Carcinoembryonic antigen-related cell adhesion ...

Cancer Metastasis Rev DOI 10.1007/s10555-013-9444-6

NON-THEMATIC REVIEW

Carcinoembryonic antigen-related cell adhesion molecules (CEACAMs) in cancer progression and metastasis Nicole Beauchemin & Azadeh Arabzadeh

# Springer Science+Business Media New York 2013

Abstract The discovery of the carcinoembryonic antigen (CEA) as a tumor marker for colorectal cancer some 50 years ago became the first step in the identification of a much larger family of 12 carcinoembryonic antigen-related cell adhesion molecules (CEACAMs) with surprisingly diverse functions in cell adhesion, in intracellular and intercellular signaling, and during complex biological processes such as cancer progression, inflammation, angiogenesis, and metastasis. The development of proper molecular and biochemical tools and mouse models has enabled bidirectional translation of the CEACAM network biology. Indeed, CEACAM1, CEACAM5, and CEACAM6 are now considered valid clinical biomarkers and promising therapeutic targets in melanoma, lung, colorectal, and pancreatic cancers. These fascinating proteins illustrate how a better understanding of the CEACAM family of cell adhesion molecules reveals their functional link to the underlying disease and lead to new monitoring and targeting opportunities. N. Beauchemin (*) : A. Arabzadeh Goodman Cancer Research Centre, McGill University, Montreal, Canada e-mail: [email protected] A. Arabzadeh e-mail: [email protected] N. Beauchemin Department of Biochemistry, McGill University, McIntyre Medical Sciences Building, 3655 Promenade Sir William Osler, Lab 708, Montreal, QC, Canada H3G 1Y6 N. Beauchemin Department of Medicine, McGill University, McIntyre Medical Sciences Building, 3655 Promenade Sir William Osler, Lab 708, Montreal, QC, Canada H3G 1Y6 N. Beauchemin Department of Oncology, McGill University, McIntyre Medical Sciences Building, 3655 Promenade Sir William Osler, Lab 708, Montreal, QC, Canada H3G 1Y6

Keywords CEACAM1 . Carcinoembryonic antigen . CEA . CEACAM6 . Cell adhesion molecule . Biomarker . Immunity . Immunotherapy

1 Introduction It is now almost 50 years since a young medical fellow by the name of Dr. Phil Gold working in the laboratory of Dr. Samuel Freedman initiated a research project to question whether tumor-specific antigens could be identified in colorectal carcinoma (CRC). Using a number of immunological techniques, they uncovered that, indeed, extracts from both primary and metastatic colorectal cancers inhibited precipitin activity formed between the tumor extract and antibodies raised to such extracts. In addition to the carcinomas, precipitin activity was also unveiled in gut, liver, and pancreas tissues of 2– 6 months human fetuses, and thus, they gave this marker the name “carcinoembryonic antigen” (CEA) [1]. Further purification of tumor extracts by Mach and Pusztaszeri and von Kleist et al. indicated that another antigen displaying neither tumor nor organ specificity was present in greater amounts in intestinal carcinomas; it was designated “nonspecific crossreacting antigen (NCA or CD66c)” [2, 3]. The fact that the CEA glycoprotein was detectable in human serum from 85 % of colon carcinomas suggested the possibility of large patient population screening using radioimmunoassays [4]. Over the next years, the number of CEA cross-reacting antigens accumulated, as did tumor sites with elevated CEA-like activity. The heterogeneity of CEA preparations raised serious issues regarding the uses and limitations of available radioimmunoassays in the clinical management of patients. However, it soon became recognized that serial elevation of CEA levels after curative surgical resection in colorectal cancer patients was indicative of a secondary local recurrence or distant metastatic disease [5]. Thus, CEA immunoassays soon became standard practice in the management of postoperative

