SAP-Regulated T Cell–APC Adhesion and Ligation-Dependent and -Independent Ly108–CD3ζ Interactions

This article is featured in In This Issue, p.3833

Signaling lymphocytic activation molecule (SLAM)–associated protein (SAP) is a Src homology (SH) domain 2–containing intracellular adaptor protein that is predominantly expressed in cells of the hematopoietic system, particular those of the T and NK lineages. The SAP deficiency causes profound defects in the germinal center (GC) response and long-term humoral immunity (1–4). At the cellular level, these defects stem from an intrinsic failure of SAP-deficient CD4⁺ T cells to stably interact with Ag-presenting B cells (5). Without stabilized adhesion, otherwise competent Th cells cannot efficiently deliver contact-dependent help signals (e.g., CD40L) to drive optimal B cell clonal expansion and GC formation. Interestingly, the SAP deficiency apparently affects only T cell interactions with B cells, but not with dendritic cells (DCs) (5), although underlying mechanisms for such distinction are yet to be defined.

The SAP SH2 domain binds to the immunoreceptor tyrosine-based switch motifs (ITSMs) (6) in the cytoplasmic domain of the SLAM receptor and its related CD2-like transmembrane (TM) protein family members (SLAMF) including Ly108 and CD84, which are highly expressed on activated T and B cells (7–9). SAP has two homologous adaptor proteins, EAT-2 and ERT, which bind to ITSMs of the SLAMF molecules in a similar fashion but are expressed only in macrophages, B cells, NK cells, and not in T cells (10–12). A majority of SLAMF receptors including SLAM, CD84, and Ly108 are homophilic molecules whose extracellular domains form homodimers in trans across cell membranes and regulate intercellular signal exchange (13–15). Homotypic engagement of SLAMF proteins leads to ITSM-mediated recruitment of SAP and additional signaling molecules to their cytoplasmic domains (16, 17). Most notably, ITSM-bound SAP can physically interact with the SH3 domain of Fyn kinase by an R78-based motif (18, 19), likely through an induced conformational change after SLAM trans engagement (20). Increased Fyn recruitment enhances SLAM phosphorylation and downstream signaling cascades that may augment T cell cytokine production (21, 22). Additional SH3-containing molecules may interact with SAP by this R78-based mechanism (23, 24), although the R78-based motif is not required for SAP-mediated stabilization of cognate T–B adhesion or GC development (5, 25). In principle, SAP could harbor R78-independent motifs that recruit signaling molecules other than Fyn to SLAMF receptors and thereby promote cognate T–B adhesion.

Biochemical analyses have also indicated that SLAMF receptors can directly bind to SH2 domain–containing phosphatases, including SHIP-1 and SH2 domain–containing tyrosine phosphatase-1 and -2 (SHP-1 and -2) (26–29). These phosphatases, once recruited to the membrane, may suppress proximal signaling “downstream” of various immune receptors (30–32). Several SLAMF receptors in NK cells were found to inhibit cytotoxicity when SAP and SAP-related adaptors are absent (33–35), leading to the notion of “switch” functions for SAP family adaptors in that they may physically or functionally compete with negative signaling modules associated with SLAMF receptors, and thereby convert inhibitory SLAMF functions into stimulatory actions (16). In support of this switching model, Veillette and colleagues (36) have shown that SAP simultaneously couples Fyn kinase to and uncouples lipid phosphatase SHIP-1 from SLAMF receptors in NK cells.

It is less clear as to how antagonism of negative signaling and/or triggering a positive signaling cascade contributes to SAP-mediated control of cognate T–B adhesion. Earlier evidence in vitro suggested that CD84 and Ly108, both highly expressed by activated T and B cells, positively promote Ag-specific T–B adhesion in an SAP-dependent fashion (9). A more recent study indicates that, in the absence of SAP, Ly108 transmits inhibitory signals into the T cells to dampen their adhesive interactions with B cells (37). This inhibitory effect of Ly108 was correlated with its association with SHP-1 and localization to the synapse, and genetic ablation of Ly108 significantly rescued the GC defect seen in SAP-deficient animals (37). The apparently contradicting effects of Ly108 in the two studies suggest more complexities in Ly108- and SAP-mediated regulation of T–B interactions. Specifically, if Ly108 can indeed positively promote T–B adhesion in an SAP-dependent manner, SAP would alone or in collaboration with Ly108 recruit positive signaling partners by structural motif(s) yet to be defined. In contrast, if Ly108 primarily exerts an inhibitory effect in the absence of SAP, the targets of such inhibition would have to impinge on Ag-specific T–B adhesion. In this study, we systematically examined these two possibilities.

B6 (Jax 664), OVA_323–339–specific TCR transgenic OT-II (Jax 4194), dsRed-expressing (Jax 6051), and HEL-specific Ig-transgenic MD4 (Jax 2595) mice were purchased from The Jackson Laboratory. SAP-deficient mice were a kind gift from Dr. Pamela Schwartzberg (38). A new Ly108-deficient strain on the B6 background was developed using TALEN technology (Supplemental Fig. 3). All mice were maintained under specific pathogen-free conditions and used in accordance of governmental and institutional guidelines for animal welfare.

