Evolution of the CD4 Family: Teleost Fish Possess Two Divergent Forms of CD4 in Addition to Lymphocyte Activation Gene-3¹

Activation of T lymphocytes follows the recognition of specific Ag presented to the T cell Ag receptor. The T cell Ag receptor complex consists of a ligand-binding heterodimer of polymorphic TCR chains (αβ or γδ) and the nonpolymorphic chains CD3γ, CD3δ, CD3ε, and TCRζ, which are involved in signaling (1). T cell Ag receptor-mediated signaling is enhanced by the coligation of MHC class I or class II molecules by CD8 and CD4 coreceptors, respectively; the polymorphic regions of MHC molecules interact with the T cell Ag receptor, whereas the nonpolymorphic regions interact with the coreceptors (2). As a result of this interaction, intracellular molecules, including lymphocyte-specific protein tyrosine kinase (LCK or p56LCK), are recruited to the signaling complex, resulting in the initiation of a cascade of phosphorylation events that define T cell effector function (2, 3). This interaction is also thought to play a role in the differentiation and selection of immature double-positive thymocytes expressing both CD4 and CD8 into the different mature single-positive T cell populations of cytotoxic T cells that express only CD8 and Th cells that express only CD4 (4). CD4 and CD8 proteins are essential for T cell development and T cell activation in vertebrates.

Mammalian CD4 is a monomeric transmembrane glycoprotein that belongs to the Ig superfamily. It has a long extracellular portion containing four Ig-like domains (D1–D4), two of which are V-like domains (D1 and D3) and the other two are C2-like domains (D2 and D4). The short intracellular C-terminal tail of CD4 contains a CXC motif that binds the N-terminal CXXC motif of LCK via noncovalent interaction in the presence of Zn²⁺ (5), resulting in the first signal for T cell activation. In general, the D1 and D3 domains of mammalian CD4 contain nine β strands (ABCC′C″DEFG) with characteristics of Ig superfamily V domains, including a pair of cysteine residues that stabilize the Ig fold via a disulfide bridge and a conserved packing Trp molecule. Domains 2 and 4 are similar to D1 and D3 in composition except that the C″ and D strands are omitted, leaving only seven β strands (ABCC′EFG). The region of human CD4 that interacts with MHC class II lies mainly within the C″ and D strands of D1, with Phe⁴³, Lys⁴⁶, and Arg⁵⁹ of human CD4 representing contact residues (6). Phe⁴³ interacts with a conserved group of hydrophobic residues within the α2 and β2 domains of the MHC class II heterodimer, and several other amino acids of CD4 (Leu⁴⁴–Lys⁴⁶) interact with the class II β2-chain. A Phe residue located in a similar position in the chicken CD4 molecule and several lysines within D1 (Lys⁴⁵, Lys⁵⁰, and Lys⁵²) have been predicted to interact with the β-chain of class II MHC (7), suggesting that conservation of these interactions may exist throughout vertebrates.

The gene for CD4 in both mammals and birds is encoded by 10 exons. Human and mouse CD4 genes have been mapped to chromosomes 12 and 6 respectively, downstream from the CD4-related gene called lymphocyte activation gene-3 (LAG-3)³ and near GAPDH. Chicken CD4 mapped to a conserved region on chromosome 1 near GAPDH, indicating a syntenic relationship for these two genes (7, 8). Specific expression for mammalian and avian CD4 is mainly in T lymphocytes, with the highest tissue-specific mRNA expression in the thymus followed by the spleen (7, 9).

LAG-3 (CD223), a member of the CD4 family, is structurally similar to CD4 in that it is composed of four extracellular Ig domains (D1–D4), with the exception that LAG-3 D1 contains an additional loop between the C and C′ β strands and the cytoplasmic tail lacks the p56LCK binding site (10). Functionally, LAG-3 interacts with MHC class II at higher affinities than CD4, suggesting a role for Ag-specific responses (11, 12). This possibility is supported by experiments demonstrating that LAG-3 impedes interactions between CD4 and MHC class II, thereby suggesting that LAG-3 plays a role in regulatory T cells (13, 14). LAG-3 expression is primarily limited to activated T and NK cell lineages, but recently LAG-3 has been found on activated murine B cells via T cell-mediated induction (15).

Because of the lack of suitable markers for T lymphocytes in fish, the characterizations of T cell populations and functionality have yet to be fully defined in teleosts. However, recent evidence suggests that mechanisms of T cell activation and functionality in fish are similar to those observed in mammals based on the identification of T cell Ag receptor (16, 17, 18), MHC (reviewed in Ref. 19), CD3 (19, 20, 21, 22), CD8α and CD8β (23, 24), and LCK (25) in bony fish. Recently, a CD4-like molecule containing two Ig domains (IgV and IgC) was described in the sea lamprey Petromyzon marinus (26), and a four-domain CD4 molecule was identified in the pufferfish Takifugu rubripes (27). Here we describe the isolation of two distinct CD4 homologs from rainbow trout, one with four domains that is similar to the CD4 of higher vertebrates and one with two domains that is similar to the CD4-like molecule of lamprey. Both of these trout CD4 homologs contain the canonical LCK association motif. In addition, we show the presence of LAG-3 in all teleosts examined, all of which lack the p56LCK binding site consistent with mammals. Equivalent genes (CD4, CD4REL, and LAG-3) were then identified in the genomes of other ectotherms (e.g., Tetraodon nigroviridis, Danio rerio, and Xenopus tropicalis) to investigate the syntenic relationships and evolutionary path that gave rise to mammalian CD4. The genomic architecture, tissue-specific expression, and cell-specific expression of the two CD4 genes and LAG-3 were investigated. These data suggest that CD4 and CD4REL (CD4 related) are both involved in cellular immunity in teleost fish.

