cea5, a Structurally Divergent Member of the Murine Carcinoembryonic Antigen Gene Family, Is Exclusively Expressed during Early Placental Development in Trophoblast Giant Cells*

The carcinoembryonic antigen (CEA) gene family encodes a large family of glycoproteins. Some are probably involved in the homeostasis/development of epithelial cells and granulocyte activation, while others e.g. the pregnancy-specific glycoproteins, are expressed in the placenta and are essential for a positive outcome of pregnancy. In this paper, we have characterized cea5, a member of the murine CEA gene family. RNase protection and in situ hybridization analyses revealed that Cea5 mRNA is exclusively synthesized in primary and secondary trophoblast giant cells of the placenta only during early stages of development. Full-length Cea5 cDNA was obtained by a reverse transcription-polymerase chain reaction using day 10.5 post-coitum placental RNA. The 1.6-kilobase pair (kb) Cea5 mRNA encodes a secreted glycoprotein with a predicted size of 30 kDa. It is composed of a leader peptide (L), one immunoglobulin (Ig) variable or N, and one Ig constant-like or A domain. This domain organization is unique within the human and murine CEA families. Two overlapping cosmid clones covering 54 kb of the cea5 gene locus were mapped.cea5 consists of three exons (L, N, A/3′-untranslated region exon) located within a 4-kb region. rnCGM2, the ratcea5 counterpart, exhibits the same restricted expression pattern. This together with their exceptional conservation within the rat and murine CEA families and their absence from the human CEA family suggests that cea5 and rnCGM2 are of functional importance for rodent placental development.

The carcinoembryonic antigen (CEA) 1 family was originally discovered in humans and was named after CEA, the first family member to be discovered and later characterized as a human tumor marker (1). Protein sequencing and gene cloning have since revealed the existence of a CEA gene family that belongs to the immunoglobulin superfamily (2). CEA-related proteins are highly glycosylated and are composed of one immunoglobulin variable (IgV)-like and between zero and six immunoglobulin constant (IgC)-like domains. They are either membrane-bound or actively secreted molecules. The human CEA gene family has been intensively studied and consists of 29 genes, 18 of which are expressible (3). Based on sequence similarities, three subgroups can be discerned. The CEA-subgroup consists of 12 genes, of which 8 are expressed, mainly in epithelial cells and/or in granulocytes (3,4). The pregnancyspecific glycoprotein (PSG) subgroup comprises 11 expressible genes, all of which are active in the syncytiotrophoblast cells of the placenta. A third subgroup contains 6 genes, none of which have been shown to be active. All 29 genes are clustered on the long arm of chromosome 19 (3,5).
Although the in vivo functions of CEA family members are still unknown, several CEA subgroup members have been shown to convey homo-and heterophilic cell adhesion, which could play a role in the development of epithelia during organogenesis (6,7). Some CEA-related molecules appear to be involved in granulocyte adhesion to activated endothelial cells in presenting sialyl-Lewis x carbohydrate structures to E-selectin and in mediating signals for granulocyte activation (8 -10). Ectopic expression in rat myeloblasts revealed that CEA subgroup members might also play a role in the regulation of differentiation (11). Down-regulation of two CEA subgroup members, the biliary glycoprotein (BGP) and CEA gene family member 2 (CGM2), during the development of colorectal carcinomas indicates a causative role of their products in tumorigenesis (12)(13)(14)(15). Indeed, it has been demonstrated that rat and murine BGP homologs can function as tumor suppressors in human prostate cancer cells, in breast tumor cell lines, and in murine colorectal tumor cell lines (16 -18). For one PSG subgroup member (PSG11), it was shown that an arginine-glycineaspartic acid (RGD) tripeptide-containing motif mediates binding to receptors on promonocytic cell lines (19). This indicates a possible role of PSG in regulating the maternal immune system during pregnancy. In addition, murine BGPs (Bgp1 and Bgp2) function as mouse hepatitis virus receptors (20,21).