CD4⁺ T cells were isolated using the CD4 (L3T4) MicroBeads; CD19⁺ B cells were isolated using the CD19 MicroBeads or the B Cell Isolation Kit (Miltenyi Biotec) according to the manufacturer’s protocols. DCs were isolated by CD11c Microbeads (Miltenyi Biotec) after digestion with 400 g/ml collagenase D and 20 g/ml DNase I for 1 h (Roche). For in vitro T cell activation, OT-II T cells were cocultivated together with splenocytes pretreated with mitomycin (Sigma) in the presence of 1 μM OVA_323–339 peptide (Genscript), whereas polyclonal CD4 T cells were stimulated with plate-bound anti-CD3 and anti-CD28 (BioXCell), and both were maintained in the presence of 10 ng/ml IL-2 (PeproTech).

Retroviruses expressing desired target genes were packaged with the Plat-E system. In brief, Plat-E cells were transfected using X-tremeGENE HP Transfection Reagent (Roche) according to the manufacturer’s protocol. For retroviral transduction, cells were spin-infected at 1500 × g with appropriate viral supernatants in the presence of 1 μg/ml polybrene (Sigma) and appropriate cytokines for 2 h at 32°C. The murine stem cell virus–based, GFP-tagged vector was custom-modified from the plasmid murine stem cell virus puro vector (Clontech) by substituting its phosphoglycerate kinase promoter–puromycin cassette with a human ubiquitin promoter-driven enhanced GFP fragment. The plasmid–internal ribosome entry site (IRES)–enhanced GFP, plasmid IRES-monomeric RFP (mRFP), and plasmid-(GGGGS)₃-linker-GFP (for expression of fusion proteins) vectors were modified from the Moloney murine leukemia virus–based retrovirus vector by substituting its IRES-blasticidin with IRES-GFP, IRES-mRFP, and (GGGGS)₃-linker-GFP, respectively.

Bone marrow reconstitution was used to obtain DCs overexpressing different SLAMF molecules. In brief, bone marrow cells were harvested 4 d after B6 donors were primed with 150 mg/kg fluorouracil (Sigma) and cultured in DMEM (Life Technologies) containing 6 ng/ml IL-3, 10 ng/ml IL-6, and 50 ng/ml stem cell factor (PeproTech). These cells were infected by appropriate retroviruses on the following day and were used to reconstitute B6 recipient mice that were lethally irradiated by x-ray (550 rad × 2) at 5 × 10⁵ bone marrow cells per recipient. Splenic DCs were isolated from these mice 8 wk after reconstitution.

OT-II T cell blasts were stained with 5 μM TAMRA, and B cells that were activated by 1 μg/ml LPS (Sigma) for 2 d or freshly isolated splenic DCs were stained with 50 μM CMF₂HC (Invitrogen). T cells (2.5 × 10⁵ per tube) were spun down at 300 × g with B cells (5 × 10⁵ per tube) or DCs (2.5 × 10⁵ per tube) that had been pulsed with Ag, and then incubated at 37°C for indicated amount of time. After vortexing for 30 s, the frequency of T–B conjugates was enumerated by flow cytometry. In certain experiments, T cells retrovirally transduced with indicated expression constructs were used. See Supplemental Fig. 1 for cytometry gating details.

T cells were fixed by 2% paraformaldehyde at 4°C for 1 h and permeabilized by 0.1% saponin (Sigma) at room temperature for 30 min. AlexaFluor 647–conjugated anti-SAP [clone 12C4, a kind gift from Dr. André Veillette (39)] was used to stained for 30 min.

Previously activated T cells were deprived of IL-2 for 12 h before biochemical analyses. When CD3 or Ly108 stimulation was involved, ∼5 × 10⁶ T cells were washed twice with ice-cold serum-free RPMI 1640 (Life Technologies), incubated with 20 μg/ml biotinylated anti-CD3 or anti-Ly108 (eBioscience) on ice for 30 min, washed, and then incubated with 10 μg prewarmed streptavidin (Roche) for indicated amount of time at 37°C. Stimulated cells were lysed in the lysis buffer containing 50 mM Tris (pH 8.0), 1% Nonidet P-40, 2 mM EDTA, and protease (Sigma) and phosphatase inhibitor (Thermo Scientific) mixtures. Lysates were precleared by fixed Staphylococcus aureus at 4°C for 1 h, before supernatants were incubated with relevant primary Abs (anti-Ly108 from eBioscience, anti-SLAM from BioLegend, anti-GFP from Abcam, or anti-CD3ζ from Santa Cruz) at 4°C for 1 h, followed by 2-h incubation with rabbit anti-mouse or anti-rat Abs (Jackson Immunoresearch) already bound to S. aureus. After repeated washings, immunoprecipitates were eluted by 1× Non-reducing Lane Marker Sample Buffer (Thermo Scientific) for 15 min at room temperature and then boiled at 100°C for 10 min. Proteins were separated by SDS-PAGE and transferred to nitrocellulose membranes. Membranes were blocked with TBS containing 5% BSA, 0.1% Tween 20. To detect target molecules by immunoblotting, we used rabbit anti–SHP-1 (Santa Cruz), mouse anti-CD3ζ (Santa Cruz), mouse anti-Ly108 (eBioscience), rabbit anti-GFP (Abcam), and 4G10 mouse anti-phosphotyrosine (Millipore). Appropriate HRP-conjugated secondary reagents were from Jackson Immunoresearch Laboratories.

