Carcinoembryonic Antigen: Characterization and Clinical Applications
Carcinoembryonic Antigen: Characterization and Clinical Applications
20
Published by Munksgaard, Copenhagen, Denmark
No part may be reproduced by any process without written permission from the author(s)
Carcinoembryonic Antigen:
Characterization
and Clinical Applications
WILLIAM D . TERRY, PIERRE A. HENKART, JOHN E . COLIGAN
& CHARLES W . TODD
I. INTRODUCTION
Embryonic and fetal tissues appear to contain substances that are absent, or
present in very much diminished concentration, in the mature organism. Some
of these substances can be recognized by immunochemical tests, and the anti-
gens detected are referred to as fetal or embryonic antigens. It is presumed
that at least some of these substances represent 'signals' or 'signal receptors'
that have been hypothesized to be required for differentiation and organo-
genesis of the developing embryo.
It is of great biological interest that malignant tumors in adult animals
frequently contain high concentrations of molecules usually found only in
fetal tissues. This was clearly demonstrated by Abelev et al. (1963), who dis-
covered that mouse liver tumors contained a high concentration of alpha feto-
protein, a substance normally found only in fetal tissue. Gold & Freedman
(1965 a, b) similarly found that tumors of the human gastrointestinal tract
contained substances absent from normal adult gastrointestinal tissues, but
present in the fetal gastrointestinal tract. Immunochemical tests showed that
this material, which was called carcinoembryonic antigen (CEA), could be
detected in the serum of patients with gastrointestinal malignancy, but not in
the serum of normal individuals or of patients with other cancers (Thomson et
al. 1969). The possibility that CEA assays might provide a diagnostic test for
From the Immunology Branch, National Cancer Institute, Bethesda, Maryland 20014,
and the Department of Immunology, City of Hope National Medical Center, Duarte,
California 91010.
This work was supported in part by grant # 12631 and contract # NCI-G-72-3877 from the
National Cancer Institute.
CARCINOEMBRYONIC ANTIGEN 101
cancer provided impetus for extensive clinical trials, and also for investigations
of the immunochemical, physical and chemical characterization of CEA.
This review attempts to summarize recent publications concerning the
structure of CEA, presents some previously unpublished structural studies,
and briefly summarizes work on the clinical applications of CEA assays.
7.7 Nomenclature
Nomenclature in this field is awkward for historical reasons. Gold & Freedman
(1965 a, b) reported that antisera prepared against human colonic cancer
detected common antigens in extracts of fetal gastrointestinal tissue, but not in
extracts of normal adult colon or other normal adult tissues. These experi-
ments were performed by immunizing rabbits with saline extracts of a primary
colon tumor and then absorbing the antiserum with a similar extract of the
adjacent normal colon tissue from the same donor. The absorbed antiserum
precipitated the immunizing extract, as well as extracts from fetal gastrointes-
tinal tissue, and the reactive materials were therefore designated carcino-
embryonic antigens. Subsequent work has led to partial characterization of a
glycoprotein that carries these antigens, but unfortunately, the carrier has
also been called carcinoembryonic antigen (CEA). Thus, the term CEA has
been used interchangeably to refer to a poorly defined group of antigenic
determinants, and to the glycoprotein (or glycoproteins) upon which these
determinants are detected. While it might be preferable to refer to the carriers
as carcinoembryonic glycoproteins (CEGP) and the antigenic determinants
as CEA, there are sufficient gaps in our information about the carriers of
CEA that a revision of the nomenclature is probably unwise at this time. We
will therefore adhere to present conventions and refer to both the antigenic
determinants and the glycoproteins as CEA.
tissues probably reflects the minute quantities of CEA present in these tissues.
CEA has been referred to as a membrane protein, and it probably should
not be considered as such. It certainly is not an integral membrane protein (Sin-
ger 1971), in that it can be removed from cell membranes by incubation with
saline. CEA can indeed be detected on cell surfaces (Gold et al. 1968, von
Kleist & Burtin 1969, Gold et al. 1970, Denk et al. 1972), but appears to be
loosely adherent and is propably present at the outer surface of the cell mem-
brane following secretion.
2.2 Purification
Most attempts at purification have started with the initial steps used by Gold
& Freedman (1965 a) who found that CEA was soluble upon treatment of a
tissue homogenate with perchloric acid. This step effects a large purification
since perchloric acid denatures and precipitates almost all proteins other than
those, like CEA, with a high sugar content. The possible physical or chemical
alteration of native CEA by this harsh treatment must be borne in mind, how-
ever.
