INTRODUCTION

Emigration of leukocytes from blood into sites of inflammation and immune responses is essential for host defense mechanisms. Local tissue irritation causes leukocytes to stick to blood vessels, to pass through them, and finally to accumulate at irritated sites. It is now known that a family of cytokines called chemokines play important roles in recruiting selected subsets of leukocytes and are involved in a wide range of acute and chronic inflammatory processes as well as other immunoregulatory and hematopoietic functions (1, 2). The known chemokines are divided into two major subfamilies based on the spacing of the first two cysteines in the conserved motif and the chromosomal localization of their genes. The  or CXC chemokine subfamily, which includes IL-8¹ (3) and IP-10 (4), is characterized by the presence of a single amino acid separating the first two cysteines, and the human genes are clustered on chromosome 4q12-21. The two cysteines are adjacent in the  or CC chemokine subfamily, which includes RANTES (5), MCP-1 (6, 7), MCP-2 (8), MCP-3 (9), MIP-1 (10), MIP-1 (11), I-309 (12), and eotaxin (13), and the human genes are located on chromosome 17q11-12. Neutrophils are preferentially attracted and activated by members of the CXC chemokine subfamily, whereas monocytes are preferentially attracted and activated by members of the CC chemokine subfamily. A number of recent studies have revealed that certain  chemokines attract basophils, eosinophils, and lymphocytes with variable selectivity (14, 15, 16, 17, 18). Recently, two novel chemokine-like molecules, mouse lymphotactin and human SCM-1, have been described (19, 20). These cytokines carry only the second and fourth of the four cysteines conserved in other chemokines, suggesting the existence of the  or C type chemokine subfamily. The chemotactic activity of lymphotactin/SCM-1, however, remains to be vigorously proven.

The specific effects of chemokines on the target cell are mediated by seven-transmembrane G-protein-coupled receptors (21). To date, at least five human CC chemokine receptors have been identified by cDNA cloning. CC CKR1 is a receptor for MIP-1, RANTES, and MCP-3 (22, 23); CC CKR2A for MCP-1 (24); CC CKR2B for MCP-1 and MCP-3 (24); CC CKR3 for eotaxin (13); CC CKR4 for MIP-1, RANTES and MCP-1 (25); CC CKR5 for MIP-1, MIP-1, and RANTES (26). In addition to these receptors, a promiscuous chemokine receptor which binds both CXC and CC chemokines is found mainly on the surface of erythrocytes (27, 28, 29).

Recently, a new cloning method aiming at selective identification of cDNA species encoding secretory proteins and type I membrane proteins was introduced (30). The method coined as signal sequence trap took advantage of the presence of NH₂-terminal signal sequences in most precursor forms of secretory proteins and type I transmembrane proteins, which are necessary for the proper orientation of the NH₂-terminal of mature forms inside endoplasmic reticulum and exocytotic vesicles. This method enables to selectively clone cDNA species encoding intercellular signal-transducing molecules without biologic assays. In the present study, we have developed an efficient signal sequence trap method based on an Epstein-Barr virus shuttle vector and applied it to phytohemagglutinin (PHA)-stimulated peripheral blood mononuclear cells (PBMC). Novel cDNA fragments encoding potential signal sequences were identified and one of them led to identification of a novel CC chemokine that is constitutively expressed in thymus and highly selective for T cells.

EXPERIMENTAL PROCEDURES

Human hematopoietic cell lines were maintained in RPMI 1640 supplemented with 10% fetal calf serum. 293/EBNA-1 (Invitrogen, San Diego, CA) and HeLa cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. PBMC were isolated from venous blood obtained from healthy adult donors using Ficoll-Paque (Pharmacia, Uppsala, Sweden). Monocytes were purified by positive selection using anti-CD14 antibody and MiniMACS (Miltenyi Biotec, Bergisch, Germany). Lymphocytes were obtained by centrifugation on an iso-osmotic Percoll gradient as described (31). Human peripheral blood T lymphocytes were purified by negative selection using anti-CD16, anti-CD14, and anti-CD20 antibody with Dynabeads (Dynal, Oslo, Norway), and were stimulated with PHA (1:100) (Life Technologies, Inc., Gaithersburg, MD) and PMA (100 ng/ml)(Sigma). Granulocytes were purified from the pellet of Ficoll-Paque gradient by dextran sedimentation and lysis of erythrocytes. Erythrocytes were obtained by centrifugation and removing of the packed leukocytes on top of erythrocytes. Cells were analyzed cytologically by staining with Diff-Quik (Baxter Scientific Products, McGaw Park, IL) and immunologically by staining with appropriate monoclonal antibodies and flow cytometory on FACStar Plus (Beckton Dickinson, Mountain View, CA). The cell populations used were always 95-98% pure.

