Abstract
By the genetic selection of mouse cDNAs encoding secreted proteins, a B7-like cDNA clone termed mouse GL50 (mGL50) was isolated encoding a 322-aa polypeptide identical with B7h. Isolation of the human ortholog of this cDNA (hGL50) revealed a coding sequence of 309 aa residues with 42% sequence identity with mGL50. Northern analysis indicated GL50 to be present in many tissues including lymphoid, embryonic yolk sac, and fetal liver samples. Of the CD28, CTLA4, and ICOS fusion constructs tested, flow cytometric analysis demonstrated only mouse ICOS-IgG binding to mGL50 cell transfectants. Subsequent phenotyping demonstrated high levels of ICOS ligand staining on splenic CD19+ B cells and low levels on CD3+ T cells. These results indicate that GL50 is a specific ligand for the ICOS receptor and suggest that the GL50-ICOS interaction functions in lymphocyte costimulation.
The costimulatory molecules B7-1 (CD80) and B7-2 (CD86) are related members of the Ig superfamily of proteins with structural similarities reflected by the ability of both of these molecules to signal through CD28, an activating surface receptor protein, and through CTLA4 (CD152), a putative inhibitory receptor protein (1). A novel surface receptor termed ICOS (Geneseq database no. X37661; Ref. 2) or JTT.1 (Geneseq database nos. V53199, V53198, V53200) was recently described that had significant sequence identity with CD28 (24%) and CTLA4 (17%). Unlike CD28, ICOS/JTT.1 was shown to be up-regulated on stimulated T cells and caused the secretion of a panel of cytokines distinct from those mediated by CD28 costimulation (2). The ligand that activates ICOS/JTT.1 has not been identified, but structural considerations suggested that an additional B7-like protein might serve this role.
A recent innovation in the isolation of cDNA encoding secreted proteins has been developed (3) permitting cDNAs encoding signal peptides to be selected and used as a probe for isolation of the full length cDNA sequence. Using lymph nodes from IL-12-treated tumor-bearing mice as a source of cDNA, a library of signal trapped clones were obtained, of which one B7-like clone was found to be ICOS ligand.
Materials and Methods
Tissue isolation
Mice (C57BL/6) injected with MB49 bladder carcinoma cells were treated with 1 μg/mouse recombinant IL-12 on days 7–11 and 14–18. Day 9, 12, and 19 lymph nodes were pooled. Wild-type spleen, yolk sac, and fetal liver were isolated from timed pregnant Swiss Webster mice. Splenic tissues were also isolated from Rag−/− mice. Human PBL were isolated by Ficoll-Paque density centrifugation of human leukopac samples.
Nucleic acid manipulation
RNA was extracted using RNAStat 60 total RNA isolation reagent (Tel-Test B, Friendswood, TX). cDNAs were synthesized with Superscript RT enzyme (Life Technologies, Rockville, MD) in 20-μl reactions. Additional cDNA sources include a previously described C3H/Hej fetal thymus library (4). 3′ Rapid amplification of cDNA end (RACE)2 was performed with 5 μg of total RNA according to the method of Frohman (5). Signal sequence trap was performed (3) followed by Rec A isolation of cDNAs based on a method by Rigas et al. (6). Full-length PCR isolation of human GL50 (hGL50; GenBank accession no. AF199028) was accomplished by using the oligonucleotide primers 5′-GGCCCGAGGT CTCCGCCCGC ACCATG-3′ and 5′-TCACGAGAGC AGAAGGAGCA GGTTCC-3′ according to previously described conditions (7).
Sequence analysis
TBlastX, FastX, pFam, Pileup, and Sigcleave of Genetics Computer Group (Madison, WI) package and Geneworks 2.5.1 were used for DNA sequence manipulation, database searching and sequence analysis. Identity scores for pileup analysis were assigned according to the following values: 1 = 1× pair; 2 = 2× pair; 3 = 3× pair; 4 = 3 of a kind; 5 = 3 of a kind plus 1× pair; 6 = 2× 3 of a kind; 7 = 4 of a kind; 8 = 4 of a kind plus 1× pair; 9 = 5 of a kind.