Cancer Metastasis Rev

(termed A or B) [15]. These extracellular domains are required for CEACAM functionality as homophilic and heterophilic intercellular adhesion molecules [17] or as human and rodent pathogen receptors [18, 19]. Human CEACAM16 within the CEACAM subgroup, which associates as oligomers, is secreted and is involved in hearing impairment [20]. The PSG subgroup encompasses mostly secreted CEACAM-like proteins in species with hemochorial placentation [16, 21] that are implicated in fetal–maternal recognition and maternal immune responses to fetal invasion during development [22, 23]. Thus, CEACAM proteins associated as dimers and oligomers may multiply associations with other partners at the membrane and consequently modulate important functions. The focus of this review is to highlight the roles of the CEACAMs in tumor progression and metastasis, and we have, therefore, chosen to restrict the discussion to CEACAM1, CEACAM5 (CEA), and CEACAM6 as they are the best characterized in cancer processes. More information regarding the functions of other family members is available in published reviews [18, 24, 25].

surveillance of colorectal cancer patients. Over the next decade, the characterization of CEA-related protein sequences and the generation of diverse monoclonal antibodies (mAbs) to CEA and CEA-like proteins [6] finally led to the cloning of the CEA cDNA by a number of research groups [7–10]. The availability of CEA cDNA sequences and cDNA libraries derived from colorectal cancer cells or tissues then led to the cloning of a number of other CEA-related family members such as NCA [11] and biliary glycoprotein [12–14], later renamed CEACAM6 and CEACAM1, respectively. This in turn led to the identification of a large family of carcinoembryonic antigen cell adhesion molecules (CEACAM) proteins in humans encoded by 22 separate genes divided among the CEACAM (12 independent genes; Fig. 1) and pregnancy-specific glycoproteins (10 PSG) subgroups clustered on human chromosome 19q13.1–13.2. The nomenclature of this gene family had become very confusing over the years and was unified in 1999 [15]. In addition to their expression in human tissues, the CEACAM gene family is highly conserved in 27 other mammalian species and is best described in mouse, rat, cattle, dog, platypus, and opossum [16]. All CEACAM proteins belong to the immunoglobulin (Ig) supergene family and generally exhibit one variable (V)-like domain, identified as the N domain (except CEACAM16 that contains two N domains) (Fig. 1). The N domain is followed by either none or up to six constant C2-like Ig domains

1.1 Structures with similarities and differences The CEA (CEACAM5) cDNA encodes a protein that exhibits one variable (V)-like domain, identified as the N domain, followed by three repeating units comprising, in total, six

N A1 N

B1

N N

B2 A3 N

N

B3

N

N A B

N A

A B

Fig. 1 Human CEACAM family members. Twelve members of the human CEA family, which belong to the CEACAM subgroup, are depicted here. CEACAM proteins generally have one variable (V)-like Ig domain, identified as the N domain (except CEACAM16 with two N domains) (blue), but they differ in the number of constant C2-like Ig domains, identified as A (pink) or B (white), as well as the membrane anchorage. CEACAM5, CEACAM6, CEACAM7, and CEACAM8 are associated with the membrane through a GPI linkage, whereas six

A1 N A

B A2

A1 A B N

A2

N

B1

A

A2

B

N

B2

CEACAM family members (CEACAM1, CEACAM3, CEACAM4, CEACAM19, CEACAM20, and CEACAM21) are anchored to the cellular membrane via bona fide transmembrane domains. CEACAM16 is a secreted version with no membrane anchorage. The CEACAM1 cytoplasmic domain has ITIM motifs (red circles), whereas CEACAM3, CEACAM4, CEACAM19, and CEACAM20 carry ITAM motifs (blue circles). All family members are highly glycosylated proteins highlighted by the stick and balls on the extracellular domains