The signal-sequence trap vector, pDREF-CD4ST, was made as follows. Signal sequence-deleted CD4 cDNA was amplified by PCR using primers CD4-Trap (+5-CGCTCTAGAAAGAAAGTGGTGCTG-3) and CD4R (5-CGCGCGGCCGCTCAAATGGGGCTACATGT-3), and ligated downstream of the EF-1 promoter in pDREF, an Epstein-Barr virus vector, which contains the EF-1 promoter (32) and the hygromycin-resistant gene. 5 portion-enriched cDNAs were made using the 5-rapid amplification of cDNA ends technique (33) coupled with random-primed first-strand cDNA synthesis. In brief, first strand cDNAs were synthesized from 5 µg of poly(A)⁺ RNA of PBMC stimulated with PHA for 72 h by using 150 ng of random hexanucleotide primers and Superscript II (Life Technologies, Inc.). After mRNA templates were destroyed by alkali-treatment, an anchor oligo(dC) sequence was added to the 3 ends of cDNAs by using terminal deoxynucleotidyl transferase and dCTP (Life Technologies, Inc.). The second strand DNA was synthesized by priming with an anchor primer which contained SalI site and oligo(dG) (5-CTACTACTACTAGGCCACGCGTCGACTAGTACGGGGGGGGGGGGGGGG-3). The double stranded DNAs were sonicated to prepare short fragments. After blunting with T4 DNA polymerase, UNI-Amp adaptors (Clontech, Palo Alfo, CA) were ligated to the fragments. DNA fragments were separated by gel electrophoresis in a 2% Nusive-agarose gel, and fragments consisting of 300-600 bp were recovered using the PCR prep kit (Promega, Madison, WI). The fragments that contained 5 non-coding regions and partial coding regions were then amplified by PCR using UNI-Amp primer (+5-CCTCTGAAGGTTCCAGAATCGATAG-3)(Clontech) and Universal amplification primer (-5-CTACTACTACTAGGCCACGCGTCGACTAGTAC-3)(Life Technologies, Inc.). The samples were subjected to 30 cycles of denaturation (45 s at 94 °C, the first cycle for 3 min), annealing (45 s at 58 °C), and elongation (2 min at 72 °C, the last cycle for 5 min) on a Thermal Cycler (Perkin-Elmer, Norwalk, CT). Amplified DNA fragments were digested with SalI and XbaI, separated by gel electrophoresis, and recovered by the PCR prep kit (Promega). The resultant fragments were introduced into the SalI and XbaI sites of the pDREF-CD4ST vector. Raji cells (1 × 10⁷/500 µl of phosphate-buffered saline) were transfected with the library by electroporation at 250 V and 500 microfarads using Gene Pulser (Bio-Rad). One day after transfection, hygromycin was added at 200 µg/ml to obtain stable transformants. In our experience about 10⁶ stable transformants were routinely obtained from 10⁷ cells. The expanded stable transformants were washed and incubated with OKT4 monoclonal antibody at 4 °C for 30 min with constant agitation. After washing, cells were incubated with anti-mouse IgG Dynabeads (Dynal)(5 beads per cells) at 4 °C for 30 min with constant rotation. Cells coated with magnetic beads were separated using Magnetic separator (Advanced Magnetics, Cambridge, MA). After three rounds of growing and sorting, plasmids were isolated from the cells by alkaline lysis procedure and recovered into Escherichia coli. The plasmids prepared from individual colonies were reintroduced into Raji and clones that were capable of directing expression of CD4 antigen on the cell surface were identified. The length of insert fragments were determined by PCR using EF seq F primer (+5-CCTCAGACAGTGGTTCAAAG-3) and CD4ST seq R primer (5-TGTACAGGTCAGTTCCACTG-3). The cDNA fragments longer than 200 bp were analyzed by sequencing.