The procedure was essentially as previously described (40). GFP-expressing SAP knockout (KO) OT-II CD4⁺ T cells that were transduced with mRFP-coding retroviral vectors that express SAP or free cytoplasmic domains of indicated SLAMF molecules were isolated by flow sorting and transferred into B6 recipient mice at 10⁶ cells/mouse. Three days later, these mice further received 10⁶ naive MD4 B cells and were immunized s.c. with 30 μg HEL-OVA conjugate Ag, which was prepared by a HydraLink conjugation kit (SoluLink) as previously described (5). Two days after immunization, draining lymph nodes were imaged using an excitation wavelength of 880–920 nm with a typical xyz dimension of 500 × 500 × 21 μm at a rate of 30 s per frame. Each image sequence lasted for 1 h. The mRFP fluorescence used for sorting transduced T cells is not optimally excited by 880–920 nm and does not interfere identification of T and B cells by the transgenic GFP and dsRed markers, respectively. Imaging data were analyzed with the Imaris software (Bitplane). Adobe After Effect was used to prepare time-lapse image sequences and movies.

Activated T cells of indicated genotypes were washed and resuspended with ice-cold serum-free RPMI 1640, placed onto Polysine Adhesion Slides (Electron Microscopy Sciences), and fixed by 4% paraformaldehyde (Electron Microscopy Sciences) for 10 min. After blocking with 1% FBS in PBS for 20 min, slides were incubated with rat anti-CD3ε (BioXCell) or -CD11a (BD Biosciences) for 20 min, washed, and then stained with goat anti-rat AF568 (Invitrogen) for 20 min. The slides were then blocked with purified rat IgG (Thermo) for 20 min before being further stained with biotinylated anti-SLAM or anti-Ly108 for 30 min in the presence of excessive rat IgG. After washings, slides were stained with streptavidin-AF488 (Invitrogen) for 20 min before mounting for imaging with the N-SIM microscope (Nikon). The plane of focus for each cell was chosen such that the cross section has a maximal diameter. Pearson’s correlation coefficients between AF488 and AF568 channels were calculated for the cell membrane by the Imaris software (Bitplane).

293T cells were cultured in a glass-bottom dish and cotransfected with the Ly108- or LSL-GFP fusion and the CD3ζ-mRFP fusion (CD3ζ was a gift from Dr. Matthew Krummel). Forty-eight hours after transfection, cotransfected cells were imaged with an A1RSi confocal microscope (Nikon) at room temperature. One picture of GFP and mRFP were separately taken before and after mRFP bleaching by 30-s illumination with the 561-nm laser at 100% power. Individual cell–cell contacts were cropped to quantitate GFP fluorescence intensity before and after the mRFP acceptor bleaching. Images were analyzed using the NIS-Elements software (Nikon). In separate experiments, activated SAP KO T cells were simultaneously transduced with CD3ζ-mRFP and Ly108- or LSL-GFP. Cells were deprived of IL-2 for 12 h before sorting for double-transduced T cells. Such cells were incubated with 20 μg/ml biotinylated anti-Ly108 on ice for 30 min, washed, and then warmed to 37°C for 5 min before cross-linking with prewarmed streptavidin for an additional 15 min at 37°C. Cells were then immediately plated onto polysine adhesion slides and fixed with 4% paraformaldehyde for 10 min. Imaging and analysis were as described earlier except that mRFP acceptor bleaching lasted for 150 s and the entire T cell membrane was included for quantitation of GFP intensity before and after the acceptor bleaching.

Statistical tests were performed with Prism 5.0 (GraphPad). Unless specifically indicated otherwise, t tests were used to compare end-point means of different groups. Error bars depict the SEM.

SAP is required for stable cognate T–B contacts in vivo, an observation that is recapitulated by a flow cytometry–based adhesion assay in which conjugates between T cells and APCs are enumerated (5). However, because the original conjugation assay involved multiple staining and washing steps, it has not been entirely clear whether SAP is required for cognate T–B adhesion soon after Ag recognition takes place (9). To address this issue, we prelabeled T cells and B cells with fluorescent cytoplasmic dyes so that the conjugate frequency could be measured at any desirable time points after the cells were brought together by brief centrifugation (Supplemental Fig. 1). As shown in Fig. 1A, the defect of SAP-deficient T cells in adhering to Ag-presenting B cells was evident 5 min after mixing, the earliest time when the T–B conjugation frequency could be reliably measured. Blocking of LFA-1 essentially abrogated T–B conjugate formation (Fig. 1B). It thus appears that SAP is somehow involved in early TCR signaling events, which can then trigger inside-out signaling to integrins to secure the cell–cell adhesion.