Subsequent steps in CEA purification by all laboratories invariably include
gel filtration on Sephadex G-200 and/or agarose. The procedure we have used
to isolate CEA from hepatic metastases consists of homogenization and per-
chloric acid extraction followed by two gel filtration columns; the first one
with Sepharose 4B, the second on Sephadex G-200 (Coligan et al. 1972).
Zbne electrophoresis before or after gel filtration is used by some groups
(Krupey et al. 1972, Turner et al. 1972) but not by others (Coligan et al. 1972,
Turberville et al. 1973). Its utility may depend on the resolution in the gel fil-
tration systems previously used in the preparation. Electrofocusing of gel-
filtration purified CEA was reported to result in little or no increase in specific
activity (Coligan et al. 1973). On the other hand. Turner et al. (1972) were
able to eliminate some impurities by electrofocusing their CEA preparation
from primary colon tumors.
An independent approach to CEA isolation not involving perchloric acid
has been taken by Rosai et al. (1972) using primary tumors. They first pre-
pared a crude membrane fraction from the colon tumors, solubilized CEA
with lithium diiodosalicylate, and removed the proteins by a phenol extrac-
tion. The resulting crude glycoprotein was further subjected to chromato-
graphy and electrofocusing to yield purified CEA. The authors claimed a
CARCINOEMBRYONIC ANTIGEN 105
higher recovery of CEA than was obtained from primary tumors by Krupey
et al. (1968) using a traditional perchloric acid extraction, but it should be
noted that CEA concentration varies markedly in different tumors.
2.3 Homogeneity
Glycoproteins are unlikely to be as homogeneous as proteins because the en-
zymatic synthesis of the sugar moieties is not as faithful as the m-RNA
directed protein synthesis, and well studied purified glycoproteins show heter-
ogeneity of their carbohydrate groups (Montgomery 1972). Thus, sensitive
criteria for purity may demonstrate heterogeneity in a glycoprotein prepara-
tion that contains a single protein backbone with a distribution of different
sugar sidechains.
In the case of CEA, analytical ultracentrifugation has shown a single sym-
metrical peak for preparations derived from various sources (Krupey et al.
1968, Coligan et al. 1972). Other criteria based on electric charge have been
less clear cut. Polyacrylamide gel electrophoresis of CEA under various con-
ditions shows a single broad band when stained for sugar or protein, with no
minor components (Rosai et al. 1972, Terry et al. 1972, Turberville et al. 1973,
Banjo et al. 1974). CEA activity shows the same broad peak with a similar Rf
(Coligan et al. 1972). As discussed below, electrofocusing experiments show
heterogeneity with preparations from liver metastases, while in two cases,
preparations from primary tumors are reported to give a single peak (Rosai
et al. 1972, Turner et al. 1972).
In spite of the heterogeneity demonstrated by electrofocusing, Terry et al.
(1973) report a single ami no-terminal sequence in their CEA preparations.
The finding of an unambiguous sequence is good evidence for the presence
of only a single protein with an unblocked amino terminus, but does not rule
out other proteins with blocked amino termini, or non-protein impurities.
Turner et al. (1972) used centrifugation in a CsCl gradient as a step in puri-
fying their radioiodinated CEA. This technique separates molecules on the
basis of their buoyant density, which in this case is probably largely based on
the ratio of carbohydrate to protein. They found two radiolabeled impurities in
their preparations, a finding which suggests that this technique should be
applied to other CEA preparations as a homogeneity criterion.
3. PHYSICAL PROPERTIES
3.1 Size
No rigorous molecular weight determinations such as by sedimentation equi-
librium have been carried out with CEA. As a rough approximation a value
106 TERRY, HENKART, COLIGAN & TODD
of 200,000 to 300,000 is suggested by the fact that CEA from various sources
normally elutes between IgM and IgG on a Sephadex G-200 column (Coligan
et al. 1972, Puztaszeri & Mach 1973). The sedimentation coefficient is in
reasonable agreement with a value in this range, but without other physical
data not yet available a molecular weight cannot be calculated. This range of
normally elutes between IgM and IgG on a Sephadex G-200 column (Coligan
molecular weight is also compatible with the yield of amino terminal amino acid
recovered during protein sequence studies (Terry et al. 1973).
An unusually large molecular species of CEA has been reported by Coligan
et al. (1972). This large CEA was found along with normal size CEA and was
not convertible to the latter by treatment with dithiothreitol.
