| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
,1
* Department of Experimental Immunology, Academic Medical Center, University of Amsterdam, Amsterdam, the Netherlands;
Department of Biochemistry, Radboud University Nijmegen, Nijmegen, the Netherlands; and
Departments of Comparative Genetics and Refinement, and Animal Science, Biomedical Primate Research Centre, Rijswijk, The Netherlands
2Correspondence: Department of Experimental Immunology, K1–144, Academic Medical Center, University of Amsterdam, P.O. Box 22700, 1100 DE Amsterdam, The Netherlands. E-mail: j.hamann{at}amc.uva.nl
ABSTRACT
The epidermal growth factor (EGF)-TM7 receptors CD97, EMR1, EMR2, EMR3, and EMR4 form a group of adhesion class heptahelical molecules predominantly expressed by cells of the immune system. These receptors bind cellular ligands through EGF-like domains, localized N-terminal to a large extracellular region. Remarkably, EMR2 possesses a chimeric structure with a seven-span transmembrane (TM7) region most related to EMR3 and an EGF domain region nearly identical to CD97. By comparing EGF-TM7 receptors in primates and dogs, we identified an intriguing pattern of concerted evolution, apparently mediated by gene conversion, among EMR2 and the oppositely orientated and physically adjacent genes CD97 and EMR3. This concerted evolution has continuously maintained the chimeric structure of EMR2 since early mammal radiation. Most highly conserved between EMR2 and CD97 is the fourth EGF domain, which mediates binding to chondroitin sulfate, a ligand specificity shared by both receptors. Another ligand, CD55, is bound effectively only by CD97. We show that different molecular mechanisms (mutations vs. alternative splicing) prevent CD55 binding by EMR2 in hominoids. Our findings illustrate how various and partially opposing evolutionary events have shaped the structure and ligand specificity of a modern mammalian gene family.—Kwakkenbos, M. J., Matmati, M., Madsen, O., Pouwels, W., Wang, Y. Y., Bontrop, R. E., Heidt, P. J., Hoek, R. M., Hamann, J. An unusual mode of concerted evolution of the EGF-TM7 receptor chimera EMR2.
Key Words: adhesion class • gene conversion • ligand specificity • mammals
THE EPIDERMAL GROWTH FACTOR (EGF)-TM7 FAMILY (1)
is a group of adhesion class seven-span transmembrane (TM7) receptors (2
, 3)
predominantly expressed by hematopoietic cells. With the completion of the human genome project, this family was found to comprise five members: CD97 (4
, 5)
, EMR1 (6)
, EMR2 (7)
, EMR3 (8)
, and EMR4 (9)
. EGF-TM7 receptors are characterized by a large extracellular region with, N-terminally, several tandemly arranged EGF-like domains that bind cellular ligands (1)
. CD97, EMR1, and EMR2 undergo alternative RNA splicing, which leads to the production of isoforms that possess variable numbers of EGF domains. A G-protein-coupled receptor proteolytic site (GPS) proximal to the first transmembrane domain gives rise to autocatalytic processing within the endoplasmic reticulum (10)
. Translated as single polypeptides, EGF-TM7 receptors are cleaved into an extracellular 
subunit and a TM7/cytoplasmic ß subunit, which noncovalently associate at the cell surface. Recent functional studies have linked these molecules to leukocyte trafficking, angiogenesis, and generation of peripheral tolerance (11
12
13)
.
Various observations provide evidence that evolution of the EGF-TM7 family is a recent and ongoing process, particularly in primates. 1) EGF-TM7 receptor genes have been identified so far only in vertebrate genomes. 2) Human genes are located in close proximity within clusters in chromosome 19p13.1 (CD97, EMR2, and EMR3) and 19p13.3 (EMR1 and EMR4) (1)
. 3) Surveys of rodent genomes failed to identify orthologs for EMR2 and EMR3 (1
, 14)
. 4) Human EMR4 is inactive due to a one-nucleotide deletion in exon 8 that is not present in the great apes (9)
. 5) Molecular cloning of human EMR2 unraveled a remarkable homology with two other human EGF-TM7 receptors (7
, 8)
. As shown in Fig. 1
, the TM7 region is most similar to that of EMR3 while the signal peptide, the EGF domains, and the most N-terminal part of the stalk region strongly resemble that of CD97. Only 6 of 236 amino acids within the EGF domain region are different.