Gene expression analysis
For analysis of commercially prepared RNA blots (Clontech, Palo Alto CA), DNA fragments encompassing nucleotides 1065–1588 of mouse GL50 (mGL50; GenBank accession no. AF199027) (494 bp) or nucleotides 84–782 of hGL50 (699 bp) were radiolabeled and hybridized to filter membranes according to the manufacturer’s protocol. Northern analysis of 5 μg of total RNA from developmentally regulated tissues was performed using the NorthernMax-Plus kit (Ambion, Austin, TX).
Cell preparation and culture
Cell suspensions were isolated from BALB/c mice (∼3 mo old). RBC were lysed, and washed splenocytes (∼1 × 107 cells/ml/well) from BALB/c mice were cultured with 25 μg/ml LPS (Sigma), while splenocytes from Swiss Webster mice were cultured with 10 ng/ml PMA and 1 μg/ml ionomycin.
Flow cytometry
The mICOS-mIgG2am protein was made at the Genetics Institute consisting of human oncostatin M leader, predicted extracellular domain of mICOS sequence, and IgG2a hinge-carboxyl tail with mutations introduced to lower Fc binding (8). For analysis of surface proteins, COS cells were transfected using lipofectamine (Life Technologies) with mGL50 in pcDEF3 (9) or kit ligand/membrane (KLM) from cDNA in pED expression vectors. Cells were blocked with PBS supplemented with 10% rabbit serum and then were stained with 0.5 μg of fusion protein in 100 μl of PBS supplemented with 2% FCS. Secondary staining was performed with PE-linked goat anti-mouse IgG. All Ab reagents were purchased from PharMingen (San Diego, CA). Splenocytes were incubated in 20 μg/ml Fc block, 10% rabbit serum, and incubated with either mICOS-mIgG2a or murine IgG2a control followed by secondary staining with biotinylated anti-mouse IgG2a and FITC-conjugated mAb as indicated. Cells were analyzed by flow cytometry using a FACScalibur and CellQuest software (Becton Dickinson, San Jose, CA).
Results and Discussion
Lymph node partial cDNAs were placed under genetic selection for signal peptides using the signal sequence trap method (3). Of a total of 333 cDNA:invertase fusion clones isolated and sequenced, one partial cDNA clone with limited sequence identity with B7-1 was identified and termed mGL50. Four additional overlapping cDNA clones were isolated from a mouse fetal thymus cDNA library. The consensus 2718 nt sequence encoded a 322-aa protein with a predicted mass of 36 kDa (Fig. 1⇓). The hydropathy plot of the open reading frame predicted a leader sequence, an extracellular domain, hydrophobic transmembrane region, and potential intracellular cytoplasmic domain. Pfam structural analysis revealed an IgV-like domain and an IgC-like domain based on domain delineation of B7-1 and B7-2 (10). Comparison of mGL50 revealed identity to the recently published mouse costimulatory molecule, B7h (11).
Pileup analysis of hGL50, mGL50, hB7-1, mB7-1, hB7-2, and mB7-2. Signal peptide, Ig-like domains, transmembrane, and cytoplasmic domains are as indicated. Predicted hydrophobic transmembrane residues are underlined. Black diamonds indicate essential residues in either B7-1 or B7-2 sequences necessary for either CTLA4 or CD28 binding. Asterisks denote residues defined to contribute to Ig structure. Extracellular cysteines and tryptophans, indicators of Ig structure, are in bold. The identity scoring system is described in Materials and Methods.
FastX sequence search of GenBank database using mGL50 yielded a cDNA isolated from human brain, KIAA clone 0653 (accession no. AB014553; Ref. 12), a 4.3-kb cDNA localized on chromosome 21, encoding a putative 558-aa protein with a molecular mass of 60 kDa. The analysis of the first 303 residues of the deduced AB014553 protein sequence indicated strong similarity with the entire length of mouse mGL50, excluding the region of the cDNA corresponding to signal peptide sequences. Because the remaining 255 residues of AB014553 did not have homology to mGL50, it was not likely that AB014553 was the human ortholog of the mouse mGL50 clone.
