CD40 Ligand Mutants Responsible for X-linked Hyper-IgM Syndrome Associate with Wild Type CD40 Ligand*

CD40 ligand (CD40L) is a 33-kDa type II membrane glycoprotein mainly expressed on activated CD4+ T cells in trimeric form. When it is mutated, the clinical consequences are X-linked hyper-IgM syndrome (XHIM), a primary immunodeficiency disorder characterized by low levels of IgG, IgA, and elevated or normal levels of IgM. Mutated CD40L can no longer bind CD40 nor provide signals for B cells to proliferate and to switch from IgM to other immunoglobulin isotypes. When considering gene therapy for XHIM, it is important to address the possibility that the mutated CD40L associates with transduced wild type CD40L, and as a consequence, immune reconstitution is not attained. In this study, we demonstrate that the various mutated CD40L species we have identified in patients with XHIM, including both full-length and truncated mutants, associate with wild type CD40L on the cell surface of co-transfected COS cells. The association between wild type and mutated CD40L was also observed in CD4+ T cell lines established from XHIM patients with leaky splice site mutations. The clinical phenotype of these patients suggests that this association between wild type and mutated CD40L species may result in less efficient cross-linking of CD40.

CD40 ligand (CD40L) 1 (CD154, gp39, or TRAP) is a type II membrane glycoprotein, consisting of 261 amino acid residues, and is expressed mainly on activated CD4 ϩ T cells (1,2). The natural receptor for CD40L is CD40, a member of the TNF receptor superfamily, expressed on a variety of cells including B cells, macrophages/monocytes, dendritic cells, vascular endothelial cells, and epithelial cells (3). The interaction between CD40L and CD40 therefore plays a crucial role in the immune system (3,4). The cross-linking of CD40 by CD40L induces a signal for B cells to undergo proliferation and immunoglobulin isotype switching and to escape apoptosis (3,4). In addition, CD40L-CD40 interaction influences many aspects of T cellmediated inflammatory responses, such as up-regulation of adhesion molecules, cell extravasation, production of inflammatory cytokines and chemokines, as well as activation of macrophage effector function (3).
The physiologic significance of the CD40L-CD40 interaction has been underscored by the observation that mutations of the CD40L gene cause X-linked hyper-IgM syndrome (XHIM) (5-9), a primary immunodeficiency disorder characterized by low or absence of IgG, IgA, and IgE and normal or elevated IgM. Mutations of the CD40L gene identified in XHIM patients are highly heterogeneous. They include missense, nonsense, and splice site mutations, and insertions or deletions (10,11), and are distributed throughout the CD40L gene which consists of 5 exons and 4 introns and spreads over 12 kilobase pairs in genomic DNA (12,13). More than 75 unique mutations have been reported to date (10,11). In most instances, the mutated CD40L on the cell surface of activated T cells is undetectable if anti-CD40L monoclonal antibodies (mAbs) or CD40-Ig, a fusion protein consisting of the extracellular domain of CD40 and the Fc portion of human immunoglobulin G, are used. However, if a polyclonal anti-CD40L antiserum is used, the expression of mutated CD40L by activated T cells is detected in the majority of XHIM patients (11).
When considering gene therapy for XHIM patients, the possibility of an association between the transduced wild type CD40L and the patient's mutated CD40L resulting in heterotrimer formation has to be addressed. Since two CD40L monomers contribute to form one functional CD40-binding site (25,26), the association of wild type CD40L monomer with mutated CD40L monomer is expected to generate decreased numbers of CD40-binding sites and, consequently, render the heterotrimer less efficient in clustering CD40. In this study, the association of wild type CD40L with various mutated CD40L species isolated from XHIM patients is demonstrated in transfected COS cells and in activated T cell lines established from XHIM patients with different splice site mutations.
Construction of Expression Plasmid-A schematic representation of the protein structures we expressed in COS cells and the arbitrary designation of each plasmid and corresponding protein are shown in Fig. 1 and Table I, respectively. All of the expression plasmids were constructed by reverse transcription-polymerase chain reaction (RT-PCR) with Pfu DNA polymerase (Stratagene, La Jolla, CA) using cDNA isolated from activated peripheral blood mononuclear cells derived from a healthy volunteer or from selected XHIM patients (11). The human CD40L cDNA consisting of nt 1-807 (nucleotide numbering is based on the sequence data of Diane Hollenbaugh et al. (1)) was amplified by RT-PCR using sense primer (SP) 1 (5ЈCGCGGATCCATTTCAACTTTA-ACACAGC3Ј, recognition sequence underlined) and antisense primer (AP) 1 (5ЈGCGCTCGAGTCAGAGTTTGAGTAAGCCAAAGG3Ј), and subsequently cloned into BamHI and XhoI sites of pcDNA3.1/Zeo(ϩ) (Invitrogen, Carlsbad, CA). Naturally occurring mutant cDNAs, exon 2-skipped cDNA, 19 nucleotides of intron 2-inserted cDNA, and exon 3-skipped cDNA, were similarly cloned into pcDNA3.1/Zeo(ϩ) using cDNA generated from the appropriate XHIM patients. The expression vector of the CD40L lacking the cytoplasmic domain (CytDel, Met 21 -Leu 261 ) was generated using SP2 (5ЈCGCGGATCCATTTCAACTTTAA-CACAGCATGAAAATTTTTATGTATTTAC3Ј) and AP1 and cloned into BamHI and XhoI sites of pcDNA3.1/Zeo(ϩ). The expression vectors of Flag-tagged wild type and the naturally occurring mutant CD40L were generated by RT-PCR using primers SP3 (5ЈCGCGGATCCATTTCAA-CTTTAACACAGCATGGATTACAAGGACGATGACGACAAGATCGA-AACATACAACCAAACTTC3Ј) and AP1, cloned into the same vector; Flag peptides consisting of DKYDDDDL were inserted between Met 1 and Ile 2 of wild type or mutant CD40L. In constructing pF-L258S, AP2 (5ЈGCGCTCGAGTCAGAGTTTGAGTGAGCCAAAGG3Ј) was designed to have a missense mutation within the antisense primer sequence and FIG. 1. Schematic presentation of the constructs including wild type CD40L, naturally occurring mutant CD40Ls, and control constructs used in this study. Shown are the contribution of each exon of the CD40L gene to the domain structure: IC, intracellular tail; TM, transmembrane domain; ECU, extracellular unique region; and TNFH, TNF homology domain. The number on the right side of the protein structure represents the amino acid number where each domain starts or the one where each protein or truncated ECU domain ends. In order to discriminate wild type CD40L both by protein size and by antigenicity from various mutant CD40Ls, a Flag peptide (DKYDDDDL) was inserted at the N terminus between the first amino acid methionine and the second amino acid isoleucine of wild type CD40L. Flag-tagged proteins are represented by the "F-" preceding each protein name. The constructs designed for this study included the following: cartoon 1, wild type CD40L; 2, CD40L lacking the intracellular tail and starting at Met 21 (CytDel); 3, Flag-tagged wild type CD40L (F-Wild); 4, Flag-tagged CD40L with the extracellular part of CD40L consisting of Gln 114 -Leu 261 (soluble CD40L) cleaved off (F-Stalk); 5, Flag-tagged mutant CD40Ls with one or two amino acid substitutions in TNFH selected from patients with XHIM, including F-DM (double mutations resulting in S128R/E129G), F-T147N, F-Y170C, F-G227V, F-A235P, F-T254M, and F-L258S; 6 and 7, Flag-tagged mutant CD40Ls with premature termination selected from patients with XHIM, representing F-W140X and F-Q186X, respectively; 8, mutant CD40L with in-frame deletion of 44 amino acids encoded by exon 2 and identified in XHIM patients with intron 2 splice donor site mutation (E2skip); 9 and 10, a truncated mutant CD40L without or with the N terminus Flag peptide (E2ins and F-E2ins, respectively), resulting from the translation of the mRNA transcripts carrying the 19 nucleotides' insertion of intron 2 and identified in XHIM patients with intron 2 splice donor site mutation; 11 and 12, a truncated mutant CD40L without or with the N terminus Flag peptide (E3skip and F-E3skip, respectively), resulting from the translation of exon 3-skipped mRNA transcripts identified in XHIM patients with intron 3 splice donor site mutation; 13, Flag-tagged control protein construct (F-HybFasL or F-HybCD30L), representing the hybrid protein in which the extracellular region of the human FasL (from Gln 103 to Leu 281 ) or of human CD30L (from Gln 63 to Asp 234 ) was fused to the IC and TM of CD40L. Both F-HybFasL and F-HybCD30L contain three extra amino acids, RDP, at the fusion site. See also Table I; the number of each cartoon shown here corresponds to that in Table I. was used for amplification instead of AP1. The Flag-tagged expression vector, pF-Stalk, expressing the remainder of the CD40L molecule after the extracellular domain of CD40L (Gln 114 -Leu 261 ) had been cleaved off (20,21) was constructed using SP3 and AP3 (5ЈGCGCTCGAGTCAC-TACATTTCAAAGCTGTTTTCTTTC3Ј). The control expression vectors, pF-HybFasL and pF-HybCD30L, expressing a Flag-tagged fusion protein consisting of the cytoplasmic and transmembrane domains of CD40L and the extracellular domain of FasL (Gln 103 -Leu 281 ) (29) and CD30L (Gln 63 -Asp 234 ) (30), respectively, were constructed as follows. The coding sequence of cytoplasmic and transmembrane domains of CD40L was amplified by RT-PCR using SP4 (5ЈATAAGAATGCGGCC-GCATTTCAACTTTAACACAGCATGGATTACAAGGACGATGACGAC-AAGATCGAAACATACAACCAAACTTC3Ј) and AP4 (5ЈCGCGGATCC-CGAAGATACACAGCAAAAAGTGCTG3Ј) and subsequently cloned into NotI and BamHI sites of pcDNA3.1/Zeo(ϩ). The coding sequences of the extracellular domain of the FasL and CD30L were amplified by PCR using SP5 (5ЈCGCGGATCCACAGCTCTTCCACCTACAGAAGGA-G3Ј) and AP5 (5ЈGCGCTCGAGTTAGAGCTTATATAAGCCGAAAAAC-GTCTG3Ј) for FasL and SP6 (5ЈCGCGGATCCACAGAGGACGGACTC-CATTCC3Ј) and AP6 (5ЈGCGCTCGAGTCAGTCTGAATTACTGTATA-AG3Ј) for CD30L, respectively, and then fused into BamHI and XhoI sites of the pcDNA3.1/Zeo(ϩ) into which the coding sequence of the cytoplasmic and transmembrane domains of CD40L had already been cloned. All expression vectors were sequenced to verify the correct nucleotide sequences.
Transfection of COS Cells-COS cells were electroporated with supercoiled expression plasmid under the following conditions: 10 g of plasmid and 4 ϫ 10 6 cells were mixed in 0.4 ml of serum-and antibioticfree DMEM and electroporated at 210 V and 960 microfarads using GenePulser (Bio-Rad). Forty eight hours after electroporation, cells were harvested after incubation in phosphate-buffered saline (PBS), 0.5% bovine serum albumin, 5 mM EDTA for 10 min and used either to examine surface expression of the transduced gene product by flow cytometry or to biochemically characterize the expressed proteins.