|
The striking homology of EMR2 with CD97 and EMR3 raises questions about the evolutionary history and the biological consequences of its chimeric structure; 97% amino acid identity within the EGF domain region between EMR2 and CD97 (7)
suggests the necessity for exchange of DNA sequences between the encoding genes during late primate evolution (15)
, as recently reported for a member of the Siglec family (16)
. Despite their unusual similarity, EMR2 and CD97 only partially share ligand specificity. Both receptors bind the glycosaminoglycan chondroitin sulfate (CS) (17)
via EGF domain 4, which is identical between EMR2 and CD97 in humans. In contrast, CD97, but not EMR2, binds CD55 (decay accelerating factor) (7
, 18)
, a glycosylphosphatidyl inositol-linked molecule that prevents complement deposition (19)
. Three altered amino acids within EGF domain 1 and 2 account for this altered ligand specificity (20)
. Assuming a regional sequence homogenization between EMR2 and CD97 in hominoids, modification of the ligand specificity of EMR2 could only have occurred afterward and so might contribute to immunity-related differences between humans and the great apes (21
, 22)
. We investigated the evolutionary history of EMR2 in relation to its ligand binding properties using comparative genomics (23)
and molecular assays. Surprisingly, we discovered a unique mode of long-lasting concerted evolution of EMR2 with CD97 and EMR3 and observed that different molecular mechanisms have shaped ligand specificity of EMR2 in hominoids.
MATERIALS AND METHODS
Animals
Primate species investigated within this study include chimpanzee (Pan troglodytes), orangutan (Pongo pygmaeus), rhesus macaque (Macaca mulatta), Hamadryas baboon (Papio hamadryas), common marmoset (Callithrix jacchus), and cotton-top tamarin (Saguinus oedipus). Peripheral blood from healthy animals and genomic DNA samples were obtained from the Biomedical Primate Research Centre (Rijswijk, The Netherlands). Peripheral blood from a healthy Beagle dog (Canis familaris) was obtained from the Central Laboratory Animal Institute (Utrecht, The Netherlands).
DNA amplification and sequence analysis
The following nucleotide sequences deposited earlier into the GenBank were used: human CD97 (NM_078481), mouse CD97 (NM_011925), rat CD97 (XM_341662), cow CD97 (NM_176661), pig CD97 (NM_213925), human EMR2 (NM_013447), and human EMR3 (NM_032571).
Genomic sequences of EMR2 from various primates were amplified by polymerase chain reaction (PCR) (35–40 cycles with 30 s at 93°C, 30 s at 58–60°C, and 20 s at 72°C) using 100–500 ng genomic DNA per reaction as template and specific primers derived from the human sequence (Supplemental Table 1). PCR products were separated on a 1.2% agarose gel, purified, and sequenced using the BigDye terminator cycle sequencing kit (Applied Biosystems, Warrington, UK).
The cDNA sequence of rhesus macaque EMR2 was derived by a combination of RT-polymerase chain reaction (RT-PCR) and 5'- and 3'-rapid amplification of cDNA ends (RACE) using the Smart RACE cDNA amplification kit (Clontech, Palo Alto, CA, USA). For RT-PCR, total RNA was isolated from peripheral blood mononuclear cells (PBMC) and transcribed into first-strand cDNA using Superscript II reverse transcriptase (Invitrogen, Breda, The Netherlands). PCR amplification was done with specific primer combinations derived from the human sequence as well as those specifically designed for rhesus macaque EMR2, for 35 cycles of 30 s at 93°C, 30 s at 58°C, and 60 s at 72°C. Sequences of the amplification products were determined. Unfinished sequences of CD97 and EMR3 were derived from the preliminary genome assembly (version 0.1) of the rhesus macaque that became available during preparation of this manuscript (http://www.hgsc.bcm.tmc.edu/projects/rmacaque). The nucleotide and deduced amino acid sequences were compared with primate homologs with the ClustalW software (http://www.ebi.ac.uk/clustalw).
Canine EGF-TM7 receptor genes were identified by a basic local alignment search tool (BLAST) survey (http://www.ncbi.nlm.nih.gov/BLAST) of genomic contigs recently made available by the Broad Institute (Cambridge, MA, USA). The mRNA sequences of CD97, EMR1, EMR2, EMR3, and EMR4 were obtained by computational analysis using BLAST and GENSCAN (http://genes.mit.edu/GENSCAN.html) algorithms in combination with molecular-biological techniques such as RT-PCR and 5'- and 3'-RACE (see above). The deduced amino acid sequences were analyzed with ClustalW.
