The Lymphotoxin-α (LTα) Subunit Is Essential for the Assembly, but Not for the Receptor Specificity, of the Membrane-anchored LTα1β2 Heterotrimeric Ligand*

The lymphotoxins (LT) α and β, members of the tumor necrosis factor (TNF) cytokine superfamily, are implicated as important regulators and developmental factors for the immune system. LTα is secreted as a homotrimer and signals through two TNF receptors of 55–60 kDa (TNFR60) or 75–80 kDa (TNFR80). LTα also assembles with LTβ into a membrane-anchored, heterotrimeric LTα1β2 complex that engages a distinct cognate receptor, the LTβ receptor (LTβR). To investigate the role of the LTα subunit in the function of the membrane LTα1β2 complex, gene transfer via baculovirus was used to assemble LTα and -β complexes in insect cells. LTα containing mutations at D50N or Y108F are secreted as homotrimers that fail to bind either TNF receptor and are functionally inactive in triggering cell death of the HT29 adenocarcinoma cell line. In contrast, these mutant LTα proteins retain the ability to co-assemble with LTβ into membrane-anchored LTα1β2 complexes that engage the LTβR and trigger the death of HT29 cells. Membrane-anchored LTβ expressed on the cell surface in absence of the LTα subunit binds the LTβR but is functionally inactive in the cell death assay. These results indicate that the TNF receptor-binding regions of the LTα subunit are not necessary for engagement of the LTβR, but the LTα subunit is required for the assembly of LTβ into a functional heteromeric ligand.

Lymphotoxins (LT) 1 ␣ and ␤ are structurally related to TNF, the prototypical member of a superfamily of type II transmembrane glycoproteins (1,2). These cytokines also exist in soluble forms, although distinct mechanisms generate secreted and membrane-bound LT␣ and TNF. Secreted TNF is generated by proteolysis of the transmembrane protein (3)(4)(5), whereas LT␣ lacks a transmembrane domain and is exclusively secreted as a homotrimer (and in this form is also known as TNF␤). Unlike TNF, LT␣ also assembles with LT␤ into heteromeric complexes and is consequently localized to the cell surface by the transmembrane domain of LT␤ (6,7). Substantial evidence indicates that membrane LT exists in two trimeric forms with either an ␣1␤2 or ␣2␤1 stoichiometry (6,8). The secreted and membranebound forms of LT are further distinguished by their distinct specificities for cell surface receptors. LT␣ and TNF both bind and signal through two receptors, the 55-60-kDa TNF receptor (TNFR60; CD120a or type 1) (9, 10) and the 75-80-kDa TNFR (TNFR80; type 2 or CD120b) (11). By contrast, the surface LT␣1␤2 complex binds a related but distinct receptor, termed LT␤R, that does not bind either LT␣ or TNF, whereas both TNFRs bind the LT␣2␤1 heterotrimer (8,12). The LT␣1␤2 complex is the most abundant form expressed by activated T cells (13), and unlike TNF, it is not produced naturally in soluble form (8,12). The existence of a LT␤ homotrimer is uncertain, since LT␤ protein is apparently always associated with LT␣ in T cells, and a direct assessment has been hindered by unsuccessful attempts at stable expression of membranebound LT␤ in mammalian cells (8).