Rigorous proof is still lacking that the in vitro properties of CEA family members are relevant for normal development and tumorigenesis. As a basis for in vivo functional investigations, e.g. through gene inactivation experiments, the rodent CEA gene family has also been analyzed. Although CEA gene families exist in both the rat and mouse, in comparison with their human counterparts, they vary at both the sequence level and in their domain structures (22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32). These differences make it impossible to determine homologs for most family members between orders and, in some instances, even between rat and mouse species (for a review see Ref. 33). Sequence analyses indicate that prior to mammalian radiation only one or two CEA-like genes existed, which have rapidly expanded through gene duplication events, but independently in the different mammalian orders, leading to these differences (33). Therefore, the CEA gene family represents a paradigm for molecular evolution in progress, both between and within mammalian orders.
Based mainly upon expression studies, the murine gene family can also be divided into a CEA subgroup, consisting to date of 3 genes and a PSG subgroup, where 15 genes have so far been characterized (31). 2 Similar subgroups exist in the rat (33). The only gene homologs so far identified within the human, rat, and mouse CEA gene families are BGP-related genes based on an identical exon arrangement and similar expression pattern (21, 34 -38). In general, orthologous gene pairs between the CEA families cannot be assigned through sequence comparison, even between rat and mouse (33). However, to date one exception is known: the putative counterparts rnCGM2 (rat) and cea5 (mouse), are highly conserved at the amino acid sequence level of their N domains (93% identity), whereas no counterpart exists in humans (26,39). Therefore, these two genes may be rodent-specific and could serve important rodentspecific function(s). Until now, the structure, expression pattern, and function of Cea5 and rnCGM2 remained unknown.
For this reason, we have isolated and characterized a fulllength cDNA for the murine CEA gene family member cea5, for which only partial genomic sequences were so far available (26). This gene encodes a CEA-related protein with a novel domain organization and cannot be unequivocally classified either to the murine CEA or PSG subgroups based on sequence comparisons. As found for PSG genes, cea5 was found to be expressed exclusively in the placenta. However, in contrast to the PSG genes, cea5 is expressed in primary and secondary trophoblast giant cells, whereas PSG transcripts are found in secondary trophoblast giant cells and in the spongiotrophoblast (40,41). Consequently, following the disappearance of the giant cells, cea5 shows an inverse expression kinetics to PSG genes in being predominantly active only during early stages of placental development. Northern blot analyses revealed a similar expression pattern for rnCGM2 in the rat, supporting the idea that cea5 and rnCGM2 represent homologs and are indeed functionally important.

MATERIALS AND METHODS
RNA Isolation, RNase Protection Assay, and Northern Blot Analysis-BALB/c mice (Charles River) or Wistar rats (animal facility of the Institute of Biochemistry and Molecular Biology, University of Freiburg) were anesthetized by inhalation of isoflurane and killed by cervical dislocation. Tissues were immediately frozen in liquid nitrogen after dissection and stored at Ϫ70°C. For mating, male and female mice or rats were kept together overnight. Noon of the next day was designated day 0.5 of gestation. Total RNA was extracted by the acid phenol method (42). For RNase protection analysis the N domain exon of cea5 was amplified by PCR (annealing: 52°C, 15 s; extension: 72°C, 3 min; denaturation: 94°C, 15 s; 30 cycles) using the previously isolated cea5 cosmid cosC3 as template and BK4 (position 103-119) and BK5 (position 461-445; see Fig. 3A) as primers, the sequences of which were based on partial genomic sequence data (26). The resulting 360-bp DNA fragment was subcloned into pBluescript KSϩ (Stratagene), yielding pBS/Cea5N. Single-stranded RNA probes were synthesized with T3 RNA polymerase from either 1 g of the HindIII-linearized pBS/Cea5N or 1 g of the 3.5-kb PstI fragment of pCEA2b (26), which contains part of the A domain and the 3Ј-untranslated region the Cea2 cDNA, in the presence of [ 32 P]UTP (800 Ci/mmol; Amersham Corp.; adjusted to 3 M by addition of unlabeled UTP) using the MaxiScript kit according to the manufacturer's instructions (Ambion, ITC Biotechnology). RNase protection assays (Ambion, ITC Biotechnology) were performed with 10 g of total RNA and 3 ϫ 10 5 dpm of the 418-nucleotide Cea5. A 325-nucleotide ␤-actin RNA probe was added and served as an internal control for sample loading and RNA integrity. The probes contain part of the multiple cloning site region of the vectors. The specific activity of the ␤-actin probe was reduced to 1/10 by addition of unlabeled UTP to compensate for the differences in abundance of ␤-actin and Cea5 mRNA. Hybridization and RNase A/RNase T1 treatment were performed following the manufacturer's protocol. The protected probe fragments were separated on 8 M urea, 6% polyacrylamide gels, which were washed with 10% ethanol, 10% acetic acid and then in 10% ethanol, dried, and exposed to Kodak XAR films.