Alternatively, spliced sequences that encode the EGF domains of CD97 and EMR2 in dog, human, and chimpanzee were amplified by RT-PCR using primers specified in supplemental Table 1. PCR products were separated on a 1.5% agarose gel, photographed, and sequenced. The 5' end of exon 3 in chimpanzee CD97 and EMR2 and the sequence encoding exon 5 in orangutan and Hamadryas baboon were separately amplified using additional primers.
Phylogenetic analyses
Nucleotide alignments were produced using ClustalW and adjusted manually for optimal orthology. Positions lacking in > 25% of the compared sequences were excluded. We separately compared the first two EGF domains (267 bp), the lower part of the stalk region (648 bp), and the TM7 region (723 bp). EGF domains 3 to 5 of CD97 and EMR2 were excluded from the comparison to preclude effects from domain number (EMR3 has only two EGF domains). The N-terminal part of the stalk region, encoded by exons 8 and 9 in EMR2 and CD97 and by exon 4 in EMR3, respectively, was excluded because this region is nearly identical in CD97 and EMR2. The best fitting model for each data set was selected using the AIC criterion in MODELTEST 3.7 (24)
and were TVMef+G for the EGF domains, TrN+I+G for the stalk region, and GTR+G for the TM7 region. The phylogenetic criterion was maximum likelihood and performed with PAUP4.0b10 (25)
. A neighbor-joining tree was used as the starting tree and the tree bisection-reconnection option was used to swap branches. Node stability was estimated with 100 replicates of nonparametric bootstrapping.
Statistical tests
Kishino-Hasegawa (26)
and Shimodaira-Hasegawa (27)
statistical tests were performed in PAUP4.0b10 (25)
with full optimization (FullOpt command) and 1000 bootstrap samplings to evaluate the following hypothesis: No concerted evolution; concerted evolution between CD97 and EMR2; concerted evolution between CD97 and EMR3; concerted evolution between EMR2 and EMR3. For each hypothesis, the best maximum likelihood tree and likelihood score were calculated and used for statistical tests. The best maximum likelihood tree was calculated by constraining the a priori hypothesis but leaving all other sequences free in the heuristic search [e.g., the constrain for no concerted evolution in EGF domains (see Supplemental Fig. 1 for abbreviations): (hCD97, mCD97, rCD97, bCD97, cCD97, pCD97) (hEMR2, rhEMR2, cEMR2, cEMR2-like 1) (hEMR3, rhEMR3, cEMR3)]. The trees used in the statistical analyses are provided as supplemental Fig. 1.
Differences in the conservation of EGF domains between homologues from different mammalian orders were calculated from amino acid alignments, generated in ClustalW, with the 1-way ANOVA test. Where the available sequences allowed for more than one comparison between two orders, the mean value was used for calculation.
Flow cytometry
Flow cytometry was performed by standard procedure on a FACSCalibur (Becton Dickinson, San Jose, CA, USA). Whole blood from human or chimpanzee was incubated with biotinylated monoclonal antibody (mAb) CLB-CD97/3 (binds to stalk region of CD97) (28)
, 2A1 (binds to stalk region of EMR2) (29)
, CLB-CD97/1 (binds to EGF domain 1 of CD97 and EMR2) (18)
, 1B5 (binds to EGF domain 4 of CD97 and EMR2) (17)
, or control IgG. Streptavidin-APC (PharMingen, San Diego, CA, USA) was used as second reagent. Prior to cytometry, erythrocytes were lysed using FACS lysing solution (Becton Dickinson).