It is possible that CEA as it is normally isolated may consist of several
glycoprotein chains held together by disulfide or noncovalent bonds. CEA,
reduced and alkylated with 10 mM dithiothreitol and 20 mM ^*C-iodoacetamide
in 9 M urea, has the same Rf in SDS acrylamide electrophoresis as untreated
CEA (M. Egan, personal communication). In spite of the anomolous behavior
of glycoproteins on SDS gels, the fact that the untreated and the reduced and
alkylated CEA have the same Rf suggests that CEA is composed of only one
polypeptide chain. Amino terminal sequence determinations (Terry et al. 1972)
show only one sequence, and as discussed above, the yield is compatible with
one protein chain per molecule. SDS gel electrophoresis cannot be used to
estimate the molecular weight of the protein chain using standard methods
because of CEA's high carbohydrate content (Weber et al. 1972, Segrest et al.
1971).
3.3 Shape
Physical studies capable of giving significant information regarding the shape
of CEA molecules have not yet been performed. The lack of a hypersharp
CARCINOEMBRYONIC ANTIGEN 107
3.4 Charge
CEA has been reported to migrate in immunoelectrophoresis as a /^-globulin
(Krupey et al. 1967), although different preparations appear to show differ-
ences in migration (Pusztaszeri & Mach 1973).
Isoelectric focusing, a technique of high resolution which 'focuses' mole-
cules at their isoelectric point in pH gradient, has been applied to CEA by
several laboratories with different results. Rosai et al. (1972) and Turner et al.
(1972) working with extracts of primary colon tumor (prepared by different
methods) report CEA activity in one peak showing little heterogeneity with an
isoelectric point of 4.8 and 4.7, respectively. However, Rule & Goleski-Reilly
(1973 a, b) show multiple CEA activity peaks on their saline extracts of pri-
mary colon carcinomas when electrofused in urea. CEA purified from hepatic
metastases of colon and rectum tumors was found by Coligan et al. (1973)
to show extensive heterogeneity of activity as measured by radioimmunoassay,
with the major peaks at the acidic end near pH 3. Isoelectric heterogeneity
of metastatic CEA has been confirmed by Turberville et al. (1973), who found
the major activity peak at pH 4.4, and by Banjo et al. (1974 b) who found two
major peaks at pH 3.0 and 3.75. Coligan et al. (1973) have shown that the
charge heterogeneity is not entirely due to sialic acid, since after its removal
with neuraminidase some heterogeneity remained. Rule & Goleski-Reilly
(1973 b) recently have shown that perchloric acid treatment of a saline tumor
extract containing CEA causes a marked shift in the complex isoelectric
focusing pattern they obtain. It seems at present difficult to reconcile the
varying results of isoelectric focusing of CEA, but at least three variables may
be important; 1) tissue source of the antigen, 2) prior treatment with perchloric
acid, and 3) individual tumor variation (e. g. Figure 2 in Coligan et al. 1973).
4. PROTHN CHEMISTRY
4.1 Chemical composition
Chemical analyses of purified CEA from hepatic metastases of bowel tumor
indicate that it is a glycoprotein containing approximately 50 per cent sugar
(Tables II and III). Elemental analyses in agreement with this have been
published by Coligan et al. (1972). It is interesting to note that the combined
carbohydrate plus protein yields do not exceed 90 per cent in most of these
108 TERRY, HENKART, COLIGAN & TODD
TABLE I
Amino Acid Composition of CEA Preparations
Moles Amino Acid/10^ g Protein
Amino Acid 1 2 3 4
preparations and in one case the yield does not exceed 50 per cent. Banjo et al.
(1974 b) have reported that CEA does not contain fatty acids. It is unclear
why the combined weight of the analyzed components does not add up to 100
per cent of the dry weight.
TABLE II
Carbohydrate composition of various CEA preparations
a b c d e f g
Fucose 37 15 57 54 49 20 55
Mannose 38 49 20 45 38 33 50
Galactose 51 62 72 71 107 26 71
N-Acetyiglucosamine 48 97 118 127 132 62 96
Sialic Acid 34 24 7.5 12 59 7 8.5
N-Acetylgal actosamine NS* NS NS 11 9.8 NS 11
Carbohydrate per cent 42 47 57 58 77 27 52
Protein per cent 40 25 20 34 16 21 38
seems to be variable, the amino acid levels expressed on a dry weight basis
can vary from preparation to preparation even though the basic protein com-
position remains constant. We therefore have expressed our amino acid com-
positions as moles of amino acid per 10^ g of protein, eliminating the dry
weight measurement as a factor in the value. These data, shown in Table I,
differ somewhat from the original compositions reported (Krupey et al. 1968),
but are similar to more recent reports from the same group (Table I,
column 3).
