LIGHT is a member of the TNF cytokine superfamily that signals through the lymphotoxin (LT)β receptor and the herpesvirus entry mediator. LIGHT may function as a costimulatory factor for the activation of lymphoid cells and as a deterrent to infection by herpesvirus, which may provide significant selective pressure shaping the evolution of LIGHT. Here, we define the molecular genetics of the human LIGHT locus, revealing its close linkage to the TNF superfamily members CD27 ligand and 4-1BB ligand, and the third complement protein (C3), which positions LIGHT within the MHC paralog on chromosome 19p13.3. An alternately spliced isoform of LIGHT mRNA that encodes a transmembrane-deleted form is detected in activated T cells and gives rise to a nonglycosylated protein that resides in the cytosol. Furthermore, membrane LIGHT is shed from the cell surface of human 293 T cells. These studies reveal new mechanisms involved in regulating the physical forms and cellular compartmentalization of LIGHT that may contribute to the regulation and biological function of this cytokine.
LIGHT is a recently identified member of the TNF cytokine superfamily closely related to lymphotoxin (LT)4 α and LTβ (1). LIGHT is a type II transmembrane glycoprotein that is transiently expressed on the surface of activated T lymphocytes (1) and dendritic cells (2). Two differentially expressed cell surface receptors mediate LIGHT signaling, the herpesvirus entry mediator (HVEM also known as HveA), prominent on T cells, and the LTβR, found on stromal cells, but absent on lymphocytes (1). LIGHT interacts with a soluble binding protein, decoy receptor 3, which also interacts with Fas ligand (FasL) (3, 4).
The interaction of LIGHT with the LTβR induces proinflammatory gene expression through activation of the transcription factor NF-κB, similar to the LTαβ complex, the ligand required for lymphoid organogenesis and development of NK (5) and NK-T cells (6). LIGHT induces apoptosis in susceptible colon carcinoma cells via the LTβR (7), and in vivo, transduction of tumor cells with LIGHT cDNA suppresses tumor outgrowth (8). Evidence is accumulating that LIGHT signaling through HVEM may function as a costimulatory molecule for T cells, including the enhancement of T cell proliferation and secretion of IFN-γ (2, 9, 10). Additionally, inhibition of LIGHT with soluble LTβR-Fc decoy or anti-LIGHT Ab suppresses graft vs host disease (10) indicating LIGHT is involved in effector functions mediated by T cells.
LIGHT may also function as a direct viral deterrent. HSV infection induces premature death of activated T cells (11) and can block maturation of dendritic cells (12, 13) potentially leading to localized immune suppression. Envelope glycoprotein D binds HVEM as one of the cellular entry routes used by HSV (14). Glycoprotein D directly competes for the binding of membrane LIGHT to HVEM, whereas LIGHT can interfere with HSV entry by down-regulation of HVEM (1, 15). These observations suggest that LIGHT may be an integral part of host immunity to herpesvirus, which is strongly supported by recent observations on the role that LIGHT and LTαβ play in resistance to CMV (16). The interaction of viral gene products with TNF superfamily (TNFSF) members may have shaped the evolutionary course of these molecules, which prompted us to examine the genetic organization of LIGHT (TNFSF14) for insights into the regulation of this cytokine.
Here, we define the molecular genetics of the human LIGHT genomic locus, revealing close linkage to CD27 ligand (CD27L) (CD70, TNFSF7), 4-1BB ligand (4-1BBL) (TNFSF9), and the third complement protein (C3) defining a novel immune response locus within an MHC-like region on chromosome (Chr) 19p13.3. A detailed analysis of LIGHT gene organization uncovered a differentially spliced transcript that encodes a novel transmembrane-deleted form of LIGHT, which is nonglycosylated, unlike the membrane form. Furthermore, membrane LIGHT is shed from the surface revealing multiple mechanisms involved in regulating the physical form and cellular compartmentalization of LIGHT.