To examine whether unrecognized structural motifs on the surface of the SAP molecule may be essential for recruiting signaling molecules to mediate its adhesion-promoting functions, we conducted alanine mutagenesis and tested SAP mutants for complementing SAP-deficient T cells after retroviral transduction. To focus on the most relevant target residues of the SAP molecule, we took advantage of the fact that EAT-2 and SAP SH2 domains share a similar overall structure and 47% amino acid identity, and are expected to bind to the same set of SLAMF ITSMs (11). As shown in Fig. 2A, both bare SAP SH2 domain (SAP^SH2) and bare EAT-2 SH2 domain (aa 1–103; EAT-2^SH2) rescued SAP-deficient T cells as efficiently as the full-length SAP molecule. We thus aligned SAP and EAT-2 SH2 domains according to Morra and coworkers (11), and chose for mutagenesis a total of 13 residues (Y7, H8, G9, L19, L20, L21, G24, G62, L83, I84, S85, V95, and L98) that are identical between SAP and EAT-2 SH2 domains, nonessential for phosphotyrosine binding, not involved in the R78-based motif [residues 75–82, according to Latour et al. (19)], but are solvent exposed on the molecular surface, and thus may potentially interact with other molecules (18, 41). These 13 residues were individually replaced by alanine. Expression of seven mutants (Y7A, H8A, L20A, G24A, L83A, I84A, and L98A) was undetectable by intracellular staining using a SAP-specific mAb 12C4 or when expressed as GFP fusions, suggesting that these SAP mutant proteins were unstable (data not shown). Consistent with this notion, some of these residues (e.g., Y7, H8, I84) were indeed mutated in the X-linked lymphoproliferative disease patients (11, 26). The remaining six mutants (G9A, L19A, L21A, G62A, S85A, V95A) were tested for rescuing SAP-deficient T cells in conjugate formation with Ag-presenting B cells, together with the loss-of-function R55L and neutral R78A mutants as positive and negative controls, respectively. As shown in Fig. 2B, SAP^G9A, SAP^L19A, SAP^L21A, SAP^S85A, and SAP^V95A mutants were able to rescue as efficiently as the intact SAP or the SAP^R78A mutant. Although the SAP^G62A failed to rescue efficiently, it was expressed at a level >10-fold lower than any of the controls (Fig. 2B); thus, the failure likely resulted from hypomorphic expression rather than alteration of a G62-based structural motif.

The C-terminal tail of the SAP molecule harbors a total of four serine and threonine residues, three of which are conserved and could potentially be phosphorylated. It is possible that these serine/threonine residues could exert self-inhibitory effects on putative SAP interactions with other molecular partners, and the bare SAP SH2 domain lacking the tail might be a constitutively active form. To rule this out, we changed all serine and threonine residues to alanine to obtain the SAP^ST mutant and found it fully functional (Fig. 2B). Finally, because the SAP molecular surface harbors a highly charged area centered on S85 and K89, we also tested a charge-inversion mutant in which S85 was changed to an arginine and K89 to an aspartic acid (SAP^S85R-K89D). As shown in Fig. 2B, this mutant was also as competent as an intact SAP molecule. Taken together, our comprehensive mutagenesis analysis reveals no loss-of-function mutants and strongly suggests that, other than the phosphotyrosine-binding groove characteristic of an SH2 domain, no additional motifs on the molecular surface are functionally essential to SAP-mediated stabilization of cognate T–B adhesion.

It has recently been shown that, in the absence of SAP, SLAMF molecules such as Ly108 are associated with inhibitory phosphatases and dampen Ag-specific T–B adhesion (36, 37, 42, 43). To examine whether the inhibitory function is a unique property of Ly108, we retrovirally transduced SLAM, CD84, and two main Ly108 isoforms into either WT or SAP KO T cells (Fig. 3A), and tested these cells for adhesion to Ag-presenting B cells. As shown in Fig. 3B, whereas SLAM and CD84 transduction did not have any appreciable effects, increased Ly108 expression significantly reduced conjugate frequencies in both types of T cells. This inhibitory effect was more pronounced in the absence of SAP but clearly detectable in its presence. Therefore, the inhibition of Ag-specific T–B adhesion is a unique function of Ly108 that can be antagonized by SAP.

Ag-specific T–DC adhesion is apparently normal in the absence of SAP in T cells, even though DCs express Ly108, albeit at a level lower than that of B cells (5). To determine whether the differential requirement of SAP for Ag-specific T–DC and T–B adhesion reflects a quantitative difference in Ly108 expression between the two APC types, we tested Ly108-overexpressing DCs derived from Ly108-transduced bone marrow stem cells. As shown in Fig. 3C, Ly108-overexpressing DCs were clearly impaired in supporting Ag-specific adhesion by SAP-deficient T cells as compared with control DCs that overexpressed GFP or SLAM or CD84. Therefore, T–DC interactions are not fundamentally different from T–B cell interactions in terms of their sensitivities to Ly108-mediated inhibition, and the level of Ly108 trans-engagement by APCs determines the strength of inhibition experienced by the T cells.