constant C2-like domains (termed A1, B1, A2, B2, A3, and B3; Fig. 1) [8–10]. The Ig domains are preceded by a 34amino-acid signal peptide. In addition, the CEA cDNA structure predicted 12 cysteine residues forming the core of the Ig folds and 28 N-linked glycosylation sites in line with the high carbohydrate content of the purified protein [7, 10]. Anchorage to the membrane is a distinctive feature of CEACAM proteins. CEACAM5 and CEACAM6 are associated with the membrane through a glycosylphosphatidylinositol (GPI) linkage. A stretch of 26 hydrophobic amino acids comprises the predicted CEA membrane anchor, later demonstrated as corresponding to a GPI anchor [26] as CEA was released from the cell surface upon treatment with phosphatidylinositolspecific phospholipase C [27]. The CEA GPI link evolved from the transmembrane attachment of the primordial CEACAM1 gene through mutation within its transmembrane-encoding exon [28]. Importantly, this represents a recent evolutionary event in this gene family as none of the GPI-linked CEACAMs are found in the mouse genome [16]. Six CEACAM family members (CEACAM1, CEACAM3, CEACAM4, CEACAM19, CEACAM20, and CEACAM21) are anchored to the cellular membrane via bona fide transmembrane domains. The CEACAM1 primary transcript is subjected to alternative splicing generating 12 different human isoforms (http://www.carcinoembryonic-antigen.de/human/ human-ceacam1-splice-variants.html; Fig. 2). Other than the addition of one, two, or three C2-like domains, CEACAM1 alternative splicing is also particularly active in generating two major cytoplasmic domains, termed the long (-L) and short (-S) tails either by inclusion or exclusion of exon 7, respectively. The presence of the long or short cytoplasmic tails is a regulated event in normal tissues and depends on both cisacting regulatory elements within the flanking introns and exons, as well as decreased expression of hnRNP A1 and L and overexpression of hnRNP M [29]. However, L and S tail expression becomes dysregulated in colorectal, breast, and non-small cell lung cancers [30–32]. CEACAM1-L is best known for its two cytosolic phosphotyrosine residues and immunoreceptor tyrosine-based inhibitory motifs (ITIM) signaling motifs implicated in the regulation of several functions that will be described in greater detail within this review. Of the 12 different CEACAM1 isoforms, 3 are secreted versions and play a significant role in the inhibition of intercellular adhesion, particularly in TAP2-deficient patients expressing high amounts of soluble CEACAM1 (sCEACAM1) in their serum [33] or in diseased situations such as obstructive jaundice [34]. Importantly, secreted CEACAM1, as discussed further, is a marker of melanoma, pancreatic, and urothelial bladder cancer progression [35–37]. Except for very similar N domains, characteristic membrane attachments, and the number of glycosylation sites, what differentiates these family members lies in the number of C2-like domains projecting from the membrane with CEA

bearing six and CEACAM6 bearing only two, whereas the most common forms of CEACAM1 will generally exhibit two or three such C2-like domains (Figs. 1 and 2), resulting from alternative splicing. Other than the CEA A3/B3 domains connecting to the reciprocal N domain of another CEA molecule as cell adhesion binding domains [38], the C2-like domains of CEACAM1, CEA, and CEACAM6 mediate the association to other cellular receptors involved in a phosphatidylinositide 3-kinase (PI3K)-elicited bacterial uptake [19]. This implies a number of possible cis or trans direct interactions of CEACAM proteins through their extracellular domains with a variety of membrane receptors, a few of which have been identified. These might contribute to the diversity of functions ascribed to these molecules. 1.2 The diversified expression pattern of CEACAM1, CEACAM5, and CEACAM6 Although the CEACAM family members were discovered decades ago and many specific mAbs have been raised to these proteins [6, 39], their very similar structures precluded proper assignment of their expression patterns in both normal and tumor tissues until recently. The availability of mammalian cells transfected with most of the CEACAM gene family members and generation of mutant proteins facilitated proper definition of their respective functions [40]. As its designation implies, CEA is present early in human embryonic and fetal development (weeks 9–14); additionally, it maintains its expression throughout life [41, 42]. In normal adult tissues, it is localized in the stomach, tongue, esophagus, cervix, sweat glands, and prostate [43]. Its main residence site is in columnar epithelial and goblet cells of the colon, particularly in the upper third of the crypt and at the free luminal surface [43]. Following malignant transformation, CEA is also detected in carcinomas of the lung, small cell lung cancer, pancreas, gallbladder, urinary bladder, mucinous ovarian, and endometrium [43] and is significantly overexpressed in colorectal [44] and gastric [45] carcinomas. Thus, 50 years after its initial discovery, CEA remains an excellent marker to identify colorectal tumors relative to their matched normal tissue (in 98.8 % of cases) and to detect positive lymph nodes [46]. In addition, in breast cancer, CEA expression enables the stratification of patients at risk of relapse with high discrimination, regardless of E-cadherin expression in these tumors [47]. CEACAM6 is more widely distributed than CEA in normal tissues, with significant expression in many epithelia, as well as in granulocytes and monocytes [48]. Its deregulation was first noticed in leukocytes of chronic myeloid leukemia [49] and in childhood acute lymphoblastic leukemias (of B cell origin) [50]. CEACAM6 overexpression in colonic hyperplastic polyps and early adenomas was also deemed to be the earliest molecular change observed in these lesions [48] and predicts