Materials and Methods
Cells, cytokines, and Abs
The human II-23 cell line (D7 subclone), a CD4+ T cell hybridoma line (17), was maintained in RPMI 1640 containing 10% FBS (HyClone Laboratories, Logan, UT) and 100 U/ml penicillin/100 μg/ml streptomycin (Life Technologies, Grand Island, NY). II-23 cells were activated with 100 ng/ml PMA or PMA and 1 μg/ml ionomycin for 4 h. Human kidney 293 cells expressing the adenovirus large T Ag (293T) (American Type Culture Collection, Manassas, VA) were grown in DMEM containing 10% FBS and antibiotics. Human PBL were isolated by Ficoll gradient centrifugation and adherent cell depletion as described previously (1). B lymphocytes were removed by passage through nylon wool, and the T lymphocytes were activated with anti-CD3 (OKT3) (1 μg/ml) and anti-CD28 (1 μg/ml) (BD Biosciences, Mountain View, CA) and cultured in RPMI 1640/FBS with IL-2 (10 ng/ml). After 7 days of culture, these T cells were activated with PMA (100 ng) and ionomycin (1 μg) in fresh medium at 2 × 106 cells/ml as described previously (1). Rat anti-human LIGHT was prepared by immunization with 50 μg of purified LIGHTt66, a recombinant truncated form with a deletion of the cytoplasmic and transmembrane regions generating a soluble protein as previously described (7). Anti-human LIGHT Omniclone is a bacterially expressed combinatorial Ab containing VH and Vκ-chains generated from a BALB/c mouse immunized with recombinant soluble human LIGHTt66 (Biosite Diagnostics, San Diego, CA) (18). This Ab reacts with natural and recombinant human LIGHT in flow cytometry and immunoprecipitation; no cross-reactivity was observed with TNF, LTαβ, or mouse LIGHT (a detailed description of the properties of this reagent is in preparation). Mouse anti-methamphetamine Omniclone was used as an isotype control and was provided by G. Valkirs (Biosite Diagnostics). Fc fusion proteins consisting of the ecto domain of HVEM or TNFR1 fused to the Fc region of human IgG1 were produced in baculovirus expression system and affinity purified as previously described in detail (1, 19, 20).
RNA was isolated from 2 × 106 cells using 1 ml of TRIzol (Life Technologies) following the manufacturer’s protocol. First-strand cDNA synthesis was performed using Superscript II (Life Technologies) as described previously (21). PCR amplification, using oligonucleotide primers derived from the human LIGHT cDNA sequence (forward 5′-TCAGTGTTTGTGGTGGATGGA-3′, reverse 5′-CTTCCTTCACACCATGAAAGC-3′) was accomplished using the following amplification schedule: 95°C for 2 min; 30 cycles of 95°C for 30 s, 58°C for 30 s, 7°C for 30 s; and 72°C for 10 min. Following amplification, products were analyzed by agarose (1.5%) gel electrophoresis. The gels were stained using 1 μg/ml ethidium bromide and photographed under UV trans-illumination with an alpha imager (Alpha Innotech, San Leandro, CA).
Human embryonic kidney 293T cells, in six-well dishes (50% confluent), were transfected with 3 μg of cDNA per well, using the calcium phosphate coprecipitation method (22). The cells were incubated with the precipitate for 12 h, and then fresh complete medium was added for 48 h to achieve maximal protein expression.
Receptor-mediated ligand precipitation and Western blot analysis
Tissue culture supernatants were harvested from transfected 293T cells, and debris was removed by centrifugation at 1000 × g for 5 min. Supernatants were then treated with a protease inhibitor mixture (1 mM PMSF, 10 mM iodoacetamide, 10 μg/ml leupeptin, 10 μg/ml aprotinin, and 0.02% azide) and adjusted to a final concentration of 1% Nonidet P-40 (NP40). Cellular extracts were prepared by treatment of cell pellets (2.5 × 106 cells) with 0.7 ml of 1% NP40 lysis buffer containing 50 mM HEPES, pH 7.5, 150 mM NaCl, 20 mM EDTA, 1 mM PMSF, 10 mM iodoacetamide, 10 μg/ml leupeptin, 10 μg/ml aprotinin, and 0.02% azide. Supernatants or detergent lysates were precleared by incubation with 10 μg/ml human IgG or mouse isotype control and 30 μl of protein G-Sepharose beads (Pharmacia, Peapack, NJ) for 2 h at 4°C with nutation. Ligands were precipitated from supernatants or lysates by incubation with 10 μg/ml HVEM-Fc, 10 μg/ml TNFR1-Fc, or 10 μg/ml mouse anti-LIGHT Omniclonal Ab for 2 h at 4°C with protein G-Sepharose beads. Washed immunoprecipitates were treated with recombinant forms of endoglycosidase H (endo H) and peptide-N-glycosidase F (PNGase F) (New England Biolabs, Beverly, MA) for 1 h at 37°C. Samples were resolved using a Tris-glycine SDS-polyacrylamide (14%) gel (NOVEX, San Diego, CA), and proteins were transferred to Polyscreen polyvinylidene difluoride transfer membrane (NEN, Boston, MA) using a semidry blotting unit (Fisher, Pittsburgh, PA). Immune complexes were detected with goat F(ab′)2 anti-rat IgG conjugated to HRP and detected by ECL (Pierce SuperSignal; Pierce, Rockford, IL) as described (7).