In the absence of SAP, Ly108 exhibits increased association with SHP-1 in activated T cells upon anti-CD3 restimulation (42) or upon Ag-specific engagement by presenting B cells (37, 43). As shown in Fig. 4A, we found that a certain level of SHP-1 was constitutively associated with Ly108 in activated T cells regardless of whether SAP is present, and that the SHP-1 association was reduced in WT T cells but increased in SAP KO T cells after restimulation. To test whether unstable T–B interactions caused by the SAP deficiency in vivo are indeed due to the inhibitory effect of Ly108–SHP-1 complexes, the Ly108 cytoplasmic domain, termed Ly108-T, was expressed as a standalone, cytosolic protein in the SAP KO T cells. We reasoned that the cytosolic LY108-T should competitively reduce SHP-1 recruitment to the Ly108 molecule on the plasma membrane. As shown in Fig. 4B, 15 min after anti-CD3 treatment of SAP KO T cells that expressed cytosolic Ly108-T, SHP-1 that can be coimmunoprecipitated with Ly108 was reduced. In contrast, expression of the CD84 or SLAM cytoplasmic domain (CD84-T, SLAM-T) did not have this effect (Fig. 4B and data not shown). Consistent with these observations, SAP KO T cells complemented with SAP or cytosolic Ly108-T, but not with CD84-T or SLAM-T, exhibited significantly improved adhesion to Ag-presenting B cells in vitro (Fig. 4C). In separate experiments, Ly108-T and SLAM-T were expressed as GFP-tagged versions to experimentally verify level of expression by GFP blotting, and similar results were obtained (Supplemental Fig. 2). To test whether membrane Ly108–SHP-1 association accounts for the inability of SAP KO T cells to stably interact with Ag-specific B cells in vivo, we conducted intravital imaging using GFP-expressing SAP KO OT-II T cells that were retrovirally transduced with vector control, SAP, Ly108-T, or SLAM-T. Two days after adoptive transfer of these T cells together with MD4 B cells into separate B6 mice that were immunized with HEL-OVA, T–B contacts were visualized in vivo. As exemplified in Fig. 4D and corresponding Supplemental Videos 1–4, SAP or Ly108-T transduction rendered SAP KO T cells capable of forming long-lasting T–B conjugates, whereas vector control or SLAM-T–transduced cells failed to do so. As quantitated in Fig. 4E, contact durations of the SAP and Ly108-T groups were comparably long (28.7 ± 2.6 versus 26.5 ± 2.7, p = 0.6) and much more durable than the vector control (9.1 ± 0.9; SAP versus vector, p < 0.0001; Ly108-T versus vector, p < 0.0001) or the SLAM-T group (7.7 ± 1.0 min). These data indicate that, when SAP is absent, Ly108–SHP-1 complexes on the plasma membrane exert a dominant inhibitory effect on Ag-specific T–B interactions in vivo.

Consistent with the fact that SAP is required for Ag-specific T–B adhesion almost as soon as the onset of Ag recognition (Fig. 1A), the baseline CD3ζ phosphorylation was significantly lower in SAP KO T cells as compared with their wild-type counterparts (73 ± 6% of the control, p < 0.05; Fig. 5A). Strikingly, the baseline CD3ζ phosphorylation could be further reduced upon increased Ly108 expression in either SAP KO (Fig. 5B) or WT T cells (Fig. 5C). Because the 3- to 5-fold increase achieved by retroviral transduction was similar to the physiological increase in Ly108 expression on follicular Th (T_FH) cells as compared with other types of T cells (Supplemental Fig. 3), these data suggest that CD3ζ could be a target for Ly108-mediated inhibition through SHP-1, particularly given the fact that Ly108 is constitutively associated with SHP-1 even in the presence of SAP (Fig. 4A, 4B). To test whether Ly108 and CD3ζ molecules are colocalized in close proximity, we used structure illumination microscopy (SIM) to analyze the distribution pattern of Ly108 in relation with CD3ε as a surrogate marker for the CD3 complex. The Pearson correlation algorithm was used to compute pixel-by-pixel correlation of Ly108 and CD3ε. For comparison, we also examined in a pairwise manner Ly108–CD11a, SLAM–CD3ε, and SLAM–CD11a correlation on the membrane of individual T cells. As exemplified in Fig. 5D and quantitated in Fig. 5E, Ly108 was highly correlated with CD3ε, but not with CD11a, in both WT and SAP KO T cells, whereas SLAM was more correlated with CD11a, but not with CD3ε, in the two types of cells. Because the Nikon N-SIM scope setup we used provides a lateral resolution of 100–200 nm, these data demonstrate Ly108, but not SLAM, is significantly colocalized with the CD3 complex in an SAP-independent manner and support the possibility that SAP and Ly108 regulate baseline CD3ζ phosphorylation by tuning locally available SHP-1 activities.