Generation of Fc fusion proteins
Fc fusion proteins were generated by placing the N-terminal part of EGF-TM7 receptors immediately upstream to the CH3-CH2 hinge region sequence of mouse IgG2b linked to a C-terminal biotinylation sequence (30)
. Generation of human CD97 and EMR2 Fc constructs has been reported (17)
. Constructs encoding the largest isoform of chimpanzee CD97 and EMR2 were made by PCR amplification of chimpanzee PBMC first-strand cDNA using the specific primers 5'-AAGCTTCCATGGGAGGCCGCGTCTTTCT-3' [a HindIII site is in italics], and 5'-CTGGAGTCCACAGCCAGACGCTTGCGGCCGC-3' [a NotI site is in italics]. A construct encoding the middle isoform of dog EMR2 was generated by PCR amplification of dog PBMC first-strand cDNA with the specific primers 5'-GCTGGTACCAGATGGCATCCTCTGTG GCTG-3' [a KpnI site is in italics] and 5'-ACGAATTCCTGAGCCAAGGCTGGCTTG-3' [an EcoRI site is in italics]. Using the introduced restriction sites, PCR products were cloned in a pcDNA3.1/Neo(+) vector containing a mouse Fc sequence, followed by a recognition sequence for the Escherichia coli biotin holoenzyme synthetase BirA (kindly provided by Dr. M. Stacey, Sir William Dunn School of Pathology, Oxford, UK).
HEK 293 cells were transfected with the different constructs and after 4 days conditioned Optimem 1 medium (Life Technologies Ltd., Paisley, Scotland) containing soluble recombinant protein was purified using a protein A (Sigma, St. Louis, MO, USA) column. After purification, recombinant protein derived from four 225 cm2 cell culture flasks was concentrated to 
0.5 ml using a 30 kDa cutoff filter (Millipore, Bedford, MA, USA), dialyzed against 10 mM Tris-HCl (pH 8) buffer, and incubated with BirA enzyme and supplied substrates (Avidity, Denver, CO, USA) overnight at room temperature. Excess biotin was subsequently removed by dialyses against 10 mM Tris-HCl (pH 7.3) buffer containing 10 mM CaCl2 and 150 mM NaCl. The biotinylated proteins were then aliquoted and stored at –80°C after quantification by Bradford assay.
Ligand binding studies
Cell binding assays using biotinylated CD97- and EMR2-Fc proteins coupled to fluorescent beads were performed as described (30)
. Briefly, 10 µl avidin-coated fluorescent beads (Spherotech Inc., Libertyville, IL, USA) were washed with PBS/0.5% BSA and incubated with saturating amounts (>1 µg) of biotinylated recombinant protein in a volume of 10 µl. After 1 h, nonbinding protein was removed by washing with PBS/0.5% BSA. The bead-protein complexes were sonicated immediately before addition to the cells (0.5x106 cells/50 µl PBS/0.5% BSA). For blocking studies, 1 µg of mAb was added to the bead-protein complexes and incubated for 10 min at 4°C before adding complexes to the cells. Cell-bead mixture in a 96-well, flat-bottom plate was spun at 1000 g at 4°C for 10 min, incubated for another 50 min at 4°C, then resuspended in 300 µl of PBS for flow cytometric analysis.
Human splenocytes (31)
and two mutants of Chinese hamster ovary (CHO) cells were used for ligand binding studies. In a mutant CHO cell line [PgsB-618 from American Type Culture Collection (ATCC), Manassas, VA, USA], glycosaminoglycan synthesis is mostly absent (32)
, whereas another mutant (PgsD-677; ATCC) lacks heparan sulfate but not other glycosaminoglycans (33)
. To generate cells that express CD55 but not CS, PgsB-618 cells were transfected with full-length CD55 cDNA in pcDNA3 (Invitrogen). After selection with G418 (Invitrogen) at 500 µg/ml, stable clones were tested for CD55 expression by flow cytometry with the mAb CLB-CD97L/1, which binds to the first short consensus repeat of CD55 (18)
. One positive clone was enriched for CD55-expressing cells by a single sort on a FACSAria (Becton Dickinson) using the anti-CD55 mAb CLB-CD97L/1. Spleen cells were treated with chondroitinase ABC (Sigma) and CLB-CD97L/1 as described recently (31)
.
RESULTS
EMR2 developed through concerted evolution
To trace the evolutionary history of EMR2, we started by assessing different primate genomes. Performing PCR with primer combinations that amplify sequences of low homology with other EGF-TM7 receptors [exon 11 (stalk region) and exon 19 (cytoplasmic tail)], we could detect EMR2 in chimpanzee, orangutan, and rhesus macaque (data not shown). Identification of the complete cDNA sequence of rhesus macaque EMR2 did not reveal essential differences with human EMR2 (Supplemental Fig. 2). 88% amino acid identity and a comparable regional similarity with CD97 and EMR3 (data not shown) imply that the chimeric structure of EMR2 originates prior to hominoid evolution.