Rainbow trout fry were obtained from Clear Springs Food and maintained at a constant temperature of 15°C in sand-filtered and UV-treated freshwater at the Western Fisheries Research Center (Seattle WA).

The amino acid sequences for the extracellular domains of mouse (GenBank accession no. M36850) and rabbit (GenBank accession no. M92840) CD4 were used in TBlastn-based (28) searches to identify orthologous sequences within the trout expressed sequence tag (EST) database at the National Center for Biotechnology Information (NCBI; Bethesda, MD) and The Institute for Genomic Research (TIGR; Rockville, MD; www.tigr.org). ESTs displaying similarity to CD4 were isolated and aligned using the Assembly Line program (version 1.0.9) from the MacVector software (Oxford Molecular Group) to predict contiguous cDNA sequences for CD4-like molecules. Primers were designed using two different CD4-like EST sequences identified in the trout EST gene index (see Table I). The nCD4-2-R4 and nCD4-2-R1 primers were used for anchored PCR with the T3 primer in sequential amplifications to obtain the missing portions of the putative tCD4 cDNA from a rainbow trout splenic unidirectional cDNA library (Stratagene). Subsequent amplification using the tCD4-F10 and tCD4-R5 primers amplified the full-length cDNA for this gene, confirming the originally derived contig sequence. EST CA382329 (CD4REL) was obtained from the National Center for Cool and Cold Water Aquaculture of the U.S. Department of Agriculture (Leetown, WV). The open reading frame was fully sequenced and used as a template for generating a PCR-amplified probe (CD4REL-F1/CD4REL-R1). This probe was used to screen a cDNA library generated from the spleen of the OSU-142 homozygous line of rainbow trout. Further primers (CD4REL-F2/CD4REL-R2) were designed from the full-length clone of CD4REL to amplify CD4REL from the Hot Creek homozygous line of rainbow trout. TBlastn using human LAG-3 as the query revealed trout EST (GenBank accession no. CA364501 (D3 through the cytoplasmic tail)) as a trout LAG-3 homolog. We amplified this region from OSU-142 splenic cDNA using LAG-3-D4-F1 and LAG-3-Cy-R1, cloned the fragment into pTOPO, and sequenced the insert to confirm the EST.

To isolate PBLs, 5 ml of blood was isolated from the caudal vein of four individual trout, diluted with 45 ml of 1× PBS containing heparin, and underlaid with Histopaque 1071 (Sigma-Aldrich). The buffy coat was removed and washed with PBS, and cells were pelleted. RLT buffer (Qiagen RNeasy kit) containing 1% 2-ME was added to cell pellets and subjected to RNA extraction procedures according to the manufacturer’s instructions, including in-column RNase-free DNase (Qiagen) treatment. One microgram of the resulting total RNA was used to synthesize first-strand cDNA in 20-μl volumes as described previously (29). First-strand cDNAs were diluted 10-fold for subsequent quantitative RT-PCR (qPCR).

Trout weighing 200–300g were euthanized in MS-222, and various tissues were removed immediately. For direct RNA extraction, the tissues were snap frozen in liquid nitrogen and stored at −80°C until required for RNA extraction. Total RNA was isolated as previously described (29) for Northern blot analysis or isolated by the Qiagen RNeasy kit for qPCR analysis.

PCR products were cloned into pTOPO2.1 (Invitrogen Life Technologies) following the manufacturer’s instructions. Plasmid DNA was isolated from colonies containing the correctly sized inserts using the QIAprep spin miniprep kit (Qiagen), and three randomly selected clones representing each product were sequenced (Applied Biosystems). Comparisons of nucleotide and amino acid sequences with the GenBank and Swiss-Prot databases were performed using BLAST. TBlastn analysis of the ENSEMBL database was used to identify related genes in other vertebrate genomes to initially derive the syntenic relationships between species. Full-length coding regions were elucidated from identified genomic scaffolds using BLAT (genome.ucsc.edu/cgi-bin/hgBlat), ENSEMBL (www.ensembl.org/index.html), and manual analysis. Genome versions used during these analyses are Gallus gallus assembly version 2 (February 2004), X. tropicalis assembly version 3 (October 2004), T. nigroviridis assembly version 7 (February 2004), T. rubripes assembly version 3 (August 2002) in BLAT and version 4 (December 2005) in ENSEMBL, and D. rerio assembly version 5 (May 2005). Direct comparisons between two sequences were performed using the NEEDLE global alignment program (30) within EMBOSS (www.uk.embnet.org/Software/EMBOSS/). Multiple sequence alignments were generated using ClustalW (version 1.74) (31), and phylogenetic trees were constructed from the ClustalW alignments (amino acid sequences) for either two Ig domains or full-length CD4 molecules that contain all four domains using the neighbor-joining method (32) from MEGA 2.1 software (33) with Poisson correction and complete deletion of gaps and bootstrapped 1000 times.

