We therefore used the absorbed serum pool only as a zero standard, and carefully
8 ~ / 0 - -
0
0
8
g Z,, ko
O
~ e / O
~2 o/ o" 01
60
120 180 MINUTESINCUBATION
240
Fig. 1. Binding kinetics of the assay components. The volumes and concentrations of the reagents were as in the immunometric assay (see text). The CEA sample concentration was 5 # g / I . After washing with 3×0.5 ml PBS on magnetic racks, the radioactivity bound to the immunobeads was counted, and calculated as the percentage of radiolabelled antibody added. ( o ) Binding of complexes of CEA and radiolabelled antibody 12-140-1, preformed by incubation overnight, to immunobeads with antibody 12-140-10; ( I ) binding to immunobeads of CEA and radiolabelled antibody, added simultaneously; (r,) binding of radiolabelled antibody to CEA, followed by a short (15 min) incubation with immunobeads.
calibrated the CEA content of the untreated serum pool before preparing the other standards. The experiments performed to determine the necessary incubation times in the new assay (Fig. 1) yielded results of practical importance. When all components (sample, tracer antibody, and immunobeads) were added simultaneously, more than 4 h were needed to reach equilibrium. On the other hand, the binding of sample CEA to tracer antibody reached equilibrium within 1 h when they were incubated alone, and the subsequent binding of the complexes so formed to the im10
1o0
%
"~ 80 123 z om 60
8
2~
0
4,,
0 2 i i i 15 45 135 CEA CONCENTRATION, /~g/I Fig. 2. Typical standard curve, and precision profile. The precision was calculated from the differences between duplicates of 830 samples during six consecutive runs, and is expressed as the coefficient of variation (CV). Assay sensitivity (Zero + 2 SD) was better than 0.1 # g / I in all runs.
176 munobeads also approached equilibrium in 1 h. Consequently, by first incubating sample and tracer antibody for 1 h, then adding immunobeads and continuing incubation for another hour, the necessary incubation time was more than halved• Evaluation of assay performance The performance of the assay was characterized in several ways. The precision and sensitivity proved to be excellent, as demonstrated by the data in Fig. 2. The overall coefficient of variation (CV) for control samples at 5 and 50 /~g/l for every 23rd patient sample in routine assays over long periods of time was 5% at both levels. The reference limit was determined by analyzing sera from 305 blood donors and patients with non-carcinomatous diseases, showing a median CEA value of 0.9 p,g/l. 5 # g / l defined the 98.4th percentile and was conservatively chosen as the practical reference limit. Combined dilution and recovery experiments were performed with sera from ten colorectal cancer patients having moderately elevated CEA values (20-71 # g / l ) . Aliquots were diluted with equal volumes of sera having a low CEA content or with assay buffer. Mean recovery was 100% (range 96-105%) in serum, 104% (range 98-109%) in assay buffer. Assay performance at extremely elevated CEA values was studied by analyzing samples with purified CEA added up to 50,000 # g / l . As expected a ' h o o k effect' occurred, so that the 50,000 ~g/1 sample was measured as 15/xg/l. The contribution of cross-reactions with N C A in the assay was evaluated by adding purified N C A (a kind gift from Dr. P. Burtin, Villejuif,
TABLE |I1 CROSS-REACTIONS WITH PURIFIED NCA a NCA added
CEA measured (/~g/I) by
(/~g/l)
IRMA
Abbott b
Roche c
750 2250 6750
0.6 0.9 1.9
0.2 0.5 1.3
0.4 0.9 3.2
a A gift from Dr. P. Burtin. h Abbott CEA-RIA monoclonai. c CEA EIA test 'Roche'.
r-.98
o o
oO
o .41•
•" o
..'"Sm,~-I
..*"
o •
o,ik. *
o O . . ."
o..+.,,~00 ,, °+°' 40
1Q
c a laor~-./i 40 r-.M
'~,3o o
o
LJ •
.-"o
~"
o
.¢," o
•
10
c o , troop,~ / 0
r,-.g3 • ..'
~'20 10 n
..'" ..'o
.'"
# ~1~ ° °° , , 10
o
•
•
o
o.~'"
•
8
~3
40
c ~ Ag8o~. ~g/i
Fig. 3. Correlation of CEA values obtained by three different assays in sera from 35 patients treated for colorectai cancer. Top panel.. The IRMA vs. Abbott CEA-RIA monocional. Middle: The IRMA vs. CEA EIA test 'Roche'. Bottom: Abbott CEA-RIA monoclonal vs. CEA-EIA test" Roche'.
