Journal of Immunological Me#,ods, 150 (I 992) 133-143
133
© 1992 ElsevierScience Publishers B.V. All rights reserved 01)22-1759/92/$05.00
J1M06336
Tumour markers in oncology: past, present and future Henri Magdel6nat Laboratoire de Radiopathologie, Institut Curie, Paris, France
(Accepted 12 February 1992)
Historically, tumour markers are substances selectively released by tumour cells in the blood stream so that they can be detected in the serum or other body fluids for clinical monitoring of various malignancies. The term is now extended to cell or tissue characteristics, such as cytogenetic markers, oncogenes or abnormally expressed proteins with various biological functions (enzymes, receptors, etc.), which help characterize the tumour type and constitute specific biological targets for new drugs. Serum markers can be classified in three main categories: (i) oncofetal antigens which are normally expressed only during fetal life, (ii) carbohydrate antigens as part of glycolipids or glycolipoproteins (mucins) and (iii) miscellaneous, including tissue specific enzymes, cytoskeletal proteins, etc. Most carbohydrate antigens which are useful in serum diagnosis of human cancers show turnout type-specific aberra.,t glycosylation, with O instead of N linkage to protein moieties. Despite a huge progression in analytical performances with immunoassays, and because of still insufficient biological sensitivity and specificity, serum assays of tumour markers are not effective in mass screening, their usefulness being restricted to the follow-up of diagnosed patients. Selective expression of specific antigens allows in vivo detection of occult turnout masses (immunoscintigraphy) and immunotargetting of drugs as therapy. Cell and tissue markers comprise markers of transformation (molecular genetics, oncogenes and anti-oncogenes, etc.), differentiation (hormone receptors, etc.), proliferation (growth factors, Ki67, etc.) and metastatic potential (proteases, etc.). Although masking the functional aspects of many of these markers, as for hormone or growth factor receptors, immunoanalytical methods, specially immunohistochemistry, are rapidly expanding in clinical oncology as tools for diagnosis, prognosis and, promisingly, treatment adjustment. Key words: Tumor marker; Tumor antigen; Hormone receptor; lmmunoanalysis;Oncology
Introduction
The first cancer marker was identified over 140 years ago, in 1846, when H. Bence-Jones described a heavy precipitate upon acidification of the urine of a patient diagnosed as having
Correspondence to: H. Magdel6nat, Laboratoire de Radiopathologie et Physiopathologie, Institut Curie, 26 rue d'Ulm, 75231 Paris Cede× 5, France.
'mollities osseum' or softening of the bones. It took more than 100 years before the 'Bence Jones protein' was identified as the monoclonal light chain of immunoglobulin, and an elevated concentration was found in the plasma of patients with multiple myeloma. The second era of tumour markers, from 1928 to 1963, included the discovery and application of hormones (human chorionic gonadotrophin (Zondek, 1930); adrenocorticotropic hormone (Cushing, 1932)), enzymes and isoenzymes (Markert, 1959) to cancer diagno-
134 sis. However, the era of turnout markers as more generally applicable tools began with the discovery of a-fetoprotein (AFP) by Abelev in 1963 and carcinoembryonic antigen (CEA) by Gold and Freeman in 1965. Gold and Freedman immunized rabbits with extracts of human colonic cancer tissue, absorbed the resulting antisera with normal human colon and found that some of the antisera reacted with the tumour extracts, but not extracts of normal tissue. The antigen so detected was named CEA since it was also found in embryonic tissue. This approach raised a great hope of finding many other 'specific' tumour antigens and is still used to generate new specific antibodies. In the past decade, glycolipids and high molecular weight glycoproteins, termed mucins, were found the most clinically useful, tumour-associated antigens. The present era is therefore characterized by intense efforts at molecular characterization of these antigens. Progress in detailed biochemical analysis of antigenic epitopes, in molecular biology, in gene mapping and in transcriptional and post-transcriptional regulation mechanisms will certainly help developing more specific probes for clinic and research. The expansion of tumour markers in clinical practice is largely due to (i) the development of sensitive radioimmunoassays by Berson and Yallow in 1968, (ii) the discovery of monoclonal antibodies by KShler and Milstein in 1975, which yield reproducible sources of antibody and provide more specific reagents. At the clinical level, the present era is characterized by the screening of new monoclonal antibodies and a shift from radioactive to non-radioactive detection systems, from manual to automated analysis, in brief, by important technological and commercial developments. From this very brief summary, a definition of a tumour marker seems obvious: a tumour marker refers to any substance that can be detected in plasma (or body fluids) and related to the presence or progress of malignancy. There are strong tenants of this unique definition but, in a clinically oriented opinion, the term applies more generally to any biological characteristics of a malignant tissue which can be assayed by biochemical or histochemical methods and provides information relevant to diagnosis, prognosis or treatment monitoring in oncology. This leads to
the distinction between serum (or plasma) markers, as already introduced, and tissue (or cellular) markers, the development and application of which constitute the new era of tumour markers. In this sense, the Philadelphia chromosome (Nowell and Hungerford, 1960) may constitute the first specific cellular marker of a clonal malignancy, i.e., chronic myeloid leukaemia. More recently, steroid hormone receptors have proved clinically useful as prognostic factors and predicto,:s of response to hormonal therapy in breast cancer (McGuire et al., 1986), whereas growth factors and oncogens constitute tissue markers of great potential interest (Borg et al., 1990; Watson et al., 1991).
Serologic tumour markers An ideal seric tumour marker would be a substance not normally present in blood, whose detection in serum reflects the presence of a minimal tumour burden (sensitivity) and whose concentration accurately reflects tumour progression or regression (low inertia). In the ideal situation, each tumour type would express a different and specific antigen (specificity). Consequently, the presence of a marker in blood is dependent on its expression and release by the tumour mass. Monoclonal antibodies have been used extensively to unravel antigen specifically expressed by tumour cells. The conclusion of these studies is that such 'neo-antigens' do not exist. Thus, tumour-assoeiated antigens represent either a quantitative overexpression of antigens normally expressed by tissues of the same embryonic origin (enzymes, peptide hormones, etc.), or the re-expression of cellular antigens selectively expressed during embryogenesis (oncofetal antigens), or, more specifically, post-transcriptional variants, for instance by abnormal glycosylation, of cell membrane components (tumour-associated glycolipids and mucins). It may, however, be foreseen that mutations or translocations in the corresponding gene could result in cancer-specific protein alteration, as already exemplified by modified p21 protein (p21 raS) or mutated p53 anti-oncogene product (Table I).
135 TABLE I CLASSIFICATION O F T U M O U R
MARKERS
Oncofetal antigens Tumour-associated antigens (carbohydrate antigens)
AFP, CEA (I) Carbohydratedeterminant: Ca 19.9.CA50, CA72.4, GD2 etc, (2) Epithelial mucins(PEM): CAI5.3, MCA, CA549, CTA etc. (3) Glycoproteins:CA125, SCC. TPA, PSA etc.
Hormonal peptides
fl-HCG, tlLP, ACTH, TCal, GH, PTH etc.
Neuromediators
Catecholamine, melatonine, serotonine etc.,
Proteins Enzymes
,82-microglobulin,ferritine thyroglobulin, PAP, PAcP, NSE, TK
Amines
Polyamines(spermidine, spermine)
Oncogene products
p21, c-myc, neu etc.
Oncofetal antigens Oncofetal antigens can be defined as antigens that are expressed during intra-uterine life, diminish strongly or disappear after birth and reappear in situations of repair or neoplastic growth in the organs where they appeared (in the fetus) during gestation. This category comprises a fetoprotein (AFP), strictly fitting the definition, and carcino-embryonic antigen (CEA) with main features of oncofetai antigens. AFP, a glycoprotein purified by Nishi in 1970 is expressed in the vitellus membrane and in the fetal liver and digestive tract. Glycosylation is variable during the fetal life and in tumours, allowing the characterization of specific subfractions (lshiguro et al., 1985). In the adult, it is essentially associated with malignant hepatoma, the immunoassay being positive in over 70% of hepatoma patients. Elevated serum levels are also found in ovarian and testicular choriocarcinomas (but not in seminomas), rarely in cancers of the digestive tract or lungs, frequently but at moderate level in chronic hepatitis or in cirrhosis. CEA or carcinoembryonic antigen represents a large family of related cell membrane glycoproteins, with extracellular domain structures similar to immunoglobulins, and thus belongs to the immunoglobulin gene superfamily (Thompson and Zimmermann, 1988). So far, at least 14 genes and 36 different glycoproteins have been identified in the C E A family. Based on sequence similarities,
the CEA gene family can be divided into two main subgroups. One of these contains the genes encoding CEA and cross-reacting antigens (NCA) and the biliary glycoprotein. The other contains the gene encoding the pregnancy-specific/31-glycoproteins (PSG), a group of glycoproteins found in placenta and tumours of trophoblastic origin. All the genes from both CEA and PSG subgroups are clustered on the long arm of chromosome 19 and apparently closely linked (Zimmermann and Thompson, 1990). It is now well established that the CEA gene family belongs to the immunoglobulin supergene family (Oikawa et al., 1987; Paxton et al., 1987). Where known, members of the lg family have receptor or cell adhesion functions, e.g., the neuronal cell adhesion molecule (N-CAM) to which CEA has been compared (Benchimol et al., 1989).
Turnout-associated mucins Alterered glycosylation appears to be a constant phenomenon associated with oneogenic transformation in experimental systems as well as in essentially all types of naturally occurring human cancers. Most of the biochemical or, more recently, immunological methods used to identify tumour-assoeiated antigens have resulted in the isolation of glycolipids or glycoproteins (mucins) with altered glycosylation patterns. Turnout antigens defined by carbohydrate determinants are related to blood group alloantigens or to differentiation carbohydrate antigens. Although these
changes are all quantitative rather than qualitative, some antibodies can recognize tumour-distinctive structural organizations of carbohydrate moieties. Other antigens refer to the protein core rather than to the sugar moiety of tumour-associated highly glycosylated, high molecular weight glycoproteins or mucins. Little is known about the detailed structure of the core protein of human mucins but the recent cloning and sequencing of the highly polymorphic gene of one such mucin differentially expressed by the lactating mammary gland and by many carcinomas has defined a new class of tumour-associated antigens, the polymorphic epithelial mucins or PEM (Gendler et al., 1991).
