Identification of SH3 domain interaction partners of human FasL (CD178) by phage display screening

Protein-protein interactions link signal transduction pathways from receptors to the nucleus and govern intracellular processes as diverse as organelle targeting, directional transport, cytoskeletal reorganization, membrane positioning, endo- and exocytosis and protein degradation. Protein-protein interactions are mostly mediated by modular domains with the best studied examples being Src homology (SH) 2 and 3 domains [1]. SH3 domains are phylogenetically highly conserved protein interaction modules that comprise 50 to 70 amino acids and are found in a variety of functionally unrelated proteins. As typical interaction modules, they fold into a similar globular structure. Most SH3 domains bind proline residues in a certain arrangement (e.g. PxxP) in so-called "proline-rich domains" (PRD) [2, 3].

Fas ligand (FasL, CD95L, Apo-1L, CD178) is a type II transmembrane protein of the tumor necrosis factor superfamily that acts as a prototypic death factor of immune cells [4, 5]. FasL is employed by cytotoxic T lymphocytes (CTLs) and natural killer (NK) cells to kill virus-infected or tumorigenic cells. It is implicated in the downregulation of immune responses by activation-induced cell death, the establishment of immune privilege, and in the modulation of T cell activation [6‐8]. FasL is stored in so-called secretory lysosomes and is recruited to the immunological synapse in an activation-restricted fashion. Its surface expression is down-modulated by shedding through a disintegrin and metalloprotease (ADAM) 10 activity and intramembrane proteolysis by the γ-secretase-like protease SPPL2a (signal peptide peptidase-like 2a). The released intracellular domain may translocate to the nucleus or be prepared for degradation [9‐11].

The FasL N-terminus comprises a unique PRD that contains several putative SH3 domain binding sites (Fig 1). Different experimental approaches have already led to the identification of several FasL-interacting proteins including Src-related tyrosine kinases (Fyn, Lyn, Lck, Hck, Fgr, Src, and Abl), adapter proteins involved in T cell receptor (TCR)-associated signal transduction (Grb2, Gads, p85 subunit of PI3 kinase, Nck) and members of the Pombe Cdc15 homology (PCH) protein family (protein kinase C and casein kinase substrate in neurons 1-3 (PACSIN1-3), Formin-binding protein 17 (FBP17), Cdc42-interacting protein 4 (CIP4), CD2-binding protein 1 (CD2BP1), Rho GTPase-activating protein 4 (ARHGAP4), Fer/CIP4-homology (FCH) and double SH3 domains 1 (FCHSD1) and SLIT-ROBO Rho GTPase-activating protein 2) [12‐16]. Many aspects of FasL biology are indeed closely linked to PRD-SH3 domain interactions: members of the PCH family regulate lysosomal association [16, 17], tyrosine kinases are involved in reverse signaling and sorting of human FasL to multivesicular bodies [18, 19], and the adapter protein Nck is crucial to bring FasL loaded vesicles to the immunological synapse [20].

The aim of the present study was to get an idea about the complete FasL-SH3 domain interactome and to define interactors that could be involved in the translocation of FasL to the nucleus or in the priming of N-terminal fragments (NTF) for degradation. We used a phage display library that covers the entire human SH3 domain proteome to screen for interactions with the FasL N-terminus. We confirmed several previously identified interactions and introduce a number of SH3 domain proteins as novel candidate FasL binding partners. These include additional non-receptor tyrosine kinases (e.g. the Tec kinases), sorting nexins and other cytosolic or nuclear adapter proteins that could be involved in the intracellular trafficking of FasL or FasL fragments.

In its proline-rich domain (PRD), the human FasL contains several putative SH3 domain binding sites. Previous studies have employed different strategies to screen for FasL interaction partners, including yeast two-hybrid assays, peptide- and recombinant SH3 domain-based experiments as well as proteomic screens of FasL precipitates from T cells and neuronal cells [12‐16]. We now screened a human SH3 domain phage display library, covering the entire human SH3 proteome with 288 conventional SH3 domains.

