Binding of the intracellular Fas Ligand (FasL) domain to the adaptor protein PSTPIP results in a cytoplasmic localization of FasL

The tumor necrosis factor family member Fas ligand (FasL) induces apoptosis in Fas receptor-expressing target cells and is an important cytotoxic effector molecule used by CTL- and NK-cells. In these hematopoietic cells, newly synthesized FasL is stored in specialized secretory lysosomes and only delivered to the cell surface upon activation and target cell recognition. FasL contains an 80-amino acid-long cytoplasmic tail, which includes a proline-rich domain as a bona fide Src homology 3 domain-binding site. This proline-rich domain has been implicated in FasL sorting to secretory lysosomes, and it may also be important for reverse signaling via FasL, which has been described to influence T-cell activation. Here we report the identification of the Src homology 3 domain-containing adaptor protein PSTPIP as a FasL-interacting partner, which binds to the proline-rich domain. PSTPIP co-expression leads to an increased intracellular localization of Fas ligand, thereby regulating extracellular availability and cytotoxic activity of the molecule. In addition, we demonstrate recruitment of the tyrosine phosphatase PTP-PEST by PSTPIP into FasL.PSTPIP.PTP-PEST complexes which may contribute to FasL reverse signaling.


The TNF family member Fas Ligand (FasL) induces apoptosis in Fas receptor-expressing target cells and is an important cytotoxic effector molecule used by CTL-and NK-cells. In these hematopoietic cells newly synthesized
FasL is stored in specialized secretory lysosomes and only delivered to the cell surface upon activation and target cell recognition. FasL contains an 80 aa long cytoplasmic tail which includes a proline rich domain as a bona fide SH3 domain-binding site. This proline rich domain has been implicated in FasL sorting to secretory lysosomes and it may also be important for reverse signaling via FasL which has been described to influence T-cell activation. Here we report the identification of the SH3 domain containing adaptor protein PSTPIP as a FasL-interacting partner which binds to the proline rich domain. PSTPIP co-expression leads to an increased intracellular localization of Fas Ligand, thereby regulating extracellular availability and cytotoxic activity of the molecule. In addition, we demonstrate recruitment of the tyrosine phosphatase PTP-PEST by PSTPIP into FasL/PSTPIP/PTP-PEST complexes which may contribute to FasL reverse signaling.
Fas Ligand (FasL, CD95/APO-1 Ligand, CD178, TNFSF6) is a 281 amino acid (aa) long type II transmembrane molecule belonging to the large TNF family of proteins which bind to and activate members of the TNF receptor family (1,2). FasL and its corresponding receptor Fas (CD95/APO-1) both interact as oligomers (3), and the activated Fas receptor complex initiates a pro-apoptotic death signal in the receptor-bearing cell (4,5). Although mainly known for its death-promoting activity, FasL has also been studied in the context of further non-apoptotic functions (1,2,6). Such experiments are motivated by the special structure of this type II transmembrane protein. In addition to its hydrophobic transmembrane domain, which anchors the molecule within the plasma membrane, and to its extracellular ectodomain, responsible for binding to its receptor, the FasL protein possesses an 80 aa long N-terminal intracellular part which is responsible for the transduction of (extracellular) signals into FasL-bearing cells and/or which may fulfill regulatory functions. Such a "reverse signaling" has been described in mouse T-cells which display an altered proliferative behavior upon triggering of their FasL surface molecules (7)(8)(9). Results obtained with mouse sertoli cells, in which FasL engagement leads to MAPK activation, as measured by an increase in Erk phosphorylation, also imply stimulatory FasL reverse signaling (10). The importance of the intracellular domain for FasL function and regulation is underlined by its high homology among different species (7,11,12). Several signaling motives within the FasL intracellular domain are highly conserved, including a tandem casein kinase I phosphorylation site (aa 17-21 (11)), a phosphorylatable tyrosine at position 7, a proline rich region (aa  with bona fide SH3/WW domain-binding sites (7), and a potential nuclear localization sequence (aa 69-76; V. Kirkin, unpublished data). These sequence elements are likely to contribute to FasL reverse signaling and regulation which, however, so far has only been described phenomenologically. No molecular mechanism for such signaling or for any FasL regulation involving the intracellular ligand domain has been identified. Apart from transducing signals into the ligandexpressing cell, the intracellular FasL domain may be involved in other regulatory processes. Interestingly, FasL is not expressed ubiquitously on the surface of hematopoietic cells (e.g. cytotoxic T-cells and NK-cells) but is rather sorted to special organelles, the so-called secretory lysosomes, from where the molecule is delivered to the cell surface upon contact with a target cell (13). This sorting to secretory lysosomes is dependent on the proline rich region within the intracellular FasL domain (14), but again no molecular interactors have been identified that regulate lysosomal targeting. We screened a Yeast-Two-Hybrid library to isolate proteins which bind to the intracellular FasL domain and which are involved in such processes. Here we describe the identification and analysis of the adaptor protein PSTPIP (proline, serine, threonine phosphataseinteracting protein; in this manuscript the human homolog CD2BP1 is for simplicity called hPSTPIP (15,16)) as a FasL interacting protein, which appears to be involved in the regulation of FasL cell surface expression versus intracellular storage. Co-expression of PSTPIP with FasL in non-hematopoietic 293T cells leads to a decline in cell surface FasL, while the amount of intracellular FasL is increased. We also show that PSTPIP simultaneously interacts with the phosphatase PTP-PEST to form a ternary complex with FasL, which may be important for the phosphorylation status of PSTPIP and FasL.