France) to the assay buffer. Similar experiments were performed also with the commercial assays CEA EIA test 'Roche' (F. Hoffmann-LaRoche, Basel, Switzerland) and Abbott C E A - R I A monoclonal (Abbott Laboratories, Chicago, IL), dissolving NCA in their respective sample diluents. The results in Table III show that all assays gave clinically insignificant cross-reactions. Finally, the correlation of the new I R M A with the two commercial assays was investigated by analyzing serum samples from 35 colorectai cancer patients having CEA values in the new assay ranging from 1.1 to 30 # g / l . The results showed satisfactory correlations (Fig. 3) but also demonstrated the analytical problems of CEA assays.
177 Discussion Our objective was to develop a rapid, precise and sensitive immunometric assay for CEA as a replacement of our former radioimmunoassay (B/Srmer, 1982). The IRMA described, requiring 2 h incubation and yielding highly satisfactory sensitivity and precision, successfully meets the required objectives. Current knowledge indicates that CEA has a well-defined protein backbone. The large antigenic heterogeneity observed is due to post-translational modifications, mainly extensive glycosylations, which obviously may also influence the presentation of protein epitopes. Such variations in epitope presentation may affect antibody affinity and consequently the results of immunoassays, since these assume that standard and sample epitopes bind with the same affinity. The emerging insight into the epitope structure of CEA may lead to a better standardization of C E A assays, and hopefully also to the identification of the epitopes that convey the most relevant clinical information. In solid-phase immunometric assays the dependency on epitope presentation can be expected to be most pronounced for the binding of the tracer antibody, which is used at a comparatively low concentration. In contrast, the large excess of solid-phase antibody usually employed renders this assay step less susceptible to such variations. Although our choice of antibodies for the IRMA was in effect determined by other factors, the antibody 12-140-1 used as radiolabelled tracer turned out to be the least susceptible to conformational changes following reduction and alkylation of CEA, as is demonstrated in Table II. Notably the antibody 12-140-10, used on the solid phase, seemed to be considerably less dependent on CEA conformation than did antibodies 12-140-5 and -7, even though these recognized epitopes in the same (Gold 4) group (Hammarstri3m et al., 1990). It is reasonable to assume that the use of antibodies with a low dependency on conformation, as tested in our experiments, will yield an assay that is comparatively resistant to variations in CEA presentation in samples from patients. Due to the presence of cross-reacting substances, particularly NCA, in several normal cells and tissues, immunohistochemical evaluation of
specificity is an important step in the characterization of anti-CEA antibodies. It should be noted that an antibody cross-reacting with NCA may have an affinity for NCA that is several orders of magnitude lower than its affinity for CEA (Wagener et al., 1983a). Our studies of antibody binding to radiolabelled and solid-phase NCA confirmed this to be true also for the non-specific antibodies 12-140-5, -7 and -10 (data not shown). This means that even an assay using cross-reacting antibodies will usually have a low sensitivity for NCA. However, the concentration of N C A in normal sera may be up to 500 #g/1 (Von Kleist et al., 1977), increasing up to several thousand # g / I in patients with chronic myeloid leukemia (Wahren et al., 1980). Consequently the CEA assays should be tested for cross-reaction at these high NCA levels. The results in Table III show cross-reactions judged to be insignificant both in the IRMA and in the two commercial assays, especially as a slight CEA contamination in the NCA preparation cannot be excluded. Some practical aspects of the new IRMA require comment. Due to the heterogeneity of CEA mentioned above, nonlinearity of the responses, obtained with dilution series of some samples from patients is a common experience with most or all CEA assays. We therefore consider the dilution and recovery results obtained with the new assays to be very satisfactory. The 'hook effect' seen at extremely elevated CEA levels can be completely avoided by running the assay with two separate incubation steps, firstly incubating standards and samples with immunobeads, next washing away unbound CEA, and then incubating with labelled antibody. This procedure requires an extra washing step. Since such extremely elevated CEA values never occur in patients who are not suspected of having advanced disease, we have chosen the more simple incubation procedure for the routine assay. The reference limit was conservatively set at 5 # g / l ; only 1.6% (5/305) of the reference population had higher serum CEA levels. However, with the excellent sensitivity and reproducibility of the assay, changes within the reference range may be highly significant during the follow-up of a cancer patient. The correlation between the results obtained
178
with the new I R M A and those obtained with the two commercial assays on serum samples from colorectal cancer patients is shown in Fig. 3. The results demonstrate a good correlation between our I R M A and the Abbott C E A - R I A monoclonal, whereas the correlation with the CEA-EIA ' Roche' and between the two commercial assays was poorer. Due to the wide scatter of the comparatively few results no regression analysis was performed. The lines drawn have a slope = unity, and the distribution of results around these indicates that our assay and the Abbott kit were similarly calibrated, while the Roche assay yielded slightly lower CEA values. With the current knowledge of the CEA epitope structure it is reasonable to believe that the scatter of CEA values found with the various assays is due to differences in the epitope specificity of the antibodies used. The results in Fig. 3 suggest that the Abbott assay employs antibodies with a specificity similar to ours, while the antibodies used in the Roche assay appear to be different. Studies are in progress to classify the specificity of several commercial CEA assays within the ISOBM Workshop epitope groups, correlating the results obtained with the assays in various patient categories.