In tissue culture systems, TNF and LT␣ homotrimers are well recognized for their abilities to elicit a similar but not identical spectrum of cellular responses, including apoptosis and proinflammatory activities (14). Purified soluble recombinant LT␣1␤2 (15) exhibits the ability to induce tumor cell death (16) and chemokine secretion (17) and activate NF-B, a transcription factor that regulates inflammatory gene expression through the LT␤R (18,19), but may be less potent than LT␣ and TNF. Interestingly, membrane-anchored TNF is more active in signaling via TNFR80 than soluble TNF (20), raising the possibility that membrane-bound and soluble ligands may diverge in some of their functions. This possibility was suspected for the different forms of LT (7) and was brought into acute focus by the characterization of mice with an inactivated LT␣ gene (21,22). LT␣-deficient mice lack most lymph nodes and Peyer's patches, a phenotype not associated with deletions of TNF (23) or either of the TNF receptor genes (24 -27). Placental transfer of an LT␣1␤2 antagonist constructed as a fusion protein between LT␤R extracellular domain and the Fc region of IgG (LT␤R-Fc) results in lymph node-deficient offspring, which established a role for membrane-bound LT␣1␤2 distinct from the LT␣ trimer (28). In addition, the formation of germinal centers during an immune response, a process critical for efficient antibody class switching, is also dramatically altered in mice that lack LT␣ (21,22), TNF (23), TNFR60 (23,29), or LT␣1␤2 (29,30) or express LT␤R-Fc as a transgene (31). Thus, characterization of the membrane-anchored LT ligands will help elucidate their physiologic functions. * This work was supported in part by National Institutes of Health Grants AI33068 and PO1 CA69381, American Cancer Society Grant IM663, and funding provided by the Institute of Molecular and Cell Biology, National University of Singapore. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
ʈ Here, we employ recombinant baculovirus to reconstitute LT␣ and LT␤ homo-and heteromeric complexes in insect cells to investigate the roles of the LT␣ and LT␤ subunits in activation of the cell death response in tumor cells. Using two loss of function mutations in LT␣ (32), aspartic acid 50 to asparagine (D50N) and tyrosine 108 to phenylalanine (Y108F), we show that these LT␣ mutant proteins co-assemble with LT␤ to form a ligand that binds LT␤R but not TNFR. The membrane-bound mutant LT␣⅐LT␤ complexes, but not the secreted homotrimers, are active at inducing death of the HT29 colon carcinoma cell line. By contrast, LT␤ when expressed alone as a membrane protein binds the LT␤R but is functionally inactive in inducing cell death. These results demonstrate that the LT␣ subunit is necessary for the LT␣1␤2 complex to activate the LT␤R cell death pathway.
Expression of LT Subunits in Insect Cells-The construction of recombinant baculoviruses expressing LT␣ or soluble LT␤ tagged with a Myc epitope (sLT␤myc) has been described (12,34). A cDNA encoding full-length membrane-bound LT␤ was isolated as an 860-base pair HindIII fragment from pCDM8/LT␤ (7), and HindIII/BamHI linkers were added. After restriction with BamHI, the LT␤ cDNA was ligated into the baculovirus transfer vector, pVL1393. Recombinant baculoviruses were produced by coinfection of pVL1393/LT␤ with baculovirus DNA as described (34). LT␣Y108F and LT␣D50N mutant cDNAs as originally constructed for expression in bacteria (32) lack the LT␣ signal sequence required for export; therefore, a 300-base pair Nsi1/Pfl M1 cassette, containing the Y108F or D50N mutation, was isolated from p8/3 and p11A/20, respectively, and then used to replace the corresponding region in wild-type LT␣. The resulting mutant cDNAs containing the LT␣ signal sequence were isolated as NotI fragments and ligated into pVL1393. The mutant constructs were confirmed by sequence analysis of the baculovirus vector (U.S. Biochemical Corp. Sequenase version 2.0 sequencing kit).
Radioimmunoassays-The concentration of LT␣ was determined by competitive radioimmunoassay using the anti-LT␣ mAbs NC2 and 125 I-LT␣. Anti-LT␣ NC2 was bound (50 ng/well) to plastic snap wells (Immulon 2, Dynatech, Chantilly, VA) precoated with goat anti-mouse Ig (500 ng/well) to capture 125 I-LT␣. The standard curve was generated with purified recombinant LT␣ (33) diluted in 100 l of phosphatebuffered saline with 1% bovine serum albumin with a 30-min binding interval. LT␣ was radioiodinated to a specific activity of 126 Ci/g by the IodoGen method (36). 125 I-LT␣ in 10 l was added to a final concentration of 0.2 nM and allowed to bind for an additional 30 min. Each well was washed five times, and the bound 125 I-LT␣ in individual wells was detected using a ␥-counter. Each data point is the mean of duplicate wells from which the LT␣ concentration in supernatants was determined from the mean of four dilutions using the radioimmune assay template in Prism (GrapdPAD Software, San Diego, CA). The range was less than 5% for duplicate determinations.
Receptor binding activity of LT␣ and mutant proteins was assessed using a solid phase competitive radioligand binding assay with TNFR60-Fc as a surrogate receptor. The format was identical to the radioimmunoassay described above except that purified TNFR60-Fc was bound at 50 ng/well to wells previously coated with goat antihuman Ig at 500 ng/well.