Cloning, sequencing, and sequence analysis
The human LIGHT locus-containing P1 clone was isolated by PCR screening a P1 library using the following LIGHT specific oligonucleotide primers: forward 5′-GCTCCTGGGAGCAGCTGATACAA-3′ and reverse 5′-TGGGTTGACCTCGTGAGACCTTCG-3′ (IncyteGenomics, Palo Alto, CA). The LIGHT23). Hypothetical proteins were identified by pBLAST search of GenBank. Consensus motifs for sequence-specific DNA binding proteins were identified using the transcription element search system (TESS) at the University of Pennsylvania (Philadelphia, PA) to search the TRANSFAC database (http://www.cbil.upenn.edu/tess/) (24).+. Bacterial artificial chromosome (BAC) DNA was analyzed for gene content using the gene finding algorithm GenScan provided by the Massachusetts Institute of Technology Computational Biology Department (http://genes.mit.edu/GENSCAN.html) (
Results and Discussion
The chromosomal location of LIGHT
In a BLAST search of the unfinished high throughput genomic sequence database, two BAC sequences (AC008760 and AC025343) from human Chr19 were identified; each contained the entire LIGHT gene (TNFSF14). The identity between LIGHT mRNA nucleotide sequences and the Chr19 genomic sequences, with the exception of introns, was absolute (E value of 0.0). This result was puzzling because a prior report assigned LIGHT to Chr16 by in situ hybridization of metaphase chromosomes (8). This discrepancy was resolved by screening a P1 phage library for the LIGHT genomic locus, and the entire P1 DNA clone was used for in situ hybridization of metaphase chromosomes (Fig. 1⇓A). In this analysis, a Chr19 specific probe cohybridized with the P1 LIGHT probe and verified that the chromosomal location of LIGHT is 19p13.3.
The LIGHT gene resides adjacent to C3 and within a T cell costimulatory locus containing CD27L and 4-1BBL
Identification of the LIGHT-containing BAC clone AC008760 yielded 200 kb of genomic DNA for analysis of closely linked genes. Sequence analysis of this BAC clone revealed tight linkage of the C3 to LIGHT separated by only 7.78 kb with the 3′ end of C3 adjacent to the 5′ end of LIGHT (Fig. 1⇑C). Several TNF family members localize in clusters, prompting a screen for related ligands in this region using the GenScan gene finding algorithm. The genomic locus for CD27L (CD70/TNFSF7) was identified on both BAC clones AC008760 and AC025343 and was mapped on NT 011098 to within 79 kb of LIGHT. In addition, the genomic locus of 4-1BBL (TNFSF9) was identified on a BAC clone (AC010503) that overlaps BAC AC025343 containing the LIGHT and CD27L genes. 4-1BBL was mapped on NT 011169 to within 235 kb of CD27L. Therefore, C3, LIGHT, CD27L, and 4-1BB-L all reside on 19p13.3 in a region that spans ∼370 kb (Fig. 1⇑, B and C). The close genetic linkage of LIGHT to CD27L and 4-1BBL supports the possibility that this gene cluster arose from localized gene duplication events (25).