A conventional cDNA library of PBMC stimulated with PHA for 72 h was constructed using cDNA synthesis kit (Life Technologies, Inc.). The library was screened with clone 98, one of the trapped fragments containing putative signal sequences, and one positive clone D3A was isolated. The full-length clone and subclones were sequenced using Autoread Sequence kit and A.L.F. sequencer (Pharmacia).

Poly(A)⁺ RNAs prepared from PBMC were separated by electrophoresis on a 1% agarose gel containing 0.66  formaldehyde. The gel was blotted onto a filter membrane (Hybond N+) (Amersham Japan, Tokyo). Multiple tissue Northern blots were purchased from Clontech. Hybridization was carried out at 42 °C with 50% formamide, 5 × SSPE, 2% SDS, and 100 µg/ml sonicated salmon sperm DNA. The probe was the SmaI-PstI fragment of clone D3A. Probes for other chemokines were prepared by reverse transcriptase-PCR and described previously (34). The -actin probe was purchased from Clontech. The fragments were labeled with ³²P using the Multiprime DNA labeling system (Amersham Japan). After washing at 60 °C with 0.2 × SSC and 0.1% SDS, filters were exposed to x-ray films at 80 °C with intensifying screens.

The fragment encoding the predicted mature form of TARC was prepared by PCR using primer A (+5-CGCGGATCCGCTCGAGGGACCAATGTG-3) and primer B (5-CGCGCGGCCGCTCAAGACCTCTCAAGGCT-3), and cloned into BamHI-NotI sites of pGEX3 (Pharmacia). The TARC-glutathione S-transferase fusion protein that was expressed in E. coli and purified as described previously (35) was injected into guinea pigs by routine protocols. The predicted mature TARC was also expressed as a non-fusion protein in E. coli using the T7 polymerase system (Promega). The fragment generated by PCR using primer C (+5-CGCGGATCCGCTCGAGGGACCAATGTG-3) and primer B was cloned into NdeI-NotI sites of pGEMEX-1 (Promega), and inclusion bodies were obtained.

HeLa cells were infected with recombinant vaccinia virus expressing T7 RNA polymerase and transfected with pSPORT1 vector or pSPORT1 containing TARC cDNA as described previously (36). After 6 h, the transfection mixture was removed. The cells were cultured in fresh Dulbecco's modified Eagle's medium without fetal calf serum for 12 h and the supernatants were harvested. The culture supernatants were concentrated by trichloroacetic acid precipitation. Samples were separated on 15% Tris-Tricine SDS-PAGE gels (PAGEL) (ATTO, Tokyo, Japan) and transferred to a membrane (Clear Blot Membrane-P) (ATTO). Immunoblot analysis was performed with anti-TARC antiserum diluted 1:1000 and horseradish peroxidase-conjugated protein A (Amersham Japan) as described previously (36). For NH₂-terminal sequence analysis, recombinant TARC was concentrated by cation-exchange Hitrap-SP column (Pharmacia).