Northern blot analyses were performed using standard procedures. For detection of Cea5 mRNA, the partial 1.3-kb Cea5 cDNA fragment starting at position 103 and covering the N and A domain and the 3Ј-untranslated region (Fig. 3A) was excised from the pUC18 plasmid (see below) with XbaI and KpnI. As an rnCGM2 probe, a 289-bp Hin-dIII/BamHI genomic DNA fragment comprising about 50 bp of intron 1 and two-thirds of the N domain exon (exon 2) was used (39). Both fragments were labeled with [␣-32 P]dATP (Megaprime TM DNA labeling system, Amersham Corp.). The final wash was performed at high stringency in 0.1 ϫ SSPE (1 ϫ SSPE: 180 mM NaCl, 10 mM sodium phosphate, pH 7.4, 1 mM EDTA), 0.1% sodium dodecyl sulfate (SDS) twice for 20 min at 65°C. The filters were rehybridized with a 32 P-labeled 1.2-kb mouse ␤-actin probe (43) to check for equal loading and the integrity of the RNAs.
The cea5 gene-containing cosmid clone 14.23 was isolated from a C57BL/6 mouse liver cosmid library in pWE15 (Stratagene). 3.6 ϫ 10 5 cosmid clones were plated in duplicate onto Hybond-N ϩ (Amersham Corp.) and hybridized with the 32 P-labeled cea5 N exon probe. The final wash was in 4 ϫ SSPE, 0.1% SDS twice for 20 min at 65°C. The extent of overlap with cosC3 (26) was determined by comparing the sizes of the DNA fragments of 14.23 and cosC3 after digestion with various restriction endonucleases.
Determination of Transcriptional Start Sites-The major transcriptional start sites of cea5 were identified using the 5Ј-RACE system for the rapid amplification of cDNA ends (Life Technologies, Inc.) according to the manufacturer's protocol. Reverse transcription of 1 g of total placental RNA (day 10.5 of pregnancy) was performed with GSP1 (position 555-535 in Fig. 3A) as primer. 5Ј-RACE products were amplified (annealing: 48°C, 30 s; extension: 72°C, 2 min; denaturation: 94°C, 1 min; 35 cycles) using Taq DNA polymerase and GSP2 (position 286 -265) as nested primer in combination with the 5Ј-anchor primer supplied by the manufacturer. Ten percent of the PCR product was reamplified under the same conditions and subcloned into pUC18 as described above. Positive clones were identified by hybridization with 32 P-labeled BK4 (Fig. 3A). Six positive clones were selected at random and partially sequenced.
Sequence Determination and Analysis-Suitable DNA fragments were subcloned into pBluescript. Genomic and cDNA fragments were sequenced as double-stranded templates on both strands with flanking universal and internal oligonucleotides as primers using a T7 polymerase sequencing kit (Pharmacia). The program PC-GENE (IntelliGenetics) was used to calculate the degree of similarity of the amino acid sequences after optimal alignment.
In Situ Hybridization-In situ hybridization analyses were performed essentially as described before (44), except that the RNA probes were not fragmented by hydrolysis. Embryos and placentae of day 9.5 of development were embedded in toto in Tissue Freezing Medium (Jung) diluted 1:1 with water by freezing in ethanol/dry ice. The tissues were cryosectioned (7-m sections) and placed on SuperFrost/Plus microscope slides (Roth). RNA probes were synthesized in the presence of digoxigenin-labeled UTP (DIG Labeling Kit; Boehringer Mannheim) after subcloning of a 0.6-kb EcoRI Cea5 cDNA fragment (EcoRI cuts in the 3Ј-untranslated region as well as in the oligo(dT) primer used for cDNA synthesis) into the pBluescript vector and linearization of the plasmid with either XhoI (antisense probe) or BamHI (sense probe) using T7 and T3 RNA polymerase, respectively.