Acknowledgements We are grateful to Ms. Tone Varaas and Ms. Kari Thrane-Steen for skilled technical assistance, to Elisabeth Paus, Ph.D., for valuable discussions, and to Prof. Alexander Pihl for advice in the preparation of the manuscript. The work was supported by The Norwegian Cancer Society.
References Bt~rmer, O.P. (1982) A direct assay for carcinoembryonic antigen in serum and its diagnostic value in metastatic breast cancer. Clin. Biochem. 15, 128. Btirmer, O.P. (1989) Interference of complement with the binding of carcinoembryonic antigen to solid-phase monoclonal antibodies. J. Immunol. Methods 121, 85. Boscato, L.M. and Stuart, M.C. (1986) Incidence and specificity of interference in two-site immunoassays. Clin. Chem. 32, 1491.
Dasovir-Kne~.evir, M., IKrrmer, O.P., Holm, R., Hoie, J., Sobrinho-SimSes, M. and Nesland, J. (1990) Carcinoembryonic antigen in medullary thyroid carcinoma. An immunohistochemical study applying six novel monoclonal antibodies. Mod. Pathol., in press. Felder, R.A., MacMillan III, R.H. and Bruns, D.E. (1987) Two monoclonal-based assays for carcinoembryonic antigen compared. Clin. Chem. 33, 700. Fracker, P.J. and Speck, J.C. (1978) Protein and cell membrane iodination with the sparingly soluble chloramide 1.3.4.6-tetrachloro-3a,6a-diphenyl glycoluril. Biochem. Biophys. Res. Commun. 80, 849. Gold. P. and Freedman, S.O. (1965) Demonstration of tumorspecific antigens in human colonic carcinomata by immunological tolerance and absorption techniques. J. Exp. Med. 121,439. Hammarstr~Sm, S., Shively, J.. Paxton, R.J., Beatty, B.G., Larsson, A., Ghosh, R., B/3rmer, O., Buchegger, F., Mach, J.-P., Burtin, P., Fuks, A., Kalthoff, H., Von Kleist, S., Grunert. F., Matsuoka, Y., Kuroki, M., Wagener, C., Weber, T., Yachi, A. and Imai, K. (1989) Antigenic sites in carcinoembryonic antigen. Cancer Res. 49, 4852. Nap, M., Mollgard, K., Burtin, P. and Fleuren. G.J. (1988) Immunohistochemistry of carcino-embryonic antigen in the embryo, fetus and adult. Tumor Biol. 9. 145. Nustad, K., Danielsen, H., Reith, A., Funderud, S., Lea, T., Vartdal, F. and Ugelstad, J. (1988) Monodisperse polymer particles in immunoassays and cell separation. In: A. Rembaum and Z.A. T~Skrs (Eds.), Microspheres: Medical and Biological Applications. CRC Press, Boca Raton, FL, p. 53. Paus, E., BiSrmer, O. and Nustad, K. (1982) Radioiodination of proteins with the lodogen method. In: Radioimmunoassay and Related Procedures in Medicine 1982. International Atomic Energy Agency, Vienna, p. 161. Price, M.R. (1988) Epitopes of carcinoembryonic antigen (CEA) defined by monoclonal antibodies. Br. J. Cancer 57, 165. Thompson, J. and Zimmermann. W. (1988) The carcinoembryonic antigen gene family: Structure, expression and evolution. Tumor Biol. 9, 63. Von Kleist, S., Troupel, S., King, M. and Burtin, P. (1977) A clinical comparison between non-specific cross-reacting antigen and CEA in patients' sera. Br. J. Cancer 35, 875. Wagener, C., Clark, B.R., Rickard, K.J. and Shively, J.E. (1983a) Monoclonal antibodies for carcinoembryonic antigen and related antigens as a model system: Determination of affinities and specificities of monoclonal antibodies by using biotin-labeled antibodies and avidin as precipitating agent in a solution phase immunoassay. J. lmmunol. 130, 2302. Wagener, C., Yang, Y.H.J., Crawford, F.G. and Shively, J.E. (1983b) Monoclonal antibodies for carcinoembryonic antigen and related antigens as a model system: A systematic approach for the determination of epitope specificities of monoclonal antibodies. J. lmmunol. 130, 2308. Wahren, B., Gahrton, G. and Hammarstrbm, S. (1980) Nonspecific cross-reacting antigen in normal and leukemic myeloid cells and serum of leukemic patients. Cancer Res. 40, 2039.