Sf9 cells and Tn5B1-4 cells (Invitrogen) were maintained at 27 °C in EX-CELL 400 medium (JRH Biosciences, Lenexa, KS). The baculovirus transfer vector pVL1393 was purchased from Invitrogen. The TARC transfer vector, pVL-TARC, was made by ligating the EcoRI-NotI fragment of the clone D3A into EcoRI-NotI sites of pVL1393. A linearized AcMNPV DNA containing a lethal deletion (Clontech) and pVL-TARC were cotransfected into Sf9 cells and the recombinant baculovirus was isolated by limiting dilution. For the expression of recombinant TARC, Tn5B1-4 cells were plated at 1.2 × 10⁷ cells onto 150-cm² flasks and infected after 2 h at multiplicity of infection of 10-20 with the recombinant baculovirus produced in Sf9 cells. The culture supernatants collected 2 days after infection were cleared by filtration through 0.22-µm membranes. The filtrate was mixed with 1/10 volume of 500 m MES (pH 6.5) and applied to 1-ml cation-exchange Resource-S column (Pharmacia) equilibrated with 50 m MES (pH 6.5), 100 m NaCl. The column was washed with 50 m MES (pH 6.5), 100 m NaCl and eluted at a rate of 1 ml/min with a 45-ml linear gradient of 0.1-1.0 NaCl in 50 m MES on the FPLC system (Pharmacia). The fractions containing recombinant TARC were identified on SDS-PAGE and pooled. After adding trifluoroacetic acid to 0.1%, the pooled fraction was injected into a reverse-phase HPLC column (12 × 220 mm Cosmocil 5C4-300) (Cosmobio, Tokyo, Japan) equilibrated with 0.1% trifluoroacetic acid. The column was eluted with a 60-ml linear gradient of 0-60% acetonitrile in 0.1% trifluoroacetic acid at a flow rate of 1 ml/min. Fractions containing recombinant TARC were pooled, vacuum dried to remove acetonitrile, and dialyzed against endotoxin-free phosphate-buffered saline. Protein concentrations were determined by the BCA kit (Pierce). Endotoxin levels were determined by the Limulus amoebocyte lysate assay (QCL-1000) (Bio Whitaker, Walkersville, MD) and were always <4 pg/µg of recombinant TARC. NH₂-terminal sequence analysis was performed on a protein sequencer (Shimazu, Tokyo, Japan).

The purified recombinant TARC was radiolabeled using ¹²⁵I-labeled Bolton-Hunter reagent (Amersham Japan) to a specific activity of 81.6 or 16.3 µCi/µg. For displacement experiments, cells were incubated for 1 h at 15 °C with ¹²⁵I-labeled TARC in the presence of increasing concentrations of unlabeled chemokines in 200 µl of RPMI 1640 containing 20 m Hepes (pH 7.4), 1% bovine serum albumin, and 0.02% NaN₃. For saturation experiments, cells were incubated with increasing concentrations of radiolabeled TARC in the presence or absence of 1 µ unlabeled TARC. After incubation, cells were separated from unbound radiolabeled ligands by centrifugation through a mixture of dibutyl phtalate/olive oil (4:1). All samples were determined in duplicate. The binding data were analyzed by the LIGAND program (37).

The chemotaxis assay was performed using a 48-well microchemotaxis chamber as described (38, 39). Samples were assayed in triplicate. For testing T cell lines, chemoattractants were diluted in HEPES-buffered RPMI 1640 supplemented with 1% bovine serum albumin and placed in the lower 25-µl wells. Cells were resuspended in RPMI 1640, 1% bovine serum albumin at 8 × 10⁶ cells/ml and 50 µl was added to the upper wells that were separated from the lower wells by a polyvinylpyrrolidone-free polycarbonate filter with 5-µm pores precoated with type IV collagen (39). The chamber was incubated for 120 min at 37 °C in 5% CO₂, 95% air. The filters were removed and stained with Diff-Quik. Migrated cells were counted in five randomly selected high-power fields (×800) per well. For testing monocytes, PBMC at a concentration of 4 × 10⁶ cells/ml were added to the upper wells separated by polyvinylpyrrolidone-free polycarbonate filter with 5-µm pores. The chamber was incubated for 90 min. For testing neutrophils, phosphate-buffered saline supplemented with 1 m CaCl₂, 1 m MgSO₄, and 0.5% bovine serum albumin was used for assay buffer. Cells were used at 2 × 10⁶ cells/ml and incubated for 90 min. In neutralizing experiments, recombinant TARC were incubated with affinity-purified polyclonal anti-TARC antibody or normal guinea pig IgG at 4 °C for 30 min and then used for chemotaxis assay. In the case of Bordetella pertussis toxin treatment, Hut78 cells were incubated for 90 min at 37 °C without or with B. pertussis toxin, washed, and used for the assay.

RESULTS

We generated a new signal sequence trapping vector, pDREF-CD4ST, which was based on an Epstein-Barr virus shuttle vector and contained the signal sequence-deleted CD4 cDNA for a reporter protein (Fig. 1A). Since the vector can be maintained as episomes in the presence of EBNA-1, this system allows an efficient stable expression of inserted cDNAs and an easy recovery of plasmids. As a pilot experiment, we inserted a fragment containing the signal sequence of CD30 (40) into pDREF-CD4ST. The CD4 antigen was detected on the surface of Raji cells transfected with the construct (Fig. 1B). No such expression of CD4 antigen was seen when Raji cells were transfected with the vector alone. Exogenous signal sequences would thus allow CD4 fusion proteins to be expressed on the cell surface if cloned in-frame.