RESULTS
Expression Pattern of Cea5-To be able to clone Cea5 cDNA, we identified Cea5 mRNA-containing tissues. To this end, we used an RNase protection assay, which allows the specific detection of Cea5 mRNA in the presence of mRNAs of other closely related CEA family members. A complementary 32 Plabeled RNA of 418 nucleotides was hybridized with total RNA from various organs of adult BALB/c mice, including those known to express other members of the murine CEA family, like colon, small intestine, and bone marrow. The probe was derived from the N domain exon (360 nucleotides) of the previously isolated cea5 gene (26) and is flanked by sequences from the multiple cloning site of the vector which was used as template for probe synthesis. The expected 360-nucleotide protected fragment could not be detected with RNA from any of the adult tissues tested (Fig. 1A, lanes 1-8; Fig. 1B [3][4][5][6][7][8][9][10][11][12][13][14][15]. Since the CEA-related PSGs are expressed in placenta, we also assayed total RNA from placentae of different developmental stages for the presence of Cea5 mRNA. Indeed, the protected Cea5 probe fragment of the expected size was obtained with RNA from placentae of early development. The largest amounts were found in placentae of day 10.5 of embryonic development (E10.5). The steady state levels of Cea5 mRNA declined continuously during pregnancy and were barely detectable at E15.5 (Fig. 1C, lanes 3 , 5, 7, 9, 11, and 13). On the other hand, RNAs from maternal decidual and uterine tissues as well as from day 13.5 fetuses were negative (Fig. 1C,  lanes 4, 6, 8, 10, 12, and 14).
Expression of rnCGM2, the Rat Cea5 Counterpart-We also investigated the expression pattern of rnCGM2 (39), the presumed cea5 homolog of the rat, to assess the functional significance of cea5 expression in placenta as well as the orthologous relationship of the two genes. Northern blot analysis using part of exon 2 of the rnCGM2 gene as a probe revealed that rnCGM2 exhibits an expression pattern very similar to that found for cea5. An rnCGM2 mRNA species of about 1.1 kb was detected predominantly in E12.5 and smaller amounts in E10.5 and E17.5 rat placentae (Fig. 2, lanes 2, 5, and 7). Fainter signals (relative to the ␤-actin content) were seen with total RNA from Total RNA (10 g) of the fetal or maternal part (decidua including the uterus) of placentae of various gestational stages, from E12.5 embryos as well as from a number of tissues from adult rats, were size-fractionated by electrophoresis on a 0.9% agarose gel in the presence of 1.1 M formaldehyde, transferred to a nitrocellulose membrane, and hybridized with a 32 P-labeled rnCGM2 exon 2 probe. The blot was rehybridized with a mouse ␤-actin cDNA fragment to demonstrate the intactness and loading of the RNAs. The presence of rnCGM2 mRNA in decidual RNA is probably due to incomplete dissection of the fetal trophoblast tissue from maternal tissue.
Cloning of Cea5 cDNA-Initial experiments failed to amplify Cea5 mRNA by polymerase chain reaction (PCR) from E10.5 placental RNA after reverse transcription (RT) using the Cea5specific primer BK3 (Fig. 4A) from the putative 5Ј-untranslated region of cea5. We, therefore, employed a Cea5-specific primer (BK4) from the 5Ј-end of the N exon (Fig. 3A) and a (dT) 14 containing 3Ј-anchor primer to amplify Cea5 cDNA. A DNA fragment of about 1.3 kb was obtained and sequenced after subcloning.