The positioning on Chr19p13.3 places LIGHT within a large genetic region paralogous to the MHC on Chr6p21.3, which is ∼8 Mb in size (26, 27). The Chr19p13.1-p13.3 MHC paralog is one of four regions thought to have arisen by chromosomal duplication, which include Chr1q21-q25/p11-p32 and Chr9q33-q34. Other TNF-related superfamily members map to these MHC paralogs, FasL and OX40 ligand to Chr1, and CD30 ligand and vascular endothelial growth inhibitor to Chr9. The evolutionary relationship of these paralogs has been the subject of much interest and controversy in understanding the origins of the MHC (28). These paralogs contain class I, complement and TNF-related genes among other conserved markers. The LIGHT locus is notably reminiscent of the TNF locus containing LTβ, TNF, and LTα closely linked to C2 and C4 within the MHC (29, 30). It should be noted that within the complement protein family, C3 displays the highest sequence similarity to C4 (31). In addition, LIGHT and LTβ share functionality and the highest sequence similarity within the TNF super family and LTβ resides near the C4 gene. During evolution, gene duplication events are often followed by translocations that either relocate the gene cluster to different chromosomes or break up the cluster altogether. Therefore, it is tempting to speculate that a translocation of the TNF-LT locus gave rise to the entire LIGHT locus or vice versa. However, CD27L and 4-1BBL have only three exons, whereas TNF, LTβ, LTα, and LIGHT have four exons suggesting that the LIGHT locus was not directly descendent from the TNF/LT locus. This result compelled an examination of the genomic organization of TNF-related ligands on Chr1, where sufficient information on gene structure could be extracted. The gene content of the genomic DNA adjacent to the FasL gene, within the contig NT 000039, which is comprised of nine P1-derived artificial chromosome clones and two BAC clones of various sizes, contained FasL (TNFSF6), GITR ligand (TNFSF18), and OX40 ligand (TNFSF4) (Fig. 1⇑D). Although the genes in the FasL locus span a greater distance than those of the LIGHT locus, the gene orientation and exon organization of FasL is strikingly similar to the LIGHT locus. Additionally, GITRL and OX40L contain three exons matching CD27L and 4-1BBL. Phylogenetic analysis of the amino acid sequence of the TNF family reveals significant sequence similarity in the proteins encoded by LIGHT and FasL loci that corresponds with their position in the gene clusters (Fig. 1⇑E). (GITRL does not fit this tree well owing to significant divergence in sequence in the initial A β strand.) However, the sequence and organizational similarities do not apply when the LIGHT locus is compared with the LTαβ/TNF locus.
At a functional level, FasL and LIGHT exhibit significant amino acid sequence homology in their ectodomains (31%), second only to the similarity between LIGHT and LTβ (34%) (1). Although binding to distinct cellular receptors, FasL and LIGHT do share the ability to bind soluble DCR3 (4). Although various possible combinations of gene duplication and translocation may explain the origin of each gene cluster, collectively, the relationship between the FasL and LIGHT loci is suggestive of a duplicative event involving the entire gene cluster, presumably occurring during a chromosomal duplication. This result would support the idea that Chr19 and Chr1 MHC paralogs are more closely related to each other than the Chr6 paralog. Given the close linkage of LIGHT and C3, which is considered the primordial gene of C5 and C4 based on both phylogenetic and functional arguments (31), LIGHT may be primordial to LTβ and FasL. The chromosomal position of the LIGHT locus is conserved in mice; however, insufficient phylogenetic analysis is available for most TNFSF members limiting further analysis.
The genomic organization of LIGHT
The LIGHT gene spans 5.1 kb (from AUG to stop codon) and is comprised of four exons with similar organization to FasL, LTβ, and other TNF family members (Table I⇓). Exon 1 of LIGHT encodes the first 73 amino acids of the polypeptide, which comprises the entire cytoplasmic tail (aa 1–38), transmembrane domain (aa 38–58), and the beginning of the extracellular stalk region. The second and third exons encode amino acids 74–85 and 86–100, respectively, which make up the stalk region and the beginning of the trimerization domain. The fourth exon encodes the remainder of the trimerization domain, the receptor binding domain (amino acids 101–240), and includes a site for N-linked glycosylation (Asn102). Based on structural homology with LTα (32) and supportive evidence from modeling and biochemical analysis, the trimerization domain of LIGHT folds into an anti-parallel β sandwich; wherein the receptor-binding sites are formed from adjacent subunits (7).