Surface Biotinylation and Immunoprecipitation of Proteins Transiently Expressed by COS Cells-Forty eight hours after electroporation, cells were harvested as described above. When co-transfection with an expression plasmid of wild type CD40L (pWild) and of a CD40L mutant was intended, 5 g of each plasmid was mixed and electroporated under the same conditions. Cells were resuspended at a concentration of 5 ϫ 10 6 /ml and kept for 30 min on ice in lysis buffer containing 1% Triton X-100, 10 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.025% NaN 3 , freshly supplemented with 200 g/ml phenylmethylsulfonyl fluoride (Sigma), 10 g/ml aprotinin (Roche Molecular Biochemicals), 10 g/ml leupeptin (Roche Molecular Biochemicals), and 10 mM iodoacetamide (Sigma). For surface biotinylation, cells were washed with PBS twice, suspended at 5 ϫ 10 6 /ml in PBS, and Sulfo-NHS-biotin (Pierce) added to a final concentration of 500 g/ml. Cells were incubated for 30 min with rotation at room temperature and then lysed. The lysate was cleared by centrifugation at 14,000 ϫ g for 10 min at 4°C. Protein concentration was measured using Bio-Rad DC Protein Assay (Bio-Rad) and bovine ␥-globulin as standard. The lysate was adjusted to 200 g of protein in 200 l of lysis buffer and precleared overnight with 20 l of protein G-Sepharose or protein A-Sepharose (50% slurry) (Pierce) at 4°C, followed by immunoprecipitation with 3 g of specific antibody or 5 g of CD40-Ig for 1.5 h at 4°C. The immune complexes were collected using 10 l of protein G-Sepharose or protein A-Sepharose. The Sepharose beads were washed three times with 0.1% Triton X-100, 10 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.025% NaN 3 , followed by a single wash with 10 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.025% NaN 3 , and with 50 mM Tris-HCl (pH 6.8). The Sepharose beads-absorbed immune complexes were treated with 20 l of 2ϫ Laemmli's sample buffer containing ␤-mercaptoethanol and were incubated for 5 min in boiling water. The eluted proteins were resolved by SDS-polyacrylamide gel electrophoresis (SDS-PAGE).
N-Glycosidase F Treatment of the Immunoprecipitates-The Sepharose beads-absorbed immune complexes were suspended in 20 l of 0.5% SDS, 0.1 M ␤-mercaptoethanol and incubated for 5 min in boiling water. After centrifugation, the supernatant was mixed with or without 5 units of protease-free N-glycosidase F (Roche Molecular Biochemicals) in a total volume of 60 l containing 50 mM sodium phosphate buffer (pH 7.2), 1% n-octyl glucoside (Roche Molecular Biochemicals) and incubated for 18 h at 37°C. The reaction mixture was then treated with 12 l of 6ϫ Laemmli's sample buffer containing ␤-mercaptoethanol for 5 min in boiling water and resolved by SDS-PAGE.
Detection of Expressed Proteins by Western Blotting-Following immunoprecipitation from the lysate of transfected COS cells, proteins were resolved on SDS-PAGE and transferred to an Immobilon-P membrane (Millipore Corp., Bedford, MA). The blotted membrane was incubated in a blocking solution containing 5% blocking non-fat milk (Bio-Rad) in 20 mM Tris-HCl (pH 7.6), 137 mM NaCl, 0.1% Tween 20 (TBS-T), for 2 h at room temperature or overnight at 4°C and then probed with a specific antibody at 2 g/ml in blocking solution for 1 h at room temperature. Membranes were washed three times with TBS-T and then incubated with TBS-T containing 1:2000-diluted streptavidinhorseradish peroxidase conjugate (Life Technologies, Inc.) or 1:5000diluted goat anti-rabbit immunoglobulin-horseradish peroxidase conjugate (BioSource International) for 1 h at room temperature. After washing three times with TBS-T, proteins recognized by the specific antibody were visualized by the ECL system (Amersham Pharmacia Biotech) according to the manufacturer's instructions.

Surface Expression of Transduced
Protein-Surface expression of wild type CD40L, mutant CD40L species, and control hybrid proteins (F-HybFasL and F-HybCD30L) by transfected COS cells was confirmed by flow cytometry (data not shown). Wild type CD40L and Flag-tagged wild type CD40L (F-Wild) were expressed abundantly by transfected COS cells and detected by mAb 5c8 as well as CD40-Ig construct, suggesting that the Flag peptide at the N terminus does not affect the structure and function of CD40L. Two unique truncated mutants, CytDel and E2skip, both found in XHIM patients (11), were detected in transfected COS cells using mAb 5c8 and the CD40-Ig construct, respectively. The expression of other mutated CD40L species, including mutants with amino acid substitutions and truncated mutants, was confirmed by the binding of a polyclonal anti-CD40L antiserum. Similarly, both F-HybFasL and F-HybCD30L expressed by transfected COS cells were detected well by mAb NOK1 and three different anti-CD30L mAb preparations (M80, M81, and M82), respectively. Since NOK1 recognizes the antigenic epitope consisting of two FasL monomers, it seems likely that F-HybFasL retains the inherent structural integrity of FasL and forms a trimer. The finding that F-HybCD30L expression was detectable by three different anti-CD30L mAbs suggests that this hybrid construct also retains the inherent structural integrities of CD30L and forms a trimer.