To amplify the complete coding region of the Cea5 mRNA, an oligonucleotide (M8-ko) was synthesized based on the sequence of the 3Ј-untranslated region of the partial Cea5 cDNA clone and used together with an oligonucleotide (DF4; Fig. 3A) located closer to the putative translational start of cea5 than BK3. The composite Cea5 cDNA sequence is shown in Fig. 3A. Cea5 mRNA codes for a putative mature protein (after removal of the leader peptide of 33 amino acids) of 201 amino acids (M r ϭ 22,582) and is composed of one IgV-like N domain and one IgC-like A domain (Fig. 3, A and B). No hydrophobic domain capable of membrane anchorage is present. The amino acid sequence of the A domain is most closely related to the A2 domains of the murine Bgps (e.g. Bgp1 A2: 57% identity) and less to the A domains of the murine PSGs (e.g. Cea2 A: 48% identity). The least similarity (30% identity) is observed with the A1 domain of Bgp1 (Fig. 3D). The Cea5 protein is predicted to contain two potential N-glycosylation sites which would re- Oligonucleotides, which were derived from partial genomic sequences (26) and used for amplification of Cea5 cDNA fragments, are indicated above the sequence. The location of the EcoRI site used to construct the probe for in situ hybridization is also shown. B, the domains of the coding region of Cea5 mRNA are depicted as boxes. The number of amino acids are given below. The vertical arrow marks the polyadenylation signal sequence. C, total placental RNA (10 g) of the indicated developmental stages was size-fractionated and processed for Northern blot analysis as described in the legend to Fig. 2. The RNA was hybridized with a 32 P-labeled Cea5 cDNA fragment (covering exons 2 and 3) as a probe. The blot was rehybridized with a mouse ␤-actin cDNA fragment. D, the amino acid sequences of the indicated domains were compared, and the fraction of identical amino acids is given in %. Sequence sources: Bgp1 a , Bgp2 (34); Cea2 (26) sult in an M r of 28,000 -30,000 assuming a mean increase in the molecular mass of 3,000 -4,000 per carbohydrate addition site (45). The major transcriptional start site of cea5 was determined by PCR (5Ј-RACE) and found to be located at position 30 of the Cea5 cDNA (as estimated by the frequency of the 5Ј-RACE clones obtained starting at this site), only 12 nucleotides upstream of the putative translational start site of Cea5 (Fig. 3A). This unusually short 5Ј-untranslated region is caused by a deletion as evidenced by comparison with the corresponding region of cea10, another member of the murine CEA gene family ( Fig. 4A; 31). A minor start site must be located in the vicinity of the region between the BK3 and DF4 primers used for PCR amplification of Cea5 cDNA (Fig. 4A), since only DF4 yielded an amplification product. Indeed, transcriptional start sites of members of the human CEA gene family (e.g. CEA and NCA) have been mapped to the corresponding regions (26). A prominent Cea5 mRNA species with a size of 1.6 kb migrating slightly below the 18 S rRNA could be detected in placenta using the 1.3-kb cDNA fragment as a probe (Fig. 3C). An additional faint hybridization signal corresponding to a 3.5-kb RNA species migrating between the 28 S and 18 S rRNAs was observed. As expected from the RNase protection experiments (Fig. 1C), the steady state levels decline during placental development (Fig. 3C). No cross-reaction was observed with PSG mRNAs, which are increasingly abundant during later stages of development.
Cea5 Is Expressed in Trophoblast Giant Cells of the Placenta-To identify the cell type that expresses Cea5 mRNA, we performed in situ hybridization analyses on uteri of BALB/c mice at day 9.5 of gestation. Only trophoblast cells were labeled when an antisense probe derived from the 3Ј-untranslated region of the Cea5 mRNA was used (Fig. 5A). Both primary and secondary trophoblast giant cells, which are readily identifiable by their large nuclei as well as their location, were stained (Fig. 5, A and B). Primary trophoblast giant cells are located distally to the region of invasive trophoblast growth, and secondary giant cells are found more proximal to this region (46). Besides these cells, a trophoblastic cell population with smaller nuclei directly underneath the layer of secondary trophoblast giant cells was also labeled (Fig. 5B). In contrast, fetal spongiotrophoblast cells were negative (Fig. 5B). The embryo, the maternal part of the placenta (decidua), and other maternal tissues did not exhibit any hybridization signal. No signal was obtained with a corresponding sense probe (Fig. 5C).