In order to obtain phage clones binding the FasL N-terminus, magnetic beads were coated with affinity-purified GST-hFasL(1-80) fusion protein or GST as a control. The beads were used to pan the phage display library as described. Following binding, the beads were extensively washed with buffer containing either 0.15% or 0.5% Tween-20. It is important to mention that phage titers differed significantly depending on the detergent concentration. For less stringent washing conditions (0.15% Tween-20), a titer of (2.85 ± 0.15) × 10⁷ PFU/ml was calculated in eluate obtained from the GST-panning, while the titer following more stringent washing was (0.185 ± 0.075) × 10⁷ PFU/ml. Under the same conditions, eluates from GST-hFasL(1-80) panned beads displayed titers of (1.59 ± 4.53) × 10⁸ PFU/ml and (3.31 ± 0.6) × 10⁷ PFU/ml, respectively, showing that the specificity of phage-displayed SH3 domain-pVIII fusion proteins binding to the hFasL(1-80) moiety was significantly increased at higher stringency.

Using the phage display library approach, 49 human SH3 domains displayed in vitro binding to hFasL(1-80) (Table 1 and Fig 2). Of note, some SH3 domains were identified only once in one of the two screenings, for instance the tyrosine kinase Itk. In contrast, for other SH3 domains, including those of CD2BP1, osteoclast stimulating factor 1 (OSTF1) and sorting nexin 33 (SNX33), several transduced clones were found. The number of hits, however, does not necessarily reflect the binding specificity of SH3 domains to the hFasL PRD, since it cannot be guaranteed that all 288 phage clones are represented in comparable amounts within the library. Of note, in the case of CD2BP1, we detected a point-mutated residue within the SH3 domain which was verified by the supplier and which might have an impact on the affinity to target motifs.

Several of the detected interactions have been previously described. These include FasL binding to SH3 domains of non-receptor tyrosine kinases such as Fyn, Hck, Lyn, and Src. Notably, the Fyn SH3 domain has not only been the first interaction module associated with FasL [12], but Fyn and Lyn have recently been implicated in tyrosine phosphorylation of FasL associated with sorting to multivesicular bodies [18]. Furthermore, SH3 domains of the adapter proteins Nck1 and Nck2 were found to bind hFasL in our assay. In both cases, the interaction was mediated by the second SH3 domain. This is partially in line with our previous observation that Nck-binding to FasL is mediated by the second and, although less efficiently, the third SH3 domain [13, 20]. The missing reactivity of the third SH3 domain in the phage display screen may be due to a lower representation of respective phages in the library. Also, SH3 domains of the NADPH oxidase subunit p47phox and the regulatory subunit of the phosphoinositide 3-kinase (PI3K) were found to bind to the FasL PRD as reported before [13]. Interactions of FasL with members of the PCH protein family have been described in different studies based on pulldown experiments, yeast two hybrid screens, and co-immunoprecipitation and co-localisation of overexpressed proteins [14, 16, 17]. Although we meanwhile know that several members of different PCH protein subfamilies interact with FasL, only the SH3 domains of CD2BP1 (Proline-serine-threonine phosphatase-interacting protein 1 (PSTPIP1) in mice) and SLIT-ROBO Rho GAP2 bound hFasL(1-80) in the present study.

Importantly, we identified several previously not described putative interaction partners of the FasL N-terminal region. In addition to the non-receptor tyrosine kinases mentioned before, the phage display library screen revealed that the Src-like kinase Yes as well the Tec family kinases Tec and Itk might bind to hFasL. Although Yes is expressed in a number of cells types, including T cells [21], its actual involvement in T cell signal transduction and especially FasL phosphorylation still remains elusive. Tec kinases constitute a well-characterized family of non-receptor tyrosine kinases that have been implicated in a variety of signaling processes, predominantly controlled by Src kinases. In particular, the involvement of Itk in T cell development and activation has been highlighted over the past few years [22]. The interaction of Tec kinases with FasL might therefore be interesting in the context of primary T cell activation and is already subject of further investigation. In a first series of pull-down experiments, we used a GST fusion protein containing the isolated Itk SH3 domain to verify FasL binding (Fig 3). As shown before for other interactions [13], FasL was precipitated with Itk SH3 domains from lysates of stably FasL expressing KFL9 cells as a doublet and from JFL39.1 cells as a single band. Whether full length Tec family kinases also interact with FasL, and whether there is a differential binding to full length versus proteolytically processed FasL remains to be assessed.