We generated 5 portion-enriched cDNA fragments from PHA-stimulated human PBMC by using the 5-rapid amplification of cDNA ends method (33) coupled with random-primed first-strand cDNA synthesis. The cDNA fragments of around 300 to 600 bp were inserted into pDREF-CD4ST, and about 10⁶ independent clones were obtained. The expression library was then transfected into Raji cells and stable transformants were selected. The initial frequency of the CD4 antigen-positive cells was approximately 0.1% as determined by flow cytometric analysis. After expansion, CD4 antigen-positive cells were sorted and enriched to 18.2%. The antigen-positive cells increased to 29.7% after the second expansion and sorting, and to 44.8% after the third expansion and sorting. No appreciable increase in the CD4 antigen-positive cells was obtained by the fourth expansion and sorting. We therefore rescued plasmids from the cells into E. coli. Plasmids were prepared from randomly selected 100 colonies and individually reintroduced into Raji cells. We identified 42 clones that were capable of directing surface expression of CD4 antigen. Considering that short fragments might fortuitously encode hydrophobic sequences in unnatural open reading frames, we only sequenced 36 clones possessing inserts longer than 200 bp. Comparison of the sequences with the data bases revealed that 12 clones were reported previously and 24 clones were unknown. Among the 12 known clones, 9 clones were positive for the anchor oligo(dC) that had been introduced at 5 ends of mRNA. The remaining 3 clones were negative for the oligo(dC) anchor and derived from mitochondrion DNA. Signal sequences were present in 6 out of 9 oligo(dC)-positive clones. These were derived from CD6 (41), MGC24 (42), TCR (43), and 2-microglobulin (3 independent clones) (44). Among 24 unknown clones, 13 clones were positive for the anchor oligo(dC). Hydrophathy analysis of the 13 oligo(dC)-positive clones revealed that at least 4 clones possessed hydrophobic profiles resembling to signal sequence.

We selected clone 98 for further study because it presented a characteristic feature of CC chemokine, i.e. a double-cysteine motif nine residues downstream of the putative signal sequence cleavage site (1, 2). In order to isolate the full-length cDNA, we screened a conventional cDNA library of PHA-stimulated PBMC with clone 98 and obtained a clone D3A (Fig. 2). The cDNA is 538 bp in length and has an identical sequence with the clone 98 in its 5-overlapping region. The clone 98 had extra 12 nucleotides in the 5 side. The sequence of clone D3A contains a single long open reading frame that starts with the 5-proximal methionine codon at nucleotide 59 and encodes a highly basic polypeptide of total 94 amino acids with a calculated molecular weight of 10,507. The deduced polypeptide sequence contains a highly hydrophobic amino-terminal region characteristic of a signal peptide with a putative cleavage site between Ala-23 and Ala-24. The predicted mature protein has an isoelectric point of 9.7 and a molecular weight of 8,083. There are no potential N-glycosylation sites. The 3-untranslated region contains a potential polyadenylation signal (ATTAAA). The ATTTA motif which is frequently found in the 3-noncoding sequences of cytokines and involved in rapid degradation of mRNA (45) is not observed. The predicted mature protein shows significant homology to CC chemokines and all the four cysteine residues conserved in the CC chemokine family are present (Fig. 2). The identity is 29% with RANTES (5), 28% with MIP-1 (11) and MCP-3 (9), 26% with MIP-1 (10), and 24% with I-309 (12), MCP-1 (6, 7) and MCP-2 (8). The protein encoded by clone D3A is thus a novel member of the CC chemokine subfamily.