N N

N N

N

A1 N

A1

A1

B

A1

B

B

A1

A2

B

A2

A2

B

A1

N

N

B

A1

A1

A2

B

B N

N

N

A1

A1

B

B

N

Fig. 2 Human CEACAM1 isoforms. Alternative splicing adds to the complexity of CEACAM proteins. CEACAM1 transcripts can be alternatively spliced in order to generate 12 different isoforms containing 1, 2, or 3 C2-like domains as well as 2 major cytoplasmic domains, termed as long and short tails. According to standardized nomenclature, the number

after CEACAM1 is indicative of the number of extracellular Ig-like domains, while the letter following this number points to the presence of either a long (L) or short (S) cytoplasmic tail, a unique terminus (C), or an Alu family repeat sequence (A) (black boxes). Graphic domain features and glycosylation sites are the same as in Fig. 1

poor overall survival, allowing the subdivision of patients into low-risk and high-risk groups [51]. In fact, CEACAM6 overexpression in colorectal cancer cells increases invasiveness which correlates with its highest expression in liver CRC metastases; thus, it could also represent an excellent tumor biomarker [52]. CEACAM6 expression is more abundant than CEA in breast, pancreatic, mucinous ovarian, gastric, and lung adenocarcinomas, whereas its abundance is similar in prostate cancer to its normal counterpart [53]. In summary, it now appears that CEACAM6 might be the most specific marker of this large protein family for a number of aggressive cancers. CEACAM1 is the most widely distributed protein within the gene family, being present in different epithelia, on endothelial cells, as well as in lymphoid and myeloid cells in normal tissues [43, 54, 55]. In normal colon epithelia, CEACAM1 is present in the functional mid-crypt compartment of colonic mucosa [54]. In polarized Mardin–Darby canine kidney cells, the transfected CEACAM1-L isoform was present on both the apical and basolateral cell surfaces [56]. Localization of CEACAM1-L at the lateral surface of the epithelium required the expression of Tyr515 within its cytoplasmic domain, and CEACAM1-L Tyr phosphorylation induced its rapid PI3Kdependent movement to the endosome/lysosome compartment [57]. CEACAM1-S was present only at the apical pole of the cells [56]. In tumor tissue, the expression of CEACAM1 is very dynamic: the protein isoforms are considerably reduced in the early phases of many cancers including colon [58, 59], prostate [60], liver [61], and breast [62] cancers. This downregulation within the epithelial compartment is an early event during disease progression as it occurs in hyperplastic lesions [60, 63]. In fact, reinsertion of various CEACAM1 isoforms in

colorectal or prostate CEACAM1-negative tumor cells demonstrated that CEACAM1-L expression was essential for the maintenance of a normal phenotype with the inhibition of allograft or xenograft tumor development in syngeneic or immune-deficient mice [64, 65]. These results suggested that, indeed, CEACAM1-L behaved as a tumor suppressor protein. However, CEACAM1-L overexpression in other types of aggressive cancers such as melanoma [66], non-small cell lung [67], gastric [68], thyroid [69], and bladder [70] cancers indicated that, on the contrary, tumors with a high abundance of CEACAM1-L correlated with metastatic spread [71]. Notably, recent analyses in advanced and metastatic colon cancer with isoform-specific antibodies highlighted that a high CEACAM1-L/CEACAM1-S ratio was associated with lymph node involvement, hematogenous metastasis, and shorter survival of patients, thus redefining a role for CEACAM1-L in CRC invasion [72] in spite of its growth inhibitory properties in early stage cancers. A comprehensive model implicating CEACAM1-L in the angiogenic switch, as highlighted in prostate cancer, reconciles these diverse expression patterns and broadens CEACAM1-L functions, as discussed further in this review [73].