Carbohydrate or blood group-related tumour antigens Many of the more clinically useful tumour markers identified over the last decade are mucins. Historically, the term mucin was used to designate biological polyanions which are stained by various dyes, including Alcian blue. More recent studies have characterized mucin tumour antigens as high molecular weight glycoproteins, from 200 kDa to greater than 1000 kDa. Mucin antigens contain from 25 to 80% carbohydrate and their density is usually greater than 1.40 g/i. Cancer carbohydrates Jelong to four major structural classes: ganglio, globo, lacto type 1 and lacto type 2 (Hakomori, 1986; Kannagi et al., 1988). Galactose and N-acetylglucosamine are the predominant sugar residues, whereas mannose is usually in low level or absent. An O-glycosidic linkage between the carbohydrate and either serine or threonine of the protein core of mucins is the most common linkage of cancer mucin antigens, in contrast with mucins of non-malignant cells or with the 'classic' tumour markers - CEA, AFP or PAP (prostatic alkaline phosphatase) which are N-linked carbohydrates. Many differentiation antigens present on cell surfaces being glycolipids and O-linked glycoproteins with qualitative and quantitative alterations during various phases of differentiation or oncogenesis (Feizi, 1981; Hakomori, 1985), the importance of O-linked carbohydrate epitopes appears less surprising. Changes in carbohydrate expression corre-
spond to changes in the enzyme activities of specific glycosyltransferases responsible for their synthesis or glycosidases responsible for their degradation. For example, the decreased expression of blood group antigens (ABH) in some tumours is due to incomplete synthesis (Schoentag et al., 1987) or masking by sialic acid (Chu et al., 1989). Alterations of carbohydrate structure can result in the generation of tumour markers without enhanced production ef the peptide core. Monoclonai antibodies that can detect subtle carbohydrate epitope alterations are thus particularly useful to discriminate malignant from benign pathological process (Itai et al., 1991) and even to bring organ specificity in malignancy. CA19.9, one of the first identified tumour mucin antigens, is the 2,3-sialylated form of the Lewis a (Lea) structure (Tempero et al., 1987), present on glycolipids of certain tumours (Magnani et al., 1982). This epitope is a normal component of certain tissues of some individual and is expressed strongly on glycolipids and glycoproteins from adenocarcinomas of the colon and pancreas but not from normal colon. The immunogenic determinant being essentially carbohydrate in nature, neuraminidase treatment of CA19.9 results in complete loss of antigenic activity. As a tumour marker, the serum level of CA19.9 is elevated in patients with pancreatic and gastrointestinal (Herlyn et al., 1982) cancers with limited sensitivity ( ~ 30%) in patients with early stage tumours. Lewis negative individuals are genetically incapable of expressing CA19.9 (Bara et al., 1986). Interestingly, the 2,6-sialylated Le a antigen, which has close structural relationship to the CA19.9 antigen, is abundantly present in non-malignant pancreas tissues and is significantly decreased in pancreatic cancer tissues, the 2,3/2,6 antigen ratio allowing a better discrimination between malignant and benign disease (Itai et a1.,1988, 1991). CASO is related to CA19.9 as the afucosyl form of sialylated Lewis antigen (Mansson et al., 1985). It is similar to CA19.9 in monitoring advanced gastrointestinal and pancreatic cancers but with broader specificity (Harmenberg et al., 1988). CAS0 does not require the fucose residue in the Lewis epitope, whereas CA19.9 does. Consequently, CAS0, like Span-l, another CA19.9-re-
137 lated monoclonal antibody, binds to tissues from Lewis negative individuals (Chung et al., 1987). CA72.4 or TAG72 is a high molecular weight ( > 10 ~ Da) mucin like molecule (Johnson et al., 1986) recognized by the monoclonal antibody B72.3, generated by immunization with membrane enriched fractions of human breast cancer metastases. The suggested antigenic determinant is the disaccharide NeuAc(2,6) GalNAc (Gold et al., 1988). It may be useful in diagnosis and monitoring some human epithelial cancers, for instance gastric cancer. Breast cancer mucin (BCM) is a recent tumour-associated mucin-like antigen defined by M85, a mouse monoclonal IgM antibody, reacting with paragloboside carbohydrate structures with specificity for Galfll,4GIcNac (lacto type 2 chain) sequence (Manderino, 1990). The antigen has been termed 'cryptic' since M85 reactivity to serum or ascites is greatly enhanced by removal of sialic acid by neuraminidase. The presence of cryptic cell surface antigens, some of them differentiation-associated, some tumour-associated, has been reported, notably by Springer 0984) and Gooi (1983). Polymorphic epithelial mucins (PEM) The polymorphism of epithelial mucins originally noted at the protein level has now been shown to be at the D N A level, and is due at least in part to the presence of differing numbers of tandem repeats (Gendler et al., 1988).The gene of PEM has been cloned (Swallow et al., 1987) and located on human chromosome 1 (Gendler et al., 1990). The utility of circulating mucins as breast cancer markers, initially proposed by Ceriani (1982), has been rapidly confirmed and expanded (Hilkens, 1984). Several antibodies have been raised against such mucins, from membrane fractions of either normal lactating (HMFG-1 or HMFG-2) or malignant mammary glands (MAM6 family, DF3, CTA). CA15.3, an important antigen in the monitoring of breast cancer, is a 400 kDa glycoprotein recognized by a pair of monoclonai antibodies, 115 D8 (Hilkens et al., 1984) against membranes of human milk fat globules (HMFG), and DF3 against a membrane enriched fraction of a metastatic breast cancer lesion (Kufe et al., 1984).
Epitopes of HMFG-derived antibodies are partially carbohydrate (lectin binding) and partially peptide (BurcheU et ai., 1987). MAb DF3 recognizes a high molecular weight glycoprotein containing sialyl oligosaccharides present on a peptide backbone. Further characterization of the carbohydrate components of the DF3 antigen has revealed that the major oligosaccharides are the p e a n u t agglutinin binding dissaccharides Gal/31,3GaINAc (lacto type l chain), also designated as the Thomsen-Friedenreich antigen and its mono- and di-sialylated derivatives (Hull et al., 1988). Although linked to breast malignancy, CA15.3 is not breast specific and can be found elevated in metastazing ovarian, colonic or bronchial carcinoma. Its sensitivity is low (2030%) in early breast cancer. However it is largely used in the follow-up of patients at risk of recurrence after treatment of primary breast cancer and in the monitoring of metastasized breast cancer (Pons Anicet et ai., 1987; Kallionemi et al. 1988). The follow-up and the monitoring of breast cancer are the major applications of turnout markers in clinical oncology, due to the high incidence of this disease and to the possibility to increase survival by early detection and treatment of recurrence. So, variants of CA15.3 and mucinlike antigens are available for routine clinical analysis from different suppliers: MCA, MSA, CAM26, CAM29, CA 549, BCM, etc. (Bieglmayer et al., 1991, Garcia et al., 1990). Unfortunately, the lack of sensitivity in early stages and a moderate specificity has limited the sensible clinical application of this markers to the monitoring of advanced stages of the disease. MCA or mucin-like cancer antigen is a 350 kDa glycoprotein (Caravatti, 1988). It is a membrane protein overexpressed in breast cancer cells. In serum MCA is measured using a sandwich immunoenzyme assay employing the same antibody as primary and secondary antibody. MCA is similar to CA15.3 in monitoring patients with breast cancer. Being elevated in non-malignant breast diseases its use for diagnosis is unlikely. CA125 is identified by a monoclonal antibody raised to an ovarian carcinoma cell line (Bast et al., 1981). The nature of the epitope is still controversial but appears to contain sialie acid (Hanisch et al., 1985). Its sensitivity is good for
the non-mucinous histological type of ovarian carc#1omas but it lacks specificity (breast, lung, benign and malignant effusions are often reactive). The half-time of its fall after initial trealment is prognostic and its elevation during follow-up of ovarian carcinomas may spare surgical 'second look'. Unfortunately, treatments of reccurrent ovary cancer are rather inefficient, which limits the therapeutic gain potentially attainable with a sensitive indicator of early receurrence. CA125 is one of the rare monoelonal antibodies used to detect occult metastasis by immunoscintigraphy in vivo.
Other antigens PSA or prostate-specific antigen, isolated by Nancy et al. in 1979, is a 34 kDa intracellular glycoprotein, exclusively secreted by the prostate gland with protease activity. PSA is elevated in patients with benign prostatic, hypertrophy and acute prostatitis, so that its usefulness for diagnosis is limited. It is very useful in monitoring the efficacy of surgery after removal of the tumours and the reccurence of prostatic cancer. Its sensitivity and specificity appear superior to that of PAP (prostate alcaline phosphatase) in this monitoring. TPA or tissue polypeplide antigen is a 47 kDa protein sharing homologies with epithelial cytokeratines 8, 18 and 19. It is actively secreted during mitosis of multiplicating cells in culture. It is therefore not cancer specific but rather related to cell turnover or proliferation. Although not useful in diagnosis, its good sensitivity makes it a useful test in patients with known malignancies, specially in the absence of more specific marker (for instance in bladder cancer). Also related to the cytoskeleton, Villin is a cytoskeleton protein of the enterocyte brush border, a differentiation marker of early embryogenesis with restricted tissue specificity in the human adult (enterocyte of small and large intestin, gall bladder and pancreas duct cells). It may be useful in diagnosis and follow-up of colorectal cancers (Dudouet et al., 1990). SCC or squamous cell carcinoma antigen, is a 48 kDa glycoprotein, sub-fraction of the complex TA-4 fraction, isolated in 1977 by Kato et al.
from an epidermoid carcinomz of the uterine cervix. It is a marker for epidermoid (or squamous) carcinomas, with a good specificity but low sensitivity (recently improved by the use of a monoclonal antibody instead of the initial polyelonal one). It can be useful in monitoring advanced epidermoid cancers of the uterine cervix, of the lung and of head and neck.
When are tumour markers clinically useful? Criteria of utility The important criteria for a marker are its sensitivity and specificity. A sensitive marker detects a high percentage of patients with disease, while a specific assay is only positive in cancer patients, not in those without cancer. Sensitivity is calculated as (true positive/(true positive + false negative)), while sensitivity is calculated as (true negative/(true negative+ false positive)). Predictive values are linked to sensitivity and specificity but also to the prevalence (P) of the disease in the population studied (Sulman et al., 1987) (Table II). This is important to consider when evaluating the worth of a tumour marker in different clinical situations, for instance in the context of mass screening (low prevalence) or in a medical oncoiogy department (high prevalence). The predictive value of a tumour marker test may therefore be augmented either by increasing the sensitivity and specificity of the test or by selecting the appropriate sub-group of population or patients to be tested. Tumour markers tests being quantitative, their sensitivity and specificity depends upon the cut-off value chosen to diserimiTABLE !1 SENSITIVITY,SPECIFICITY,PREDICTIVE VALUES Cancer Marker
Positive
TP
No Cancer FP
Negative
FN
TN
Sensitivity
Specificity
Positive predictive value Negative predictive value
139 nate between 'positive' and 'negative' tests, here again this cut-off may be chosen according to the clinical end point to reach, i.e., whether sensitivity or specificity is more important in the population tested. The use of ROC curves facilitates the choice of a cut-off value and allows an objective evaluation of the respective worths of different tests (Sulman et al., 1987; Bieglmayer et al., 1991).