Biosynthetic Labeling and Immunoprecipitation-Baculovirus-infected insect cells were labeled with [ 35 S]methionine and [ 35 S]cysteine as described (34). Briefly, 24 h after infection (multiplicity of infection was 10 at 10 5 cells/cm 2 ), the cells were washed with buffered saline and incubated in medium deficient in methionine for 2 h before adding [ 35 S]methionine and [ 35 S]cysteine labeling mixture at 0.2 mCi/ml. After 20 h, the supernatants were cleared by centrifugation for 15 min at 23,000 ϫ g, treated with protease inhibitors, and dialyzed against saline. Protein cross-linking was carried out by the addition of 10 l of BSCOES, freshly dissolved in Me 2 SO at 100 mM, to 1.0 ml of serum-free culture supernatants from baculovirus-infected Tn5B1-4 cells (37). After a 30-min incubation on ice, the reaction was stopped by the addition of 25 l of glycine (1 M) and subjected to immunoprecipitation as described below. The cellular fraction was extracted with Nonidet P-40 (1%) nonionic detergent in buffer with 50 mM Tris, pH 7.4, containing 10 mM iodoacetamide and protease inhibitors. The detergentsoluble fraction, obtained after centrifugation, was subjected to immunoprecipitation as described (13). Briefly, detergent extracts were precleared by the addition of 10 g of normal mouse or rabbit IgG and 20 l of protein G-Sepharose beads followed by the addition of 10 g of either mouse anti-LT␣ or LT␤ antibodies or polyclonal rabbit anti-LT␣ and protein G beads. Proteins were resolved by SDS-polyacrylamide gel electrophoresis (12% acrylamide) and detected by PhosphorImager analysis (Molecular Dynamics).
Flow Cytometry-Tn5B1-4 cells were harvested, washed, and incubated on ice in Hanks' balanced salt solution with 10% bovine calf serum, 0.1% sodium azide containing the antibodies or TNFR-Fc chimeras at 10 g/ml. Phycoerythrin-conjugated affinity-purified goat anti-mouse or anti-human IgG (5 g/ml) was used to stain for mAb or TNFR-Fc, respectively. Controls for nonspecific binding included normal mouse or human IgG and inclusion in the buffer of human (or mouse) heat-aggregated IgG at 10 g/ml to block nonspecific binding when staining for mouse IgG. Immunofluorescence staining was detected by flow cytometry (FACScan, Becton-Dickenson) using forward and side scatter parameters to identify infected and noninfected cells. Each fluorescence histogram represents 1 ϫ 10 4 events gated on infected cells. Fluorescence intensity ϭ (mean fluorescent channel) ϫ (percentage of positive fluorescent events), where a positive event has a fluorescence value Ͼ98% of the value for normal IgG. Specific fluorescence intensity represents the fluorescence intensity after subtraction of the value for normal IgG.
Cytotoxicity Assays-Cytotoxicity of soluble LT␣ produced by insect cells was determined using a colorimetric assay with (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide (MTT) as described (38). Briefly, the HT29.14S (a TNF/LT-sensitive subclone of HT29 human adenocarcinoma) (16) or murine L929 fibrosarcoma cells (10 4 cells/well in 96-well flat bottom microtiter plates) were incubated in medium with serial dilutions of supernatants from infected insect cells. For HT29.14S cells, human interferon-␥ was included in the medium at 80 units/ml. After 3 days of incubation, viable cells were detected by the addition of the MTT dye. Cytotoxicity assays with insect cells were performed using paraformaldehyde-fixed insect cells. Tn5B1-4 cells were infected with LT␣ and/or LT␤ recombinant baculoviruses at a multiplicity of infection of 10 for each virus. After 2 days, Tn5B1-4 cells were harvested, washed twice in phosphate-buffered saline, and incubated for 15 min on ice with 1% paraformaldehyde in phosphate-buffered saline and then washed four times with RPMI 1640 containing 10% fetal bovine serum. The fixed insect cells were added at various ratios to HT29.14S cells (10 4 cells/well) and incubated for 3 days, and viability was detected by reduced MTT dye. The percentage of cell viability was calculated as a ratio of the absorbance of reduced MTT dye at 570 nm for cytokine (or insect cell)-treated cells to the absorbance of dye by cells in medium (with interferon-␥) times 100. Each data point represents the mean Ϯ S.D. of triplicate wells. D50N and Y108F mutations in LT␣ were identified as cytotoxicity loss mutants that failed to bind to L929 cells (32). Both LT␣D50N