The LIGHT transcript has been previously reported as 2.5 or 2.7 kb by Northern blot analysis, although two considerably smaller cDNAs have been cloned of 1169 nt (1) and 1491 nt (9). Therefore, it is likely that the full-length transcript has yet to be defined. Analysis of the nucleotide sequences 5′ of the LIGHT gene up to the 3′ region of the C3 gene revealed three potential consensus TATAA elements at nucleotide positions −1979, −1762, and −839 relative to the ATG of the first exon. In addition, a combination of consensus motifs for sequence-specific DNA binding proteins, consistent with the inducibility of LIGHT by TCR signaling, are also present. These motifs, among others, include AP-1, NF-κB, and Oct-1 binding sites, which are all present in the highly TCR-inducible IL-2 promoter (33). Usage of the TATAA element at nucleotide position −1762 would yield transcripts of ∼2.8 kb; this transcript size would be more consistent with those observed by Northern blot analysis. The finding that CD70, LIGHT, and 4-1BBL all colocalize to a small region on human Chr19 p13.3 supports the possibility that these molecules may be commonly regulated by higher order chromosomal regulation involving chromatin remodeling. However, each ligand displays a different cellular expression pattern, LIGHT is produced by T cells, whereas CD27L is expressed by B cells and 4-1BBL resides on macrophages, yet all of their receptors are expressed by T cells, underscoring their functional roles in T cell activation.
An alternate spliced form of LIGHT
RT-PCR analysis of LIGHT expression in the human T cell hybridoma (II-23.D7) revealed a second transcript ∼100 nt smaller in size than the full-length LIGHT message that was also inducible with PMA and ionomycin treatment (Fig. 2⇓). DNA sequencing revealed that the smaller transcript contained an internal deletion of 36 amino acids that removes the entire transmembrane domain, referred to as LIGHT ΔTM (accession number AY028261). LIGHT ΔTM transcripts were observed at levels considerably lower than the full-length LIGHT transcript (Fig. 2⇓B). A similar pattern of inducible expression was also observed in human PBLs following activation with PMA and ionomycin (Fig. 2⇓C). The detection of LIGHT ΔTM in activated peripheral blood T lymphocytes is supportive of a possible biological role for this isoform of LIGHT in vivo.
Analysis of the genomic sequence of LIGHT, using a splice donor and acceptor consensus site prediction model, revealed a splice donor consensus sequence at the exact nucleotide corresponding to the deletion in LIGHT Δ TM. Therefore, this alternative transcript is generated by joining the cryptic splice donor in exon 1, at nucleotide position 111, to the splice acceptor that defines the beginning of exon 2, at nucleotide position 218, resulting in the removal of 107 nucleotides including the transmembrane domain in exon 1 (Fig. 3⇓). Thus, direct genetic evidence from splice donor-acceptor prediction algorithms supports alternative splicing of LIGHT as the mechanism for the generation of the smaller LIGHT transcript detected by RT-PCR.
A cDNA encoding LIGHT ΔTM was cloned into the mammalian expression vector (pCDNA3.1+) to study its biologic activity. 239T cells transfected with this construct expressed a 28-kDa protein that reacts with a polyclonal rat anti-human LIGHT antiserum by Western blot analysis (Fig. 4⇓A). In addition, HVEM-Fc fusion protein, but not TNFR1-Fc, binds a portion of the expressed LIGHT ΔTM suggesting that it assembles as a trimer.
Typical of other TNF-related ligands, LIGHT is a type II (cytoplasmic amino terminus) transmembrane glycoprotein. However, in the absence of a transmembrane domain, LIGHT ΔTM no longer contains the necessary stop-transfer signal for proper membrane topology and may either be routed to the cytosol or secreted. To determine the cellular location of LIGHT ΔTM, nonionic detergent lysates or spent supernatants from LIGHT ΔTM-transfected 293T cells were analyzed by precipitation with HVEM-Fc. LIGHT ΔTM was not detectable in the culture supernatant (Fig. 4⇑A, lane 3), but was precipitated from the cell extract (Fig. 4⇑A, lane 7) as a single band of ∼28 kDa. In contrast, an N terminus-truncated soluble form of LIGHT (LIGHTt66), replaced with the signal peptide from VCAM1 to direct its secretion, was readily detected in the supernatant (Fig. 4⇑A, lane 1) and cell lysate (Fig. 4⇑A, lane 4) as multiple bands, which is suggestive of glycosylation intermediates.