Tumour markers and clinical phases of the disease The clinical history of cancer is schematized in Fig. 1. The information expected from serological tumour markers varies according to the stage of the disease. Screening before apparent disease. A tumour marker with almost absolute specificity (at least > 95%) and good sensitivity would be required for mass screening of an apparently healthly population, due to the low prevalence of cancer in the general population. The lack of high specificity and sensitivity for available markers in cancers where early detection would be beneficial (breast, colon) precludes such tests for mass screening at reasonable cost-efficiency ratios. A simple calculation shows that in ideal conditions (1% incidence of the disease, 99% sensitivity and specificity), the frequency of false positive is already 50%. The only successful tumour marker used in
screening is human chorionic gonadotrophin (HCG) in choriocarcinoma. The relatively high prevalence of tumour (5-10%) in women who have had a hydatiform mole, the excellent assay sensitivity and specificity, together with the high chemosensitivity of the tumour, has led to a significant reduction in mortality through screening of this population. Diagnosis. The assay of a tumour marker may be interesting at the time of diagnosis as a confirmation of origin or to assess the extension of the neoplasm although no biological marker has been yet integrated in the staging of the disease. Even at this symptomatic stage of disease, the sensitivity of current tumour markers is too low (20-30% for CA15.3 in primary breast cancer) for a serum assay to be really useful for diagnosis. It may however be interesting to consider an elevated pre-treatment value in order to subsequently monitor tile efficacy ef the initial treatment through., for instance, the half-life clearance of the marker. Specific antibodies may also be useful to confirm a diagnosis by immunocytochemical staining of tumour biopsies. Follow up. Serological tumour markers are useful in the follow-up of patients after initial therapy. A continuous elevation of the serum level is strongly indicative of reccurrence of the disease, with in many cases, a lead time of several
Primary treatment clinical = Idetection level ~11~1.. I'atlures
Tumor birden
/ Secon~"trea}ment
. ~ _~_ Tumor / ~ - .... -~. . . . . . . . . . . . . . . . ~- ..... -~ . . . . . . . . . / g r ~ . ~ . : i ....... ~ ""~""t ""'~'-" ~ General ...................... population ~init ~promotion Remission I second remission C!inical =Clinical n~ malignancy diagnosis recurrence High risk
.......... Screening
Therapeutic staoino efficiency Nb Meta + Therapeutic prognosis Residual ÷ efficiency which therapy? disease Detection of recurrence
time •
Fig. 1. Informationexpected from tumour markers in relation with the clinicalhistoryof cancer.
months (viz., CA15.3 in breast cancer) before clinical signs (Pons-Anicet et al., 1987; Kallionemi et al,. 1988). The pessimistic view which opposed the lack of efficient treatment of recurrent or metastasized cancer to the possibility of early detection is less valid since treatments now become efficient at this disseminated stage. Treatment monitoring. The major interest of tumour markers resides in the monitoring of treatment of advanced disease since they all show a good correlation with the evolution (progression or regression) of the disease and the efficacy of treatment. In conclusion, while the serological markers available today can provide very useful information when used sensibly, they are unlikely to assist in early diagnosis or in screening. The low prevalence of the disease definitely reduces the predictive value to a very low level as soon as sensitivity and specificity do not both approach 100%. Predictive value could be greatly enhanced if serological screening tests applied to high risk groups, primarily identified by epidemiological a n d / o r genetic analysis.
Assay of serologic tumour markers Assays of tumour markers are designed for the detection of antigens highly repressed in the healthy individual, with analytical sensitivity around 10 -~° or 10 -~' M. Most immunoassays are performed in vitro in heterogeneous phase (i.e., with separation of Ab-Ac complexes), assays in homogeneous phase not being yet enough sensitive. Two types of reaction can be considered: (i) competition (RIA) or (ii) extraction-saturation (IRMA). RIA requires a single antibody, monoor polyclonal, and purified labeled and unlabeled antigen preparations. IRMA requires at least two antibodies (at least one monoclonal) directed against different epitopes of the antigen, but no purified antigen. IRMA is more sensitive and more specific than RIA (Ekins, 1981). Finally the detection signal must be optimized for sensitivity and practicability. Radioisotope labelling (125I) of the antibodies has been used for long, but the technical and legal restrictions for their use have hampered the diffusion of radioactivity-based assays. New detection systems now include: fluores-
cence, time-resolved fluorescence, enzymatic fluorogenic or chromogenic reaction. They are all competitive and adaptable to automatic analyzers for routine analysis of large series of clinical samples. The competition is presently at the commercial level.
Tissue markers Besides diagnostic criteria, oncoiogists need prognostic criteria to decide between more or less agressive treatments. Ideally they require predictors of response to treatment. Response to treatment is largely dependent on the biological characteristics of a tumour, even if the precise biological feature is not yet known for the majority of treatment modalities. For instance, the presence of steroid receptors in breast tumours reliably predicts response to endocrine therapy. It would be most important to know what is determining the response to radiotherapy or to chemotherapy. Finally specific biological characteristics of tumours constitute potential targets for new types of therapies. For instance, growth factor receptors are currently under clinical evaluation for their prognostic worth and as targets for biotherapies. In this respect, tumour markers can be classified as (1) markers of differentiation, (2) markers of proliferation, (3) markers of metastatic potential, (4) oncogenes and anti-oncogenes. Representative of the first category are oestrogen and progesterone receptors in breast tumours, the presence of which is highly predictive of response to hormonal therapy. Biological markers of proliferation include cell cycle phase-associated antigens such as Ki67 (Gerdes et al., 1983), and growth factors and their specific receptors. Intensive research aims at the characterization of molecular determinants of the metastatic potential of tumours. Recently, several antigens have been proposed for clinical evaluation: proteases such as urokinase-plasminogen activator (Duffy, 1990) and cathepsin D; the rim23 gene product identified as a nucleotide diphosphokinase (Wallet et al., 1990). Cell adhesion molecules (fibronectin, integrins) have great potential interest in this field. Finally, the discovery of oneogenes (myc, H-ras, erbB2, etc.) and anti-onco-
141 genes (p53, Rb, DCC) has opened new avenues for basic and clinical investigations of a new generation of tumour markers. It is beyond the scope of this review to describe the biochemical and physiopathological characteristics of the above mentioned cellular and tissular parameters, but it is interesting to emphasize the methodological evolution which rapidly develops in the determination of such parameters, generally endowed with a functional biological activity. This will be illustrated by the case of receptors for steroid hormones and growth factors.
Assays of cellular receptors Radioligand binding assays in tissue extract Receptivity is the interaction of a iigand with a cellular receptor in the nucleus (steroid receptors), in the cytoplasm or in the cell membrane (growth factor receptors). Thus the first step of a biochemical assay is the extraction procedure. The assay of a receptor can be done by radioligand binding or by immunoenzymo-detection. Radioligand assays have the advantage of specific and absolute quantification of receptors since they determine: the affinity (or dissociation) constant of the specific binding; - the absolute number of specific sites; the specificity of the binding. However, the correct determination of binding sites imply a number of contraints: - rapid freezing of the tissue, due to the lability of binding sites; incubations with several concentrations of radiolabeled ligands in order to determine the affinity constant, through Scatchard analysis of saturation or competition curves; - the accurate determination of the specific radioactivity of the ligand; - controlled conditions for the separation of free from bound ligands; accurate radioactivity counting; - correct mathematical treatment of ligand binding data, specially in the presence of receptor sites with different binding affinities. One clue in clinical practice is the small quantity of tissue available: a Scatchard analysis re-
-
-
-
quires 500 mg of tissue, which is not always available (conservative treatment). One can eventually reduce the number of incubation concentrations down to one, provided an excess of ligand concentration (single saturation concentration) is used.
Immunological detection Monoclonal antibodies can detect epitopes of the receptor molecule which are not related to the binding sites, so that the receptor function is not evaluated (for instance, the receptor may be detected even when the binding site is blocked by endogenous ligands or antagonistic drugs). Antibody based assays of receptors provide however many advantages: they are robust since an epitope is generally more stable than the receptor site; sensitivity is excellent; they require a tiny amount of tissue extract (one incubation test), down to 20 mg of tissue or a few thousands cells; - colorimetric measurement is easy and inexpensive. The main flaws are the following: - the binding caracteristics of the receptor are not determined; - they do not provide absolute values of binding sites, and require an external calibration, generally through binding assay. The only alternative to the radioligand calibration problem is the availability of purified or recombinant receptor molecules. Despite the above limitations, immunoenzymatic assays of receptors are rapidly expanding: they are accessible to a greater number of laboratories, thus to a greater number of patients, than radioligand assays. -
-
-
lmmunohistochemical detection The clinical biochemist is faced with the rapid expansion of immunohistochemical methods for the determination of receptors, such as steroid hormone receptors. It requires even less tissue than biochemical assays. For instance, a transcutaneous fine needle aspirate of a breast can be reliably tested. Above all, it allows the direct visualization of the cellular heterogeneity within the tumour tissue. If an accurate evaluation of
142 the percer, tage of positive cells ten be done, the s t a i n i n g i n t e n s i t y is n o t er, sily q u a n t i t a t e d d u e t o the lack of reproducibility. Image analysis can partially overcem,~ this issue. Immunohistochemical determinations of rec e p t o r s f o r c l i n i c a l r o u t i n e a r e still in t h e e a r l y a g e o f a d e v e l o p m e n t , w h i c h is e x p e c t e d t o b e important. Quality control studies and clinical evaluation should be urgently implemented.