Organization of the Cea5 Locus-The cosmid clone cosC3, where the leader and N exon of the cea5 gene have already been identified (26), was used to characterize the complete exon/ intron structure of cea5. Sequence analysis of genomic EcoRI/ BamHI DNA fragments downstream of the N domain exon revealed the presence of one additional exon that encodes the A domain and the 3Ј-untranslated region ( Fig. 6A; Table I). The transcription unit of cea5 is relatively small and comprises 4 kb. Sequence comparison of the A domain exon of cea5 with the corresponding exons of other members of the CEA family revealed a potential splice donor site directly upstream of the stop codon of cea5 which is similar to the one found at the 3Ј-end of the A2 domain exon in the murine and human BGP genes (Fig. 6B). The latter is used to join the A2 domain to the transmembrane domain exon of the murine bgp1 (34). The homologous site in the human BGP gene can be used for splicing or read-through yielding membrane-bound and secreted BGP isoforms, respectively ( Fig. 6B; Refs. 36 and 47). We,

FIG. 4. Comparison of the transcriptional start site sequence (A) and the amino acid sequence of the N domain exon (B) of cea5 and other rodent CEA gene family members.
A, the nucleotide sequences (numbered at the right margin) upstream of the boxed start codon of cea5 and cea10 (31) were aligned. Dashes were inserted to allow for optimal alignment. Identical nucleotides are indicated by asterisks. The transcriptional start sites of cea5 are marked by an arrow. The region homologous to the footprint 1 of the human CEA gene promoter (52) is underlined in the cea5 nucleotide sequence. The consensus binding sequence for the transcription factor, upstream regulatory factor (USF) is underlined in the cea10 sequence. The positions of two oligonucleotides used for the amplification of Cea5 cDNA are marked. Since Cea5 cDNA could be amplified by RT-PCR using the DF4 but not with the BK3 oligonucleotide as 5Ј-primer, additional transcriptional start site(s) must be located between the positions of the two oligonucleotides. B, the amino acid sequences in one-letter code (numbered at the right margin) deduced from the exon 2 nucleotide sequences of cea5 and its rat homolog rnCGM2 (39) were aligned and identical amino acids marked by asterisks. The presumed start of the mature protein is indicated by an arrow. The amino acids (R and D) possibly involved in formation of the salt bridge are marked by a bracket. The non-conservative amino acid changes in the Cea5 sequence in the close neighborhood of the salt bridge are indicated by ϩ and Ϫ above the sequence.
FIG . 5. Expression of cea5 in trophoblast cells. Unfixed uteri of pregnant BALB/c mice (day 9.5 of development) were cryosectioned and hybridized in situ with a digoxigenin-labeled antisense (A and B) or with a corresponding sense Cea5 RNA probe as a negative control (C). The hybridized RNA probe was visualized by reaction with an antidigoxigenin alkaline phosphatase-conjugated antibody and subsequent incubation with 5-bromo-4-chloro-3-indolyl phosphate and nitro blue tetrazolium salt which results in a dark blue staining of Cea5 mRNA-containing cells. A, only fetal primary (pgi) and secondary trophoblast giant cells (sgi) are stained. e, embryo; d, decidua; mv, dilated maternal blood vessel; uw, uterine wall. B, at higher magnification of a different E9.5 placenta, an additional positive cell population probably of trophoblastic origin located between the secondary trophoblast giant cells (sgi; arrowheads) and the spongiotrophoblast (sp) is visible. They are characterized by their smaller nuclei (arrows). C, no staining was seen on a near-adjacent section with a sense probe. Scale bars, 1 mm (in A); 0.4 mm (in B and C). therefore, searched for the presence of potential mRNA splice variants of cea5 in placental RNA by RT-PCR using various oligonucleotides from the N and A domain region in combination with the oligo(dT) anchor primer. No additional PCR products to the one described above were found (data not shown).
To analyze whether cea5 is closely linked to other members of the CEA gene family we screened a BALB/c cosmid library for overlapping clones using the N domain exon sequence as a probe. One clone (14.23) with an insert of 41 kb was identified and shown to extend 20 kb further upstream (Fig. 6A). Digestion of 14.23 DNA with BamHI, EcoRI, HindIII, and SstI, respectively, and subsequent hybridization with the exon 2 Cea5 probe (see above) under non-stringent conditions did not reveal additional N domain-related exons (not shown). Therefore, no Cea-related gene seems to be present in the 5Ј-vicinity of cea5.   (48) of members of the murine CEA family revealed that cea5 and rnCGM2 are old genes, the exons of which diverged from bgp and from the PSG subgroup ancestors more than 65 million years ago, around the time of primate and rodent separation. The expansion of the CEA and PSG subgroups started some 20 million years later (33).