and Y108F mutants, like wild type LT␣, are secreted by Tn5B1-4 cells, typically to 20 -40 g/ml (34). When treated with BSOCOES (1 mM), a homobifunctional protein cross-linking reagent, LT␣ and the two mutant proteins formed a ladder of three bands consistent with predicted sizes for trimers, dimers, and monomers of LT␣ (Fig. 1a). The ladder is created by the incomplete cross-linking of LT␣ subunits by BSCOES (37). The less selective cross-linker glutaraldehyde (0.1%) forms a 65-70-kDa adduct in similar preparations (data not shown, and see Ref. 33), indicating that the majority of the LT␣ subunits exist as trimers. Both LT␣ mutants were ineffective as competitors for binding to TNFR60-Fc (K i ϭ Ͼ700 nM) when compared with wild type LT␣ (K i ϭ 10 nM) (Fig. 1b). Wild type LT␣ induces death in HT29 adenocarcinoma cells (IC 50 ϭ 100 -200 pM), but both mutants were inactive when tested on HT29 cells (Fig. 1c) or L929 cells (data not shown). D50N and Y108F LT␣ mutants were tested for their ability to assemble into membrane LT␣␤ complexes by coinfection of Tn5B1-4 cells with recombinant LT␤ baculovirus. Reciprocal co-immunoprecipitations with antibodies to individual ␣ and ␤ subunits were used to detect the formation of LT␣␤ complexes. Anti-LT␣ specifically immunoprecipitated major bands at 18, 21, and 22-23 kDa, consistent with precursor and glycosylated forms of LT␣ from baculovirusinfected Tn5B1-4 cells labeled with [ 35 S]methionine and [ 35 S]cysteine (Fig. 2, lanes 1-4). A band at 31-33 kDa expected for LT␤ was also immunoprecipitated by anti-LT␣. Similarly, anti-LT␤ co-immunoprecipitated two major bands: LT␣ at 22-23 kDa and LT␤ at 31-33 kDa (Fig. 2, lanes 5-7). The LT␣ mutants associated with LT␤ equally as well as wild type LT␣ as judged by the volume-density of the phosphor image. These results indicate that these mutations do not disrupt assembly of LT␣␤ heteromers. Also, note that LT␤ is not associated with the secreted form of LT␣ (see Fig. 1a), indicating that insect cells, like mammalian T lymphocytes, do not cleave LT␤.

Reconstitution of LT␣␤ Heteromers on the Surface of Tn5B1-4 Cells-The
In the Tn5B1-4 cells, anti-LT␤ primarily immunoprecipitated the 22-23-kDa mature form of LT␣ and not the smaller 18-and 21-kDa forms recognized by anti-LT␣ (Fig. 2, lanes  4 -7). These two smaller LT␣ bands match reasonably well with the predicted sizes of the nascent and signal peptidase-cleaved LT␣ polypeptides of 22.2 and 18.6 kDa, respectively. LT␣ produced by insect cells is glycosylated (39), which indicates that LT␤ assembles with LT␣ soon after the initial processing steps. Precursor-product analysis of LT␣␤ synthesis in T-lymphocytes by pulse-chase methods revealed that LT␤ initially associates with a 21-22-kDa LT␣ precursor that matures to a 25-kDa form (6,8). Glycosylation (N-and O-linked) of LT␣ also occurs in mammalian cells (40), although in T lymphocytes LT␣ shows more extensive change in molecular mass compared with the protein produced by Chinese hamster ovary cells (21 kDa).
Tn5B1-4 cells singly infected with LT␤ baculovirus express LT␤ protein on the cell surface as detected by immunofluorescence staining with anti-LT␤, whereas anti-LT␣ did not stain (Fig. 3, a and b). Surface expression of LT␣ requires infection with both LT␣ and LT␤ recombinant baculoviruses (Fig. 3c). The level of LT␤ protein is approximately the same on both singly and coinfected cells, indicating that the presence of the LT␣ subunit does not modify surface expression of LT␤ protein.
Receptor binding function, assessed by staining with Fc fusion proteins as surrogate receptors, revealed that LT␤R-Fc, but not TNFR60-Fc, stained cells expressing LT␤ (Fig. 3d). However, coinfection with the LT␣ baculovirus dramatically increased the LT␤R-Fc-specific fluorescence staining. Half-maximal binding of the LT␤R-Fc to LT␣␤-expressing cells occurred at 0.8 g/ml (ϳ6 nM), similar to the binding to LT␤ alone (Fig. 4). However, the total LT␤R-Fc bound is substantially greater (ϳ100-fold) in cells coinfected with LT␣ and LT␤. This result would be consistent with LT␣ increasing the number of binding sites for LT␤R. Specific binding of TNFR60-Fc occurred only with LT␣ coinfection, a result consistent with the formation of the LT␣2␤1 ligand (Figs. 3e and 4) with half-maximum binding at 0.6 g/ml, although the total TNFR60-Fc bound was sub- stantially less (ϳ10-fold) than the LT␤R-Fc.