The extracellular domain of LIGHT contains a single predicted site for N-linked glycosylation at amino acid position Asn102. Thus, digestion with selected glycohydrolases should provide a diagnostic picture of whether LIGHT ΔTM is processed as secreted LIGHTt66 or membrane LIGHT (Fig. 4⇑B). Endo H cleaves high mannose structures and some hybrid oligosaccharides characteristic of glycoproteins that are in progress through the secretory pathway, whereas PNGase F cleaves nearly all types of N-glycans at the asparagine residue of N-linked glycoproteins. LIGHT ΔTM displayed no shift in mobility when treated with either endo H or PNGase F (Fig. 4⇑B, lanes 1–3). By contrast, LIGHTt66 in the cell-associated immunoprecipitates (lanes 7–9) was shifted in apparent molecular mass by these glycohydrolases. Digestion of full-length LIGHT by either endo H or PNGase F shifted the apparent mass indicating that membrane-bound LIGHT was glycosylated (Fig. 4⇑, lanes 11–13). Together these results indicate that LIGHT ΔTM is likely moving into a cytosolic compartment, whereas LIGHTt66 and membrane LIGHT are processed normally through the secretory pathway. The identification of LIGHT ΔTM is the first example of the production of a soluble TNF family ligand by alternative splicing. Likewise, it is unique that LIGHT ΔTM is not destined for secretion, but is probably retained in the cell cytosol and translated on free ribosomes, unavailable for processing by glycosyltransferases in the endoplasmic reticulum, and thus matures in a compartment distinct from transmembrane LIGHT. The function of LIGHT ΔTM in the cytosol remains to be ascertained.
LIGHT is shed
A metalloprotease cleavage site in FasL (34) is also present in LIGHT (residues 81–84). Although previous immunoprecipitations failed to detect a shed form of LIGHT produced by activated II-23 T cells (1), the possibility could not be dismissed that other cell types might be capable of shedding LIGHT. Consistent with the size of the predicted shed protein, 293T cells transfected with full-length transmembrane LIGHT displayed a ∼26-kDa form in the supernatant (Fig. 5⇓). Furthermore, only the 30-kDa full-length form of LIGHT was present in the cellular fraction. This pattern is predicted for a molecule cleaved on the external side of the membrane, as occurs for TNF (35). The shed form of LIGHT binds HVEM-Fc suggesting that it is a functional ligand. However, with FasL the soluble form generated by shedding is not capable of inducing apoptosis (34). Bearing on this result, recombinant LIGHT truncated near the cleavage site is relatively unstable compared with the LIGHTt66 form (C. Ware, unpublished observations), suggesting the possibility that shedding of membrane LIGHT could be a mechanism of inactivation.
Together, these results demonstrate that LIGHT exists in several distinct molecular forms, which are directed to distinct cellular compartments, including the extracellular space, the membrane, and the cytosol. The realization that LIGHT is tightly linked to molecules of immunologic importance, including 41BBL, CD27L, and C3, which reside in an MHC paralog, suggests that this region could play an undisclosed role in genetically linked immune pathology, which may now be possible to define.
The assistance provided by the General Clinical Research Center of Scripps Research Institute and Randi Vita for isolation of T cells and G. Valkirs and T. Mikayama for the anti-LIGHT Omniclone are gratefully appreciated.
↵1 This work was supported in part by U.S. Public Health Service, National Institutes of Health Grants CA69381, AI03368, and AI48073.
↵2 This is publication 428 from the La Jolla Institute for Allergy and Immunology.
↵3 Address correspondence and reprint requests to Dr. Carl F. Ware, Division of Molecular Immunology, La Jolla Institute for Allergy and Immunology, 10355 Science Center Drive, San Diego, CA 92121. E-mail address:
↵4 Abbreviations used in this paper: LT, lymphotoxin; BAC, bacterial artificial chromosome; FasL, Fas ligand; HVEM, herpesvirus entry mediator; TNFSF, TNF superfamily; NP40, Nonidet P-40; Chr, chromosome; PNGase F, peptide-N-glycosidase F; CD27L, CD27 ligand; 4-1BBL, 4-1BB ligand; C3, third complement protein; endo H, endoglycosidase H.
- Received June 29, 2001.
- Accepted August 24, 2001.
- Copyright © 2001 by The American Association of Immunologists