References Abelev, G.I. (1968) Production of embryonal serum alpha globulin by hepatomas. Review of experimental and clinical data. Cancer Res. 28, 1344-1350. Zabaleta, E.H. and Mollicone, R. (1986) Distibution of GICA in normal gastrointestinal and endocervical mueosae and in mucinous ovarian cyst using CA19.9. Am. J. Clin. Pathol. 85, 152-159. Bast, R.C.. Feeney, M., Lazarus, H., Nadler. L.M., Colvin, R.B. and Knapp, R.C. (1981) Reactivity of monoclonal antibodies with human ovarian carcinoma. J. Clin. Invest. 68, 1331-1337. Bence-Jones, H. (1847) Papers on chemical pathology. Lecture II1. Lancet 2, 269. Benchimol, S., Fucks, A., Jothy, S., Beauchemin, N., Shirota, K. and Stanners, C.A. (1989) Carcinoembryonic antigen functions as an intercellular adhesion molecule. Cell 57, 327-334. Bieglmayer, C., Szepezi, T., Kopp, B., Hoffmann, G. and Petrik, W. (1991) CAI5.3, MCA, CAM26, CAM29 are members of a polymorphic family of mucin-like glycoproreins. Tumor Biol. 12, 138-148. Bolla, M. and Martin, P.M. (1989) Les Marqueurs Tumouraux. Masson, Paris. Borg, A., Tandon, A., Sigurdsson, H., Clark, G., Fern6, M., Fuqua, A.W., Willander, D. and McGuire, L. (1990) HER2 / n e u amplification predicts poor survival in node-positive breast cancer. Cancer Res. 50, 4332-4337. Burchell, J., Gendler, S., Taylor-Papadimitriou et al. (1987) Development and characterization of breast cancer reactive monoclonal antibodies directed to the core protein of human milk mucin. Cancer Res. 47, 5476-5482. Caravani, M. (1988) Biochemistry and quantification of MCA in serum. Tumor Diag. Ther. 9, 39. Chu, T.M., Co~atantine, R. and Nemoto, T. (1989) Serum level of cryptic ;.umour antigens in breast cancer patients as determined by two monoclonal antibodies (M85/F36) and its comparison with CAI5.3. J. Clin. Lab. Anal. 3, 267-272. Chung, Y.S., Ho, J.J.U, Kim, Y.S. et al. (1987)The detection of human pancreatic cancer associated atttigen in the serum of cancer patients. Cancer 60, 1636-1643. Dudouet,B.. Jacob, L., Beuzeboc, P., Magdel~nat, H. et al. (199fl) Presence of villine, a tissue-specific cytoskeletal protein, in sera of patients and an initial clinical evalua-
tion of its value for the diagnosis and follow-up of colorectal cancers. Cancer Res. 50, 438-443. Duffy, M.I. (1990) Plasminugen activators and cancer. Blood Coag. Fibrinolysis 1,681-687. Ekins, R. (1981) Merits and disadvantages of different labels and methods of immunoassay. In: P. Voller (Ed.), Immunoassays for the 80's. MTP Press, Lancaster, pp. 5-16. Feizi, T. (1981) Carbohydrate differentiation antigens. Trends Biochem. Sci. 6, 333-335. Feizi, T. (1984) Monoclonal antibodies reveal saccharide structures of glycoproteins and glycolipids as differentiation and tumour associated antigens. Biochem. Soc. Trans. 12, 545-549. Feller, W.F. (1991) Mucin glycoproteins as tumour markers. In: R.B. Herberman and D.W. Mercer (Eds.), Immunodiagnosis of Cancer, 2rid edn. Marcel Dekker, New York, pp. 631-672. Garcia, M.B., Blankenstein, M.A., Van der Wall, E., Nortier, J.W.R., Schornagel, J.H. and Thijssen, J.H.H. (1990) Comparison of breast cancer mucin (BCM) and CATS.3 in human breast cancer. Breast Cancer Res. Treat. 17, 69-76. Gendler, S., Taylor-Papadimitriou, Duhig, T., Rothbard, J. and Burchell, J. (1988) A highly immunogenic region of a human polymorphic epithelial mucin expressed by carcinomas is made up of tandem repeats. J. Biol. Chem. 263, 12820-12824. Gendler, S.J., Cohen, E.P., Craston, A., Duhig, T., Johnstone, G. and Barnes, D. (1990) The locus of the polymorphic epithelial mucin (PEM) tumour antigen on chromosome lq21 shows a high frequency of alteration in primary human breast tumours. Int. J. Cancer 45, 431-435. Gerdes, J., Schwab, U. and Lemke, H. (1983) Production of a mouse monoelonal antibody reactive with a human nuclear antigen associated with cell proliferation. Int. J. Cancer 31, 13. Gold, P. and Freeman, S.O. (1965) Demonstration of tumour-specific antigens in human colonic carcinomata by immunological tolerance and absorption techniques. J. Exp. Med. 121,439-445. Gold, D.V. and Mattes, M.J. (1988) Monoclonal antibody B72.3 reacts with a core region structure of O-linked carbohydrates. Tumor Biol. 9, 137-142. Gool, H.C., Uemura, K.-I., Edwards, P., Foster, C.S., Pickering, N. and Feizi, T. (1983) Two mouse hybridoma antibodies against human milk fat globules recognize the ! (MA) antigenic determinant. Carbohydrate Res. 120, 293302. Hakomori, S. (1985) Aberrant glycosylation in cancer in cancer cell membranes as focused on glycoSpids: overview and perspectives. Cancer Res. 45, 24(15-2414. Hakomori, S. (1986) Tumor-associated glycolipid antigens, their metabolism and organization. Chem. Phys. Lipids 42, 209. Hanisch, F.G., Uhlenbruck, G., Dienst, M., Stottrop, M., hippauf, E. (1985) CA125 and CAI9-9: two cancer-associated sialylsaccharide antigens on a mucus glycoprotein from human milk. Eur. J. Bioehem. 149, 323. Harmemberg, U., Wahren, B. and Wiechel, K,-L. (1988) Tumor markers carbohydrate antigens CA19.9 and C~,~;0 and
CEA in pancreatic cancer and benign diseases of the panereatobiliary tract. Cancer Res. 48, 3856-3863. Herlyn, M., Sears, H.F., Steplewski, Z. and Koprowski, H. (1982) Monoctonal antibody detection of a circulating turnout-associated antigen. Presence in sera of patients with colorectal, gastric and pancreatic eareinoma. J. Clin. lmmunobiol. 2, 135-140. Hilkens, J., Buijs, F., Hilgers, J., Hageman, P., Calafat, J., Sonnenberg, A. and Van der VaIL M. (1984) Monoclonal antibodies against human milk fat globule membrane detecting differentiation antigens of the mammary gland and its tumors. Int. J. Cancer 34, 197. Hull, S., Bright, M., Abe, M., Kufe, D. and Carraway, L. (1988) Oligosaceharides of the DF3 antigen of the BT-20 human breast carcinoma line. J. Cell. Biochem. 12E (suppl.), 130-133. Ishiguro, T., Sugitachi, I. and Sakaguchi, H. (1985) Serum alpha protein substractions in patients patients with primary hepatoma or hepatic metastasis of gastric cancer. Cancer 55, 156. Itai, S., Arii, S., Tobe, R. et al. (1988) Significance of 2,3 and 2,6 sialylation of Lewis A antigen in pancreas cancer. Cancer 61,775-787. Itai, S., Nisbikata, J., Yoneda, T., Ohmori, K. and Kannagi, R. (1991) Tissue distribution of 2,3 and 2,6 sialyl Lewis A antigens and significance of the ratio of two antigens for the differential diagnosis of malignant and benign disorders of the digestive tract. Cancer 67, 1576-1587. Johnson, V.G. and Schlom, J. (1986) Analysis of human tumour-assoeiated glycoprotein (TAG72) identified by monoelonal antibody B72.3. Cancer Res. 46, 850. Kallionemi, O.-P., Oksa, H., Aaran, R.-K., Hietaner, T., Lehtinen, M. and Koivula, T. (1988) Serum CA15.2 assa~ in the diagnosis and follow up of breast cancer. Br. J. Cancer 58, 213-215. Kannagi, R., Hakomori, S.I. and Imura, H. (1988) Cancer-associated serum glycoprotein antigens expressing type I and type 2 chain polylactosamines. In: C.L. Reading, S.I. Hakomori and D.M. Marcus (Eds.), Al',ered Glycosylation in Tumor Cells. Alan R. Liss, New York, pp. 279-294. Kato, H. and Moioka, H. (1983) Prognostic significance of the tumor antigc;~ TA-4 in squamous cell carcinoma of the uterine cervix. Am. J. Obstet. Gynecol. 145, 350-354. K6hler, H. and Millstein, C. (1975) Continuous cultures of fused cells secreting antibody of pre-defined specificity. Nature 256, 495. Kufe, D. (1984) Differential reactivity of a novel monoclonal antibody (DF3) with human malignant versus benign breast tumors. Hybridoma 3, 223. Magnani, J.L., Nilsson, B. and Brockhaus, M. (1982) A monoclonal antibody defined antigen associated with gastrointestinal cancer is a ganglioside containing sialylated lactoN-fueopentaose II. J. Biol. Chem. 257, 14367-14369. Manderino, G. (1990) Characterization of the M65 MAb used in the BCM-IMx: satellite symposium, breast cancer mucin - new tumor marker in breast cancer. 3rd Int. Symp. Biol. Clin. Usefulness of Tumor Markers, Madrid, September 1990.
Manssonn, J.E., Frcdman, P., Nilsson, O. et al. (1985) Chemical structure of carcinoma ganglioside antigens defined by monoclonal antibody CAS0 and some allied gangliosides of human pancreatic adenocarcinoma. Biochim. Biophys. Acta 834. 11(I-117. Mant, D. and Fowler, G. (1990) Mass .screening: theory and ethics. Br. Med. J. 3(10, 916-918. McGuire, W.L., Clark G. and Dressier, L. (1986) Role of steroid hormone receptors as prognostic factors in primary breast cancer. NCI Monogr. I, 19-23. Nishi, S. (1970) Isolation and characterization of a human fetal alphaglobulin from the sera of fetuses and a heparoma patient. Cancer Res. 30, 2507. Nowell, P.C. and Hungerfold, D.A. (1960) A minute chromosome in human chorionic granulocytic leukemia. Science 132, 1497. Oikawa, S., Nakazato, H. and Kosaki, G. (1987) Primary structure of human eareinoembryonic antigen (CEA) deduced from cDNA sequence. Biochem. Biophys. Res. Commun. 142 511-518. Paxton, R., Mooser, G., Panoe, H., Lee, T.D., Shiley, J.E. and Terry, H.P. (1987) Sequence analysis of carcinoembryonic antigen: identification of glycosylation sites and homology with the immunoglobulin supergene family. Proc. Natl. Acad. Sci. USA 84, 920-924. Pons-Anicet, D., Krebs, B. and Namer, M. (1987) Value of CAI5.3 in the follow-up of breast cancer patients. Br. J. Cancer 55, 567. Roulston, J.E. (1990) Limitations of tumour markers in screening. Br. J. Surg. 77, 961-962. Sell, S. (1991) Cancer markers of the 1990s: Tumor markers in diagnostic pathology. Clin. Lab. Med. 10, 1-37. Springer, F. (1984) T and Tn, general carcinoma antigens. Science 224. 1198-1206. Sulman, C., Carpentier, P., Lemaire J., Gosselin, P. and Cazin, J.-L. (1987) Seuil de d~cision en biologie. Trait d'Union 5, 23-29. Swallow, D.M., Gendler. S., Griffiths, B., Corney, G., TaylorPapadimitriou and Bramwell, M.E. (1987) The human tumor-associated epithelial mucins are coded by an expressed hypervariable gene locus PUM. Nature, 328, 82. Tempero, M.A., Uchida, E., Takasaki, H. et al. (1987) Relationship of carbohydrate antigen 19.9 and Lewis antigens in pancreatic cancer. Cancer Res. 47, 5501-5503. Thompson, J. and Zimmermann, W. (1988) The carcinoembryonie gene family: Structure, expression and evolution. Tumor Biol. 9, 63. Wallet, V., Mutzel, R., Troll, H., Barzu, O. Wurster, B., W:ron, M. and Lacombe, M.L. (1991) Dictyostelium nucleoside diphosphate kinase is highly homologous to Nm23 and Awd proteins involved in mammalian tumor metastasis and Drosophila development. J. Natl. Cancer Inst. 82, 1199-1202. Watson, D.M., Elton, R.A., Jack, WJ., Dixon, J.M., Cherty, U. and Miller, W.R. (1991) The H-ras oncogene product p21 and prognosis in human breast cancer. Breast Cancer Res. Treat. 17, 161-169.