qPCR was used to estimate differences in tCD4 and +CD4REL mRNA expression in different tissues of and isolated PBLs of naive rainbow trout. Tissues and PBL cDNAs were generated from total RNA isolated using the RNeasy kit (Qiagen) as described above. Primer and probe sets for qPCR were designed using the Primer Express software (ABI version 2.0; Applied Biosystems) from cDNA sequences of tCD4, tCD4REL, and LAG-3. Primers were positioned so that they would span an intron, and the resulting amplicons were confirmed by sequencing. Acidic ribosomal phosphoprotein (ARP) was used for normalization as described previously (34). Each probe was synthesized, dual-labeled (5′-FAM and 3′-TAMRA), and purchased from Integrated DNA Technology. qPCR assays were performed using the ABI PRISM 7900HT sequence detection system with standard cycling conditions of 50°C for 2 min and 95°C for 10 min followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. The reactions were performed in 12-μl volumes containing 11 pmol of each primer, 2 pmol of labeled probe, 1× ABI Universal PCR master mix (Applied Biosystems), and ∼25 ng of first-strand cDNA. Each gene target was run in individual reactions in duplicate to minimize variation.

Pronephros (anterior kidney) tissue was taken from a naive, adult rainbow trout and teased apart using forceps. A single-cell suspension was obtained by pressing the pronephros through a 40-μm cell strainer (BD Falcon) using the plunger from a 3-ml syringe (BD Falcon). Peripheral blood was drawn from the caudal vein using a heparinized syringe and diluted 1/25 with PBS. The pronephros single-cell suspension and diluted blood were layered onto 10 ml of Histopaque 1077 (Sigma-Aldrich) and centrifuged (400 × g) for 30 min at 10°C using 50-ml conical centrifuge tubes (BD Biosciences). Leukocytes were then collected from the interface layer, washed 3 times with PBS-FBS (2%), and resuspended to a concentration of 1 × 10⁷ cells/ml in PBS-FBS containing 2 μg/ml biotinylated mAb 1-14. mAb 1-14 is specific for rainbow trout IgM (35). Cells were incubated on ice for 30 min, washed 3 times with PBS-FBS, and stained with Neutralite avidin-FITC (Southern Biotechnology Associates) at 1/200 in PBS-FBS for 45 min on ice. Following the staining procedure, the cells were washed 4 times with PBS-FBS, adjusted to 2.5 × 10⁷ cells/ml, and sorted from the lymphoid gate using a FACSAria flow cytometer (BD Biosciences). Cells (1 × 10⁶) were sorted under stringent conditions, and those that were clearly positive for sIgM (sIgM⁺) and negative for surface IgM (sIgM⁻) were collected into PBS-FBS (10%), pelleted by centrifugation, and directly lysed using the RLT buffer from the RNeasy kit (Qiagen). RNA was isolated from the cell pellets using the RNeasy kit in accordance with the manufacturer’s instructions, including an in-column DNase step before cDNA synthesis as described previously (29). RT-PCR analyses were performed for several genes, including CD4 (CD4-1093F/CD4-1213R), CD4REL (BBCD4-882F/BBCD4-1002R), LAG-3 (LAG3-Q-F1/LAG3-Q-R1), TCR-α-chain (TCR-23743/TCR-23744), and the membrane-bound form of IgM (IgM-MEM-F1/R1) in addition to the housekeeping gene ARP (ARP-For/ARP-Rev) (see Table I for all primer sequences). Standard PCR conditions used 1 U of AmpliTaq DNA polymerase (Roche), 12 pmol each of forward and reverse primers, 400 μM dNTP mix (Promega), 1× PCR buffer containing MgCl₂ (1.5 mM) (Roche), and 1 μl of cDNA template in 25-μl reactions with sterile H₂O. The cycling protocol was one cycle at 94°C for 2 min, 35–38 cycles at 94°C for 30 s, 50–60°C for 30 s, and 72°C for 1 min, with a final extension step of 5 min at 72°C. Amplified products were analyzed on a 2% agarose gel containing ethidium bromide (100 ng ml⁻¹).

Genomic DNA was isolated from a single rainbow trout tail fin as previously described (36). Using 0.5 μg of genomic DNA as template, PCR was performed to obtain the sequences of the tCD4 genes over the peptide-coding region. For tCD4REL, two pairs of primers, tCD4REL-F10/tCD4REL-R4 and tCD4REL-F6/tCD4-REL-R6, were used to obtain two overlapping PCR products. Similarly, the genomic sequence for tCD4 was obtained using primers, tCD4-F10/tCD4-R9 and tCD4-F8/tCD4-R5. All primers are listed in Table I.