To determine whether Ly108 TM and/or cytoplasmic domain controls its distribution pattern with relation to the CD3 complex, we compared Ly108 with domain-swapped Ly108-SLAM hybrid molecules, LSL, LLS, and LSS (see schematic depiction in Fig. 5F). To avoid interference from endogenously expressed Ly108, we transduced these hybrid molecules into T cells from an Ly108-deficient strain that was newly made directly on the B6 background (Supplemental Fig. 4). As shown in Fig. 5G and 5H, the substitution of the Ly108 TM with SLAM TM domain (LSL) did not change its distribution pattern, but replacement of the cytoplasmic domain with or without concomitant TM swapping (LSS and LLS) led to significantly reduced CD3ε colocalization. Therefore, the cytoplasmic domain of Ly108 is important not only for recruiting SHP-1, but also for colocalizing the Ly108 molecule with the CD3 complex.

CD3ζ dephosphorylation by surrounding Ly108–SHP-1 complexes would set a higher threshold for TCR activation of SAP KO T cells and could, in principle, explain why SAP KO T cells fail to form stable Ag-specific conjugates with B cells. In contrast, the same SAP KO T cells differentially adhered to Ag-presenting DCs that express different levels of Ly108 (Fig. 3C), indicating that the degree of surface Ly108 ligation in trans determines the strength of inhibition that T cells experience. A provocative possibility is that Ly108 trans-engagement promotes spatially closer colocalization and physical interactions with the CD3 complex, and thereby more efficiently suppresses CD3ζ phosphorylation.

At the light-microscopic resolution, SLAMF receptors including both Ly108 and SLAM are accumulated to the T–B interface soon after conjugate formation (Fig. 6A), presumably in the trans-engaged form. However, it is difficult to ascertain whether Ly108 trans engagement precedes, coincides with, or follows TCR triggering in this assay system, and it is not clear whether the interface accumulation of SLAMF molecules is partly a consequence of TCR- and/or SAP-dependent signaling. To create a system where Ly108 trans engagement can be analyzed in insolation from TCR and SAP, we cotransfected 293T cells with a construct expressing Ly108-GFP fusion protein and a construct expressing SAP-mRFP fusion protein. By chance, some Ly108-expressing cells would be in contact with other Ly108-expressing cells. In such cell couples or clusters, as shown in Fig. 6B, Ly108 was predominantly accumulated at the interface between contacting cells that both expressed Ly108. SAP could be recruited to the interface, as expected, if Ly108 had at least one intact ITSM. Importantly, wild-type or ITSM-mutated Ly108 exhibited identical interface distribution patterns regardless of SAP recruitment. Therefore, trans-engaged Ly108 automatically accumulate toward the cell–cell contact interface without requiring TCR- or SAP-mediated signaling. To examine how close trans-engaged Ly108 could colocalize and/or interact with CD3ζ, we cotransfected 293T cells with Ly108-GFP and CD3ζ-mRFP constructs. As shown in Fig. 6C, a significant fraction of CD3ζ molecule was distributed to the contact interface where trans-engaged Ly108-GFP accumulated. We then took advantage of the fact that GFP and mRFP form a fluorescence resonance energy transfer (FRET) pair and tested GFP fluorescence recovery after bleaching the mRFP acceptor. As shown in Fig. 6C and quantitated in Fig. 6D, mRFP bleaching led to a significant recovery of GFP fluorescence in 30 s. As a positive control for comparison, the Ly108-GFP recovery was ∼5–10% after bleaching SAP-mRFP in similar cotransfection experiments (data not shown). These data suggest a fraction of trans-engaged Ly108 and CD3ζ molecules must be within 10 nm from each other, a distance that likely permits efficient molecular interactions. Strikingly, when the Ly108 TM domain was substituted with the SLAM TM domain, the resulting LSL-GFP hybrid molecule can still accumulate to the interface area together with the CD3ζ-mRFP molecule, but the two molecules failed to exhibit FRET as efficiently (Ly108 versus LSL: 3.8 ± 0.5 versus 0.0 ± 0.4%, p < 0.0001). These data suggest that trans-engaged Ly108 interact with the CD3ζ molecule in a TM domain–dependent manner.