Table I shows amino acid compositions of 4 different CEA preparations
from three laboratories. All show generally similar features - high levels of
carboxyl and hydroxyl amino acids, rather low levels of basic amino acids,
and very low levels of sulfur-containing amino acids. Some glycoproteins with
a high carbohydrate content, e. g., blood group substances (Watkins 1972),
show an amino acid content which is quite atypical of proteins in general, but
such is not the case with CEA.
no TERRY, HENKART, COUGAN & TODD
TABLE n i
Carbohydrate composition of individual CEA preparation^
Micromoles/lOO mg CEA
Fucose'' 63 42 46 57 63 57 58
Mannose'' 57 53 61 44 50 44 39
Galactose'' 56 67 81 63 82 70 75
N-Acetyl-
glucosamine'' 77 98 113 102 98 87 99
Sialic Acid^ ND« 6.6 8.8 0.6 0.2 14 21
N-acetyl-
galactosamine*^ 8.3 7.5 7.4 22 8.9 10 13
% carbohydrate 45 49 57 53 53 50 56
% protein 46 45 38 37 35 35 30
Total 91 94 95 90 88 85 86
a Unpublished data, Coligan, Egan & Todd. All CEA preparations were isolated from
individual tumors as described by Coligan et ai. (1972). All the preparations except
Mi and SiBl were isolated from liver metastases. Mi was a primary colon and SiBl
was a primary liver tumor.
I" Neutral sugars were determined by gas chromatography according to the method of
Clamp etal. (1972).
"^ Determined on the Beckman 121 amino acid analyzer using the hydrolysis method of
Liu & Chang (1971).
•J Determined by the thiobarbituric acid assay (Warren 1959) and/or by gas chromato-
graphy (Clamp et al. 1972).
«i Not determined.
5 10
LYS-LEU-THR-ILE^GLU-SER-THR-PRO-PHE-ASN
15 20
VAl^ALA-GLU-GLY-LYS-GLU-VAL-LEU-LEU-LEU
24
VAU-HIS-ASN-LEU
Figure 1. N-terminal amino acid sequence of CEA. TTie sequence was determined using
an automatic protein sequencer (Beckman 890). From Terry et al. (1972).
5. CARBOHYDRATE CHEMISTRY
that it might account for some or all of the antigenic sites detected by anti-
bodies to CEA.
5.7 Composition
As can be seen from Tables II and III, the carbohydrate compositions reported
for CEA isolated from different tumors vary widely. Significant differences
are seen even for individual preparations isolated within the same laboratory
(Krupey et al. 1968, Mach & Pusztaszeri 1972, Banjo et al. 1972, 1974 a, b,
Egan et al. 1974 b). Heterogeneity of the carbohydrate portion of CEA
(discussed in Sections 2.1 and 5.2) may account for some of these differences.
Table III gives a sampling of the values obtained from analyses of many
CEA preparations. Values for all the sugars present in CEA can be obtained
TABLE IV
Carbohydrate composition of individual CEA" preparations
experimental data we have obtained (Coligan et al. 1974, Coligan et al. 1973,
Egan et al. 1974 a) indicate that fucose and sialic acid are expected to be termi-
nal sugars. Due to the heterogeneity commonly experienced in mammalian
glycoproteins (Spiro 1970), these sugars would be expected to vary in different
CEA preparations.
N-acetylgalactosamine has been detected in relatively low amounts in CEA
preparations (Mach & Pusztaszeri 1972, Burkhard et al. 1973, Coligan et al.
1973, Banjo et al. 1974 b, Egan et al. 1974 b. Tables III and IV) and in some
cases is not detectable at all (Krupey et al. 1968, Banjo et al. 1972, 1974 a, b,
Burkhard et al. 1973). The significance of this will be discussed in more detail
later.
Some investigators (Krupey et al. 1967, Haverback & Dyce, 1974) have
noted the presence of glucose in their CHA preparations. In one instance
(Krupey et al. 1967), this was partially attributed to contamination by Sepha-
dex gel. We have observed the presence of trace amounts of glucose in several
of our samples. Whether this glucose is actually present in CEA or represen-
tative of some contamination or impurity is not known. Sephadex and cellulose
dialysis tubing represent likely sources of contamination. In some instances
we have observed that glucose is present in the perchloric acid extract prior
to column chromatography. It is therefore possible that trace amounts of glu-
cose-containing molecules are sometimes isolated along with CEA.