2 CEACAM5 (CEA), a major player in tumor progression and metastasis 2.1 CEA in primary tumors Although CEA is the “founding” member of this large family of proteins and commonly used in the clinic, knowledge of


its function and signaling is limited. It has been a recognized CRC tumor marker for the last 50 years, but novel imaging techniques and targeting strategies continuously rephrase the question as to whether CEA remains a suitable biomarker for identifying CRC patients with respect to tumor sensitivity and specificity. In 2003, Jantscheff et al. indicated that, in contrast to its closely related CEACAM6 cousin, CEA expression on CRC primary tumors of patients had no prognostic value and could not predict overall survival of patients based on relative intratumoral CEA expression levels [51]. However, a recent study has again evaluated CEA relative to candidate CRC markers such as tumor-associated glycoprotein-72 (TAG-72), epidermal growth factor receptor (EGFR), and folate receptor-α. Whether comparing tumor versus normal expression of antigens or assessing lymph node expression in a cohort of 280 CRC patients, CEA was by far the most sensitive biomarker (93.7 sensitivity), as well as being highly specific (96.1 %) [46]. These differential assessments of CEA might be due to antibody specificity (the latter used a A5B7 humanized anti-CEA antibody validated in clinical trials of targeted therapies) or enhanced recent imaging systems. In addition, a recent study using the Surveillance Epidemiology and End Results database (SEER—a large nationwide population-based database in the USA), including 17,910 patients, clearly demonstrated that elevated preoperative serum CEA levels is an independent predictor of overall survival in patients with colon cancers, regardless of clinical stage [74]. Consequent with findings in human patients, transgenic CEA mouse models have been used to define whether the expression of CEA bears any consequences on tumor development. Eades-Perner et al. and Clarke et al. essentially used the same transgene to produce mice that expressed normal levels of CEA under its natural promoter [42, 75]. CEA expression mimicked that observed in humans, and no primary tumors developed throughout the normal life span of the animals. None of the CEA-positive compound mice demonstrated a higher tumor burden relative to those that were CEAnegative in three different models of cancer, namely, breast cancer regulated by the rat Neu oncogene under the transcriptional control of the mouse mammary tumor virus, intestinal cancer as defined through the ApcMin/+ mice, and finally, lung cancer driven by the surfactant C gene under the SV40 large T antigen promoter. Thus, CEA expression does not appear to influence tumorigenesis [76]. On the other hand, in another available mouse model (the CEABAC model), tumor development results were more mitigated: this transgenic mouse was developed using a large 187-kb bacterial artificial chromosome transgene that encompassed all of the regulatory and coding regions of the human CEACAM5, CEACAM6, and CEACAM7 genes. A number of animal models are available to study colorectal cancer [77]. One of these relies on the induction of colon cancer with the colon-specific carcinogen azoxymethane; in this setting, CEABAC mice doubled the