To test whether Ly108 TM domain–mediated Ly108–CD3ζ interactions take place in T cells, we transduced Ly108-GFP or LSL-GFP and CD3ζ-mRFP into SAP KO T cells and mimicked Ly108 trans engagement by Ab cross-linking. As shown in Fig. 6E, more pronounced FRET signals were observed between Ly108-GFP and CD3ζ-mRFP than that between LSL-GFP and CD3ζ-mRFP, indicating that trans-engaged Ly108 indeed interacts with CD3ζ in a TM domain–controlled fashion. To test whether this TM domain–directed Ly108–CD3ζ interaction is indeed important for Ly108-mediated inhibitory effects on cognate T–B adhesion in the absence of SAP, Ly108 or LSL was retrovirally transduced into SAP KO T cells. Because this led to a 3- to 5-fold increase in surface Ly108 expression (Supplemental Fig. 2), the exogenous Ly108 or LSL accounts for 70–90% of the surface Ly108 on transduced T cells. Ly108 and LSL are associated with comparable levels of SHP-1 as assayed by immunoprecipitation (data not shown) and are correlated with similar levels of the background CD3ζ phosphorylation (Fig. 6G, time 0). Strikingly, however, cross-linking of Ly108, but not LSL, led to dephosphorylation of the CD3ζ molecule (Fig. 6G, 6H). In a remarkable correlation, when Ag-presenting B cells were used to engage the transduced SAP KO T cells, only increased Ly108, but not LSL, expression led to a significant reduction in Ag-specific T–B adhesion (Fig. 6I). We therefore conclude that the inhibitory effect of Ly108 on cognate T–B adhesion in the absence of SAP depends on Ly108 trans engagement that, in a unique TM domain–dependent manner, facilitates Ly108–CD3ζ interactions that promote SHP-1–mediated CD3ζ dephosphorylation and curtail proximal TCR signaling.

A disorganized synapse, particularly in the form of insufficient actin and SHP-1 clearance from its central region, has been suggested as an underlying mechanism for why SAP-deficient T cells cannot efficiently adhere to Ag-presenting B cells (37, 43). However, organized synapses with clear cSMAC and pSMAC segregation take shape 5–10 min after the onset of TCR signaling and increase in LFA-1–mediated adhesion (44–47). Given our observations that the background CD3ζ phosphorylation in T cells is reduced in the absence of SAP and that the requirement for SAP is already evident within the first 5 min after the onset of Ag recognition, SAP likely plays a role in promoting proximal TCR signaling of an optimal strength, a notion also supported by a recent study of human SAP-deficient T cells (48), and the reduced adhesion and synapse deformation are consequences of impaired TCR signaling in its absence. To fulfill this function, SAP may recruit additional signaling intermediates to the SLAMF receptors to amplify the proximal TCR signaling cascade. The Fyn kinase can directly bind to SAP and become associated with the SAP–SLAM complex (18, 19). However, R78A mutant SAP that essentially loses the ability to interact with Fyn can still promote stable T–B adhesion in vitro (5) and promote GC formation in vivo (25, 49). SLAM is also not required for GC formation in vivo (49). In contrast, genetic evidence indicates that CD84 is required for optimal cognate T–B adhesion and GC formation in a protein immunization model (9). These results prompted us to analyze the SAP molecule by systematic alanine mutagenesis for potential structural motifs that could recruit signaling intermediates to CD84 or other SLAMF receptors. Our data reveal that a bare SAP SH2 domain is sufficient for promoting cognate T–B adhesion, and that no alanine replacement or charge inversion of its surface-exposed residues impairs its function while maintaining the structural stability. They do not support the model of SAP-mediated recruitment of signaling intermediates but are compatible with the possibility that SAP functions by blocking ITSM-mediated SLAMF association with inhibitory phosphatases (6, 26).

Crotty’s group (37) was the first to report the fascinating finding that genetic ablation of Ly108 largely nullifies the need for SAP to stabilize T–B conjugation in vitro and to promote GC formation in vivo, suggesting Ly108 is an inhibitory SLAMF receptor that suppresses cognate T–B adhesion in the absence of SAP. This work, together with two other studies (42, 43), also indicates that SHP-1 is the inhibitory phosphatase associated with Ly108 in T cells. These findings are confirmed in this study, because neither SLAM nor CD84 but Ly108 is found to associate with SHP-1 and inhibit T cell adhesion to Ag-presenting B cells in vitro. Importantly, by two-photon intravital imaging, we have verified that reduction of Ly108–SHP-1 association on the T cell plasma membrane indeed renders SAP-deficient T cells capable of forming stable mobile conjugate pairs with cognate B cells, thus demonstrating that the inhibitory effect of Ly108 is responsible for the lack of productive T–B interactions in the absence of SAP in vivo.

A long-standing mystery is how the SAP deficiency impinges on T–B interactions without affecting T–DC interactions (5, 9). The model of SAP-antagonized inhibition by Ly108, rather than a model of SAP-mediated recruitment of positive signaling intermediates, suggests that T–DC adhesion might be spared of crippling inhibition simply because of the moderate Ly108 level on DCs as compared with that on B cells (9). Consistent with this, our work demonstrates much less efficient conjugation between SAP-deficient T cells and Ag-presenting DCs expressing increased levels of Ly108. Curiously, when we measured TCR downregulation, a more proximal consequence of TCR signaling than adhesive conjugation with APCs, even the moderate Ly108 level on normal DCs was demonstrably inhibitory (data not shown). Therefore, the qualitatively different regulation of T–DC and T–B interactions by SAP as originally reported (5) most likely results from quantitatively differential engagement of the Ly108-mediated inhibitory pathway. It has long been appreciated that the duration of T–DC interactions in vivo evolve in distinct phases, sequentially from a form of stable conjugation to more of a short but swarming configuration (50–52). Kinetic changes in Ly108 expression on DCs and/or T cells may play a role in these phase transitions.