TABLE V
Carbohydrate Composition of
Isoelectric Focusing Pools
Micro moles/100 ng
Initial
Fucose 41 52 47 41 40 35
Mannose 48 36 49 50 54 57
Galactose 50 78 56 47 46 29
N-Acetyl-
glucosamine 87 96 83 79 87 84
Sialic Acid 7.5 14 92 7.2 5.5 4.1
N-Acetyl-
galactosamine 3.9 11 5.9 4.4 2.5 3.9
ph Range 1.5 -«—• 3.3 *—> 3.8 -.—.- 4.1 -—• 4.4 *—* 5.2
Carbohydrate per cent<: 43 52 45 41 42 38
Protein per cent 54 44 40 43 41 43
Glucose per cent by wt. 0.86 11.8 12.4 11.0 10.5 17.4
Specific activity^, a 1.23 0.69 1.16 1.54 1.39 1.45
(Coligan et al. 1973, Banjo et al. 1974 b). An interesting observation is that
of the neutral sugars, fucose and galactose decrease and mannose increases
with increasing electronegativity. Of the amino sugars, glucosamine remains
relatively constant in all the pools, whereas the majority of the galactosamine
is present in the most electronegative pools. Comparison of the precolumn
sample with the resultant pools indicates that the electrofocused samples
were contaminated with glucose. This can probably be attributed to the sucrose
used to form the electrofocusing column density gradiant. The tenacity of this
binding is remarkable considering that these pools were dialyzed against 3M
urea for 4 days, 2M urea plus 2M NaCI for 2 days, 2M NaQ for 3 days and
deionized water for 1 week.
Despite the chemical differences that exist among the column pools, the
antigenic determinants recognized by goat anti-CEA appear to be similar in all,
as evidenced by the lines of identity obtained in Ouchterlony analysis (Coligan
et al. 1973). The data in Table V show that pool 1 has a specific activity
(radioimmunoassay value divided by dry weight) approximately 50 per cent
less than that of the other pools. This low specific activity correlates with the
high quantity of galactosamine in the most acidic pool. Since previous studies
(Coligan et al. 1973) have shown galactosamine not to be necessary for CEA
activity, and since other authors have reported the absence of galactosamine
in CEA (Krupey et al. 1968, Banjo et al. 1972, Burkhard et al. 1973), this
suggests that the acidic fractions possess impurities containing galactosamine.
Another indication of the contribution of carbohydrate to the heterogeneity
of CEA comes from studies of the interaction of purified CEA with various
lectins (Terry et al. 1974). CEA is precipitated by Conconavalin A, wheat
germ agglutinin, and ricin which specifically bind mannose, N-acetylglucos-
amine and galactose, respectively, as well as by leukoagglutinating phyto-
hemagglutinin, which binds a more complex determinant. The double diffu-
sion precipitin patterns show that the reaction between CEA and anti-CHA
spurs over the reactions between CEA and each of the lectins, which in turn
spur over one another in a manner which varies with the CEA preparation.
Thus, each of these lectins appears to bind a different subpopulation of CEA
molecules, reflecting the heterogeneity of carbohydrate on these molecules.
TABLE VI
Fragments from Nagase Hydrolysis of CEA *
A B C D E
50% Assay % Starting
Inhibition Ratio equivalents Material
Weight activity B/2.7 of CEA 100 D/305
(mg) (ng) (mg)
Nagase — Nondialyzable
>• C E A 34 3.7-ia* 11.9-103 2.86-10-3 0.0009
Dialyzabie
G-25
.. GP-2 6.6 197 63.1 0.105 0.034
-^ GP-3 24.2 225 72.1 0.336 0.11
-^ GP-4 32.7 175 56.1 0.583 0.191
-^ GP-5 5.1 143 45.8 0.11 0.036
^ GP-6 2.0 AA.A 14.2 0.I4I 0.046
-* GP-7 3.5 105.0 33.7 0.104 0.034
GP-1 75 31.5 10.1 7.43 2.4
Cellulose
powder—
*- OSF" 20
^ GPl-
DW" 40 15.75 5.05 7.92 2.56
AG50W-X4
-^ GPl-A 14 55.0 17.6 0.79 0.26
-^ GPl-B 5.5 71.4 22.9 0.24 0.08
-^ GPl-C 1.0 37.5 12.0 0.08 0.03
-^ G P I - D 1.2 24.4 7.8 0.15 0.05
-^ G P l - E 8.0 35.6 11.4 0.70 0.23
-* G P l - F 1.4 37.5 12.0 0.12 0.04
TAliLE VII
Chemical composition of Be-CEA and blood group substances
et al. (1974) suggests that the CEA and blood group determinants are both
on the same molecule.