Procedures for in situ hybridization on trout chromosomes have been previously described (37). Bacterial artificial chromosomes (BACs) positive for tCDREL and LAG-3 were identified by screening the homozygous OSU-142 4.5X BAC library using PCR generated probes specific for tCD4 (CD4-1093F/CD4-1213R), tCD4REL (CD4REL-F1/R1), and LAG-3 (LAG3-D4F1/R1) on high-density filters. A CD4⁺ BAC was isolated using PCR super pools specific for the Swanson 10X BAC library (38) Direct sequencing and secondary PCR were used to confirm BAC clones before use as probes.

BLAST-based screening of the trout EST database at the TIGR and NCBI sites identified partial sequences (GenBank accession nos. CA352276, BX875992, and CA382329) with high similarity to mammalian and avian CD4 genes. The first two ESTs contained overlapping sequences that produced a fragment analogous to the 3′ end of mammalian CD4. The latter EST encodes a second distinct CD4-like sequence. The full-length cDNA sequences for both tCD4-like molecules were isolated from trout splenic cDNA libraries. The first CD4 molecule is designated tCD4 (GenBank accession no. AY973028) and encodes a 489-aa open reading frame. The second transcript (designated tCD4REL; GenBank accession no. AY973029) is shorter (1091 bp) and encodes a peptide of 325 aa. Further examination of the trout EST database revealed a sequence (GenBank accession no. BX877326) with >90% nucleotide identity to tCD4REL that likely represents a second tCD4REL gene or an allelic variant. In addition, a trout EST (GenBank accession no. CA364501) was identified that displayed moderate amino acid similarity (28%) to human LAG-3 from the end of D3 through the cytoplasmic tail. This sequence was used as the basis for BAC cloning and expression analysis of trout LAG-3.

Following the isolation of the tCD4 cDNA sequences, we determined their genomic organization (GenBank accession nos. AY973030 (tCD4) and AY973031 (tCD4REL)). The intron positions of both tCD4 genes reveal similarities in the gene structure of tCD4 and tCD4REL to those of human and chicken CD4 genes (Fig. 1). The CD4 gene is composed of 10 exons in both humans and chickens, with each exon containing a separate domain of the CD4 molecule: exons 1 and 10 encode the 5′ and 3′ untranslated regions, respectively, exon 2 codes for the signal peptide, domain 1 is divided between exons 3 and 4, extracellular domains 2, 3 and 4 are encoded on exons 5, 6, and 7, and exons 8 and 9 code for the transmembrane and intracellular regions, respectively (8, 39). Although the exons containing the 5′ untranslated region of tCD4 have not been sequenced, it is clear that the coding region of tCD4 is divided in an identical manner as the other vertebrate CD4 genes, supporting its analogy with CD4. The split first variable domain (D1) is significant, as this is a unique feature that occurs in both the CD4 and the LAG-3 genes of higher vertebrates. The intervening intron occurs between sequences coding for the C′ and C″ strands of D1 in all cases. A similar situation is observed for tCD4REL, as the first domain is encoded by exons 1 and 2 and the second domain and transmembrane region are each encoded by single exons. The main difference between tCD4 and tCD4REL lays in the absence of domains 3 and 4 (as discussed below), which in CD4REL are replaced with a short exon encoding a hinge-like region between domain 2 and the transmembrane region. Nevertheless, tCD4REL is clearly related to vertebrate CD4 genes for its genomic organization.

Based upon multiple sequence alignment analysis (Fig. 2), tCD4 possesses the same general peptide structure as mammalian and chicken CD4 molecules (7, 40). In short, a signal peptide comprising 21 aa precedes four Ig domains (D1–D4) of the predicted extracellular portion of the trout coreceptor followed by a transmembrane region and a C-terminal intracellular region that contains a CXC (LCK binding) motif (5, 41). Despite the highest similarity of tCD4REL to mammalian CD4 molecules (∼25% amino acid identity over D1 and D2; Table II) in initial BLAST searches, this gene encodes only two Ig domains, a variable domain (D1) followed by a constant domain (D2). From multiple sequence alignment and pairwise alignments (Fig. 2 and Table II), the Ig-like domains of tCD4REL resembles the D1 and D2 of other CD4 molecules, implying that CD4REL D1D2 may associate with MHC class II (6). Structural similarities between the tCD4-like molecules and other vertebrate CD4s are described in more detail in Discussion. In addition, fugu, trout, and Tetraodon CD4s share a single N-linked glycosylation site within strand D3 (E strand) and two N-linked sites within the D3 B strand (conserved with human) and between the F and G strands of domain 4. Fish CD4REL (fugu, Tetraodon, and trout) each contain a single N-linked sited within the second domain, therefore indicating that both CD4 molecules, tCD4 and CD4REL, are likely glycosylated.