We examined expression of D3A mRNA in PBMC by Northern blot hybridization (Fig. 3A). The expression of D3A was undetectable in fresh PBMC. After stimulation with PHA, the transcripts of about 0.8 kilobases accumulated to maximum levels at 24 h and returned to low levels by 72 h. On the other hand, the transcripts for MIP-1 were present in fresh PBMC, rapidly increased by PHA-stimulation with a peak at 4 h, and returned to resting levels by 24 h. Next, we examined the expression of D3A mRNA in various tissues (Fig. 3B). The transcripts were detected strongly in thymus, and weakly in lung, colon, and small intestine. Compared to other five chemokines examined in parallel, D3A is quite unique for its constitutive expression in thymus and lack of such expression in spleen or PBL. From the constitutive expression in thymus and stringently regulated induction in PBMC by activation, we designated this novel member of the CC chemokine family as TARC (Thymus and Activation-Regulated Chemokine).

To demonstrate TARC as a secretory protein, the TARC cDNA was expressed in HeLa using the vaccinia/T7 system. The culture supernatants from transfectants were analyzed by immunoblotting using polyclonal antiserum prepared against glutathione S-transferase-TARC fusion protein produced in E. coli. A single protein of approximately 8 kDa was detected by the antiserum in the supernatant from HeLa transfected with the TARC cDNA but not in the supernatant from HeLa transfected with the vector alone (Fig. 4). The size of the secreted protein was close to the predicted value of the mature TARC. Amino acid sequence analysis demonstrated that the NH₂ terminus of recombinant TARC secreted from HeLa started at Ala-24 of the predicted sequence. These results indicate that the predicted signal sequence and the cleavage site of TARC are functional and correct, respectively.

Recombinant TARC was produced in Tn5B-1 insect cells infected with a recombinant baculovirus and purified from culture supernatants by cation-exchange chromatography and reverse-phase HPLC. Recombinant TARC was eluted from the reverse-phase column as a single peak (Fig. 5A). The yield of purified protein was typically 1-1.5 µg/ml of starting culture supernatant. When analyzed by SDS-PAGE and silver staining, the purified protein migrated as a single band of 8 kDa (Fig. 5B). Amino acid sequence analysis demonstrated that the NH₂ terminus of recombinant TARC started at Ala-24 of the predicted sequence, indicating that insect cells correctly removed the signal peptide (Fig. 5C). Immunoblot analysis showed that TARC derived from HeLa or insect cells comigrated with bacterial recombinant TARC lacking the signal sequence (Fig. 4). These results indicate that there are no post-translational modifications such as glycosylation or proteolytic processing in mature TARC.

As a first step for elucidation of the biological activity of TARC, we determined the distribution of receptors for TARC on various cell types (Fig. 6). A considerable specific binding of TARC was detected on a number of T cell lines, PBL, and peripheral blood T cells activated by PHA/PMA. On the other hand, myelomonocytoid cell lines and peripheral monocytes showed only a marginal, if any, specific binding of TARC, while Raji, 293/EBNA-1, and peripheral granulocytes showed virtually no specific binding. To characterize the TARC receptor on T cells, further binding experiments were performed with Jurkat. Binding of radiolabeled TARC reached equilibrium by 1 h at 15 °C. When the binding was performed with increasing concentrations of ¹²⁵I-TARC (Fig. 7A), a single class of receptor with a K_d of 2.1 n and 603 sites/cell was observed (Fig. 7B). Competition binding experiments showed that unlabeled TARC fully competed the binding of ¹²⁵I-TARC (Fig. 7C). Scatchard analysis of the competition data showed a single class of receptor with a K_d of 2.1 n and 948 sites/cell. None of the tested CXC and CC chemokines (IL-8, RANTES, MCP-1, and MIP-1) showed significant competition for ¹²⁵I-TARC (Fig. 7D). Similar results were obtained with Hut 78 cells and peripheral lymphocytes activated by PHA/PMA (data not shown).

It is known that erythrocytes possess a promiscuous chemokine receptor (28, 29). Radiolabeled TARC indeed bound to erythrocytes (Fig. 8). Competition experiments showed that not only unlabeled TARC but also heterologous chemokines with the exception of MIP-1 also inhibited ¹²⁵I-TARC binding to erythrocytes with similar dose-response profiles. The K_d for TARC was 17 n which is comparable to the reported values of other chemokines (28, 29). Collectively, these results indicate that TARC binds to a highly specific receptor(s) expressed mainly, if not exclusively, on T cells and to the promiscuous chemokine receptor expressed on erythrocytes.