The slightly higher similarity of the Cea5 A domain to the A2 domains of the murine bgp genes, bgp1 and bgp2, would suggest assignment to the CEA subgroup. However, no murine or human CEA subgroup member is exclusively expressed in placenta as is cea5. This is characteristic for members of the PSG subgroup of both species (33,41). In contrast to PSGs, Cea5 mRNA levels decrease during placental development. This is also found for cea10 which codes for a soluble member of the murine CEA subgroup and is closely related to bgp1 and bgp2 (31). This decline of expression follows the disappearance of the trophoblast giant cells located at the maternal/fetal interphase (49) which indicates expression of cea5 (and cea10) in this cell population. Indeed, in situ hybridization studies confirm expression of cea5 in primary and secondary trophoblast giant cells and in a trophoblast cell population with smaller nuclei located in the vicinity of the secondary trophoblast giant cells. Thus, Cea5 mRNA complements the set of known markers of trophoblast development (50). In contrast, rodent PSGs are expressed in the secondary but not in the primary trophoblast cells and additionally in spongiotrophoblast cells (40). In contrast to cea5, cea10 is also transcribed in the gastrointestinal tract and bone marrow, which might be due to the presence of an upstream stimulatory factor (USF) binding motif (CACGTG). This motif is present in all members of the murine and human CEA family that are expressed in intestinal epithelia and has been shown to bind USF and to be important for the expression of the human CEA and BGP genes (51,52). There is no conserved USF motif in the homologous promoter region of cea5 (Fig. 4A), which supports the hypothesis that USF is involved in directing expression to intestinal epithelial cells (31) but is not necessary for trophoblast expression.
The chromosomal location of cea5 also stresses its uniqueness within the murine and human CEA gene families. It is displaced several centimorgans, probably as a single gene (no CEA-related gene has been identified in the immediate neighborhood (Fig. 6A)) from the region where other CEA subgroup members (cea10/bgp1) reside on the centromere-proximal region of mouse chromosome 7 (53). The latter region is syntenic to the human CEA/PSG locus at 19q13.2, where all gene family members are clustered within a region of about 1 megabase pair (54). It is interesting to note, that in mice the PSG gene cluster also became detached from the CEA subgroup (53). Taken together, cea5 seems to represent a separate entity, belonging to neither the CEA nor the PSG subgroup.
Function of Cea5-The function of Cea5 is still unclear. It probably differs, however, from that of the placentally expressed PSGs for two reasons. First, there is no RGD-like motif in the N domain of Cea5 (33), and second, Cea5 exhibits inverse expression kinetics and reveals a different expression pattern with regard to the cell population during placental development ( Fig. 1C and Fig. 5 and Refs. 40 and 41). Cea5-like molecules seem to be important for both mouse and rat based on the conservation of the N domain amino acid sequence in the rat Cea5 homolog, rnCGM2 (39), which has 93% of its amino acid sequence in common with Cea5 (26). This extremely high conservation is unparalleled in other rat and murine CEA family members. In comparison, Bgp1 a /Bgp1 b and their rat homologs C-CAM1 a /C-CAM1 b share only 49 -54% of their N domain amino acid sequences. Interestingly, the few amino acid substitutions observed in Cea5 and rnCGM2 represent non-conservative replacements of uncharged amino acids by charged ones. However, they probably do not lead to major structural changes because they seem to be neutralized by an opposite charge in their presumed direct spatial neighborhood (Fig. 4B). The extremely high conservation of Cea5 and rnCGM2 together with their apparent rodent specificity and their similar expression patterns indicates a function that is exclusively needed during early placental development in rodents. Indeed, a number of structural and functional differences exist between rodent and primate placentae, necessitating differences at the molecular level. The study of placentally expressed members of evolutionary young gene families, like the CEA gene family, is expected to further our understanding of the molecular mechanisms involved in the establishment of structural and functional differences in mammal placentation.