Genes for CD4, CD4REL, and a further member of the CD4 gene family, LAG-3, were identified in the genome databases of nonmammalian vertebrates using ENSEMBL and BLAT. These genes and their relative locations are listed in Table III, and indicate that different teleost species possess all three CD4 family members, CD4, LAG-3, and CD4REL. All three CD4-related molecules exist on a single chromosome (chromosome. 8) of the Tetraodon genome in a region containing GAPDH, LPREL2, and the tapasin-related (TAPBP-R) gene (Fig. 3). Tetraodon CD4REL (bases 9167339–9168871) resides close to CD4 (bases 9140427–9143009) on chromosome 8 (Fig. 3,A) relative to LAG-3 (bases 8944024–8942311). The chromosomal location of zebrafish CD4 differs from that of LAG-3 and CD4REL. LAG-3 and CD4REL both reside on chromosome 16 in zebrafish (bases 10121775–10125183 and 5908545–5910812, respectively). Domain 2 through the intracellular tail (exons 4–8) of zebrafish CD4 was identified on chromosome 2 in version 5 of the zebrafish genome (this gene was not present in zebrafish version 4). However, the signal peptide region and the exons encoding the first Ig domain of CD4 were found in two gene copies on chromosome 16 (bases 6427255–6429589 and 6395451–6398903) close to CD4REL and LAG-3. In addition to the genomic sequences, two cDNA sequences were identified in the zebrafish EST database (GenBank accession nos. C999671 and DR725676) that encode the signal peptide domain through the second Ig domain of zebrafish CD4, thus containing sequences from the partial CD4 genes on both chromosomes 2 and 16. To extend the syntenic relationship of the CD4 family in salmonid fish, we physically mapped BAC clones harboring CD4REL and LAG-3 to trout chromosomes using in situ hybridization. Both CD4REL and LAG-3 hybridized to the short arm of chromosome 2 (LG27), with CD4REL staining closer to the centromere (Fig. 4). A separate BAC clone that contained the CD4 gene and a second (partial) CD4REL sequence was identified by screening the Swanson BAC library, as we were unable to identify CD4 positive BAC clones from the OSU-142 library. The CD4⁺ BAC mapped to chromosome 9 (LG21) of rainbow trout (Fig. 4), which represents a much smaller chromosome than the chromosome encoding CD4-REL (chromosome 2). The CD4 BAC probe colocalized with the centromere probe 10H19, which is specific for chromosomes 7, 9 and 11, but did not colocalize with additional probes specific for either chromosomes 7 or 11 (data not shown).

To further analyze the relationships of the CD4 family, the amino acid sequences for fish CD4 and CD4REL were compared with the equivalent regions of mammalian and avian CD4 molecules in addition to LAG-3 using the neighbor-joining method (Fig. 5). Three different ways of analyzing the sequence data were used. First, those sequences containing full-length molecules that possess all four extracellular domains were compared (Fig. 5,A), demonstrating a tight clustering of fish CD4 to mammalian CD4 that is supported by bootstrap analysis. The fugu LAG-3 homolog (26% identity to mouse LAG-3) clusters in a separate clade with mammalian LAG-3 and the predicted LAG-3 orthologs encoded in chicken and Xenopus. The smaller CD4REL molecules were compared with the four-domain CD4 and LAG-3 molecules using only two Ig domains, either D1D2 only (Fig. 5,B) or D3D4 only (Fig. 5 C). In these analyses, fish CD4 and CD4REL molecules both group with mammalian CD4, whereas fugu LAG-3 forms a distinct clade with mammalian LAG-3.

Transcripts for tCD4, tCD4REL, and LAG-3 were most abundant in the thymus of naive rainbow trout, as initially assessed by Northern blot analysis (data not shown) and then by qPCR (Fig. 6). Northern analysis resulted in a single band for all three genes predominantly within the thymus. Moderate expression was observed in the spleen and pronephros (bone-marrow equivalent in fish), which are rich in lymphocytes, with no detectable expression in nonlymphoid tissues such as liver, heart, muscle, and testis. qPCR was performed on tissues positive for CD4, CD4REL, and LAG-3 during Northern analysis expression, using liver as a control with negligible expression and including an additional tissue, PBL. qPCR indicated expression of tCD4, tCD4REL, and LAG-3 in PBLs, although levels were low relative to the levels in the other lymphocyte-rich tissues. To more formally address the cell types that express these genes, lymphocytes were isolated from the pronephros and PBLs of naive trout and separated by FACS sorting into populations expressing sIgM or lacking sIgM expression. Transcripts for both CD4 and CD4REL were restricted to sIgM⁻ lymphocytes for both tissues, directly reflecting the expression pattern for TCR-α in these sorted populations of cells (Fig. 7). LAG-3 was expressed in both sIgM⁻ and sIgM⁺ lymphocytes, suggesting that the cells were activated. The same cell populations were analyzed for expression of the membrane-bound form of trout IgM, which was found predominantly in sIgM⁺ lymphocytes, although weaker expression was also apparent in the sIgM⁻ fractions. These latter results for the pronephros can be easily explained because the pronephros is the bone marrow equivalent in fish, thereby harboring pre-B cells and plasma cells that lack surface IgM expression.

T cell functionality depends heavily upon the interactions of the T cell coreceptors CD4 and CD8 with the MHC for mediating cellular immune responses. Our laboratory has previously described components of the MHC and the CD8 coreceptor in trout (23, 37, 42) as a means of defining the evolutionary origins of adaptive immunity and for providing key markers for assessing cell-mediated immunity in fish.