Biochemical Characterization of Wild Type CD40L Transiently Expressed by COS Cells-The biochemical characteristics of proteins expressed by COS cells transfected with pWild, pCytDel, and pF-Wild were analyzed by immunoprecipitation followed by Western blotting. From COS cells transfected with pWild, CD40-Ig immunoprecipitated two CD40L species, one with an apparent molecular mass of 33 kDa (p33) and the other of 31 kDa (p31) when probed with mAb bio-106. CD40Ls obtained from pCytDel-transfected cells were found to have an apparent molecular mass of 29 (p29) and 27 kDa (p27), and those obtained from pF-Wild-transfected COS cells had a molecular mass of 34 (p34), 32 kDa (p32), and p29, respectively ( Fig. 2A). Because CD40L has two potential N-glycosylation sites, one in the cytoplasmic domain and the other in the C terminus of the TNF homology (TNFH) domain, N-glycosidase F treatment was performed to determine whether glycosylation contributes to the existence of multiple species of CD40L. After deglycosylation, p31 and p27 were detected in pWild-transfected cells, a single band, p27, in pCytDel-transfected cells, and p32 and p27 in pF-Wild-transfected cells ( Fig. 2A). These results suggest that COS cells express full-length CD40L as well as CD40L lacking the cytoplasmic domain and that both species exist in glycosylated (p33, p29, respectively) and unglycosylated forms (p31, p27, respectively) when transfected with pWild. This interpretation was further supported by the results of probing the same preparation with Rb784 (Fig. 2B), an antiserum recognizing only the cytoplasmic domain of CD40L; p29 and p27 were no longer detectable, and immunoprecipitation of pCytDel-transfected cells failed to identify a band. The production of F-Wild and CytDel both in glycosylated and in unglycosylated forms was also noted in pF-Wild-transfected cells. As reported by others (27), the CytDel form of CD40L occurs and associates with full-length CD40L in Jurkat cells that constitutively express CD40L and in COS cells transfected with the expression plasmid carrying a full-length CD40L cDNA. Interestingly, Rb784 identified a band of 16 kDa (p16) in the immunoprecipitates of pWild-transfected cells and an 18-kDa (p18) band in pF-Wild-transfected cells (Fig. 2B). These small proteins appear to be the N-terminal fragment of CD40L consisting of Met 1 -Met 113 generated after cleavage of the extracellular domain known as soluble CD40L (20,21); p18, observed in the immunoprecipitate of pF-Wild-transfected COS cells, was detected by anti-Flag mAb bio-M5 (Fig. 2C). As shown later (Fig. 5), a protein derived from COS cells to which pF-Stalk, a plasmid expressing Flag-tagged Met 1 -Met 113 , was transfected migrated to the same position as the p18 band. Since the mobility of p16 of pWild-and p18 of pF-Wild-transfected cells on SDS-PAGE did not change following N-glycosidase F treatment (Fig. 2, B and C), the postulated N-glycosylation site of the intracellular domain of CD40L does not seem to be glycosylated. These finding demonstrate that CD40L forms a multiheteromeric complex consisting of full-length CD40L (wild type CD40L or F-Wild) with or without glycosylation, a CD40L lacking the cytoplasmic domain (CytDel) with or without glycosylation, and the N terminus fragment of CD40L (Stalk or F-Stalk) and that CD40-Ig can co-precipitate even Stalk and F-Stalk fragments that lack the extracellular TNFH domain required for the binding of CD40 (5,10,19,31).
Naturally Occurring Mutants of CD40L Co-precipitate with Wild Type CD40L-To study whether wild type CD40L and Flag-tagged naturally occurring mutant CD40L species can associate with each other, immunoprecipitation with CD40-Ig or with the anti-Flag mAb M5 was performed using lysates from COS cells to which pWild and various plasmids express-FIG. 2. Detection of proteins in transfected COS cells by immunoprecipitation and subsequent Western blotting. COS cells were transfected with either pWild, pCytDel, or pF-Wild, and cell lysates prepared 48 h after electroporation were subjected to immunoprecipitation with CD40-Ig. Subsequently, each immunoprecipitate was incubated with or without protease-free N-glycosidase F (see "Experimental Procedures"); Ϫ represents untreated and ϩ treated immunoprecipitate. Samples were resolved by 13% SDS-PAGE, transferred onto Immobilon P membrane, and detected by either mAb bio-106 (A), Rb784 antisera (B), or mAb bio-M5 (C). The sizes of the molecular mass standard proteins are indicated as kDa on the left of each panel.
ing different Flag-tagged mutant CD40L species were co-transfected. To test the system, we co-transfected COS cells with pWild and either pF-Wild, pF-HybFasL, or pF-HybCD30L and examined their association. As shown in Fig. 3A, F-Wild was detected by bio-M5 in CD40-Ig-generated immunoprecipitates of lysates from pWild ϩ pF-Wild co-transfected cells, but nei- Note that all Flag-tagged mutant CD40Ls with amino acid substitutions tested could be detected in the immunoprecipitates obtained with mAb M5 (A) but not in those obtained with CD40-Ig (B). However, F-T147N mutant, which is slightly larger than other Flag-tagged proteins due to the generation of a cryptic glycosylation site by the missense mutation, showed minimal binding with CD40-Ig. C, lysates from COS cells co-transfected with pWild and either pF-Wild or a plasmid carrying Flag-tagged mutant CD40L with an amino acid substitution were subjected to immunoprecipitation with CD40-Ig. The immunoprecipitates were resolved by 12% SDS-PAGE, transferred onto Immobilon P membrane, and probed with mAb bio-M5. The sizes of the molecular mass standard proteins are indicated as kDa on the left of each panel. Note that all mutant CD40Ls with amino acid substitutions tested could be detected in the immunoprecipitates obtained with CD40-Ig when co-expressed with wild type CD40L (C), although they could not be immunoprecipitated with CD40-Ig when expressed alone in COS cells (see B).
ther F-HybFasL nor F-HybCD30L was detected in lysate from pWild ϩ pF-HybFasL or pF-HybCD30L co-transfected cells, respectively. When a membrane with similarly treated samples was probed with mAb bio-106, we found wild type CD40L in all lysates (Fig. 3B), demonstrating that neither F-HybFasL nor F-HybCD30L associate with wild type CD40L. Similarly, wild type CD40L was detected by mAb bio-106 in the mAb M5generated immunoprecipitates only if pWild ϩ pF-Wild were co-transfected but not if pWild ϩ pF-HybFasL or pF-HybCD30L were co-transfected (Fig. 3C), whereas F-Wild, F-HybFasL, and F-HybCD30L were successfully detected by mAb bio-M5 in the appropriate co-transfectant (Fig. 3D). These results demonstrate that F-Wild associates with wild type CD40L and that Flag peptide at the N terminus of CD40L does not affect trimer formation of CD40L monomers.