As expected, both the LT␣D50N and Y108F mutant proteins were retained on the surface of insect cells coinfected with LT␤ baculovirus, consistent with the ability of these mutants to assemble with LT␤ into heteromers (Fig. 5, a and b). Tn5B1-4 insect cells infected with mutant LT␣ baculoviruses specifically bound to the LT␤R-Fc fusion protein (Fig. 5, d and f). However, binding interactions with TNFR60-Fc were dramatically reduced by the LT␣ mutants (Fig. 5, e and g). The D50N mutant retained a some capacity to bind TNFR60-Fc compared with Y108F, with a half-maximum binding at 80 -100 nM for D50N and Ͼ200 nM for Y108F (data not shown). TNFR80-Fc binding to insect cells expressing either LT␣ mutants was also decreased (data not shown).
These results indicated that the D50N and Y108F residues in the LT␣ subunit are not directly involved in interactions with the LT␤R, although the LT␣ subunit dramatically enhanced binding of the LT␤R-Fc. The binding of LT␤R-Fc to LT␤ expressed alone suggested the possibility that, as a resident membrane protein, LT␤ could activate the LT␤R.
Cell Death-inducing Activity of Membrane-anchored LT␣␤ Complexes-To investigate the role of the LT␣ and LT␤ subunits in the activation of the LT␤R, insect cells infected with LT␤ or coinfected with LT␣ baculoviruses were fixed and used as effector cells in cytotoxicity assays to measure the functional capacity of the surface ligands. LT␤-infected cells displayed no significant cytotoxic activity for HT29.14S cells when compared with uninfected Tn5B1-4 cells (Fig. 6a). By contrast, LT␣ and LT␤ co-expressing cells were highly effective at killing HT29.14S cells, typically with a 50% reduction in viability at an effector:target cell ratio of 0.5 (Fig. 6b). Supernatants from fixed cells were not active in this assay, demonstrating that cell death requires cell contact. In striking contrast to the soluble LT␣ mutants, insect cells infected with either D50N or Y108F mutants and LT␤ were completely functional in this assay (Fig.  6, c and d). That the effect of cell death was mediated by LT␣1␤2 is indicated by the ability of the LT␤R-Fc, but not TNFR60-Fc, to block the death-inducing activity of these killer insect cells. Together, these results indicate that the LT␣ subunit is required for functional conformation of the LT␣1␤2 but not for specificity of binding to the LT␤R. DISCUSSION The crystal structures of LT␣ and TNFR60 provide a conceptual framework to model interactions between LT␣␤ ligands and their receptors (41,42). Aspartic acid 50 located in the A-AЉ  loop and Y108 in the D-E loop are solvent-exposed residues positioned on opposite sides of the LT␣ monomer, although in the native trimer both residues from different subunits localize to the same receptor binding site (Fig. 7, a and b). Our results indicate that the D50N and Y108F mutations probably cause a local distortion of the TNFR binding site, and not disruption of trimeric architecture, that results in the loss of cytotoxic activity. Additional support for this conclusion is seen in the ability of these LT␣ mutants to assemble with membrane-bound or soluble forms of LT␤ (15). Furthermore, these LT␣ mutants form a functional ligand with LT␤ that activates the LT␤R cell death pathway.
The TNFR60 binding site lies along the cleft formed by adjacent LT␣ subunits, an "␣␣" cleft (42). Based on this model, the interface between two adjacent LT␤ subunits is hypothesized to create an analogous ␤␤ (z-x) site that forms the major LT␤R binding site within the LT␣1␤2 complex (Fig. 7c). This model is consistent with high affinity binding observed between the LT␤R and the LT␣1␤2 complex and LT␤ but not the LT␣2␤1 complex, which lacks a ␤␤ interface. Similarly, the major TNFR60 binding site on LT␣2␤1 would be at the ␣␣ interface (Fig. 7d). The single LT␣ subunit within the LT␣1␤2 complex creates two nonequivalent ␣␤ interfaces (x-Y108 and z-D50), where the D50N and Y108F mutations reside in different ␣␤ clefts.