We next searched for EMR2 in phylogenetically more distant genomes by performing BLAST surveys of the currently known mammalian genome drafts. In the dog genome, we identified EMR2 together with CD97, EMR1, EMR3, EMR4, and two EMR2-like genes on chromosome 20. The complete cDNA sequence of all canine EGF-TM7 receptors was determined (Supplemental Table 2). Comparison of human and canine EMR2 revealed an amino acid identity of 70% (Supplemental Fig. 3). Unexpectedly, regional homology of EMR2 with CD97 and EMR3 in the dog closely matches the situation in humans (Fig. 2
A). As in humans, the five EGF domains of EMR2 and CD97 in the dog are >95% identical (Fig. 2B
) compared with 70% amino acid identity between the human-canine orthologs. For the TM7 region, intra- and interspecies homology are in the same range (85% amino acid identity). The two EMR2-like genes likely arose by duplications of EMR2. Impaired EGF domain regions call their functionality into question (Supplemental Table 2).
|
An "inverted " homology by which paralogs are more similar than orthologs is mostly due to the exchange of DNA sequences within gene families, also known as concerted evolution (34
, 35)
. To test the possibility that EMR2, CD97, and EMR3 in mammals did not evolve independently, we performed a phylogenetic analysis of the nucleotide sequences encoding the EGF domains, the lower stalk region, and the TM7 region. We included all available sequences that have been confirmed experimentally so far. Maximum likelihood trees for the compared sequences are shown in Fig. 3
. These trees strongly support concerted evolution of EMR2 and CD97 for the EGF domain region and, more weakly, of EMR2 and EMR3 for the TM7 region in primates. The stalk region, in contrast, evolved monophyletically. We next evaluated for all sequences the following a priori hypotheses: 1) no concerted evolution, 2) concerted evolution between CD97 and EMR2, 3) concerted evolution between CD97 and EMR3, and 4) concerted evolution between EMR2 and EMR3. As depicted in Table 1
, concerted evolution of CD97 and EMR2 is the only reasonable explanation of the high similarity of both receptors in the EGF domain region. For the TM7 region, there is no significant statistical support for concerted evolution of EMR2 and EMR3 in primates and dog, but it is the most likely explanation. In support of this notion, we observed that some introns in the TM7 regions are also highly conserved, with up to 80% nucleotide identity between different intronic sequences (data not shown). Comparisons of the lower stalk region clearly support independent evolution. These findings imply that DNA transfer between EMR2 and CD97 occurred frequently and recently in both primates and carnivores, while DNA transfer between EMR2 and EMR3 has been less frequent and less recent. In conclusion, EMR2 has coevolved in a regional mode and exchanged DNA in a different frequency with CD97 and EMR3. This process of concerted evolution has been essentially the same in the ancestors of humans and dogs since their divergence during mammal radiation 94 million years ago (36)
.
|
|
Two mechanisms of concerted evolution have been proposed: unequal crossing-over and gene conversion (34)
. Gene conversion, defined as nonreciprocal exchange of genetic information, is very precise and requires little identity in flanking sequences (37
, 38)
. The absence of homologous flanking sequences (data not shown) indicates that EMR2 has evolved by gene conversion. This assumption is also supported by the opposite transcriptional orientation of EMR2 and CD97 in humans, which leaves gene conversion as the only option (39)
. The EMR2 gene consists of 20 exons (7)
. The EMR2/CD97 conversion tract includes the region from exon 1 to exon 9 (in dogs exon 2 to exon 9), while the EMR2/EMR3 conversion tract overlaps with the region from exon 14 to exon 18 (data not shown). In humans, 700 bp upstream of the start codon are also nearly identical between EMR2 and CD97. The parts of EMR2 that encode the lower stalk region (exon 10 to exon 13) and the cytoplasmic tail (exon 19 to exon 20) evolved independently from other EGF-TM7 receptor genes as indicated by a much higher inter- as opposed to intraspecies homology (Fig. 3
and Table 1
).