By using the same technique, naturally occurring mutants of CD40L were co-transfected with wild type CD40L to test whether mutants can associate with wild type CD40L. Seven mutants with a single or two amino acid substitutions were studied (Table I). F-DM which has two amino acid substitutions (S1128R/E129G) and F-T147N are Flag-tagged mutants with amino acid substitutions at positions that directly compromise the CD40-binding site (31). F-Y170C and F-G227V are Flag-tagged mutants of CD40L with amino acid substitutions at positions that are important for trimer formation (31). F-A235P, F-T254M, and F-L258S are Flag-tagged mutants whose amino acid substitutions interfere with monomer packaging and folding (31). As shown in Fig. 4A, all seven Flagtagged missense mutants were expressed by singly transfected COS cells and immunoprecipitated with anti-Flag mAb M5 from the lysate. In contrast, six of seven Flag-tagged mutants failed to be immunoprecipitated with CD40-Ig (Fig. 4B) as expected since these mutants were isolated from patients with XHIM. F-T147N, which has a larger molecular mass than F-Wild and other Flag-tagged missense mutants due to the generation of a cryptic N-glycosylation site by the amino acid substitution, was the only mutant detected in the CD40-Iggenerated immunoprecipitate, although only as an extremely faint band (Fig. 4B). In contrast, all mutants with amino acid substitutions studied were detected (Fig. 4C) in immunoprecipitates obtained with CD40-Ig if COS cells were co-transfected with pWild and expression plasmids of Flag-tagged missense mutants. Similarly, wild type CD40L was readily demonstrated in immunoprecipitates obtained with mAb M5 from the lysate of each co-transfectant (data not shown). These findings clearly demonstrate that all mutants with amino acid substitutions studied associate with wild type CD40L and that the resultant complexes, although they include mutants that cannot bind CD40-Ig by themselves, can bind CD40-Ig. The inability of these mutants to bind CD40-Ig appears not to be due to their inability to form complexes by themselves; when plasmids carrying Flag-tagged mutants and those carrying corresponding plain mutants were co-transfected to COS cells, both Flag-tagged and plain mutants were detected in the immunoprecipitates obtained with mAb M5 using an anti-CD40L polyclonal antiserum (data not shown). This observation implies that these mutants can form complexes themselves but fail to generate binding sites for CD40.
To investigate if truncated CD40L can associate with wild type CD40L, we transfected COS cells with the following Flagtagged truncated mutants (Table I)

. Immunoprecipitation and Western blotting of the lysates from COS cells expressing a truncated CD40L or from COS cells co-expressing wild type CD40L and a truncated CD40L. A and B, COS cells transfected with a plasmid expressing either F-Wild or
Flag-tagged truncated mutant CD40L. The lysates were immunoprecipitated (IP) with either M5 (A) or CD40-Ig (B), resolved by 12% SDS-PAGE, and transferred onto Immobilon P membrane. mAb bio-M5 was used for detection. The sizes of the molecular mass standard proteins are indicated as kDa on the left of each panel. Note that all Flag-tagged truncated mutant CD40Ls tested could be detected in the immunoprecipitates obtained with mAb M5 but not in those obtained with CD40-Ig. C, the lysates of COS cells co-transfected with pWild and either pF-Wild or a plasmid expressing a truncated mutant CD40L were subjected to immunoprecipitation with CD40-Ig. The immunoprecipitates were resolved by 12% SDS-PAGE, transferred onto Immobilon P membrane, and probed with mAb bio-M5. The sizes of the molecular mass standard proteins are indicated as kDa on the left of each panel. Note that all truncated mutant CD40Ls tested could be detected in the immunoprecipitates obtained with CD40-Ig when co-expressed with wild type CD40L (C), although they could not be immunoprecipitated with CD40-Ig when expressed alone in COS cells (see B).
(10); F-Q186X, a truncated mutant carrying a part of TNFH domain, found in XHIM patients (11); F-E2ins, the dominant CD40L mutant found in XHIM patients with intron 2 splice donor site mutations (11); and F-E3skip, the dominant CD40L mutant found in XHIM patients with intron 3 splice donor site mutations (11). Although all truncated CD40L mutants tested were expressed by singly transfected COS cells and detected in the immunoprecipitates obtained with mAb M5 (Fig. 5A), none was immunoprecipitated with CD40-Ig, as expected (Fig. 5B). However, all truncated mutants tested could be demonstrated in the CD40-Ig-generated immunoprecipitates (Fig. 5C) if each expression plasmid of truncated mutant CD40L was co-transfected with pWild. These results suggest that the truncated CD40L species tested are able to associate with wild type CD40L. This was further confirmed by the observation that wild type CD40L was detected in the immunoprecipitates obtained with mAb M5 from the lysates of co-transfectants (data not shown).
The Mutant CD40L Lacking the Exon 2-encoded Stalk Is Less Efficient in forming a Complex with Wild Type CD40L-The mutant E2skip, generated by exon 2-skipping, is a unique truncated CD40L that can still bind CD40-Ig since the entire TNFH domain is preserved. When pE2skip and pF-E2ins (both products generated from patients with intron 2 splice donor site mutations) were co-transfected into COS cells, we failed to demonstrate an association between these two mutants. F-E2ins could not be identified by mAb bio-M5 in the CD40-Iggenerated immunoprecipitates from lysates of pE2skip ϩ pF-E2ins co-transfected cells (Fig. 6, left panel), whereas the as-sociation of wild type CD40L with F-E2ins was observed (see Fig. 5C). Similarly, if COS cells were co-transfected with pE2skip and pF-E2ins, followed by the immunoprecipitate with mAb M5, E2skip could not be detected by mAb bio-106 (data not shown). These observations suggest that the stalk region, which is mainly encoded by exon 2 of the CD40L gene, as well as the TNFH domain, play an important role in the association of CD40L monomers. This conclusion is further supported by the observation that the amount of E2skip found in the mAb M5-generated immunoprecipitate from a pF-Wild ϩ pE2skip co-transfectant was very low (Fig. 6, right lane in right panel) when compared with the amount of E2skip found in the immunoprecipitate obtained with CD40-Ig by which wild type CD40L and E2skip are independently immunoprecipitated regardless of association (Fig. 6, left lane in right panel).
Another interesting feature of the E2skip mutant is the fact that the efficiency of E2skip to transduce a signal through CD40 appears to be less than that of wild type CD40L if E2skip is anchored on cell surface, although the binding of E2skip and CD40-Ig in the cell lysate is normal. Although E2skip is efficiently recovered in CD40-Ig-generated immunoprecipitates from transfected COS cells (Fig. 6, middle panel and left lane in right panel), the staining intensity of pE2skip-transfected COS cells with CD40-Ig (as shown by flow cytometry) is much lower than that of pWild-transfected cells: whereas the mean fluorescence intensity (MFI) of E2skip-expressing COS cells was 60.4, the MFI of wild type CD40L-expressing COS cells was 159.6 when detected by CD40-Ig (immunostained histograms obtained by flow cytometry not shown). On the other hand, similar MFI were obtained for E2skip-and wild type CD40Lexpressing COS cells when mAb 5c8 was used for detection (421.7 and 406.8, respectively). We hypothesize that the weaker intensity of CD40-Ig binding by membrane-expressed E2skip, when compared with the bright staining with mAb 5c8, is due to steric hindrance resulting from the loss of the extracellular stalk which is encoded mainly by exon 2.