In this report we extend our analysis of teleost adaptive immunity to include the description of three members of the CD4 family that are representative of teleosts, including the analysis of rainbow trout, zebrafish, Tetraodon, and fugu genomes. One of the CD4 genes encodes four Ig-like domains in its predicted extracellular portion that are in the order VCVC, whereas the other (CD4REL) possesses only two domains (VC) that are similar to the first two domains (D1D2) of mammalian CD4, which interacts with MHC class II. The first domain (D1) of tCD4 contains 101 aa arranged in nine β strands. It has a conserved Cys in the F strand but lacks the second Cys residue in the B strand for disulfide bridge formation in contrast to mammalian and bird CD4 molecules. This particular Cys is also absent in the Fugu CD4 ortholog (27) and may result in the two β-sheets of the fish CD4 molecules being further apart in D1, as was observed for D3 of human CD4 (6, 43). Whether these differences affect the three-dimensional structure and/or function of the trout molecule remains to be determined. In contrast, the Cys pair required for the Ig fold is conserved in both trout and fugu CD4REL for D1.

Mammalian CD4 binds to MHC class II primarily through the interaction of the phenol ring of Phe⁴³ (human CD4) located at the C′ and C″ strand junction of D1 (6). In tCD4, the corresponding region is not well conserved (Fig. 8) and there is not a Phe residue available for binding MHC class II in either fugu or trout, suggesting that other contact residues are required for the interaction of fish CD4 with class II if this molecule is the ligand of fish CD4. For example, a Trp residue located between the C″ and D strands of both teleost CD4 sequences aligns well with a Phe residue in mouse CD4. Trp, like Phe, contains a phenol ring that could fit into the hydrophobic “groove” of class II. Additionally, in tCD4 an Arg residue occurs three amino acids away, forming a motif (WXXR) that resembles the FXXK motif of mammalian and avian CD4 (Fig. 8). There are also several lysine and arginine residues that may interact with the class II molecule. Fugu CD4REL possesses an FXXK motif and tCD4REL encodes a shorter version (FXK), implying that they may associate with class II via similar mechanisms. Interestingly, CD4REL also possesses residues similar to those that are required for the association of LAG-3 with MHC class II. Human LAG-3 has four residues implicated in MHC class II binding, Asp³⁰, His⁵⁶, Tyr⁷⁷, and Arg¹⁰³ (44), which occur on the top surface of the molecule. In tCD4REL, an Asp residue is found between the B and C strands of D1, a Tyr residue is present in or near the C′ strand, and an Arg residue is located in or near the D strand. In addition, there is a high degree of conservation between the C′ and E strand for CD4REL D1 within teleosts, suggesting a functional role. Although the loop containing His⁵⁶ of human LAG-3 is absent in fish CD4REL, another His residue located in strand C is conserved in human LAG-3 and fish CD4REL molecules. These residues, in addition to the shorter length of CD4REL, suggest that the top face of CD4REL may be involved in the interaction with class II similar to that for LAG-3.

The presence of features similar to both CD4 and LAG-3 in tCD4REL is consistent with the hypothesized two-domain primordial molecule that gave rise to both CD4 and LAG-3, thereby supporting the theory that the four-domain CD4 molecule evolved by tandem duplication of two Ig-like domains (10). In support of this theory, a CD4-like molecule composed of two Ig-like domains was recently described from the most primitive vertebrate, the lamprey (26). Because the predicted CD4 binding regions in the α2 and β2 domains of trout class II are highly conserved with respect to mammalian class II (45), similar interactions would be expected to occur for accessory molecules. Domain 3 (D3) of tCD4 also resembles the first IgV domain (D1) in size (nine β strands), in agreement with mammalian, chicken, and pufferfish CD4 molecules. LAG-3 molecules, however, are shorter over this domain such that D3 is more similar to a C2 type Ig domain (10) (Fig. 2).

The second domains (D2) for both tCD4 and tCD4REL resemble Ig superfamily C2 domains in the number of β-sheets and the conserved pair of cysteines required for the Ig fold. The conserved WXC motif in the F strand of D2 is unique to CD4 molecules (10) and represents another feature of tCD4 and tCD4REL that supports the theory that these molecules are related to the ancestral gene that gave rise to mammalian CD4 and LAG-3. This feature is repeated again in D4 within vertebrate CD4 and LAG-3.

One interesting aspect of CD4REL is a stretch of amino acids immediately adjacent to the transmembrane region that is Ser-, Thr-, and Pro-rich and contains a pair of cysteines. This portion of tCD4REL resembles the hinge region of mammalian and fish CD8α (23, 46), including the presence of several potential O-linked glycosylation sites with the motif XPXX (where at least one X = S or T), implying that this region forms a rigid structure that allows it to reach class II while also indicating the potential of tCD4REL to form dimers based upon the CXXC motif that is also conserved in fugu and Tetraodon. A similar “stem” is not found in the four-domain CD4, and the CXXC motif is absent from zebrafish CD4REL (Table III).