Human T cell lines, Hut78, Hut102, and Jurkat, as well as fresh monocytes, neutrophils, and lymphocytes were assessed for their migration across a polycarbonate filter in response to TARC (Fig. 9). Consistent to little expression of the TARC receptor, no chemotactic response was elicited in monocytes or neutrophils by TARC. These cells, however, efficiently respond to respective positive controls, MCP-1 and IL-8 (Fig. 9). TARC induced migration in T cell lines, Hut78 and Hut102, with a typical bell-shaped curve with a maximum effect at 100 ng/ml. Notably, no other chemokines tested in parallel (IL-8, RANTES, MCP-1, MIP-1, MIP-1, and SCM-1/lymphotactin) were capable of inducing such chemotactic responses in Hut78 or Hut102 (Fig. 10A and data not shown). In the case of Jurkat, TARC induced migration at least in some experiments. So far, we could not demonstrate significant migration above background levels of fresh or PHA/PMA-activated peripheral T cells to TARC. It is conceivable that our chemotactic assay is not proper for normal T cells. It is also possible that, in addition to expression of the TARC-specific receptor, distinct phenotype and/or activation status of T cells represented by Hut78 and Hut102 cells may be required for the chemotactic response to TARC.

A checkerboard analysis confirmed that the migration of Hut 78 cells toward TARC was chemotactic rather than chemokinetic (data not shown). Anti-TARC antibody raised by immunizing guinea pigs with recombinant TARC almost completely neutralized TARC in induction of chemotaxis in Hut78 (Fig. 10A). Finally, pretreatment of Hut78 with pertussis toxin abolished chemotactic response to TARC in a dose-dependent manner with IC₅₀ at 2 n (Fig. 10B), indicating that the chemotactic response to TARC was mediated by a G_i- or G_o- subclass G-protein-coupled receptor. We, however, have not detected any Ca²⁺ mobilization in response to TARC in T cells including Hut78 and Hut102 so far. The sensitivity of the assay may be too low to detect elevation of cytosolic calcium in restricted areas proximal to activated TARC receptors, or TARC-induced signal transduction may be independent of Ca²⁺ mobilization as suggested for some other chemokines (46, 47).

DISCUSSION

Signal sequence trap is a recently developed method for selectively cloning cDNAs for secretory molecules and receptors without use of specific functional assays (30). In the present study, we have developed a modified version of signal sequence trap using an Epstein-Barr virus trapping vector, pDREF-CD4ST (Fig. 1) and applied this system to PHA-stimulated PBMC. By using one of the trapped signal sequences as a probe, we have identified a novel CC chemokine that we designate as TARC (Fig. 2). Even though several CC chemokines are known to induce chemotaxis in T cells, TARC appears to be the first CC chemokine highly selective for T cells (Fig. 6). High levels of specific TARC binding sites were detected only in some T cell lines, PBL and peripheral blood T cells activated by PHA/PMA. On the other hand, little if any specific binding was detected on myelomonocytoid cell lines, monocytes, or granulocytes. Our recent studies also showed little specific binding on peripheral B cells or NK cells.² In contrast to other known CC chemokine receptors that usually bind more than one chemokine, the TARC receptor on T cells appears to be highly specific for TARC. Only TARC competed for the binding of radiolabeled TARC, while IL-8, RANTES, MCP-1, and MIP-1 had virtually no effect on the TARC binding (Fig. 7). Furthermore, IL-8, RANTES, MCP-1, MIP-1, MIP-1, and SCM-1/lymphotactin failed to induce significant chemotactic responses in the T cell lines that are capable of responding to TARC (Fig. 10A). We have also confirmed that TARC does not bind to HEK293 cells transfected with CC CKR1 (22, 23), CC CKR2B (24), CC CKR3 (13), EBI1 (49), BLR1 (50), LESTR (39), and two other potential chemokine receptor cDNAs.² These results indicate the presence of a unique receptor for TARC, which is most probably a G_o- or G_i- subclass G-protein-coupled, seven-transmembane domain receptor (Fig. 10B) as most other CC chemokine receptors. Molecular cloning of the TARC receptor will help to examine whether its expression is indeed selective for T cells and what biochemical events are required for T cells to respond to TARC. TARC also binds to DARC on erythrocytes (Fig. 8). Recently, DARC has been shown to be expressed on endothelial cells lining postcapillary venules (51, 52). It remains to be seen whether DARC mediates the effect of TARC on these cells.