Association Occurs When Mutant and Wild Type CD40L Are Co-expressed and the Resultant Complex Is Present on the Cell
Surface-To demonstrate that the association of mutated CD40L with wild type CD40L occurs exclusively when they are co-expressed, we compared the immunoprecipitation of lysates from co-transfectants with that of lysate mixtures prepared from corresponding single transfectants. We selected F-T147N and F-W140X, representing a CD40L with an amino acid substitution and a CD40L that is truncated, respectively, since both are expressed well when transfected into COS cells. Following preclearance of lysate mixture with protein G-Sepharose beads overnight at 4°C and further incubation for 1 h at either 4 or 37°C, immunoprecipitation with mAb M5 was performed. Since wild type, F-Wild, and F-T147N are equally well recognized by mAb 106 (11) and expressed by COS cells in glycosylated and unglycosylated form, the discrimination between wild type, F-Wild, and the mutant is difficult. We therefore treated the immunoprecipitates with N-glycosidase F before resolving with SDS-PAGE. When compared with the immunoprecipitate of COS cells co-transfected with pWild and pF-Wild, only a trace (at 4°C) or a very low amount (at 37°C) of wild type CD40L was recovered from the immunoprecipitate of the mixture of the lysates of each single transfectant (Fig.  7A), suggesting that association of wild type CD40L and F-Wild occurs during co-expression by COS cells but not during the preclearing or immunoprecipitation process. However, a very small amount of wild type CD40L did associate with F-Wild at 37°C (3rd lane from the left in Fig. 7A). To test the association of mutants with wild type CD40L, similar experiments were performed using F-T147N and F-W140X. Similar to the obser- vation made with wild type CD40L and F-Wild, a strong association between wild type CD40L and F-T147N occurred only in co-transfected cells but not if lysates of singly transfected COS cells were mixed in vitro at 4°C, and only at very small quantities when mixed at 37°C (Fig. 7B). No association between wild type CD40L and a truncated mutant F-W140X was observed if mixed in vitro at 4 and at 37°C, respectively (Fig. 7C).
To demonstrate that the complexes formed between wild type and mutated CD40L are anchored in the cell membrane, surface biotinylation of the transfectant was performed using membrane-impermeable Sulfo-NHS-biotin, followed by immunoprecipitation and Western blotting (Fig. 8). An association between wild type CD40L and F-T147N or F-W140X, respectively, was observed on the cell surface in both CD40-Ig-and mAb M5-generated immunoprecipitates (Fig. 8, A and B, respectively) from surface-biotinylated co-transfected COS cells.

Association of Mutant and Wild Type CD40L Occurs in CD4 ϩ T Cell Lines Established from XHIM Patients with Leaky Splice
Site Mutations-The association of wild type CD40L with E3skip were further confirmed in CD4 ϩ T cell lines established from two XHIM patients with different intron 3 splice donor site mutations, nt 367G 3 A in one patient (Fig. 9A, lane 3) and nt 367 ϩ 5g 3 a in the other (Fig. 9A, lane 4). CD4 ϩ T cells from both patients are able to generate normally spliced and exon 3-skipped mRNA (11). When metabolically labeled activated cultured CD4 ϩ T cells were lysed and immunoprecipitated with CD40-Ig, E3skip was clearly detected in lysates from both patients (Fig. 9A, lanes 3 and 4) and in a lysate of pWild ϩ pE3skip co-transfected COS cells (Fig. 9A, lane 7). On the other hand, no association of E2skip with E2ins was observed in a CD4 ϩ T cell line from an XHIM patient with the intron 2 splice donor site mutation nt 309 ϩ 2t 3 a (11) (Fig. 9B), as expected from the finding in co-transfected COS cells (Fig. 8). E2ins could not be detected in the immunoprecipitate obtained with CD40-Ig from the metabolically labeled lysate of the patient's CD4 ϩ T cells (Fig. 9B, lane 2) nor in the lysate of pE2skip ϩ pE2ins co-transfected COS cells (Fig. 9B, lane 6). In contrast, the immunoprecipitates obtained with Rb784 from the patient's CD4 ϩ T cell line (Fig. 9B, lane 4) and from pE2skip ϩ pE2ins co-transfected COS cells (Fig. 9B, lane 8) demonstrated that E2ins was expressed but did not co-immunoprecipitate with E2skip; the short protein band in Fig. 9B, lanes 4 and 8, has the same position as the single band shown in the immunoprecipitate obtained with Rb784 from pE2ins-transfected COS cells (Fig. 9B, lane 10). DISCUSSION CD40L, a member of the TNF superfamily, is expressed on the cell surface as a trimer (5,19) similar to most other members of this family, including TNF, CD30L, and FasL (14). In addition to the full-length wild type CD40L, other derivatives of CD40L participate in forming a heteromultimeric complex. In this study of activated CD4 ϩ T cells and CD40L-transfected COS cells, we have demonstrated that this heterocomplex con-

FIG. 7. Lysates from co-transfected COS cells but not mixtures of lysates from singly transfected COS cells show an association between wild type CD40L and mutants.