The cytoplasmic domains of both tCD4 molecules resemble mammalian CD4 in that the LCK binding motif is well conserved between tCD4 (RRICRC) and tCD4REL (NDYCQC) when compared with the human CD4 LCK association motif (KKTCQC), including the presence of several basic (K/R) residues in the helical region immediately preceding the CXC motif. The CXC LCK motif has been well described in mammals and is required for binding the CXXC motif in the N terminus of p56LCK in the presence of zinc cations to initiate the first signal for T cell activation. A dileucine motif (LL) associated with CD4 internalization of human CD4 is absent from the trout and fugu CD4 sequences, as are two Ser residues implicated in this function. Down-regulation of mammalian CD4 has been attributed to phosphorylation of at least one of these serine residues by protein kinase C, allowing the dissociation of LCK from the cytoplasmic tail of CD4 (47). Trout CD4REL only possesses one leucine in this region, whereas fugu CD4REL maintains the conserved LL motif. In addition, both of the CD4REL molecules of these two fish are Ser- and Thr-rich in their putative cytoplasmic tails and, thus, have the potential for modes of regulation similar to those of mammalian CD4. Similar to tCD4, chicken CD4 lacks the two serines and is down-regulated at a much slower rate than mammalian CD4 (7, 48), likely reflecting the importance of these residues. Interestingly, though, teleost CD4 possess a conserved stretch of amino acids (PKGFYR) immediately following the LCK binding motif, and chicken CD4 contains a related motif (+XXY+, where + represents a basic residue) in this region (Fig. 2), suggesting a possible phosphorylation site.

To resolve the relationships and evolutionary history of the fish CD4-like molecules, systematic syntenic and phylogenetic comparisons were undertaken. To facilitate this comparison, related genes were identified in the genome databases for G. gallus, X. tropicalis, T. nigrovidis, F. rubripes, and D. rerio. Our analysis of the Tetraodon genome shows that all three CD4-related genes exist on a single chromosome (chromosome 8) in a region containing GAPDH, LPREL2, and the tapasin-related gene TAPBP-R, which are also syntenic in the CD4 locus of mammals and birds (Fig. 3) (8). CD4 and CD4REL are closer to each other than to LAG-3, further supporting the belief that fish CD4 and CD4REL genes are both structural orthologs of mammalian CD4, whereas a separate gene encodes LAG-3. Tetraodon and zebrafish both contain multiple tandem copies of whole and partial gene sequences for CD4REL, with at least three different genes being present within each of these fish. Two distinct copies of tCD4REL were reported by Dijkstra et al. (49) during the preparation of this manuscript in a different clonal line of trout, and at least two genes for CD4REL were detected during PCR analysis of CD4REL BAC clones (data not shown). These observations likely reflect ancient duplication events that led to the formation of the four-domain CD4 molecule that is common to all vertebrates. One curious feature observed during our in silico analysis of the zebrafish genome was the presence of two chromosomal locations of CD4, which showed that the first three exons (duplicated) were on chromosome 16 with CD4REL and that LAG-3 and the remaining exons mapped to chromosome 2. In contrast, zebrafish ESTs for CD4 contain sequences from exon 1 (signal peptide) to exon 4 (D2), suggesting that the partial gene mapping to chromosome 2 is, in fact, a continuation of one partial CD4 copy on chromosome 16, thereby suggesting assembly errors in the zebrafish version 5.0 draft near the CD4 locus. However, when observing the syntenic relationships of the tCD4 gene family, CD4 (and a partial CD4REL gene) physically mapped to a distinct chromosome (chromosome 9) from LAG-3 and CD4REL (chromosome 2), demonstrating that the CD4 gene in some teleost fish may have been duplicated and translocated to additional chromosomes. The presence of a partial CD4REL gene on the tCD4 BAC nonetheless confirms linkage between CD4 and CD4REL at one time during the evolutionary history of salmonids. Finally, in further support for the conserved syntenic relationship of the CD4 family among vertebrates, we recently mapped the trout TAPBP-R near LAG-3 (50), consistent with Tetraodon, fugu, mammals, and avians.

To further assess the relationships within the CD4 family, phylogenetic comparisons were performed using amino acid sequences of either the full-length four domain molecules where appropriate or by using two Ig domains (Fig. 5). When comparing molecules containing four Ig domains, teleost CD4 groups tightly with mammalian CD4 whereas fish LAG-3 clusters with mammalian LAG-3, lending support to the possibility that the two distinct CD4-like molecules with four Ig domains identified in fish represent true homologs of CD4 and LAG-3, with tCD4 being closely related to mammalian CD4. For addressing the relationships of the smaller molecules, the two Ig domains were compared with either D1 and D2 only (Fig. 5B) or with D3 and D4 as a unit (Fig. 5,C) because these represent the tandem IgV and IgC domains that are repeated in the four domain molecules. Although the bootstrap support is not as robust as in Fig. 5,A, the sequences still clearly divide into separate clades. Teleost CD4 and CD4REL both group with mammalian CD4, whereas teleost LAG-3 again forms a distinct clade with mammalian LAG-3. CD2 was included in these comparisons to provide a root for these latter trees, as it possesses a similar structure with two extracellular domains, IgV followed by IgC. We also included two “primitive” sequences in the D1D2 and D3D4 analyses that displayed similarity to the CD4 family, namely that of a two-Ig domain sequence found during in silico analysis of Ciona intestinalis (Table III) using CD4REL as the query and a recently described prototypical CD4-like molecule possessing two tandem Ig domains from lamprey (26). Because of the high level of divergence between these latter molecules, we were unable to fully resolve their relationship(s) within the overall CD4 family.