Several previously identified FasL interactors regulate the subcellular localization of the death factor, and have been implicated in its association with the lysosomal compartment or the transport of FasL-loaded secretory vesicles to the immunological synapse [6, 23]. In our phage display screen, we identified the sorting nexins SNX9 and SNX33, as well as endophilin A3 and intersectin-2 as potential binding partners for FasL. All four proteins are involved in membrane or vesicle trafficking and might therefore act as co-regulators of FasL transport in T and NK cells. The closely related SNX9 and SNX33 are members of the large sorting nexin protein family which share a phosphoinositide-binding phox-homology (PX) domain and are associated with various aspects of endocytic and endosomal sorting [24]. Together with SNX18, which was identified to interact with FasL in neuronal cells [17], SNX9 and SNX33 share a similar structural organization and are thus grouped into the SNX9 subfamily of sorting nexins [25]. The three proteins are widely expressed, even though expression levels may vary between individual tissues and cell types. In the context of FasL, the sorting nexins could play a role during internalization and processing of FasL N-terminal fragments following shedding and/or intramembrane proteolysis. In the case of the sorting nexins, we verified the interaction with the cytosolic region of FasL in pull-down experiments performed from lysates of individually transfected 293 T cells. In this case, all three HA-tagged sorting nexins (9, 18 and 33) were readily detected by an anti-HA antibody in whole cell lysates and in precipitates with GST-hFasL(1-80) fusion proteins but not in GST control precipitates (Fig 4). As for Tec kinases and PCH proteins, future experiments will address the potential differential binding of individual interactors to full length versus proteolytically processed FasL.

Endophilins are cytosolic proteins that also contain N-terminal Bin-Amphiphysin-Rvs (BAR domains) and C-terminal SH3 domains and are implicated in membrane tubulation during endocytosis [26]. Endophilin A3 enhances amyloid precursor protein (APP) α-secretase cleavage, possibly by inhibition of APP endocytosis [27]. Its role in endocytosis and APP shedding as well as its probable association with ADAM proteases could make endophilin A3 an interesting interaction partner of FasL in the context of "re-internalization". Intersectin-2 is a cytosolic adapter protein that, among other motifs, harbors five SH3 domains and a pleckstrin homology (PH) domain for membrane association. It was also implicated in clathrin-mediated endocytosis and apparently links Wiskott-Aldrich syndrome protein (WASP)-initiated actin polymerization to TCR-induced endocytic processes [28, 29].

The PCH proteins SLIT-ROBO Rho GAP1, SLIT-ROBO Rho GAP2 and SLIT-ROBO Rho GAP3 are coordinators of cytoskeleton-associated membrane or organelle transport processes. Both Rho and Cdc42 GTPase-activating proteins (GAPs) were associated with migratory signal transduction in neuronal cells [30]. Similarly, the Rho GTPase-activating protein 9 (ARHGAP9), which exhibits substantial GAP activity for Rac1 and Cdc42, was discussed to regulate adhesion of hematopoietic cells to extracellular matrix proteins [31]. More recently, it was shown that ArhGAP9 binds to ERK and p38 kinases and keeps these proteins in an inactive state [32]. In view of recent data from our group showing that reverse FasL signaling leads to an inhibition of T cells associated with decreased MAPK activity [8], the putative interaction of FasL with ArhGAP9 might be relevant for this aspect of FasL biology.