3′ RACE analysis of human PBL with oligonucleotides primers corresponding to extracellular domains of AB014553 resulted in four RACE products that encoded a cDNA divergent from the AB014553 cDNA sequence and that resulted in a truncated 3′ coding region encoding 9 aa, a termination codon, and a short untranslated domain. By RT-PCR of leukocyte RNA using primers specific for the full-length coding sequence predicted by RACE, an amplified product (hGL50) was isolated encoding a 309-aa protein that shared 42% sequence identity with mGL50 (Fig. 1⇑), predicting this sequence to be the human ortholog of mGL50. Sequence identity levels in the 40% range are also found in orthologous sequence comparisons between mouse and human B7-1 (41%) and mouse and human B7-2 (48%), whereas paralog sequence comparison of GL50, B7-1, and B7-2 demonstrate homology levels at ∼20% identity.
Northern blot analysis using probes specific to 3′ untranslated regions of mGL50 revealed hybridization to a 2.7-kb message in brain, lung, liver, skeletal muscle, and testis with dominant expression in spleen, heart, and kidney samples (Fig. 2⇓A). Northern analysis was performed on 5 μg total RNA extracted from either day 11.5 yolk sac, day 14.5 fetal liver, Rag-2−/− spleen, activated splenocytes, and WEHI 231 (7). With the exception of WEHI 231 samples, 2.7-kb hybridization signals were detected clearly in all samples with the highest levels of signal in activated splenocytes (Fig. 2⇓B). Phosphoimage quantitation revealed the 2.7-kb hybridization signal to be 2-fold greater in activated splenocytes than whole spleen samples. Rag-2−/− mouse splenocytes, genetically devoid of mature B and T cells, also exhibited levels of transcripts commensurate to wild-type splenocytes, suggesting the presence of mGL50 on cells other than mature B and T cells.
Northern analysis of GL50 transcript distribution. A, Northern blot of GL50 transcripts detected in mouse multiple tissue RNA panel. B, Northern blot of GL50 transcripts detected in RNA derived from developmental and PMA-ionomycin-activated tissues. C, Northern blot of GL50 transcripts detected in human multiple tissue RNA panel. Molecular size markers are indicated in kb.
mGL50 transcripts were detected very early on in embryonic hemopoiesis starting from day 11.5 yolk sac at levels similar to that found in adult spleen, with less during fetal hemopoiesis in the fetal liver. The presence of mGL50 transcripts in the embryo-derived yolk sac tissue leads to the intriguing possibility that mGL50 may function to skew the cytokine profiles of maternal T cells toward immunoprotective Th2 phenotype during pregnancy (7, 13).
Northern hybridizations of human multiple tissue panels indicated the presence of a number of transcripts found in all tissues with approximate molecular sizes of 2.4, 3.0, and 7.0 kb, with the highest levels of signal present in brain, heart, kidney, and liver samples, while low hybridization signals were detected in colon and thymus. Additional transcripts of 8.5 and 3.8 kb in certain samples were also detected (Fig. 2⇑C). A unique 1.1-kb transcript detected only in PBL samples corresponded to the predicted size of hGL50. None of the obvious transcripts correlate with the 4.3-kb published AB014553 cDNA, and, in addition, no PCR products were obtained from any cDNA source when amplified using primers specific to the sequences of hGL50/AB014553 and the 3′ sequences of AB014553 not shared with hGL50 (data not shown).
To determine the extent of relatedness between mGL50, hGL50, and human and mouse B7-1 and B7-2, protein sequence alignments were performed. From Pileup analysis (Fig. 1⇑), 18 aa locations aligned identically between all six molecules within the extracellular domain. Of the 32 positions that define the predicted IgV-like and IgC-like folds of the B7 molecule, 13 are identically conserved between all six molecules, including the four cysteines that allow intramolecular folding of domains. The identities of mGL50 and hGL50 sequences in some locations aligned more closely with B7-1 (e.g., mGL50 Y87) and in other locations more similarly with B7-2 (e.g., mGL50 V86). Of the 16 positions in GL50 with identity scores of 8, five positions were shared by B7-1, four positions were shared by B7-2, and six positions are shared between B7-1 and B7-2. Five locations in the GL50 sequences exhibited complete identity with B7 residues critical for CTLA4/CD28 binding (10), while six sites exhibited similarity with respect to conservative amino acid substitutions. These results suggest that the GL50 sequences occupies a phylogenetic space nearly parallel to the B7 family of proteins.