In mammals, expression of CD4 is largely restricted to T and NK cell lineages. In our study, the highest expression of CD4, CD4REL, and LAG-3 mRNA was detected in the thymus of naive trout followed by moderate expression in the spleen and pronephros, consistent with the expression of these genes in 1° and 2° lymphoid tissues directly reflecting CD4 and LAG-3 expression in mammalian tissues. The tissues expressing these genes in trout also correspond with the expression pattern of CD8α and the T cell Ag receptor in trout (23, 51). High expression levels of CD4 and CD4REL in the thymus are consistent with the role of the thymus for T cell lymphopoiesis and education in all vertebrates (52, 53). Furthermore, when assessing the expression of CD4 and CD4REL transcripts in sIgM⁻ and sIgM⁺ populations of trout lymphocytes, both CD4-like genes were restricted to the sIgM⁻ fraction. An identical expression pattern was also observed for trout TCR-α, strongly suggesting the expression of CD4 and CD4REL within trout T cells. Of note, CD4REL was clearly expressed at much lower levels in the sIgM⁻ lymphocytes from PBL than in the pronephros (the site of hematopoiesis in teleost fish), suggesting a potential role for CD4REL in immature T cell function or selection and indicating that CD4REL⁺ lymphocyte subsets in peripheral tissues are found at lower frequencies in comparison to those expressing CD4. Trout LAG-3 mRNA expression was present in both lymphocyte pools, reflecting expression in both B and non-B lymphocytes in fish. This finding is in line with recent reports demonstrating that mammalian LAG-3 is expressed on B cells as well as T and NK cells (15), although murine B cells required T cell-mediated activation before LAG-3 expression.

By using comparative bioinformatics, this study outlines the likely evolutionary path that gave rise to CD4 while also implying novel differences between fish and “higher vertebrates” in regard to T cell development and functionality. The phylogenetic comparisons, together with the syntenic relationships, chromosomal proximity, and presence of tandem copies of CD4REL in fish, suggest that the two-domain CD4REL evolved into the four-domain CD4 molecule by simple tandem duplication and then the four domain molecule duplicated and diverged to generate LAG-3 (Fig. 9, theory 1). An alternate model of LAG-3 evolution directly from CD4REL (or the precursor to CD4REL) would involve two duplication events with D2 that provide three IgC domains (Fig. 9, theory 2), although sequence features common to D1 and D3 suggest this model is less likely. In support of our hypothesis (theory 1), Pancer et al. (26) recently speculated on the origin of CD4 based upon their analyses of CD4 and T cell Ag receptor “prototypes” that are expressed at high levels within lamprey lymphocytes. The putative lamprey CD4-like molecule is composed of two extracellular Ig domains (V and C), a transmembrane region, and a long cytoplasmic tail (>80 aa) that is devoid of the p56LCK association motif. In addition, the second domain of the lamprey CD4 molecule encodes the unique WXC sequence tag that is found in the F strand of D2 and D4 for all of the gnathosome CD4 family members including LAG-3, which lends additional support to the notion that the lamprey molecule represents an ancestral form of CD4. Furthermore, the first portion of the lamprey CD4 cytoplasmic domain encodes an abundance of basic amino acids similar to that for CD4, CD4REL, and LAG-3, suggesting that these are related molecules. Our study also shows that all vertebrate LAG-3 molecules (teleosts, amphibians, avians, and mammals) lack the p56LCK association site. This finding suggests that fish CD4 and CD4REL are not equivalent to CD4 and LAG-3 of higher vertebrates, respectively, as was recently speculated (47). Based upon our results we contend that CD4 and CD4REL are more related to CD4 itself, with CD4REL likely representing an ancestral evolutionary step (Fig. 9) for this gene family that was later lost from the genomes of higher vertebrates.

In conclusion, our results provide strong evidence that the modern CD4 gene family (CD4 and LAG-3) was derived through both duplication and selection (via interaction with MHC II) events, as representatives of the primordial and modern CD4 genes were extracted from a variety of teleost fish. Future studies in this area will include CD4-LCK association analyses, determination of whether the teleost CD4 molecules actually bind MHC, and investigation of the involvement of CD4 and CD4REL in T cell development and cell-mediated immunity in teleosts. The presence of the extra CD4-like molecule, CD4REL, may impact T cell signaling in that it may take the place of the conventional CD4 molecule during T cell activation and/or selection, or it may control T cells via distinct mechanisms. We currently do not know the answers to these questions and cannot derive them from mammalian studies because mammals do not possess this molecule. Obviously, one exciting facet of the presence and expression of two CD4-like molecules lays with the possibility that fish possess three distinct T cell subsets, namely those of CD4, CD4REL, and CD8 single positive lineages, which would present new paradigms for T cell lineage commitment in vertebrates.

We thank Scott LaPatra for providing fish, Caird Rexroad III for supplying EST clones, Pierre Boudinot for assistance with zebrafish CD4REL, Stephen Kaattari for providing biotinylated mAb 1-14, Kimberly Keatley for technical assistance during physical mapping procedures, and Charles Cunningham for suggestions in regard to this manuscript.

The authors have no financial conflict of interest.