133
© 1992 ElsevierScience Publishers B.V. All rights reserved 01)22-1759/92/$05.00
J1M06336
Tumour markers in oncology: past, present and future Henri Magdel6nat Laboratoire de Radiopathologie, Institut Curie, Paris, France
(Accepted 12 February 1992)
Historically, tumour markers are substances selectively released by tumour cells in the blood stream so that they can be detected in the serum or other body fluids for clinical monitoring of various malignancies. The term is now extended to cell or tissue characteristics, such as cytogenetic markers, oncogenes or abnormally expressed proteins with various biological functions (enzymes, receptors, etc.), which help characterize the tumour type and constitute specific biological targets for new drugs. Serum markers can be classified in three main categories: (i) oncofetal antigens which are normally expressed only during fetal life, (ii) carbohydrate antigens as part of glycolipids or glycolipoproteins (mucins) and (iii) miscellaneous, including tissue specific enzymes, cytoskeletal proteins, etc. Most carbohydrate antigens which are useful in serum diagnosis of human cancers show turnout type-specific aberra.,t glycosylation, with O instead of N linkage to protein moieties. Despite a huge progression in analytical performances with immunoassays, and because of still insufficient biological sensitivity and specificity, serum assays of tumour markers are not effective in mass screening, their usefulness being restricted to the follow-up of diagnosed patients. Selective expression of specific antigens allows in vivo detection of occult turnout masses (immunoscintigraphy) and immunotargetting of drugs as therapy. Cell and tissue markers comprise markers of transformation (molecular genetics, oncogenes and anti-oncogenes, etc.), differentiation (hormone receptors, etc.), proliferation (growth factors, Ki67, etc.) and metastatic potential (proteases, etc.). Although masking the functional aspects of many of these markers, as for hormone or growth factor receptors, immunoanalytical methods, specially immunohistochemistry, are rapidly expanding in clinical oncology as tools for diagnosis, prognosis and, promisingly, treatment adjustment. Key words: Tumor marker; Tumor antigen; Hormone receptor; lmmunoanalysis;Oncology
Introduction
The first cancer marker was identified over 140 years ago, in 1846, when H. Bence-Jones described a heavy precipitate upon acidification of the urine of a patient diagnosed as having
Correspondence to: H. Magdel6nat, Laboratoire de Radiopathologie et Physiopathologie, Institut Curie, 26 rue d'Ulm, 75231 Paris Cede× 5, France.
'mollities osseum' or softening of the bones. It took more than 100 years before the 'Bence Jones protein' was identified as the monoclonal light chain of immunoglobulin, and an elevated concentration was found in the plasma of patients with multiple myeloma. The second era of tumour markers, from 1928 to 1963, included the discovery and application of hormones (human chorionic gonadotrophin (Zondek, 1930); adrenocorticotropic hormone (Cushing, 1932)), enzymes and isoenzymes (Markert, 1959) to cancer diagno-
134 sis. However, the era of turnout markers as more generally applicable tools began with the discovery of a-fetoprotein (AFP) by Abelev in 1963 and carcinoembryonic antigen (CEA) by Gold and Freeman in 1965. Gold and Freedman immunized rabbits with extracts of human colonic cancer tissue, absorbed the resulting antisera with normal human colon and found that some of the antisera reacted with the tumour extracts, but not extracts of normal tissue. The antigen so detected was named CEA since it was also found in embryonic tissue. This approach raised a great hope of finding many other 'specific' tumour antigens and is still used to generate new specific antibodies. In the past decade, glycolipids and high molecular weight glycoproteins, termed mucins, were found the most clinically useful, tumour-associated antigens. The present era is therefore characterized by intense efforts at molecular characterization of these antigens. Progress in detailed biochemical analysis of antigenic epitopes, in molecular biology, in gene mapping and in transcriptional and post-transcriptional regulation mechanisms will certainly help developing more specific probes for clinic and research. The expansion of tumour markers in clinical practice is largely due to (i) the development of sensitive radioimmunoassays by Berson and Yallow in 1968, (ii) the discovery of monoclonal antibodies by KShler and Milstein in 1975, which yield reproducible sources of antibody and provide more specific reagents. At the clinical level, the present era is characterized by the screening of new monoclonal antibodies and a shift from radioactive to non-radioactive detection systems, from manual to automated analysis, in brief, by important technological and commercial developments. From this very brief summary, a definition of a tumour marker seems obvious: a tumour marker refers to any substance that can be detected in plasma (or body fluids) and related to the presence or progress of malignancy. There are strong tenants of this unique definition but, in a clinically oriented opinion, the term applies more generally to any biological characteristics of a malignant tissue which can be assayed by biochemical or histochemical methods and provides information relevant to diagnosis, prognosis or treatment monitoring in oncology. This leads to
the distinction between serum (or plasma) markers, as already introduced, and tissue (or cellular) markers, the development and application of which constitute the new era of tumour markers. In this sense, the Philadelphia chromosome (Nowell and Hungerford, 1960) may constitute the first specific cellular marker of a clonal malignancy, i.e., chronic myeloid leukaemia. More recently, steroid hormone receptors have proved clinically useful as prognostic factors and predicto,:s of response to hormonal therapy in breast cancer (McGuire et al., 1986), whereas growth factors and oncogens constitute tissue markers of great potential interest (Borg et al., 1990; Watson et al., 1991).
Serologic tumour markers An ideal seric tumour marker would be a substance not normally present in blood, whose detection in serum reflects the presence of a minimal tumour burden (sensitivity) and whose concentration accurately reflects tumour progression or regression (low inertia). In the ideal situation, each tumour type would express a different and specific antigen (specificity). Consequently, the presence of a marker in blood is dependent on its expression and release by the tumour mass. Monoclonal antibodies have been used extensively to unravel antigen specifically expressed by tumour cells. The conclusion of these studies is that such 'neo-antigens' do not exist. Thus, tumour-assoeiated antigens represent either a quantitative overexpression of antigens normally expressed by tissues of the same embryonic origin (enzymes, peptide hormones, etc.), or the re-expression of cellular antigens selectively expressed during embryogenesis (oncofetal antigens), or, more specifically, post-transcriptional variants, for instance by abnormal glycosylation, of cell membrane components (tumour-associated glycolipids and mucins). It may, however, be foreseen that mutations or translocations in the corresponding gene could result in cancer-specific protein alteration, as already exemplified by modified p21 protein (p21 raS) or mutated p53 anti-oncogene product (Table I).
135 TABLE I CLASSIFICATION O F T U M O U R
MARKERS
Oncofetal antigens Tumour-associated antigens (carbohydrate antigens)
AFP, CEA (I) Carbohydratedeterminant: Ca 19.9.CA50, CA72.4, GD2 etc, (2) Epithelial mucins(PEM): CAI5.3, MCA, CA549, CTA etc. (3) Glycoproteins:CA125, SCC. TPA, PSA etc.
Hormonal peptides
fl-HCG, tlLP, ACTH, TCal, GH, PTH etc.
Neuromediators
Catecholamine, melatonine, serotonine etc.,
Proteins Enzymes
,82-microglobulin,ferritine thyroglobulin, PAP, PAcP, NSE, TK
Amines
Polyamines(spermidine, spermine)
Oncogene products
p21, c-myc, neu etc.
Oncofetal antigens Oncofetal antigens can be defined as antigens that are expressed during intra-uterine life, diminish strongly or disappear after birth and reappear in situations of repair or neoplastic growth in the organs where they appeared (in the fetus) during gestation. This category comprises a fetoprotein (AFP), strictly fitting the definition, and carcino-embryonic antigen (CEA) with main features of oncofetai antigens. AFP, a glycoprotein purified by Nishi in 1970 is expressed in the vitellus membrane and in the fetal liver and digestive tract. Glycosylation is variable during the fetal life and in tumours, allowing the characterization of specific subfractions (lshiguro et al., 1985). In the adult, it is essentially associated with malignant hepatoma, the immunoassay being positive in over 70% of hepatoma patients. Elevated serum levels are also found in ovarian and testicular choriocarcinomas (but not in seminomas), rarely in cancers of the digestive tract or lungs, frequently but at moderate level in chronic hepatitis or in cirrhosis. CEA or carcinoembryonic antigen represents a large family of related cell membrane glycoproteins, with extracellular domain structures similar to immunoglobulins, and thus belongs to the immunoglobulin gene superfamily (Thompson and Zimmermann, 1988). So far, at least 14 genes and 36 different glycoproteins have been identified in the C E A family. Based on sequence similarities,
the CEA gene family can be divided into two main subgroups. One of these contains the genes encoding CEA and cross-reacting antigens (NCA) and the biliary glycoprotein. The other contains the gene encoding the pregnancy-specific/31-glycoproteins (PSG), a group of glycoproteins found in placenta and tumours of trophoblastic origin. All the genes from both CEA and PSG subgroups are clustered on the long arm of chromosome 19 and apparently closely linked (Zimmermann and Thompson, 1990). It is now well established that the CEA gene family belongs to the immunoglobulin supergene family (Oikawa et al., 1987; Paxton et al., 1987). Where known, members of the lg family have receptor or cell adhesion functions, e.g., the neuronal cell adhesion molecule (N-CAM) to which CEA has been compared (Benchimol et al., 1989).
Turnout-associated mucins Alterered glycosylation appears to be a constant phenomenon associated with oneogenic transformation in experimental systems as well as in essentially all types of naturally occurring human cancers. Most of the biochemical or, more recently, immunological methods used to identify tumour-assoeiated antigens have resulted in the isolation of glycolipids or glycoproteins (mucins) with altered glycosylation patterns. Turnout antigens defined by carbohydrate determinants are related to blood group alloantigens or to differentiation carbohydrate antigens. Although these
changes are all quantitative rather than qualitative, some antibodies can recognize tumour-distinctive structural organizations of carbohydrate moieties. Other antigens refer to the protein core rather than to the sugar moiety of tumour-associated highly glycosylated, high molecular weight glycoproteins or mucins. Little is known about the detailed structure of the core protein of human mucins but the recent cloning and sequencing of the highly polymorphic gene of one such mucin differentially expressed by the lactating mammary gland and by many carcinomas has defined a new class of tumour-associated antigens, the polymorphic epithelial mucins or PEM (Gendler et al., 1991).