A, comparing immunoprecipitates using expression plasmid pWild and pF-Wild. COS cells were either co-transfected with pWild and pF-Wild, singly transfected with pWild, or singly transfected with pF-Wild, and lysates from each transfectant prepared (see "Experimental Procedures"). The lysates from co-transfected COS cells were precleared overnight with protein G-Sepharose at 4°C and subsequently immunoprecipitated (IP) with mAb M5. The lysates from COS cells transfected with either pWild or pF-Wild were mixed, precleared overnight at 4°C, and the supernatant further incubated either at 4°C for 1 h (2nd lane from the left) or at 37°C for 1 h (3rd lane), and immunoprecipitated with mAb M5. To indicate the position of wild type CD40L and F-Wild, respectively, on SDS-PAGE, lysates of COS cells transfected with pWild and with pF-Wild were immunoprecipitated with mAb 106 and M5, respectively. All immunoprecipitates were subsequently treated with N-glycosidase F to simplify the identification of wild type CD40L and F-Wild, resolved by 12% SDS-PAGE, and detected by mAb bio-106. Note that wild type CD40L could barely be detected in the mixture of lysates from each corresponding single transfectant when incubated at 4°C, whereas a small amount of wild type CD40L could be identified when incubated at 37°C; in contrast a significant amount of wild type CD40L was immunoprecipitated together with F-Wild in lysates from the co-transfectant. B, comparing immunoprecipitates using expression plasmid pWild and pF-T147N. This experiment is similar to that shown in A. We used pF-T147N instead of F-Wild to represent a CD40L mutant with an amino acid substitution. The preparation and treatment of cell lysates was the same as those described in A. Note that wild type CD40L could barely be detected in the mixture of lysates from each corresponding single transfectant when incubated at 4°C, although a small amount of wild type CD40L was found when incubated at 37°C. In contrast, larger amount of wild type CD40L was immunoprecipitated together with F-T147N in lysates from the co-transfectant. C, comparing immunoprecipitates using expression plasmid pWild and pF-W140X. This experiment is similar to those shown in A and B. We used pF-W140X as the representative for a truncated CD40L caused by a nonsense mutation. The preparation and treatment of cell lysates were the same as those in A and B except that N-glycosidase F treatment was not performed. Although wild type CD40L could be detected in lysates from the co-transfectant, there was no wild type CD40L demonstrable in the mixture of lysates from each corresponding single transfectant even when incubated at 37°C.
tains not only the full-length wild type CD40L but also a CD40L derivative lacking the intracellular tail (CytDel) and a truncated CD40L ("Stalk") generated by cleaving off the extracellular TNFH region to form soluble CD40L (20,21). The formation of a heteromultimeric complex consisting of fulllength CD40L, CytDel, and soluble CD40L, the latter being a minor fraction of the complex, has previously been reported (27). It has been proposed that CytDel and soluble CD40L are generated by proteolytic cleavage on the cell surface after the full-length CD40L has been transported from the intracellular compartment containing CD40L to the surface (32). Although we have not clearly detected soluble CD40L as one of the constituents in the immunoprecipitates obtained with CD40-Ig or with anti-Flag mAb, we have seen a very faint band, approximately 18 kDa in size (the 1st lane from the left in Fig. 8A) in an immunoprecipitate obtained with CD40-Ig from COS cells that were transfected with pWild and followed by surface biotinylation before lysis. This band may represent soluble CD40L, although we have no direct evidence. The proteolytic cleavage leading to a CD40L trimer containing significant molecules of "Stalk" may play a role in attenuating the CD40L-CD40 interaction in vivo by decreasing the binding site for CD40 and, as a consequence, make CD40 clustering less efficient. This may be an important strategy to limit CD40L activity under physiologic conditions. Several lines of evidence suggest that the extracellular TNFH domain is of importance for trimer formation. A soluble form of CD40L exists as a trimer (19,21) and was found to be biologically active in stimulating B cell proliferation and immunoglobulin class switch (21,33). Based on x-ray crystallography (19) and computer-based structural analysis (31), several amino acid residues in the TNFH domain have been shown to be of importance for trimer formation. Mutations occurring at those residues have been identified in patients with XHIM (10,11). It is of interest that CD40L mutants that have amino acid substitutions at positions Y170C and G227V, known to be important for trimer formation, were still able to associate with wild type CD40L if transfected together with wild type CD40L into COS cells. Furthermore, our data suggest that the Stalk region (His 47 -Met 113 ), which is mainly encoded by exon 2 of the CD40L gene, must also be important for complex formation since all mutants with intact extracellular Stalk region tested were able to associate with wild type CD40L, even if a mutant lacks the entire TNFH domain (Stalk), has a truncated TNFH domain (F-W140X and F-Q186X), or has cryptic amino acid residues after Lys 96 (F-E2ins and F-E3skip). The mutated CD40L lacking the exon 2-encoded Stalk due to an in-frame deletion from Ile 53 to Lys 96 (E2skip) was able to associate with wild type CD40L, although much less efficiently compared with the complex formation between wild type CD40L and either Flag-tagged wild type CD40L or Flag-tagged mutants carrying amino acid substitutions. Thus, the integrity of the entire extracellular part of CD40L, including the extracellular unique region as well as the TNFH domain, is of importance for efficient complex formation of CD40L. Similar to the mechanism involved in TNF receptor activation by TNF-␣ and TNF-␤, ligand-induced receptor clustering or aggregation has been associated with CD40 signaling. Several observations support this hypothesis as follows: (i) both agonistic and non-agonistic anti-CD40 mAbs have been identified (34 -36); (ii) non-agonistic mAb can be rendered bioactive when cross-linked with a secondary antibody (34,35); (iii) agonistic activity of a mAb is not dependent on the epitope to which it binds, and mAbs bound to different epitopes of CD40 can elicit strong CD40 signaling (35); (iv) anti-CD40 mAb BL-C4, a pentameric IgM mAb, has been reported to generate a strong signal that induces monocytes to produce interleukin-1 (37), whereas the anti-CD40 mAb G28-5, known to be agonistic to stimulating B cells (36), did not.