To date, three receptors of the CD28 family of proteins have been reported to share similar structure: CTLA4, CD28, and ICOS (2). Flow cytometric assays were performed to determine whether GL50 proteins were able to bind to either CTLA4, CD28, or ICOS. COS cells transiently expressing mGL50 were stained with either mICOS-mIgG2am, mCD28-IgG2am, or mCTLA4-Ig2am fusion proteins (7), resulting in detection of binding by mICOS-mIg2am (20%) fusion protein, but not by either mCD28-mIg2am or mCTLA4-Ig2am fusion proteins (Fig. 3⇓), although we cannot rule out the possibility of very-low-affinity binding and signaling of mGL50 through CD28 or CTLA-4 (14). No binding of any fusion protein was detected for the control KLM cDNA transfectants, suggesting that mGL50 is a ligand specific for mICOS-Ig2am.
Identification of ICOS ligand. mGL50 or negative control KLM was expressed in COS cells followed by staining with either mICOS-IgG2am, mCD28-IgG2am, or mCTLA4-IgG2am and followed by detection with anti-mouse IgG-PE. Binding to mGL50 was only detected in samples stained with ICOS fusion proteins.
Flow cytometric analysis of splenocytes revealed differential mICOS-mIgG2am binding to phenotypic subsets of lymphoid cells (Fig. 4⇓). CD19+ B cell populations were highly positive for mICOS-mIgG2am staining (84%), consistent with RNA expression data for mGL50 (11). Initial immunohistochemical characterization of ICOS expression using the anti-ICOS Ab F44 revealed anti-ICOS reactivity in the apical regions of germinal center light zones, an area linked to B cell maturation into terminal plasma or memory cells (2). Flow cytometric analysis suggested that germinal T cells express ICOS late in activation and would bind to a B7-like protein on B cells. Consistent with these observations, we demonstrated high levels of surface expression of ICOS ligand on freshly isolated CD19+ splenocytes, suggesting that mGL50 is constitutively expressed on splenic B cells.
Phenotypic analysis of mICOS-IgG2am+ splenocytes. Double staining with Abs against phenotypic markers CD3 (145-2C11, T cell), CD19 (1D3, B cell), CD11b (M1/70 macrophage), Gr-1 (RB5-8C5, granulocyte), CD69 (H1.2F3, activated T/B/NK cells), and pan-NK (DX5, NK cell).
A small subset of CD3+ T cell populations were positive for mICOS-mIgG2am staining (15%). These observations are similar to the findings that B7-2 is expressed at low levels on resting T cells, while B7-1 is detected on T cells only after activation (15, 16). ICOS ligand appears to be present on unactivated splenic T cells at fluorescence levels of binding similar to CD19+ B cells. Binding of T cell GL50 to ICOS on T cells could modulate cytokine profiles of interacting T cell population.
Low to negligible staining was observed on cells expressing markers for macrophage, activated lymphoid cell, granulocyte, or NK cell lineages. It should be noted that the presence of 21% of CD19−, ICOS ligand+ cells cannot be accounted for by the CD3+, ICOS ligand+ T cell component of ICOS-Ig staining, suggesting additional subpopulations of lymphoid cells display surface expression of ICOS ligand.
Our data indicate that GL50, like CD40, is constitutively expressed on B cells, suggesting that production of Th2 cytokines such as IL-4 and IL-10 (2) could be regulated by the induction of ICOS on germinal center T cells. Swallow et al. (11) also found that B7h (mGL50) expression is induced in fibroblasts and nonlymphoid tissues by TNF, suggesting that the ICOS-GL50/B7h interaction may play a unique role in inflammatory sites. Thus, there may be two scenarios by which GL50/B7h signals are transduced: one in which costimulation dependent T cell ICOS interacts with constituitively expressed GL50/B7h on B and other lymphoid cells, and a second that is dependent upon TNF induction of nonlymphoid cells.
Footnotes
-
1 Address correspondence and reprint requests to Dr. Vincent Ling, Department of Immunology, Genetics Institute, 87 CambridgePark Drive, Cambridge, MA 02081. E-mail address: VLing{at}genetics.com
-
↵2 Abbreviations used in this paper: RACE, rapid amplification of cDNA end; hGL50, human GL50; mGL50, mouse GL50; KLM, kit ligand/membrane.
- Received October 20, 1999.
- Accepted December 13, 1999.
- Copyright © 2000 by The American Association of Immunologists
References
In this issue