Blood group A substance isolated from hog gastric mucin and from human
ovarian cyst fluid has been found to be a relatively inefficient inhibitor when
compared to CEA in the radioimmunoassay described by Egan et al. (1974 b).
Amounts 460,000 and 1,300 times greater than CEA on a weight basis, re-
spectively, for hog gastric mucosa A-substance and human A-substance, were
required to produce equivalent inhibition in the radioimmunoassay (Egan,
personal communication). As can be seen in Table VII, the carbohydrate com-
positions of blood group A substances and CEA are quite different. More-
over, the distribution of the amino acids in the protein has also been found
to be remarkably different in CEA compared to those of the blood group A
substances (Table I and Watkins 1972). Thus, CEA and blood group sub-
stances appear to differ in the protein to which the carbohydrate moiety is
attached, the way it is attached (Watkins 1972), and the monosaccharide con-
tent of the material that is attached. While blood group substances may be
capable of crossreacting in the CEA assay at very low efficiency, clearly they
are not CEA, nor are they closely related chemically.
Holbum et al. (1974) have recently demonstrated that a number of CEA
preparations are reactive with anti-blood group sera (either anti-A, anti-B,
anti-Lewis b or anti-Lewis a). The blood group reactivity of the CEA appears
to correlate with the genetic background of the donor, i. e., CEA from a type
CARCINOEMBRYONIC ANTIGEN 121
B donor reacts specifically with anti-B sera. Quantitative studies indicate that
the blood group determinants and the CEA determinants are present on the
same molecule, but that the CEA immunodominant antigenic determinants
are distinct from the major blood group antigens. The presence of blood group
specificities can be attributed to the adventitious addition of the characteristic
groupings on the CEA molecule, but other explanations have been proposed
(Holbum etal. 1974).
Considering the distinct chemical differences between CEA and blood group
substances, the relative inefficiency of blood group substances in inhibiting
CEA in the radioimmune assay, and the correlation between CEA blood
group activity and the blood group of the donor, it may be conceivable that
the blood group substances are extracted and purified along with CEA as
noncovalently attached molecules. The stickiness of CEA has been previously
mentioned in relation to the relatively irreversible binding of sucrose to CEA
during isoelectric focusing. It has also been noted that ^^^I tagged bovine
serum albumin in the presence of a relatively large excess of CEA assumes
the isoelectric focusing characteristics of CEA (Coligan & Todd, unpublished
observation). Hughes (1973) has also reported the existence of noncovalently
attached substances being isolated along with CEA. Thus, the possibility
must be considered that blood group substances may be extracted inadvertently
with CEA.
6. CLINICAL IMPLICATIONS
Ever since the demonstration that the serum of some patients with malignancy
contained elevated levels of CEA (Thomson et al. 1969), attempts have been
made to investigate the usefulness of assays for CEA in the detection and diag-
nosis of malignant disease.
6.1 Assays
A number of different assays are presently being used to measure the concen-
tration of CEA in plasma or serum, or to monitor the purification of CEA
(reviewed in Kupchik et al. 1974). Most are radioimmunoassays utilizing the
general principle of inhibition of binding (Egan et al. 1974 b). Antiserum
containing antibodies to CEA is incubated with patient's serum and a known
amount of radiolabelled purified CEA. Any unlabelled CEA in the patient's
senim competes with the labelled CEA for binding to the antibodies. Antibody
and bound CEA are precipitated either chemically (Thomson et al. 1969,
Hansen et al. 1971) or with an anti-immunoglobulin (Egan et al. 1972). The
amount of label left in the supernatant is related to the concentration of CEA
122 TERRY, HENKART, COLIGAN & TODD
TABLE VIII
Occurrence of Elevated CEA Concentration in Various Clinical States
A - Healthy volunteers 10
Non-smokers 3
Smokers^ ^^
Former smokers ^
B - Non-malignant Disease 34
Pulmonary Emphysema 57
Alcoholic Cirrhosis 70
Ulcerative Colitis 30
Gastric Ulcer 45
Breast Disease 15
C - Endotermal Carcinoma (Colerectal,
Pulmonary, Gastric, Pancreatic) 73
D - Non-entodermal Carcinoma
Breast, Head and Neck, Other) 59
E - Non-Carcinoma
(Leukemias, Lymphomas, Sarcomas) 40
F - Suspected Malignancy, Not confirmed H
1 -Compiled from CEA-Roche: A Clinical Monograph (1974). These data are all based
on the Z-gel assay, with the upper limit of normal taken as 2.5 ng/ml.