The fourth EGF domain of EMR2 and CD97 is highly conserved in mammals
The EGF domains of EMR2 and CD97 mediate interactions with other cell surface molecules, implying the possibility that ligand binding formed the driving force for the coevolution of EMR2 and CD97. If true, sequence conservation should be especially high in EGF domains that harbor ligand binding sites. We thus systematically compared the EGF domains of CD97 and EMR2 between species from different mammalian orders (Fig. 4
A). This comparison clearly revealed EGF domain 4, which mediates CS binding in primates, to be significantly more conserved than any other EGF domain.
|
Based on the high conservation of EGF domain 4, we predicted that canine EMR2 and CD97 will also interact with CS. Because the composition of the EGF domain region of both receptors in humans is determined through alternative RNA splicing (5
, 7)
, we first examined the composition of isoforms, expressed in the dog, by RT-PCR and sequence analysis. As shown in Fig. 4B
, the dominantly expressed isoform of both receptors contains EGF domains 1, 2, 4, and 5, and thus might potentially interact with CS. To test the functionality of the CS binding site in the canine receptors, we used an approach that has been instrumental in the investigation of EGF-TM7 receptor-ligand interactions (17
, 20
, 31)
. A multivalent fluorescent probe loaded with recombinant Fc protein of the extracellular part of the dominant, middle isoform of canine EMR2 (dog CD97 has an identical fourth EGF domain) was generated and tested for binding to glycosaminoglycan-deficient (PgsB-618) and exclusively CS-expressing (PgsD-677) CHO cell mutants (32
, 33)
. As shown in Fig. 4C
, canine EMR2 binds CS in a manner comparable to its human homologue.
Different mechanisms prevent the interaction of EMR2 with CD55 in hominoids
We next set out to investigate the occurrence of the minor sequence differences that determine ligand specificity of CD97 and EMR2 in humans. We investigated both receptors in our closest relative, the chimpanzee. Flow cytometry analysis confirmed expression of CD97 and EMR2 on leukocytes (supplemental Fig. 4). Since ligand specificity of CD97 and EMR2 depends on the arrangement and sequence of the EGF domains (7
, 17
, 20
, 40)
, we analyzed the EGF domain region of both molecules. Using RT-PCR, differences in the isoform ratio between humans and chimpanzees for both CD97 and EMR2 were detected (Fig. 5
A). Whereas in humans the majority of CD97 transcripts encodes the smallest isoform, containing three EGF domains (41)
, a more nearly equal distribution of isoforms was found in chimpanzees. A striking difference was detected for EMR2. Other than in humans, chimpanzee transcripts dominantly encode the largest, five EGF domain-containing, isoform.
|
We then sequenced the EGF domain region of CD97 and EMR2 in chimpanzees (Fig. 5B
). Chimpanzee CD97 and EMR2 differ at one amino acid in EGF domain 2, at four amino acids in EGF domain 3, and at two amino acids in EGF domain 4. Comparison of the receptor orthologs revealed that human and chimpanzee CD97 differ at two amino acids in EGF domain 4, the binding site for CS. Human and chimpanzee EMR2 differ at two amino acid in EGF domain 1, at two amino acids in EGF domain 2, and at one amino acid in EGF domain 3. Remarkably, the three amino acid substitutions within the first two EGF domains that prevent CD55 binding by human EMR2 are not present in chimpanzees. Comparison of the nucleotide sequences of CD97 and EMR2 in primates revealed that these substitutions arose only after human speciation and excludes the possibility that they have been changed back in the chimpanzee by a recent gene conversion (supplemental Fig. 5).
Based on the amino acid sequence of the EGF domains, we wondered whether chimpanzee EMR2 possesses specificity for CD55. To answer this question, we generated multivalent fluorescent probes loaded with the extracellular part of the largest isoform of chimpanzee CD97 and EMR2 and tested them in parallel with probes derived from the human receptors. Figure 6
A shows binding to the glycosaminoglycan-deficient CHO cell mutant PgsB-618 (32)
transfected with CD55. These cells overexpress CD55 in the absence of the alternative ligand CS. As predicted by the sequence of EGF domain 1 and 2, chimpanzee EMR2–5EGF has the ability to bind CD55. Previous studies of human CD97 have shown that affinity for CD55 inversely correlates with the size of the EGF domain region (20
, 40)
. Therefore, the question remained of whether chimpanzee EMR2–5EGF will also interact with CD55-expressing primary cells. We thus tested probe binding to splenocytes, a rich source of B cells, which express both CD55 and the specific type of CS bound by human CD97 and EMR2 (31)
(Fig. 6B
). In concurrence with a prime specificity for CS (17
, 31)
, binding of human and chimpanzee CD97–5EGF and EMR2–5EGF was efficiently blocked by preincubation of the cells with chondroitinase ABC, but pretreatment with CD55 mAb had little effect. In contrast, binding of human CD97–3EGF, which exclusively binds CD55 (17
, 20
, 40)
, was not affected by chondroitinase ABC treatment but was abrogated by CD55 mAb.