Theoretically, binding to one or both of the heteromeric ␣␤ interfaces must occur in order for LT␣1␤2 to cluster (aggregate) receptors. Receptor clustering is necessary to recruit signaling molecules, such as TRAF3, to activate the cell death pathway (43), or TRAF5 that can activate NF-B (18). The degree of receptor clustering appears to have a profound effect on the type of cellular responses. In the LT␤R system, NF-B activation, but not cell death, occurs when bivalent anti-LT␤R monoclonal antibody is added to the culture medium (16,19), although cell death occurs when the same monoclonal antibody is immobilized to a surface (16) or the receptor is ligated with soluble polyclonal antibodies (43). Presumably, a higher ordered aggregation of receptors, or stabilization of the receptor signaling complex, sufficient for cell death is achieved with immobilized or polyclonal antibodies. This implies that occupation of all binding sites on the LT␣1␤2 ligand is important to signal cell death and predicts that both ␣␤ and the ␤␤ binding sites are important to achieve this conformation. Either of these LT␣ mutants should create a ligand with two normal binding sites, which might not be sufficient to form complexes capable of signaling cell death. Our results, in fact, show that neither mutation affects cell death signaling by LT␣1␤2. This indicates that LT␤R clustering sufficient to signal death of HT29.14S cells depends upon contact with LT␤ subunit, and not the LT␣ subunit. An alternate possibility is that residues other than Asp 50 and Tyr 108 in LT␣ might be involved in binding to LT␤R. Further mutational analysis may distinguish between these possibilities.
The D50N and Y108F mutations dramatically affected the binding of TNFR60-Fc to secreted LT␣ and membrane-anchored LT␣2␤1. As a soluble protein, LT␣2␤1 does not elicit cellular responses akin to TNF and LT␣; rather, it functions as a weak antagonist for TNF (19). Presumably, the reason for the inability of LT␣2␤1 to activate TNFR60 is that the ␣␤ inter- faces are not sufficient to promote TNFR60 clustering in the same way as LT␣ or TNF homotrimers. Our results show no significant gain (or loss) of cytotoxic activity by killer insect cells that express the membrane-anchored LT␣ mutant-LT␤ complexes. This result indicates that LT␣2␤1 as a membrane protein is unlikely to activate the TNFR60 or LT␤R cell death pathways. This result is further supported by the complete blocking effect of the LT␤R-Fc, which should have revealed any putative killing activity by the LT␣2␤1 complex if this ligand could independently activate TNFR60. Thus, the LT␣2␤1-TNFR60 interaction does not mirror the receptor-activating function of LT␣1␤2 binding with LT␤R and suggests a significant distinction in the way these two receptors are activated by their respective ligands.
The physiologically relevant location of LT␣1␤2 complex is presumed to be at the cell surface, since soluble forms are not naturally produced by lymphocytes. Soluble LT␣ and TNF bind to cell surface receptors with high affinity, typically with an observed K d of 10 -100 pM (37), whereas LT␣1␤2 binding to LT␤R is in the 1-10 nM range (15). 2 The restricted diffusion of a membrane-bound ligand should enhance binding to cell surface receptors so that relatively weak interactions (K d ϭ ϳ10 -100 nM) may become highly relevant in the context of cell to cell contact. The finding that LT␤ expressed alone does not induce cell death, although it is expressed in a form capable of binding the surrogate LT␤R-Fc, indicates that the presence of the LT␣ subunit is critical for the conformation that activates the LT␤R. In previous studies (12), soluble LT␤ protein (generated by deletion of the cytosolic and transmembrane domains) bound to LT␤R-Fc, but weakly compared with LT␣1␤2. Soluble LT␤ is polydisperse in the absence of LT␣, forming aggregates of high molecular mass based on elution through gel filtration matrix (15) or by protein cross-linking. 3 These biochemical findings suggest that the LT␣ subunit may restrict the assembly of LT␤ oligomers to dimers. The critical role of the LT␣ subunit is further revealed by the lymph node deficiency and germinal center failure in mice genetically deficient in LT␣ subunit (21,22). These observations indicate that LT␤ as a single subunit ligand is insufficient to signal these developmental and physiologic processes. Rather, LT␣ is essential to form the biologically active LT␣1␤2 ligand.