2 " Elevations of CEA concentrations in smokers have not been carefully studied with
assays other than the Z-gel assay.
present in the serum. Assays either use serum or plasma, and for some assays,
it is necessary that the test serum be extracted first with perchloric acid (in-
direct assays) rather than utilizing intact serum or plasma (direct assays).
Careful comparative studies between different assays and the reagents used
are needed to determine whether any have special advantages. Published com-
parative studies indicate 70-80 per cent agreement (Kupchik et al. 1973,
Laurence et al. 1972).
Most of the clinical studies of CEA have been concerned with correlations
between serum CEA concentrations and the clinical status of the patient.
CEA is detected in the serum or plasma of most normal people at a relatively
low concentration. Quantitative studies in normal subjects and patients with
various diseases are summarized in Table VIII. About 10 per cent of normal
people have 'elevated' CEA levels (Table VIII, A), although this percentage
shifts depending on the value chosen as the upper limit of normal. CEA is
elevated in carcinoma of the gastrointestinal tract (Table VIII, C), but is also
elevated in other entodermal (Table VIII, C) and non-entodermal (Table VIII,
D and E) malignancies. In general, relatively large amounts of tumor are
required before CEA values are elevated, and in different series, 50 to 80
per cent of patients with established bowel cancer limited to the bowel wall
had normal CEA values (Dhar et al. 1972, Lo Gerfo et al. 1972, Laurence
et al. 1972, Mach et al. 1974). Similarly, 83 per cent of patients with primary
breast cancer without metastases had normal levels of CEA (Steward et al.
1974). Moreover, 17 per cent of patients suspected of malignancy, but where
no lesion could be detected, had CEA concentrations above normal (Table
vni, F).
These data are important in considering the potential use of CEA assays
mentioned above. It is generally agreed that CEA assays are impractical as a
general population screen. The major reason for this is that about 10 per cent
of the population has elevated CEA serum concentrations. It is not definitely
proven, but it seems reasonable to assume that these are truly false positive
tests, i. e., these individuals do not have cancer or any of the benign conditions
associated with elevated CEA. This 10 per cent of the population would have
to have a total 'cancer' examination to rule out the presence of malignancy
since elevated serum concentrations of CEA are seen with cancers of many
different organs (Lo Gerfo et al. 1971, Moore et al. 1971, Laurence et al.
1972). The resulting load on diagnostic medical facilities would be unaccept-
able. This conclusion may be modified if smoking can be consistently shown
to cause elevated levels, since the per cent of false positive tests is apparently
lower in non-smokers (Table VIII, A). Even if the per cent of false positive
tests can be decreased, the usefulness of this assay as a screen for early cancer
is limited by the fact that 50 to 80 per cent of individuals with small gastro-
intestinal cancers or non-metastatic breast cancers have normal CEA tests;
thus two-thirds to three quarters of those individuals who already have estab-
lished but small cancers would be missed by this assay (Dhar et al. 1972,
Laurence et al. 1972, Lo Gerfo et al. 1972, Steward et al. 1974).
There is no statistically valid proof at the present time that the CEA test
is useful as a diagnostic adjunct when cancer is suspected; that is, no reported
studies have demonstrated that the accuracy of the differential diagnosis of
124 TERRY, HENKART, COLIGAN & TODD
CONCLUDING COMMENTS
There are, however, other considerations, and a full exploration of the value
of CEA and related antigens in the detection, diagnosis, and management of
human cancer is of the first importance. At the present time, there appears
to be a relatively small role for CEA in detection and diagnosis. This pessi-
mistic opinion may be altered by future developments, including the develop-
ment of more sensitive assays. For example, it is possible that serial assays on
the same normal individual will be of value in detection or diagnosis, parti-
cularly if the assays can be made 100 times more sensitive. A CEA concen-
tration that rises steadily from 0.01 ngm/ml to 1.0 ngm/ml may be very im-
portant, while the change from 'not detectable' to 1.0 ngm/ml does not help
very much. The observation of a rise over time may distinguish true positive
from false positive results, but to be of real value in detection or diagnosis of
small cancers, it will have to be coupled with the use of more sensitive assays.
More information concerning the chemistry and nature of the antigenic site
of CEA should make it possible to design such assays.