|
DISCUSSION
In Greek mythology, the Chimera has the head of a lion, the body of a goat, and the tail of a dragon. The EGF-TM7 receptor EMR2 resembles this three-piece creature by possessing an outer extracellular part nearly identical with CD97 and a membrane-spanning moiety most similar to EMR3. We here demonstrate molecular mechanisms that have been involved in the evolution of EMR2.
Duplicative transposition of genomic fragments, ranging in size from a few to hundreds of kb, has been a central mechanism in the development of mammalian gene families (42)
. We propose that the five EGF-TM7 receptor genes located on human chromosome 19 (1)
and canine chromosome 20, respectively, have evolved through a series of segmental intrachromosomal duplications (Fig. 7
). In a first step, two genes with a tail-to-tail arrangement originated from a common ancestral EGF-TM7 receptor gene. Different numbers of EGF domain-encoding exons (2 vs. 
5) manifested in these two genes. A second duplication then led to two pairs of tail-to-tail arranged genes with corresponding numbers of EGF domain-encoding exons (two in EMR3 and EMR4, five to twelve in CD97 and EMR1). Finally, EMR2 arose by a third duplication event.
|
Transcriptional orientation, number of exons, and chromosomal clustering provide evidence in support of this model. The identification of orthologs of all five human EGF-TM7 receptor genes in the dog genome indicates that the three duplication events must have occurred before the diversification of primates and carnivores during early mammal radiation 94 million years ago (36)
. In addition, phylogenetic analysis of the independently evolving stalk region (Fig. 3B
) implies that EMR2 arose by duplication of a CD97 progenitor. Loss of EMR2 and EMR3 in mice and rats, whose ancestors diverged later from early primates, was a subsequent event caused by a chromosome break (see http://www.ensembl.org). Different from humans and dogs, EGF-TM7 receptor genes in the mouse are localized on two chromosomes (CD97 on chromosome 8, EMR1 and EMR4 on chromosome 17).
Comparing EGF-TM7 receptor sequences from humans and dogs led to the unexpected finding that the chimeric structure of EMR2 did not develop by a recent incidental DNA transfer toward a previously duplicated ancestor gene in primates. Rather, the evolution of EMR2 has been continuously linked with that of CD97 and EMR3 independently in different mammalian orders since their divergence. The process that leads to intraspecies homogenization of DNA sequences within multigene families is known as concerted evolution (34)
. Concerted evolution has been identified in numerous ribosomal as well as protein-coding genes from bacteria to mammals. What makes the concerted evolution of EMR2 unusual is the exchange of DNA sequences with two related genes, arrayed either tail-to-tail (CD97) or in tandem (EMR3), in a regional mode. Furthermore, it is unlikely that the frequency of DNA transfer has been the same for EMR2-CD97 and EMR2-EMR3, which can be deduced from a higher homology in the EGF domain region compared with the TM7 region. In primates, the last gene conversion of EMR2 and CD97 occurred before the split of humans and chimpanzees, whereas the last gene conversion of EMR2 and EMR3 predates the divergence of hominoids and Old World macaques.
Although it is reasonable to assume that EMR2 is the acceptor gene, unraveling the direction of gene conversion will depend on the identification (and existence) of more ancient mammalian EMR2, CD97, and EMR3 genes that have evolved independently. Several genomes from ancestral mammals are currently sequenced within the Mammalian Genome Project (http://www.broad.mit.edu/mammals). A first inspection of the preliminary genome assembly (version 0.5) of the gray short-tailed opossum (Monodelphis domestica) revealed evidence for the existence of CD97- and EMR3-, but not EMR2-like sequences (data not shown).