Carbohydrate or blood group-related tumour antigens Many of the more clinically useful tumour markers identified over the last decade are mucins. Historically, the term mucin was used to designate biological polyanions which are stained by various dyes, including Alcian blue. More recent studies have characterized mucin tumour antigens as high molecular weight glycoproteins, from 200 kDa to greater than 1000 kDa. Mucin antigens contain from 25 to 80% carbohydrate and their density is usually greater than 1.40 g/i. Cancer carbohydrates Jelong to four major structural classes: ganglio, globo, lacto type 1 and lacto type 2 (Hakomori, 1986; Kannagi et al., 1988). Galactose and N-acetylglucosamine are the predominant sugar residues, whereas mannose is usually in low level or absent. An O-glycosidic linkage between the carbohydrate and either serine or threonine of the protein core of mucins is the most common linkage of cancer mucin antigens, in contrast with mucins of non-malignant cells or with the 'classic' tumour markers - CEA, AFP or PAP (prostatic alkaline phosphatase) which are N-linked carbohydrates. Many differentiation antigens present on cell surfaces being glycolipids and O-linked glycoproteins with qualitative and quantitative alterations during various phases of differentiation or oncogenesis (Feizi, 1981; Hakomori, 1985), the importance of O-linked carbohydrate epitopes appears less surprising. Changes in carbohydrate expression corre-
spond to changes in the enzyme activities of specific glycosyltransferases responsible for their synthesis or glycosidases responsible for their degradation. For example, the decreased expression of blood group antigens (ABH) in some tumours is due to incomplete synthesis (Schoentag et al., 1987) or masking by sialic acid (Chu et al., 1989). Alterations of carbohydrate structure can result in the generation of tumour markers without enhanced production ef the peptide core. Monoclonai antibodies that can detect subtle carbohydrate epitope alterations are thus particularly useful to discriminate malignant from benign pathological process (Itai et al., 1991) and even to bring organ specificity in malignancy. CA19.9, one of the first identified tumour mucin antigens, is the 2,3-sialylated form of the Lewis a (Lea) structure (Tempero et al., 1987), present on glycolipids of certain tumours (Magnani et al., 1982). This epitope is a normal component of certain tissues of some individual and is expressed strongly on glycolipids and glycoproteins from adenocarcinomas of the colon and pancreas but not from normal colon. The immunogenic determinant being essentially carbohydrate in nature, neuraminidase treatment of CA19.9 results in complete loss of antigenic activity. As a tumour marker, the serum level of CA19.9 is elevated in patients with pancreatic and gastrointestinal (Herlyn et al., 1982) cancers with limited sensitivity ( ~ 30%) in patients with early stage tumours. Lewis negative individuals are genetically incapable of expressing CA19.9 (Bara et al., 1986). Interestingly, the 2,6-sialylated Le a antigen, which has close structural relationship to the CA19.9 antigen, is abundantly present in non-malignant pancreas tissues and is significantly decreased in pancreatic cancer tissues, the 2,3/2,6 antigen ratio allowing a better discrimination between malignant and benign disease (Itai et a1.,1988, 1991). CASO is related to CA19.9 as the afucosyl form of sialylated Lewis antigen (Mansson et al., 1985). It is similar to CA19.9 in monitoring advanced gastrointestinal and pancreatic cancers but with broader specificity (Harmenberg et al., 1988). CAS0 does not require the fucose residue in the Lewis epitope, whereas CA19.9 does. Consequently, CAS0, like Span-l, another CA19.9-re-
137 lated monoclonal antibody, binds to tissues from Lewis negative individuals (Chung et al., 1987). CA72.4 or TAG72 is a high molecular weight ( > 10 ~ Da) mucin like molecule (Johnson et al., 1986) recognized by the monoclonal antibody B72.3, generated by immunization with membrane enriched fractions of human breast cancer metastases. The suggested antigenic determinant is the disaccharide NeuAc(2,6) GalNAc (Gold et al., 1988). It may be useful in diagnosis and monitoring some human epithelial cancers, for instance gastric cancer. Breast cancer mucin (BCM) is a recent tumour-associated mucin-like antigen defined by M85, a mouse monoclonal IgM antibody, reacting with paragloboside carbohydrate structures with specificity for Galfll,4GIcNac (lacto type 2 chain) sequence (Manderino, 1990). The antigen has been termed 'cryptic' since M85 reactivity to serum or ascites is greatly enhanced by removal of sialic acid by neuraminidase. The presence of cryptic cell surface antigens, some of them differentiation-associated, some tumour-associated, has been reported, notably by Springer 0984) and Gooi (1983). Polymorphic epithelial mucins (PEM) The polymorphism of epithelial mucins originally noted at the protein level has now been shown to be at the D N A level, and is due at least in part to the presence of differing numbers of tandem repeats (Gendler et al., 1988).The gene of PEM has been cloned (Swallow et al., 1987) and located on human chromosome 1 (Gendler et al., 1990). The utility of circulating mucins as breast cancer markers, initially proposed by Ceriani (1982), has been rapidly confirmed and expanded (Hilkens, 1984). Several antibodies have been raised against such mucins, from membrane fractions of either normal lactating (HMFG-1 or HMFG-2) or malignant mammary glands (MAM6 family, DF3, CTA). CA15.3, an important antigen in the monitoring of breast cancer, is a 400 kDa glycoprotein recognized by a pair of monoclonai antibodies, 115 D8 (Hilkens et al., 1984) against membranes of human milk fat globules (HMFG), and DF3 against a membrane enriched fraction of a metastatic breast cancer lesion (Kufe et al., 1984).
Epitopes of HMFG-derived antibodies are partially carbohydrate (lectin binding) and partially peptide (BurcheU et ai., 1987). MAb DF3 recognizes a high molecular weight glycoprotein containing sialyl oligosaccharides present on a peptide backbone. Further characterization of the carbohydrate components of the DF3 antigen has revealed that the major oligosaccharides are the p e a n u t agglutinin binding dissaccharides Gal/31,3GaINAc (lacto type l chain), also designated as the Thomsen-Friedenreich antigen and its mono- and di-sialylated derivatives (Hull et al., 1988). Although linked to breast malignancy, CA15.3 is not breast specific and can be found elevated in metastazing ovarian, colonic or bronchial carcinoma. Its sensitivity is low (2030%) in early breast cancer. However it is largely used in the follow-up of patients at risk of recurrence after treatment of primary breast cancer and in the monitoring of metastasized breast cancer (Pons Anicet et ai., 1987; Kallionemi et al. 1988). The follow-up and the monitoring of breast cancer are the major applications of turnout markers in clinical oncology, due to the high incidence of this disease and to the possibility to increase survival by early detection and treatment of recurrence. So, variants of CA15.3 and mucinlike antigens are available for routine clinical analysis from different suppliers: MCA, MSA, CAM26, CAM29, CA 549, BCM, etc. (Bieglmayer et al., 1991, Garcia et al., 1990). Unfortunately, the lack of sensitivity in early stages and a moderate specificity has limited the sensible clinical application of this markers to the monitoring of advanced stages of the disease. MCA or mucin-like cancer antigen is a 350 kDa glycoprotein (Caravatti, 1988). It is a membrane protein overexpressed in breast cancer cells. In serum MCA is measured using a sandwich immunoenzyme assay employing the same antibody as primary and secondary antibody. MCA is similar to CA15.3 in monitoring patients with breast cancer. Being elevated in non-malignant breast diseases its use for diagnosis is unlikely. CA125 is identified by a monoclonal antibody raised to an ovarian carcinoma cell line (Bast et al., 1981). The nature of the epitope is still controversial but appears to contain sialie acid (Hanisch et al., 1985). Its sensitivity is good for
the non-mucinous histological type of ovarian carc#1omas but it lacks specificity (breast, lung, benign and malignant effusions are often reactive). The half-time of its fall after initial trealment is prognostic and its elevation during follow-up of ovarian carcinomas may spare surgical 'second look'. Unfortunately, treatments of reccurrent ovary cancer are rather inefficient, which limits the therapeutic gain potentially attainable with a sensitive indicator of early receurrence. CA125 is one of the rare monoelonal antibodies used to detect occult metastasis by immunoscintigraphy in vivo.
Other antigens PSA or prostate-specific antigen, isolated by Nancy et al. in 1979, is a 34 kDa intracellular glycoprotein, exclusively secreted by the prostate gland with protease activity. PSA is elevated in patients with benign prostatic, hypertrophy and acute prostatitis, so that its usefulness for diagnosis is limited. It is very useful in monitoring the efficacy of surgery after removal of the tumours and the reccurence of prostatic cancer. Its sensitivity and specificity appear superior to that of PAP (prostate alcaline phosphatase) in this monitoring. TPA or tissue polypeplide antigen is a 47 kDa protein sharing homologies with epithelial cytokeratines 8, 18 and 19. It is actively secreted during mitosis of multiplicating cells in culture. It is therefore not cancer specific but rather related to cell turnover or proliferation. Although not useful in diagnosis, its good sensitivity makes it a useful test in patients with known malignancies, specially in the absence of more specific marker (for instance in bladder cancer). Also related to the cytoskeleton, Villin is a cytoskeleton protein of the enterocyte brush border, a differentiation marker of early embryogenesis with restricted tissue specificity in the human adult (enterocyte of small and large intestin, gall bladder and pancreas duct cells). It may be useful in diagnosis and follow-up of colorectal cancers (Dudouet et al., 1990). SCC or squamous cell carcinoma antigen, is a 48 kDa glycoprotein, sub-fraction of the complex TA-4 fraction, isolated in 1977 by Kato et al.
from an epidermoid carcinomz of the uterine cervix. It is a marker for epidermoid (or squamous) carcinomas, with a good specificity but low sensitivity (recently improved by the use of a monoclonal antibody instead of the initial polyelonal one). It can be useful in monitoring advanced epidermoid cancers of the uterine cervix, of the lung and of head and neck.
When are tumour markers clinically useful? Criteria of utility The important criteria for a marker are its sensitivity and specificity. A sensitive marker detects a high percentage of patients with disease, while a specific assay is only positive in cancer patients, not in those without cancer. Sensitivity is calculated as (true positive/(true positive + false negative)), while sensitivity is calculated as (true negative/(true negative+ false positive)). Predictive values are linked to sensitivity and specificity but also to the prevalence (P) of the disease in the population studied (Sulman et al., 1987) (Table II). This is important to consider when evaluating the worth of a tumour marker in different clinical situations, for instance in the context of mass screening (low prevalence) or in a medical oncoiogy department (high prevalence). The predictive value of a tumour marker test may therefore be augmented either by increasing the sensitivity and specificity of the test or by selecting the appropriate sub-group of population or patients to be tested. Tumour markers tests being quantitative, their sensitivity and specificity depends upon the cut-off value chosen to diserimiTABLE !1 SENSITIVITY,SPECIFICITY,PREDICTIVE VALUES Cancer Marker
Positive
TP
No Cancer FP
Negative
FN
TN
Sensitivity
Specificity
Positive predictive value Negative predictive value
139 nate between 'positive' and 'negative' tests, here again this cut-off may be chosen according to the clinical end point to reach, i.e., whether sensitivity or specificity is more important in the population tested. The use of ROC curves facilitates the choice of a cut-off value and allows an objective evaluation of the respective worths of different tests (Sulman et al., 1987; Bieglmayer et al., 1991).