number of colonic adenomas and adenocarcinomas relative to their wild-type FVB littermates with no changes in tumor volumes [78]. It remains possible that the differences seen between the various mouse models might be due to the different genetic backgrounds: the CEA transgenics were generated on a C57Bl/6 background but still contained BALB/c genetic information, whereas the CEABAC model was on an FVB background, known to be more susceptible to carcinogen-induced tumorigenesis [79]. Importantly, in the CEABAC model, CEACAM6 expression was 20-fold higher and CEA was 2-fold higher in tumors relative to levels observed in normal colonocytes. This discrepancy is not an unusual finding [80], but it remains to be demonstrated whether the higher susceptibility of CEABAC mice was not due to the increased CEACAM6 overexpression rather than that of CEA. Unfortunately, no independent CEACAM6 transgenic mice are yet available to establish a worthwhile comparison. Transfection of tumor cells with the CEA cDNA immediately suggested a potential function for this protein. Indeed, Stanners’ group defined that CEA behaves as an intercellular adhesion molecule connecting adjacent epithelial cell membranes, particularly in both embryonic intestine and colonic tumors (Fig. 3a). This function is inhibited by treating cells with anti-CEA mAbs [81]. CEA-mediated intercellular adhesion occurs in an antiparallel reciprocal manner through interactions between the N and A3B3 domains [38, 82], which is unique in this family. The A3B3 adhesion subdomains have not been well characterized. In spite of its highly glycosylated structure with 28 asparagine-linked glycosylation sites, these posttranslational modifications do not appear to play a significant role in cell adhesion, other than modifying adhesion strength [83]. In addition to its function as a cell adhesion molecule, CEA plays a significant role in other cellular processes including the inhibition of differentiation programs [84, 85], inhibition of anoikis and apoptosis in colon cells [86], and disruption of cell polarization and tissue architecture [85]. CEA regulates these activities by activating integrin signaling pathways in lipid raft subdomains where it colocalizes and clusters with the α5β1 integrin, thus triggering integrin-linked kinase, PI3K and AKT activities [87]. As discussed previously, CEA expression on epithelial cells may directly influence tumor development by CEA–CEA bridges between tumor cells or tumor–stromal cells. In addition, it may do so indirectly through distinct features that contribute to its recognition by the immune system. Colon cancer-associated CEA, in contrast to that of normal colon, presents high expression of the blood group-related carbohydrates, Lewis X and Lewis Y, known ligands for the dendritic cell-specific intercellular adhesion molecule 3-grabbing nonintegrin (DC-SIGN) (Fig. 3b). Differential binding of the PHA-L lectin also suggested that CRC-elicited CEA had increased branched N-glycans on its structure; these correspond


a

d

c

b

CEAR N A1 B1 A2 B2 A3 B3

hnRNP M (CEAR)-CEA interaction

A3

A1

B2

B1

A2

A2

B1

B2

A1

A3

N

B3

A3 B2

d. Kupffer cell

IL-1 TNFICAM-1, V-CAM, E-selectin

A2 B1 A1 N

N

B3 B3

N

c. NK

A2 CEA-CEA bridges

CEA

B2 A2 B1 A1 N A1

A1 B1

A3

N

B3

B

EC CEACAM1

A2

B2 A3 B3

Heterophilic interactions of CEA with DR5 and TGF- R1

b. DC

Cell adhesion, decreased anoikis, resistance to TGF -mediated growth inhibition

a. Tumor cell DC-SIGN-CEA interaction Tolerance induction

CEACAM1CEA-interaction Inhibition of NK cell cytotoxicity

Fig. 3 Various CEA interactions engage different immune and nonimmune cell types and receptors to support tumor progression and metastasis. Membrane-bound CEA on a cancer cell can directly engage a cell surface receptor on a target cell (depicted in a, b, and c), while secreted CEA can act in a paracrine manner stimulating secretion of pro-tumorigenic and pro-metastatic cytokines (illustrated in d). a CEA– CEA bridges between tumor cells or tumor–stromal cells can directly influence tumor development. In addition, heterophilic association of CEA with death receptor 5 (DR5) or TGF-βR1 on tumor cells leads to decreased anoikis or resistance to TGF-β-mediated growth inhibition, respectively. These effects enhance the survival of tumor cells and, consequently, increase metastasis. b CEA carbohydrates, Lewis X and Lewis Y, function as ligands for the DC-SIGN expressed on the surface of dendritic cells (DCs). This interaction has been suggested to induce

DC tolerance to tumor cells. c CEA can also interact with CEACAM1 expressed on the surface of NK cells. Such heterophilic binding via N domains of CEA and CEACAM1 results in MHC-independent inhibition of NK cell killing abilities. d CEA secreted by colon tumor cells is recognized by a putative CEA receptor (CEAR) identified as heterogeneous nuclear ribonucleoprotein M (hnRNP M) (orange Y) on the surface of liver macrophages (Kupffer cells). CEA–CEAR binding activates Kupffer cells, inducing them to secrete pro-inflammatory cytokines (e.g., IL-1β, TNF-α). These cytokines in turn act on sinusoidal endothelial cells (ECs). Hepatic ECs then upregulate the expression of a number of cell adhesion molecules such as ICAM-1, VCAM-1, and E-selectin that in turn increase the binding of circulating tumor cells to the endothelium and favor metastatic development