The most striking finding in this study is that Ly108-mediated inhibition targets CD3ζ directly by two concatenated colocalization processes. First, Ly108 and the CD3 complex are distributed in a markedly correlated manner on the membrane of T cells even without overt stimulation. It is increasingly appreciated that the plasma membrane is partitioned into cytoskeletal corrals by cortical actin filaments (53). In such picket-and-fence compartments, heterologous TM proteins might thermodynamically diffuse in a correlated manner as a result of compatible biophysical features of their TM and cytoplasmic domains (54), creating clusters of distinct proteins apparent on a length scale of few tens of nanometers when viewed by low-speed or static imaging methods (55). The correlated distribution of Ly108 and the CD3 complex at the static SIM imaging resolution of 100–200 nm is consistent with this dynamic model. This correlation is uniquely achievable by the Ly108 cytoplasmic domain, but not that of SLAM, although we do not yet understand the precise biophysical and biochemical properties that underlie these differences between Ly108 and SLAM. Nonetheless, the Ly108–CD3 distributional correlation is likely the reason why CD3ζ phosphorylation is reduced when Ly108 expression is increased, given the fact that Ly108 is constitutively associated with SHP-1 even in SAP-competent cells, a phenomenon observed in this study for mouse and previously reported for human T cells (42). Interestingly, T_FH cells express ∼5-fold more Ly108 than other activated T cells (9), and T_FH–B cell interactions inside the B cell follicle are considerably less durable than those between activated T and B cells at the T–B border (5, 56). We speculate that, by being constitutively distributed in correlation with the CD3 complex, Ly108 might be involved in setting a threshold for T cell sensitivity to B cell–mediated Ag presentation inside the follicle and GCs in SAP-sufficient animals. In contrast, Ly108-deficient mice do not show exaggerated GC formation or enhanced humoral responses in a model of viral infection (37), suggesting other inhibitory molecules could substitute for dampening baseline CD3ζ phosphorylation when Ly108 is missing.

A second Ly108-CD3 colocalization process reported in this article depends on Ly108 trans engagement. NTB-A molecules, the human Ly108 homolog, form homodimers in solution with an equilibrium dissociation constant of ∼2 μM (14). Whereas this modest affinity would not allow Ly108 to mediate cell–cell adhesion, prolonged membrane juxtaposition in the absence of TCR or SAP-dependent signal transduction can drive Ly108 into the interface of cell couples, likely because of the reduced diffusibility after trans engagement. After cross-linking, Ly108 colocalize and interact with CD3ζ within the FRET distance, and thus more effectively put the latter within the radius of effective inhibition by Ly108-associated SHP-1. Crucially, this interaction specifically requires the Ly108 TM domain. The SLAM TM domain permits the hybrid LSL molecule to be normally expressed, associated with SHP-1, and distributed in correlation with the CD3 complex, but does not support efficient CD3ζ interactions. As a result, LSL loses the inhibitory effect of Ly108 on cognate T–B adhesion. Similarly, substitution with the TM domain of CTLA-4, PD-1, or BTLA, three inhibitory molecules that can associate with phosphatases and dampen T cell activation (57–61), fails to preserve the Ly108-mediated inhibition (C. Chu and H. Qi, unpublished observations). Possibly, a peculiar conformational change has to be induced either in the Ly108 TM domain itself or, through its mediation, in the cytoplasmic domain of the Ly108 to permit efficient CD3ζ interaction. Both CTLA-4 and PD-1 have been shown to associate with and inhibit CD3ζ phosphorylation (58, 62), and mice deficient in CTLA-4 and PD-1 succumb to autoimmune inflammation of various levels. No autoimmune condition has so far been reported in the two lines of Ly108-deficient mice previously made (37, 43) or in the new line reported in this article. It is not yet clear how CTLA-4 and PD-1 are colocalized with the CD3 complex. An interesting implication of our data are that different biophysical mechanisms of colocalization and interaction with CD3ζ may generate different patterns of inhibitory modulation of proximal TCR signaling, leading to divergent T cell behaviors tailored to specific contexts. In the case of cognate T–B adhesion, Ly108 potentially sets up an intercellular negative feedback in that TCR signaling and integrin-mediated adhesion promote cell–cell conjugation, which potentiates Ly108 trans engagement and accumulation to the contact interface, forcing on CD3ζ an increased exposure to SHP-1–mediated dephosphorylation. In SAP-deficient T cells, this negative feedback is unopposed and terminates TCR signaling-dependent T–B adhesion prematurely. In the wild-type T cells, in contrast, SAP recruitment to Ly108 cannot only antagonize such negative regulation but also may be a key kinetic parameter that is modulated upon TCR triggering to control the strength and duration of cognate T–B interactions in a spatiotemporally specific manner.

We thank Drs. Xinquan Wang and Nieng Yan for advice on the structure of the SAP molecule.

The authors have no financial conflicts of interest.