Tumour markers and clinical phases of the disease The clinical history of cancer is schematized in Fig. 1. The information expected from serological tumour markers varies according to the stage of the disease. Screening before apparent disease. A tumour marker with almost absolute specificity (at least > 95%) and good sensitivity would be required for mass screening of an apparently healthly population, due to the low prevalence of cancer in the general population. The lack of high specificity and sensitivity for available markers in cancers where early detection would be beneficial (breast, colon) precludes such tests for mass screening at reasonable cost-efficiency ratios. A simple calculation shows that in ideal conditions (1% incidence of the disease, 99% sensitivity and specificity), the frequency of false positive is already 50%. The only successful tumour marker used in
screening is human chorionic gonadotrophin (HCG) in choriocarcinoma. The relatively high prevalence of tumour (5-10%) in women who have had a hydatiform mole, the excellent assay sensitivity and specificity, together with the high chemosensitivity of the tumour, has led to a significant reduction in mortality through screening of this population. Diagnosis. The assay of a tumour marker may be interesting at the time of diagnosis as a confirmation of origin or to assess the extension of the neoplasm although no biological marker has been yet integrated in the staging of the disease. Even at this symptomatic stage of disease, the sensitivity of current tumour markers is too low (20-30% for CA15.3 in primary breast cancer) for a serum assay to be really useful for diagnosis. It may however be interesting to consider an elevated pre-treatment value in order to subsequently monitor tile efficacy ef the initial treatment through., for instance, the half-life clearance of the marker. Specific antibodies may also be useful to confirm a diagnosis by immunocytochemical staining of tumour biopsies. Follow up. Serological tumour markers are useful in the follow-up of patients after initial therapy. A continuous elevation of the serum level is strongly indicative of reccurrence of the disease, with in many cases, a lead time of several
Primary treatment clinical = Idetection level ~11~1.. I'atlures
Tumor birden
/ Secon~"trea}ment
. ~ _~_ Tumor / ~ - .... -~. . . . . . . . . . . . . . . . ~- ..... -~ . . . . . . . . . / g r ~ . ~ . : i ....... ~ ""~""t ""'~'-" ~ General ...................... population ~init ~promotion Remission I second remission C!inical =Clinical n~ malignancy diagnosis recurrence High risk
.......... Screening
Therapeutic staoino efficiency Nb Meta + Therapeutic prognosis Residual ÷ efficiency which therapy? disease Detection of recurrence
time •
Fig. 1. Informationexpected from tumour markers in relation with the clinicalhistoryof cancer.
months (viz., CA15.3 in breast cancer) before clinical signs (Pons-Anicet et al., 1987; Kallionemi et al,. 1988). The pessimistic view which opposed the lack of efficient treatment of recurrent or metastasized cancer to the possibility of early detection is less valid since treatments now become efficient at this disseminated stage. Treatment monitoring. The major interest of tumour markers resides in the monitoring of treatment of advanced disease since they all show a good correlation with the evolution (progression or regression) of the disease and the efficacy of treatment. In conclusion, while the serological markers available today can provide very useful information when used sensibly, they are unlikely to assist in early diagnosis or in screening. The low prevalence of the disease definitely reduces the predictive value to a very low level as soon as sensitivity and specificity do not both approach 100%. Predictive value could be greatly enhanced if serological screening tests applied to high risk groups, primarily identified by epidemiological a n d / o r genetic analysis.
Assay of serologic tumour markers Assays of tumour markers are designed for the detection of antigens highly repressed in the healthy individual, with analytical sensitivity around 10 -~° or 10 -~' M. Most immunoassays are performed in vitro in heterogeneous phase (i.e., with separation of Ab-Ac complexes), assays in homogeneous phase not being yet enough sensitive. Two types of reaction can be considered: (i) competition (RIA) or (ii) extraction-saturation (IRMA). RIA requires a single antibody, monoor polyclonal, and purified labeled and unlabeled antigen preparations. IRMA requires at least two antibodies (at least one monoclonal) directed against different epitopes of the antigen, but no purified antigen. IRMA is more sensitive and more specific than RIA (Ekins, 1981). Finally the detection signal must be optimized for sensitivity and practicability. Radioisotope labelling (125I) of the antibodies has been used for long, but the technical and legal restrictions for their use have hampered the diffusion of radioactivity-based assays. New detection systems now include: fluores-
cence, time-resolved fluorescence, enzymatic fluorogenic or chromogenic reaction. They are all competitive and adaptable to automatic analyzers for routine analysis of large series of clinical samples. The competition is presently at the commercial level.
Tissue markers Besides diagnostic criteria, oncoiogists need prognostic criteria to decide between more or less agressive treatments. Ideally they require predictors of response to treatment. Response to treatment is largely dependent on the biological characteristics of a tumour, even if the precise biological feature is not yet known for the majority of treatment modalities. For instance, the presence of steroid receptors in breast tumours reliably predicts response to endocrine therapy. It would be most important to know what is determining the response to radiotherapy or to chemotherapy. Finally specific biological characteristics of tumours constitute potential targets for new types of therapies. For instance, growth factor receptors are currently under clinical evaluation for their prognostic worth and as targets for biotherapies. In this respect, tumour markers can be classified as (1) markers of differentiation, (2) markers of proliferation, (3) markers of metastatic potential, (4) oncogenes and anti-oncogenes. Representative of the first category are oestrogen and progesterone receptors in breast tumours, the presence of which is highly predictive of response to hormonal therapy. Biological markers of proliferation include cell cycle phase-associated antigens such as Ki67 (Gerdes et al., 1983), and growth factors and their specific receptors. Intensive research aims at the characterization of molecular determinants of the metastatic potential of tumours. Recently, several antigens have been proposed for clinical evaluation: proteases such as urokinase-plasminogen activator (Duffy, 1990) and cathepsin D; the rim23 gene product identified as a nucleotide diphosphokinase (Wallet et al., 1990). Cell adhesion molecules (fibronectin, integrins) have great potential interest in this field. Finally, the discovery of oneogenes (myc, H-ras, erbB2, etc.) and anti-onco-
141 genes (p53, Rb, DCC) has opened new avenues for basic and clinical investigations of a new generation of tumour markers. It is beyond the scope of this review to describe the biochemical and physiopathological characteristics of the above mentioned cellular and tissular parameters, but it is interesting to emphasize the methodological evolution which rapidly develops in the determination of such parameters, generally endowed with a functional biological activity. This will be illustrated by the case of receptors for steroid hormones and growth factors.
Assays of cellular receptors Radioligand binding assays in tissue extract Receptivity is the interaction of a iigand with a cellular receptor in the nucleus (steroid receptors), in the cytoplasm or in the cell membrane (growth factor receptors). Thus the first step of a biochemical assay is the extraction procedure. The assay of a receptor can be done by radioligand binding or by immunoenzymo-detection. Radioligand assays have the advantage of specific and absolute quantification of receptors since they determine: the affinity (or dissociation) constant of the specific binding; - the absolute number of specific sites; the specificity of the binding. However, the correct determination of binding sites imply a number of contraints: - rapid freezing of the tissue, due to the lability of binding sites; incubations with several concentrations of radiolabeled ligands in order to determine the affinity constant, through Scatchard analysis of saturation or competition curves; - the accurate determination of the specific radioactivity of the ligand; - controlled conditions for the separation of free from bound ligands; accurate radioactivity counting; - correct mathematical treatment of ligand binding data, specially in the presence of receptor sites with different binding affinities. One clue in clinical practice is the small quantity of tissue available: a Scatchard analysis re-
-
-
-
quires 500 mg of tissue, which is not always available (conservative treatment). One can eventually reduce the number of incubation concentrations down to one, provided an excess of ligand concentration (single saturation concentration) is used.
Immunological detection Monoclonal antibodies can detect epitopes of the receptor molecule which are not related to the binding sites, so that the receptor function is not evaluated (for instance, the receptor may be detected even when the binding site is blocked by endogenous ligands or antagonistic drugs). Antibody based assays of receptors provide however many advantages: they are robust since an epitope is generally more stable than the receptor site; sensitivity is excellent; they require a tiny amount of tissue extract (one incubation test), down to 20 mg of tissue or a few thousands cells; - colorimetric measurement is easy and inexpensive. The main flaws are the following: - the binding caracteristics of the receptor are not determined; - they do not provide absolute values of binding sites, and require an external calibration, generally through binding assay. The only alternative to the radioligand calibration problem is the availability of purified or recombinant receptor molecules. Despite the above limitations, immunoenzymatic assays of receptors are rapidly expanding: they are accessible to a greater number of laboratories, thus to a greater number of patients, than radioligand assays. -
-
-
lmmunohistochemical detection The clinical biochemist is faced with the rapid expansion of immunohistochemical methods for the determination of receptors, such as steroid hormone receptors. It requires even less tissue than biochemical assays. For instance, a transcutaneous fine needle aspirate of a breast can be reliably tested. Above all, it allows the direct visualization of the cellular heterogeneity within the tumour tissue. If an accurate evaluation of
142 the percer, tage of positive cells ten be done, the s t a i n i n g i n t e n s i t y is n o t er, sily q u a n t i t a t e d d u e t o the lack of reproducibility. Image analysis can partially overcem,~ this issue. Immunohistochemical determinations of rec e p t o r s f o r c l i n i c a l r o u t i n e a r e still in t h e e a r l y a g e o f a d e v e l o p m e n t , w h i c h is e x p e c t e d t o b e important. Quality control studies and clinical evaluation should be urgently implemented.