to DC-SIGN and galectin-3 ligands. Thus, it is not surprising to observe a twofold increase in CEA binding to these lectins [88]. Intratumoral immature DCs, but not peripheral mature DCs, interact with tumor cells by binding of the DC-SIGN lectin to the CEA carbohydrates, possibly to suppress DC functions [89]. It is postulated that DC-SIGN-positive immature DCs are incapable of priming strong T cell responses and may be involved in tolerance induction to the CRC cells by virtue of CEA engagement by DC-SIGN. Additionally, since DC-SIGN is an internalization receptor on immature DCs [90], secreted tumor CEA may be taken up by these cells and peptides presented to T cells, a finding which may have consequences for the development of better CEA vaccines [89].

2.2 CEA as a modulator of anticancer immunity Examining the adhesion binding domains of CEA for therapeutic intervention has led to some pertinent findings. In addition to the CEA–CEA-mediated homophilic interactions described earlier, heterophilic CEA–CEACAM1 binding via their N domains produces functional interactions resulting in MHC-independent inhibition of NK cell killing abilities. In addition to the N domains of either protein, the presence of the A and B domains of CEA increase the intensity of CEA to CEACAM1 binding [91] (Fig. 3c). Blocking the binding of CEA to CEACAM1 on purified human NK cells via a CEA-specific humanized PR1A3 mAb recognizing the


membrane-bound CEA form produced extensive killing of human CRC cells expressing this molecule. Importantly, lysis occurred very efficiently with the naked antibody through antibody-dependent cell-mediated cytotoxicity (ADCC); this approach could, therefore, become an effective cytotoxic therapy for CRC [92]. Another anti-CEA mAb named CC4 exhibiting similar properties has also been described [93]. Another therapeutic approach actively pursued in the last 20 years is the development of CEA-based antitumor vaccines [94]. Various formulations have been tested over the years, some based on the presentation of predicted CEA epitopes by purified DCs or others on the expression of the full-length CEA molecule included in replication-defective recombinant viruses such as canarypox or fowlpox combined to replication-competent recombinant vaccinia viruses. These procedures are deemed to be safe for patients and, although administered to very advanced patients, there has been documentation of disease stabilization in some cases and even objective responses in others [94]. Addition of T cell co-stimulatory molecules to the viruses such as B7-1, intercellular adhesion molecule-1 (ICAM-1), and lymphocyte function-associated antigen-3 (LFA-3), designated as TRICOM, induced strong T cell responses specific to CEA, leading to increased survival of patients tested [95, 96]. The quality of the preparations of mature DCs have been improved using various combinations such as OK432 consisting of Streptococcus pyogenes preparation, prostanoid, and interferon-α (OPA-DCs) [97] or Toll-like receptor-activated DCs which induce more effective antitumor immune responses [98]. A noticeable problem results from the suboptimal CEA T cell epitopes usually presented; most of these represent short sequences within the central portion of the CEA molecule [99]. To counter this less favorable selection, a recent approach has been to vaccinate mice bearing MC38-CEA-expressing tumors with a recombinant form of the IgV-like N domain (residues 1–132). This retarded growth of subcutaneously implanted tumors precluded the development of lung metastases and permitted a strong immune response with antibodydependent tumor lysis by ADCC and complement-dependent cytotoxicity (CDC), thereby representing a worthwhile novel approach for future therapies [100]. The same group has recently reported another successful approach in that two DNA aptamers capable of inhibiting CEA-mediated homophilic adhesion through its N domain were also effective at blocking MC38-derived tumor cell implantation in vivo without affecting immune responses [101]. 2.3 CEA as a marker of metastasis Colorectal cancer initiation, progression, and metastasis has been associated with a number of genetic alterations implicating 40 % mutations of the K-Ras oncogene or activation