Concerted evolution homogenizes paralogs within a genome. It allows advantageous changes to rapidly spread within multigene families, and thus represents a mechanism of molecular coevolution (34)
. Our study suggests that concerted evolution can also increase the structural diversity within a gene family. Concerted evolution of EMR2 has generated a chimeric EGF-TM7 receptor that might combine the ligand specificity of CD97 with the signaling capacity of EMR3. While signal transduction modes have not been unraveled yet for members of the EGF-TM7 family (1)
, ligand binding has been directly accessible for experimental confirmation.
Our findings emphasize the importance of EGF domain 4 of CD97 and EMR2, which provides binding to CS in humans and dogs. Para- and orthologous comparison clearly shows that the fourth EGF domain is significantly more conserved then any other EGF domain. Strong conservation of this domain might also relate to the necessity to bind a unique glycosaminoglycan structure that does not allow much room for receptor sequence alterations. We recently showed that only CS, processed by a few cell types like fibroblasts and B cells, is recognized through the fourth EGF domain (17
, 31)
. In contrast, CD55 interactions are characterized by phylogenetic restriction (19
, 43)
, where the genetic drift in CD55 will result in drift in CD97. This might explain the rather low conservation of EGF domain 1 and 2 between different mammals. A lower conservation of the CD55 binding EGF domains therefore does not contradict biological significance of this interaction. Recent functional studies provided evidence for involvement of CD97 in leukocyte trafficking and angiogenesis (11
, 12)
. How the different ligand interactions contribute to the biological function(s) of CD97 and EMR2 still needs to be shown.
Although CS binding may have driven the concerted evolution of CD97 and EMR2, CD55 binding evolved differently in the two receptors. In humans, three amino acid differences within EGF domains 1 and 2 prevent CD55 binding by EMR2 (7
, 20)
. We here demonstrate that these mutations occurred only after human speciation. This makes EMR2 one of the very few known proteins that possess a potential functional difference in humans and chimpanzees (21
, 22)
. Different from its human ortholog, chimpanzee EMR2 has the ability to bind CD55. We actually found that chimpanzee leukocytes predominantly transcribe the largest EMR2 isoform, possessing five EGF domains. In analogy to CD97, this isoform binds CS, but only weakly CD55. Human CD97 isoforms differ substantially in their affinity for CD55 (40)
, which is 10-fold lower for the largest isoform than the smallest isoform (20)
. Accordingly, unlike the smaller isoforms, CD97–5EGF has little capacity to bind CD55 when expressed at (patho)physiological levels on erythrocytes, leukocytes, or fibroblast-like synoviocytes (18
, 31
, 44)
. Predominant transcription of the largest isoform of chimpanzee EMR2 (Fig. 5)
suggests that binding to CD55 is diminished through alternative RNA splicing. Thus, different mechanisms (mutations vs. alternative splicing) in humans and chimpanzees reduce the ability of EMR2 to interact with CD55.
In summary, the evolutionary history of EMR2 is an amazing combination of molecular mechanisms including gene duplication, exon shuffling, gene conversion, alternative splicing, functional mutations, and even (in murids) deletion of the gene. Due to the late appearance of the EGF-TM7 family only in mammals, we can follow these events from an unusually close perspective.
ACKNOWLEDGMENTS
We thank professors René A. W. van Lier, M. Edward Medof, and Ronald Plasterk for comments and suggestions. This work was supported by the Royal Netherlands Academy of Arts and Sciences, the Netherlands Organization for Scientific Research (NWO, 901–07-208), and the Landsteiner Foundation for Blood Transfusion Research (LSBR, 0109).
FOOTNOTES
Note: Nucleotide sequence data reported are available in the DDBJ/EMBL/GenBank databases under accession numbers DQ227271 (rhesus macaque EMR2), DQ227272 (chimpanzee CD97), DQ227273 (chimpanzee EMR2), DQ227274 (dog CD97), DQ227275 (dog EMR1), DQ227276 (dog EMR2), DQ227277 (dog EMR3), and DQ227278 (dog EMR4).
1 These authors contributed equally to this work. 
Received for publication May 22, 2006. Accepted for publication August 22, 2006.
REFERENCES
This article has been cited by other articles:
|
|
 |
|
  J. Q. Davies, G.-W. Chang, S. Yona, S. Gordon, M. Stacey, and H.-H. Lin The Role of Receptor Oligomerization in Modulating the Expression and Function of Leukocyte Adhesion-G Protein-coupled Receptors J. Biol. Chem., September 14, 2007; 282(37): 27343 - 27353. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