References Abelev, G.I. (1968) Production of embryonal serum alpha globulin by hepatomas. Review of experimental and clinical data. Cancer Res. 28, 1344-1350. Zabaleta, E.H. and Mollicone, R. (1986) Distibution of GICA in normal gastrointestinal and endocervical mueosae and in mucinous ovarian cyst using CA19.9. Am. J. Clin. Pathol. 85, 152-159. Bast, R.C.. Feeney, M., Lazarus, H., Nadler. L.M., Colvin, R.B. and Knapp, R.C. (1981) Reactivity of monoclonal antibodies with human ovarian carcinoma. J. Clin. Invest. 68, 1331-1337. Bence-Jones, H. (1847) Papers on chemical pathology. Lecture II1. Lancet 2, 269. Benchimol, S., Fucks, A., Jothy, S., Beauchemin, N., Shirota, K. and Stanners, C.A. (1989) Carcinoembryonic antigen functions as an intercellular adhesion molecule. Cell 57, 327-334. Bieglmayer, C., Szepezi, T., Kopp, B., Hoffmann, G. and Petrik, W. (1991) CAI5.3, MCA, CAM26, CAM29 are members of a polymorphic family of mucin-like glycoproreins. Tumor Biol. 12, 138-148. Bolla, M. and Martin, P.M. (1989) Les Marqueurs Tumouraux. Masson, Paris. Borg, A., Tandon, A., Sigurdsson, H., Clark, G., Fern6, M., Fuqua, A.W., Willander, D. and McGuire, L. (1990) HER2 / n e u amplification predicts poor survival in node-positive breast cancer. Cancer Res. 50, 4332-4337. Burchell, J., Gendler, S., Taylor-Papadimitriou et al. (1987) Development and characterization of breast cancer reactive monoclonal antibodies directed to the core protein of human milk mucin. Cancer Res. 47, 5476-5482. Caravani, M. (1988) Biochemistry and quantification of MCA in serum. Tumor Diag. Ther. 9, 39. Chu, T.M., Co~atantine, R. and Nemoto, T. (1989) Serum level of cryptic ;.umour antigens in breast cancer patients as determined by two monoclonal antibodies (M85/F36) and its comparison with CAI5.3. J. Clin. Lab. Anal. 3, 267-272. Chung, Y.S., Ho, J.J.U, Kim, Y.S. et al. (1987)The detection of human pancreatic cancer associated atttigen in the serum of cancer patients. Cancer 60, 1636-1643. Dudouet,B.. Jacob, L., Beuzeboc, P., Magdel~nat, H. et al. (199fl) Presence of villine, a tissue-specific cytoskeletal protein, in sera of patients and an initial clinical evalua-
tion of its value for the diagnosis and follow-up of colorectal cancers. Cancer Res. 50, 438-443. Duffy, M.I. (1990) Plasminugen activators and cancer. Blood Coag. Fibrinolysis 1,681-687. Ekins, R. (1981) Merits and disadvantages of different labels and methods of immunoassay. In: P. Voller (Ed.), Immunoassays for the 80's. MTP Press, Lancaster, pp. 5-16. Feizi, T. (1981) Carbohydrate differentiation antigens. Trends Biochem. Sci. 6, 333-335. Feizi, T. (1984) Monoclonal antibodies reveal saccharide structures of glycoproteins and glycolipids as differentiation and tumour associated antigens. Biochem. Soc. Trans. 12, 545-549. Feller, W.F. (1991) Mucin glycoproteins as tumour markers. In: R.B. Herberman and D.W. Mercer (Eds.), Immunodiagnosis of Cancer, 2rid edn. Marcel Dekker, New York, pp. 631-672. Garcia, M.B., Blankenstein, M.A., Van der Wall, E., Nortier, J.W.R., Schornagel, J.H. and Thijssen, J.H.H. (1990) Comparison of breast cancer mucin (BCM) and CATS.3 in human breast cancer. Breast Cancer Res. Treat. 17, 69-76. Gendler, S., Taylor-Papadimitriou, Duhig, T., Rothbard, J. and Burchell, J. (1988) A highly immunogenic region of a human polymorphic epithelial mucin expressed by carcinomas is made up of tandem repeats. J. Biol. Chem. 263, 12820-12824. Gendler, S.J., Cohen, E.P., Craston, A., Duhig, T., Johnstone, G. and Barnes, D. (1990) The locus of the polymorphic epithelial mucin (PEM) tumour antigen on chromosome lq21 shows a high frequency of alteration in primary human breast tumours. Int. J. Cancer 45, 431-435. Gerdes, J., Schwab, U. and Lemke, H. (1983) Production of a mouse monoelonal antibody reactive with a human nuclear antigen associated with cell proliferation. Int. J. Cancer 31, 13. Gold, P. and Freeman, S.O. (1965) Demonstration of tumour-specific antigens in human colonic carcinomata by immunological tolerance and absorption techniques. J. Exp. Med. 121,439-445. Gold, D.V. and Mattes, M.J. (1988) Monoclonal antibody B72.3 reacts with a core region structure of O-linked carbohydrates. Tumor Biol. 9, 137-142. Gool, H.C., Uemura, K.-I., Edwards, P., Foster, C.S., Pickering, N. and Feizi, T. (1983) Two mouse hybridoma antibodies against human milk fat globules recognize the ! (MA) antigenic determinant. Carbohydrate Res. 120, 293302. Hakomori, S. (1985) Aberrant glycosylation in cancer in cancer cell membranes as focused on glycoSpids: overview and perspectives. Cancer Res. 45, 24(15-2414. Hakomori, S. (1986) Tumor-associated glycolipid antigens, their metabolism and organization. Chem. Phys. Lipids 42, 209. Hanisch, F.G., Uhlenbruck, G., Dienst, M., Stottrop, M., hippauf, E. (1985) CA125 and CAI9-9: two cancer-associated sialylsaccharide antigens on a mucus glycoprotein from human milk. Eur. J. Bioehem. 149, 323. Harmemberg, U., Wahren, B. and Wiechel, K,-L. (1988) Tumor markers carbohydrate antigens CA19.9 and C~,~;0 and
CEA in pancreatic cancer and benign diseases of the panereatobiliary tract. Cancer Res. 48, 3856-3863. Herlyn, M., Sears, H.F., Steplewski, Z. and Koprowski, H. (1982) Monoctonal antibody detection of a circulating turnout-associated antigen. Presence in sera of patients with colorectal, gastric and pancreatic eareinoma. J. Clin. lmmunobiol. 2, 135-140. Hilkens, J., Buijs, F., Hilgers, J., Hageman, P., Calafat, J., Sonnenberg, A. and Van der VaIL M. (1984) Monoclonal antibodies against human milk fat globule membrane detecting differentiation antigens of the mammary gland and its tumors. Int. J. Cancer 34, 197. Hull, S., Bright, M., Abe, M., Kufe, D. and Carraway, L. (1988) Oligosaceharides of the DF3 antigen of the BT-20 human breast carcinoma line. J. Cell. Biochem. 12E (suppl.), 130-133. Ishiguro, T., Sugitachi, I. and Sakaguchi, H. (1985) Serum alpha protein substractions in patients patients with primary hepatoma or hepatic metastasis of gastric cancer. Cancer 55, 156. Itai, S., Arii, S., Tobe, R. et al. (1988) Significance of 2,3 and 2,6 sialylation of Lewis A antigen in pancreas cancer. Cancer 61,775-787. Itai, S., Nisbikata, J., Yoneda, T., Ohmori, K. and Kannagi, R. (1991) Tissue distribution of 2,3 and 2,6 sialyl Lewis A antigens and significance of the ratio of two antigens for the differential diagnosis of malignant and benign disorders of the digestive tract. Cancer 67, 1576-1587. Johnson, V.G. and Schlom, J. (1986) Analysis of human tumour-assoeiated glycoprotein (TAG72) identified by monoelonal antibody B72.3. Cancer Res. 46, 850. Kallionemi, O.-P., Oksa, H., Aaran, R.-K., Hietaner, T., Lehtinen, M. and Koivula, T. (1988) Serum CA15.2 assa~ in the diagnosis and follow up of breast cancer. Br. J. Cancer 58, 213-215. Kannagi, R., Hakomori, S.I. and Imura, H. (1988) Cancer-associated serum glycoprotein antigens expressing type I and type 2 chain polylactosamines. In: C.L. Reading, S.I. Hakomori and D.M. Marcus (Eds.), Al',ered Glycosylation in Tumor Cells. Alan R. Liss, New York, pp. 279-294. Kato, H. and Moioka, H. (1983) Prognostic significance of the tumor antigc;~ TA-4 in squamous cell carcinoma of the uterine cervix. Am. J. Obstet. Gynecol. 145, 350-354. K6hler, H. and Millstein, C. (1975) Continuous cultures of fused cells secreting antibody of pre-defined specificity. Nature 256, 495. Kufe, D. (1984) Differential reactivity of a novel monoclonal antibody (DF3) with human malignant versus benign breast tumors. Hybridoma 3, 223. Magnani, J.L., Nilsson, B. and Brockhaus, M. (1982) A monoclonal antibody defined antigen associated with gastrointestinal cancer is a ganglioside containing sialylated lactoN-fueopentaose II. J. Biol. Chem. 257, 14367-14369. Manderino, G. (1990) Characterization of the M65 MAb used in the BCM-IMx: satellite symposium, breast cancer mucin - new tumor marker in breast cancer. 3rd Int. Symp. Biol. Clin. Usefulness of Tumor Markers, Madrid, September 1990.
Manssonn, J.E., Frcdman, P., Nilsson, O. et al. (1985) Chemical structure of carcinoma ganglioside antigens defined by monoclonal antibody CAS0 and some allied gangliosides of human pancreatic adenocarcinoma. Biochim. Biophys. Acta 834. 11(I-117. Mant, D. and Fowler, G. (1990) Mass .screening: theory and ethics. Br. Med. J. 3(10, 916-918. McGuire, W.L., Clark G. and Dressier, L. (1986) Role of steroid hormone receptors as prognostic factors in primary breast cancer. NCI Monogr. I, 19-23. Nishi, S. (1970) Isolation and characterization of a human fetal alphaglobulin from the sera of fetuses and a heparoma patient. Cancer Res. 30, 2507. Nowell, P.C. and Hungerfold, D.A. (1960) A minute chromosome in human chorionic granulocytic leukemia. Science 132, 1497. Oikawa, S., Nakazato, H. and Kosaki, G. (1987) Primary structure of human eareinoembryonic antigen (CEA) deduced from cDNA sequence. Biochem. Biophys. Res. Commun. 142 511-518. Paxton, R., Mooser, G., Panoe, H., Lee, T.D., Shiley, J.E. and Terry, H.P. (1987) Sequence analysis of carcinoembryonic antigen: identification of glycosylation sites and homology with the immunoglobulin supergene family. Proc. Natl. Acad. Sci. USA 84, 920-924. Pons-Anicet, D., Krebs, B. and Namer, M. (1987) Value of CAI5.3 in the follow-up of breast cancer patients. Br. J. Cancer 55, 567. Roulston, J.E. (1990) Limitations of tumour markers in screening. Br. J. Surg. 77, 961-962. Sell, S. (1991) Cancer markers of the 1990s: Tumor markers in diagnostic pathology. Clin. Lab. Med. 10, 1-37. Springer, F. (1984) T and Tn, general carcinoma antigens. Science 224. 1198-1206. Sulman, C., Carpentier, P., Lemaire J., Gosselin, P. and Cazin, J.-L. (1987) Seuil de d~cision en biologie. Trait d'Union 5, 23-29. Swallow, D.M., Gendler. S., Griffiths, B., Corney, G., TaylorPapadimitriou and Bramwell, M.E. (1987) The human tumor-associated epithelial mucins are coded by an expressed hypervariable gene locus PUM. Nature, 328, 82. Tempero, M.A., Uchida, E., Takasaki, H. et al. (1987) Relationship of carbohydrate antigen 19.9 and Lewis antigens in pancreatic cancer. Cancer Res. 47, 5501-5503. Thompson, J. and Zimmermann, W. (1988) The carcinoembryonie gene family: Structure, expression and evolution. Tumor Biol. 9, 63. Wallet, V., Mutzel, R., Troll, H., Barzu, O. Wurster, B., W:ron, M. and Lacombe, M.L. (1991) Dictyostelium nucleoside diphosphate kinase is highly homologous to Nm23 and Awd proteins involved in mammalian tumor metastasis and Drosophila development. J. Natl. Cancer Inst. 82, 1199-1202. Watson, D.M., Elton, R.A., Jack, WJ., Dixon, J.M., Cherty, U. and Miller, W.R. (1991) The H-ras oncogene product p21 and prognosis in human breast cancer. Breast Cancer Res. Treat. 17, 161-169.