mutations of the serine/threonine–protein kinase B–Raf (BRAF) and 75 % deletions in the p53 tumor suppressor protein. Several pathways downstream of EGFR, such as mitogen-activated protein kinase (MAPK)/MEK, PI3K/AKT, and JAK/STAT, mediate functions in proliferation, apoptosis, adhesion, and differentiation [102]. CEA has been used as a CRC tumor marker for decades and, although its use remains controversial relative to CRC detection, a large-scale analysis of the literature has revealed that preoperative CEA expression levels between 50 and 200 μg/ml constitutes one of the 11 best determinants identified by 5 major studies as independently influencing the prognosis of patients undergoing liver resection for CRC metastases [103]. Indeed, as discussed extensively in a recent review by P. Thomas, CEA plays an active role in CRC liver metastasis development [104] (Fig. 3d). Jessup et al. have clearly demonstrated that, irrespective of its function as a homophilic cell adhesion molecule, pretreatment of mice with soluble CEA favors enhanced development of experimental liver metastatic nodules in nude mice, even when using weakly metastatic CRC cells negative for CEA expression [82]. This is due to the fact that CEA binds to a putative CEA receptor (CEAR) (Fig. 3d) identified as heterogeneous nuclear ribonucleoprotein M (hnRNP M) at the surface of liver Kupffer cells (Fig. 3d) through interactions with a PELPK motif present at the hinge region between the CEA N and A1 domains. Activation of the liver-specific Kupffer macrophages through CEA–CEAR binding leads to pro-inflammatory cytokine secretion (Il-1α and IL-1β, IL-6, TNF-α) in the hepatic sinusoid (Fig. 3d) [105, 106]. These in turn upregulate a number of cell adhesion molecules such as ICAM-1, vascular cell adhesion molecule-1 (VCAM-1), and E-selectin on the adjoining endothelium, thus increasing the binding of circulating colon cancer cells (Fig. 3d) [107]. Arrested tumor cells in the microvasculature normally cause ischemic injury with concomitant increases in nitric oxide (NO) and reactive oxygen species, but those expressing CEA stimulate the release of the antiinflammatory cytokine interleukin-10 (IL-10) responsible for the inhibition of hepatic injury cytotoxicity, therefore increasing the survival of weakly metastatic cells exhibiting enhanced metastasis [106]. In addition, direct association of CEA with the DR5 receptor (TRAIL-R2) (Fig. 3a) expressed on CRC cells leads to decreased anoikis (cell death upon detachment from the matrix) and, as a consequence, increased metastasis [108]. Therapeutic approaches to block CEA expression by means of immunotherapy or vaccines are consequently very much warranted to decrease metastatic tumor load in patients. In addition to CEA’s metastatic signaling abilities through the DR5 receptor, CEA has recently been hypothesized to enhance liver metastasis through direct binding to the type I transforming growth factor-β receptor 1 (TGF-βR1) (Fig. 3a) [109]. Given that CRC cells become resistant to TGF-β-


mediated growth inhibition but can acquire a pro-invasive phenotype in response to TGF-β stimulation [110], the association of CEA and TGF-βR1 imprints a strong synergistic effect in liver metastasis (Fig. 3a). The association of CEA and TGF-βR1 impairs TGF-β-stimulated SMAD3 activity. Inhibiting CEA activities through either a CEA-specific antibody or using siRNA-mediated silencing of CEA reinstated TGF-β–pSMAD3-mediated inhibition of proliferation in various CRC cell lines. In summary, CEA on its own functions as a homophilic or heterophilic adhesion molecule or, in association with signaling receptors such as DR5 receptor and TGF-βR1, can influence either epithelial CRC cells or the surrounding stromal and immune compartments to shift their signaling programs in order to support metastasis progression. As such, monitoring disease evolution with CEA as a biomarker or exploiting CEA as a target for immunotherapy or vaccinebased therapeutics has come of age.

3 CEACAM6, a prognostic indicator and a metastatic target 3.1 CEACAM6 in primary cancers and metastasis A wealth of information has been made available in the last few years on the high potential of CEACAM6 as a prognostic marker and therapeutic target. Jantscheff et al. first examined whether the expression of CEACAM6, CEACAM1, and CEA on 243 paraffin-embedded samples of colorectal cancer patients treated with adjuvant 5-fluorouracil-based chemotherapy was predictive of outcome [51]. Although CEACAM6 expression was enhanced in only 55 % of patients relative to CEA found in 94 % of patients, multivariate Cox analyses clearly indicated that CEACAM6 overexpression was an independent predictor of poor overall survival (p