Experimental Procedures
Cell Culture and transfections HEK 293T, COS-1, RBL 2H3 and HeLa cell lines were grown in Dulbecco's modified Eagle's medium (DMEM) with 10% (v/v) FCS plus penicillin/streptavidine (Pen/Strep; Invitrogen). L12.10 mouse thymoma cells were stably transfected with p B a b e -m F a s L by electroporation and hygromycin selection in RPMI medium plus 10% FCS and Pen/Strep. HEK 293T and COS-1 cells were transfected with 5-10 µg of total plasmid DNA using Fugene reagent (Roche) according to the manufacturer´s protocol. For transient HeLa and RBL-2H3 cell transfections 1x10 7 cells were electroporated with 20µg plasmid DNA using a Gene Pulser II RF model (Bio-Rad).

Antibodies and reagents
The polyclonal rabbit antiserum against murine PSTPIP was a gift from L. Lasky (Department of Molecular Oncology, Genentech, Inc., South San Francisco, California 94080, USA) and the polyclonal rabbit antiserum against PTP-PEST from N. Tonks (Cold Spring Harbor Laboratory, Cold Spring Harbor, New York, USA). Human PSTPIP was detected with rabbit antiserum from Carol Wise (Texas Scottish Rite Hospital for Children, Sarah M. and Charles E. Seay Center for Musculoskeletal Research, Dallas, Texas, USA) or with anti-V5 antibody (Invitrogen), and human PSTPIP lacking the SH3 domain with anti-Myc 9B11 (Cell Signaling). The anti-Flag M2 monoclonal antibody was obtained from Sigma. Cell surface human FasL was stained for FACS analysis with the anti-FasL antibody Nok-1 (PharMingen), and total hFasL was probed in Western Blot analysis with the anti-FasL antibody G-247 (P h a r M i n g e n) .