The generation of expression plasmids from cDNA derived from XHIM patients with diverse, informative mutations of CD40L provided us with the opportunity to co-transfect COS cells with wild type and mutated CD40L to investigate heterotrimer formation. We found that various naturally occurring CD40L mutants can physically associate with wild type CD40L and form a complex on the cell surface if simultaneously translated. It is intriguing to consider the possible physiologic function of these complexes, although the stoichiometry of these heterotrimers has not been determined in this study. Binding of CD40 takes place in three grooves formed by the three monomers (25,26), and each monomer contributes to form the functional groove for CD40 binding. A homotrimer consisting of wild type CD40L is expected to bind three CD40 molecules resulting in the clustering of CD40 and signal transduction. In contrast, a heterotrimer consisting of two wild type CD40L molecules and one mutated CD40L molecule is expected to bind only one CD40 molecule. A heterotrimer consisting of one wild type and two mutated CD40L molecules can no longer bind CD40. Thus, heterotrimer formation is expected to decrease the number of binding sites for CD40 and may no longer be able to ligate CD40 sufficiently to initiate signal transduction in CD40-expressing cells such as B lymphocytes, monocytes, and biotinylated using Sulfo-NHS-biotin which is membrane-impermeable (see "Experimental Procedures"). After cells were lysed, expressed proteins were immunoprecipitated (IP) with either CD40-Ig (A) or mAb M5 (B) and resolved by 13% SDS-PAGE. Following transfer, the membrane was probed with streptavidin-horseradish peroxidase conjugates. Note that F-T147N and F-W140X are immunoprecipitated with CD40-Ig only when co-transfected with pWild (A) and that wild type CD40L is immunoprecipitated with mAb M5 only when co-transfected with pF-T147N or pF-W140X (B), suggesting that the association occurs on the cell surface of transfected COS cells. follicular dendritic cells. As previously reported (11), activated T cells from most patients with XHIM have mutated CD40L mRNA transcript levels that are comparable to those found in normal activated T cells and, following activation, express mutated CD40L on the cell surface. If the association between wild type and mutated CD40L species, as we have observed in this study, leads to heterotrimer formation and to a less efficient CD40 engagement on CD40-expressing cells, gene therapy for patients with XHIM may face additional difficulties.
A dominant negative effect due to heterotrimer formation may be responsible for the immunodeficiency in some XHIM patients with splice site mutations. As previously reported (11) some splice site mutations of CD40L are "leaky" and allow T cells to generate both normally and abnormally spliced mRNA transcripts. This is a naturally occurring in vivo situation where the association of wild type CD40L and mutant CD40L is likely to take place with physiologic consequences. By using an RNase protection assay, we have determined the ratio of mis-spliced to normally spliced mRNA transcripts and found in most instances that the dominant mRNA species among multiple splicing products are the mis-spliced mRNAs (11). In a patient with intron 3 splice donor site mutation (nt 367G 3 A) (patient AM in Ref. 11), the amount of exon 3-skipped transcript was approximately twice that of the normally spliced transcript. A similar ratio of the transcript carrying a 19 nucleotide insertion of intron 2 (producing E2ins) to the exon 2-skipped transcript, which generates a mutated CD40L (E2skip) that can still bind CD40-Ig, was found in a patient with the intron 2 splice donor site mutation nt 309 ϩ 2t 3 a (patient PS in Ref. 11). In both cases, the amount of normally spliced and the amount of exon 2-skipped transcripts, respectively, were quantitated as being 20 -30% of normal control, which by itself may be too high to be entirely responsible for the clinical phenotype of XHIM. Interestingly, the clinical pheno-type of patient PS, whose mutation generates two mutated CD40L species (E2skip and E2ins) that do not associate with each other, is milder than that of patient AM (nt 367G 3 A) and patient JE (11) (nt 367 ϩ 5g 3 a) whose intron 3 splice site mutations generate wild type CD40L and E3skip that complex with each other. Whereas patient PS was completely healthy until he developed parvovirus B19-induced anemia at 17 years of age, patient AM presented at 5 years of age with recurrent infections, and patient JE developed Pneumocystis carinii pneumonia at 10 months of age (11). Based on these clinical observations that suggest a correlation between the physiologic co-expression of functional CD40L and non-functioning mutated CD40L in XHIM patients with splice site mutations, it is likely that the association of wild type CD40L with mutated CD40L has a disadvantage in the in vivo signal transduction through CD40. To carry out functional assays to assess signal transduction via CD40 by heterotrimers consisting of wild type CD40L and mutant CD40L, it is necessary to establish stably transfected cell lines. If such experiments prove a dominant negative effect on signaling, gene therapy for XHIM will be difficult to achieve. Nevertheless, in considering gene therapy for XHIM, it is important to determine the type of mutation responsible for XHIM in a given family, the amount of expressed mRNA transcript, the quantity of mutant CD40L generated, and the ability of a given mutant CD40L to form complexes with wild type CD40L. The sizes of the molecular mass standard proteins are indicated as kDa on the left. B, a truncated E2skip mutant does not associate with a truncated E2ins mutant in a CD4 ϩ T cell line from an XHIM patient with intron 2 splice site mutation (nt 309 ϩ 2t 3 a). An activated CD4 ϩ T cell line derived from the patient with the mutation nt 309 ϩ 2t 3 a was metabolically labeled and then lysed. Similarly, COS cells were either co-transfected with pE2skip and pE2ins or transfected with pE2ins, metabolically labeled, and then lysed as a control experiment. Cell lysates are as follows: lanes 1, 2, 3, and 4, CD4 ϩ T cell line from the XHIM patient with nt 309 ϩ 2t 3 a; lanes 5, 6, 7, and 8, COS cells co-transfected with pE2skip and pE2ins; lanes 9 and 10, COS cells transfected with pE2ins. As indicated, control human IgG 1 was used for the immunoprecipitation in lanes 1 and 5, CD40-Ig in lanes 2 and 6, control rabbit sera in lanes 3, 7, and 9, and Rb784 in lanes 4, 8, and 10. The sizes of the molecular mass standard proteins are indicated as kDa on the left. In contrast to A which shows that wild type CD40L and E3skip can associate with each other and be immunoprecipitated with CD40-Ig, E2skip, a truncated mutant which can still bind CD40-Ig, does not associate with E2ins; CD40-Ig was able to immunoprecipitate only E2skip although E2ins was co-expressed by CD4 ϩ T cells effectively as demonstrated by the immunoprecipitation with Rb784, an antiserum recognizing the intracellular domain of CD40L.