Consistent to little specific binding, TARC did not induce chemotactic response or Ca²⁺ mobilization in monocytes or granulocytes at all. On the other hand, TARC clearly induced chemotaxis in two human T cell lines, Hut78 and Hut102 (Fig. 9). TARC, however, had little significant effect on the migration of peripheral blood resting or activated T cells above background levels in spite of considerable levels of specific binding sites for TARC (Fig. 6). The possibility that activated T cells were desensitized to TARC by an autocrine or paracrine mechanism was excluded because T cells do not produce TARC (see below) and we did chemotactic assays using T cells that were activated in the absence of other accessory cells. First of all, our chemotactic assay for normal T cells probably needs further improvements. We should mention that even RANTES that was used as a positive control had trouble in demonstrating significant migration of normal T cells above high levels of background in our assay. It is, nevertheless, noteworthy that substantial levels of TARC binding were observed only in some of the T cell lines examined (Fig. 6). Thus, TARC receptor may be expressed only in a particular subset of normal T cells. In addition, the status of activation may affect post-receptor signaling events. Hut78 and Hut102 are cutaneous T cell lymphoma lines having unique properties of mature and activated T cells. It remains to be seen whether phenotypic heterogeneity of T cells such as CD4⁺ versus CD8⁺, naive versus memory, resting versus activated, and Th subtypes critically affect responses of normal T cells to TARC.

TARC is induced in PBMC by PHA-stimulation (Fig. 3). It remains to be determined which types of cells produce TARC and whether induction occurs directly by mitogenic stimulation or secondarily by secreted cytokines. In this context, TARC was not induced in Jurkat by PHA-treatment nor in U937 by treatment with interferon-, tumor necrosis factor-, IL-1, or IL-4.² Similarly, murine TARC was not induced in mouse thymocytes or splenocytes by stimulation with ConA, anti-Thy1, anti-CD3, or PHA/PMA.³ Interestingly, however, our recent results showed that TARC was strongly induced in monocytes by cytokines which are known to be produced by Th2 type T cells, suggesting that TARC may play roles in humoral immunity.⁴ TARC is also constitutively and selectively expressed in thymus (Fig. 3). For comparison, we examined expression of other chemokines in the same multiple tissue Northern blots. RANTES was expressed strongly in PBL, spleen, and small intestine. MIP-1 was expressed strongly in spleen and PBL. MCP-1 was expressed strongly in most tissues except for PBL and brain. IP-10 was detected in spleen, thymus, PBL, and lung at low levels (Fig. 3). Gattass et al. (53) reported constitutive expression of murine IP-10 in lymphoid organs (spleen, thymus, and lymph nodes) as well as in liver. Thymic and splenic stromal cells were found to produce IP-10. Our recent studies have revealed that murine TARC is expressed in thymic stromal cells.³

In addition to leukocyte trafficking, some chemokines probably play important roles in the regulation of hematopoiesis and myelopoieis. Several chemokines have been shown to have inhibitory or stimulatory effects on the proliferation of hematopoietic stem cells or myeloid progenitor cells (54, 55, 56, 57). Mice lacking the IL-8 receptor homologue exhibited abnormal expansion of neutrophils and B lymphocytes, suggesting a role of this receptor in the regulation of expansion and development of neutrophils and B cells as well as in chemotaxis of neutrophils (58). A murine CXC chemokine PBSF/SDF-1 was shown to be produced by bone marrow stromal cell lines and to augment growth of pre-B cells in the presence of IL-7 (30, 59). The constitutive and selective expression of TARC in thymus suggest that TARC is a factor involved in the function of thymus such as T cell development or homing. It remains to be seen whether TARC acts on any thymocytes or only on particular subsets of thymocytes. Isolation of the mouse homologue of TARC and production of the recombinant murine protein now in progress will help us answer these questions. Generation of TARC-knockout mice will also be useful for elucidation of physiological functions of TARC.