The application of CEA assays to the management of patients with malig-
nancy is a more promising area for the immediate future. Appropriately de-
signed prospective trials will be required to determine a) whether changes in
CEA levels immediately after therapy distinguish reliably between 'cure' and
residual disease, and if so, b) whether additional therapy based on CEA results
improves survival or prolongs remission. Similar prospective trials will be
required to determine a) whether recurrence of an elevation in CEA level after
it had fallen to normal following therapy reliably indicates recurrent disease
and b) whether initiation of therapy based on CEA results improves survival
or prolongs remission. Studies of this type can be carried out now and within
the next year answers to these questions should be available.
REFERENCES
Abelev, G. I., Perova, S. D., Khramkova, N. I., Postnikova, Z. A. & Irlin, I. S. (1963)
Production of embryonal alpha-globulin by transplantable mouse hepatomas. Trans-
plantation 1, 174.
Abeyounis, C. J. & Milgrom, F. (1972) Studies on carcinoembryonic antigen. Int. Arch.
Allergy 43, 30.
Banjo, C , Gold, P., Freedman, S. O. & Krupey, J. (19723 Immunologically active hetero-
saccharides of carcinoembryonic antigen of human digestive system. Nature New
Biol. (Lond.) 238, 183.
Banjo, C , Gold, P., Freedman, S. O. & Krupey, J. (1973) Structure of the immuno-
dominant tumor grouping of carcinoembryonic antigen (CEA) of the human diges-
tive system (Abst.). Fed. Proc. 32, 4444.
Banjo, C , Gold, P., Gehrke. C. W., Freedman, S. O. & Krupey, J. (1974 a) Preparation
and isolation of immunologically active glycopeptides from carcinoembryonic anti-
gen (CEA). Int. J. Cancer. In press.
126 TERRY, HENKART, COLIGAN & TODD
Sorokin, J. J., Sugarbaker, P. H., Zamcheck, N., Pisick, M., Kupchik, H. Z. & Moore,
F. D. (1974) Serial CEA assays: use in detection of recurrence following resection
of colon cancer. 3AMA. In press.
Spiro, R. G. (1970) Glycoproteins. Ann. Rev. Biochem. 39, 599.
Steward, A. M,, Nixon, D., Zamcheck, N. & Aisenberg, A. (1974) Carcinoembryonic
antigen in breast cancer patients: serum levels and disease progress. Cancer. In press.
Terry, W. D., Henkart, P. A., Coligan, J. R & Todd, C. (1972) Structural studies of the
major glycoprotein in preparations with carcinoembryonic antigen activity. /. exp.
Med. 136, 200.
Terry, W. D., Henkart, P. A., Das, S., Coligan, J. & Todd, C. W. (1974) Characterization
of human carcinoembryonic antigen. In: Proc. VII Mites lnternat. Symp. In press.
Thomson, D. M. P., Krupey, J., Freedman, S. O. & Gold, P. (1966) The radioimmuno-
assay of circulating carcinoembryonic antigen of the human digestive system. Proc.
nat. Acad. Sci. (Wash.) 64, 161.
Turberville, C , Darcy, D. A., Laurence, D. J. R., Johns, E. W. & Neville, A. M. (1973)
Studies on carcinoembryonic antigen (CEA) and a related glycoprotein, CCEA-2.
Preparation and chemical characterization. Immunochemistry 10, 841.
Turner, M. D., Olivares, T. A., Harwell, L. & Kleinman, M. S. (1972) Further purifica-
tion of perchlorate-soluble antigens from human colonic carcinomata. 3. Immunol.
108, 1328.
von Kleist, S. & Burtin, P. (1969) Localization cellulaire d' un antigene embryonnaire
de tumeurs coligues humaines. Int. 3. Cancer 4, 874.
von Kleist, S., Chavanel, G. & Burtin, P. (1972) Identification of an antigen from normal
human tissue that crossreacts with the carcinoembryonic antigen. Proc. nat. Acad.
Set. (Wash.) 69, 2492.
Warren, L. (1959) The thiobarbituric acid assay of sialic acids. /. biol. Chem. 234, 1971.
Watkins, W. M. (1972) Blood-group specific substances. In: Glycoproteins, ed. Gott-
schalk, A., p. 830, Hsevier, New York.
Weber, K., Pringle, J. R. & Osborn, M. (1972) Measurement of molecular weights by
electrophoresis on SDS-acrylamide gel. Meth. Enzymol. XXVI, 3.
Zamcheck, N., Moore, T. L., Dhar, P. & Kupchik, H. (1972) Immunologic diagnosis
and prognosis of human digestive-tract cancer: carcinoembryonic antigens. New
Engl. J. Med. 286, 83.