The myosin XV SH3 domain also bound FasL in this phage display screen. Since myosins are actin-based motor proteins, an association with the FasL N-terminus could again suggest an impact on the transport of FasL or FasL loaded vesicles. However, as shown for T cells from myosin XV deficient mice, myosin XV seems to be dispensable for cytoxicity of murine cytotoxic T-lymphocytes (CTLs) [33]. Since it is currently still a matter of debate whether FasL is actually stored in granzyme-/perforin-loaded secretory lysosomes or in a separate cytotoxic compartment [34, 35], it will be interesting to determine whether myosin XV plays a role in the Nck-dependent transport of FasL and its storage granules to the cell surface.

SPIN90 (SH3 protein interacting with Nck, 90 kDa) is a cytoskeletal adapter protein with a single SH3 domain that binds to the first and third SH3 domain of Nck via its own proline motif. It contains a putative nuclear localization signal (NLS) but is functionally associated with cytosolic (re-)organization of the actin cytoskeleton, leading, for instance, to lamellopodia formation in COS-7 cells [36]. It also associates with dynamin, WASP, and the Arp2/3 (actin-related proteins 2/3) complex and plays a key role for myofibril and sarcomere assembly. The SH3 and PX domain containing adapter protein 2B (SH3PXD2B) is a so far very poorly defined gene product. It interacts with hFasL in vitro via its third and its fourth SH3 domain. Only recently, this protein has been implicated in podosome formation of Src-transformed fibroblasts [37].

We also identified the SH3 domains of Fyn-binding protein (FYB)/adhesion and degranulation promoting adapter protein (ADAP) and of Src kinase-associated phosphoprotein 2 (SKAP2)/Src kinase-associated phosphoprotein 55-related (SKAP55R) as putative FasL-interacting domains. FYB/ADAP is an adaptor protein involved in TCR signaling and 'inside-out' activation of T cell integrins and is expressed in various hematopoietic cells except B cells [38‐40]. Upon TCR engagement, FYB/ADAP is tyrosine-phosphorylated by Fyn and is recruited via SH2 domain binding to SLP-76 (SH2 domain containing leukocyte protein of 76 kDa) in a signaling cluster with phosphorylated linker for T cell activation (LAT) and Gads. Via an internal PRD, ADAP associates with SKAP1/SKAP55 or SKAP2/SKAP55R. This complex allows for membrane-recruitment of Rap1 which then triggers integrin activation crucial for lymphocyte extravasation as well as formation of immunological synapses. In addition, FYB/ADAP also links the TCR signaling complex to the actin cytoskeleton [38‐40]. Given that hFasL(1-80) potentially interacts with FYB/ADAP as well as the SKAP2/SKAP55R SH3 domains, it will be of particular interest to test whether this interaction can also be observed in CTLs and what functional implications such an interaction might have for FasL recruitment and expression.

In conclusion, the performed phage display library screening clearly confirmed the previous notion that the N-terminal PRD of FasL is rather promiscuous in terms of SH3 binding. Meanwhile, more than 50 putative interactors have been named. Importantly, several protein families clearly stick out of the gross of SH3 domain containing proteins. These include classical cytosolic adapter proteins, non-receptor protein tyrosine kinases, PCH proteins, and sorting nexins. On the other hand, given the unusual length of the PRD, it may not be surprising to find so many potential interactors for this region. In order to understand the role of individual proteins in the context of FasL biology, we already biochemically verified the interaction of FasL with several of the novel interactors, including the Tec kinase Itk and the sorting nexins 9, 18 and 33. It is important to highlight that most if not all studies performed so far on putative FasL interactors revealed clear morphological phenotypes. We are thus confident that the present study provides a new basis for the further analysis of protein-protein interactions around FasL. Besides defining the role of individual interactions, one of the major tasks will be to determine which interaction "dominates" in a given physiological context or cell type. Here, we expect new insights into processes of endo- and exocytic vesicular transport and the modulation of signal transduction through FasL. Potential functions of newly identified SH3 domain proteins interacting with FasL are depicted in Fig 5. Given the stepwise proteolytic processing of FasL and its translocation to the nucleus, it has to be determined whether certain FasL interactors selectively regulate the "late functions" of FasL.