Yeast-Two-Hybrid Screening
The intracellular domain of mouse Fas ligand (aa 1-80) was cloned into the plasmid pAS2.1 (pAS2.1 mFasL-intra; Clontech) in frame with the Gal binding domain. A murine T-cell lymphoma cDNA library (MATCHMAKER, cat.no. #ML4001AE, Clontech) with a diversity of 3x10 6 individual cDNA clones was used. The yeast strain Y-190 (Clontech) was sequentially transformed first with pAS2.1 mFasL-intra and then with the library. Prior to library screening, yeasts were tested for expression of the intracellular FasL domain by Western Blot analysis. Yeast cells were plated and incubated on agar plates lacking leucine, tryptophane and histidine for 5-10 days at 30 °C. Altogether 6.5x10 6 transformed yeast clones were screened for FasL interaction partners. Colonies appearing were restreaked onto histidine minus plates and subsequently tested for β-galactosidase activity (Bartel et al., 1995). The cDNA inserts in the pACT prey vector were isolated from yeast colonies with β-galactosidase activity and sequenced using standard procedures.
In vitro translation and GST pull-down hFasL, hFasL Δ1-39, hFasL Δ40-80, mPSTPIP, hPSTPIP and hPSTPIP ΔSH3 were used as templates for in vitro transcription using T7 RNA polymerase and in vitro translation in the presence of 35 [S]-labeled methionine and rabbit reticulocyte lysate (Promega). The in vitro translated proteins were employed for GST pulldown assays with GST alone or with GST fusion proteins.

In vitro binding-assay with GST fusion proteins
All GST fusion constructs were subcloned into p G E X -A H K (Pharmacia), resulting in the addition of N-terminal GST tags. GST fusion proteins were expressed in the exponentially growing Escherichia coli BL21C+ bacterial strain (Stratagene) by induction for 3 hours with 1 mM isopropyl-1-thio-β-D-galactopyranoside (Roth) at 30 °C. Bacteria were harvested by centrifugation and the fusion proteins were isolated using gluthatione-sepharose (Sigma). GST fusion proteins were incubated with the appropriate in vitro-translated and 35 [S]m e t h i o n i n e -l a b e l e d p r o t e i n s i n immunoprecipitation buffer (20 mM Tris, pH 7.9, 0.02% NP-40, 10% glycerol, 100 mM NaCl, 1 mM EDTA, 4mM MgCl, and 0.2mg/ml lysozyme, 5 mM β-mercaptoethanol, 0.5 mM PMSF plus 1 freshly added tablet Protease Inhibitor Mix Mini per 10 ml (Complete, Roche) for 90 min at 4 °C. The beads were washed three times with 1 ml of lysis buffer, resuspended in 40 µl of SDS sample buffer, and boiled for 5 min at 95 °C. The GST pull-down reactions were resolved by 12.5% SDS-PAGE and visualized by autoradiography.

Immunprecipitations and analysis of tyrosine phosphorylation
Transfected 293T or COS cells were lysed in 1% Triton X-100, 50 mM Hepes, pH 7.2, 10% glycerol, and 5 mM EDTA containing 1 protease inhibitor tablet (Complete, Roche) plus 1 mM sodium vanadate. Some cells were pretreated with 0.1 mM pervanadate for 4 hours before lysis. Immunoprecipitations were performed in vanadate-containing lysis buffer using 1 µg/ml of appropriate antibodies and prewashed protein G Dynabeads (Dynal) at 4°C overnight. The immune complexes were washed three times with lysis buffer without detergent. The proteins were then subjected to SDS-PAGE and transferred to PVDF-membranes (Millipore). The membranes were blocked in PBS-T (PBS, 0.05% Tween 20) containing 5% nonfat dry milk for 60 min at room temperature. Western Blot analysis was performed with 1 µg/ml affinitypurified anti-mPSTPIP, anti-PTP-PEST, anti-FasL G247.4, anti-V5 or anti-Myc 9B11 or 2 µg/ml of 4G10 anti-phosphotyrosine antibodies, followed by horseradish peroxidasecoupled secondary antibodies (Amersham). The primary antibodies were incubated with the membrane in blocking buffer for 60 min at room temperature or overnight at 4 °C. The polyclonal anti-PSTPIP and anti-PTP-PEST antibodies were used at a 1:1000 dilution in PBS-T with 5% milk. T h e m o n o c l o n a l a n t i b o d y a g a i n s t phosphotyrosine, 4G10-HRP, was used at a 1:1000 dilution in PBS-T with 1% bovine serum albumin (BSA). The commercial antibodies were diluted according to the manufacturersŕ ecommendations. After incubation with the appropriate secondary antibody conjugated to horseradish peroxidase, signals were detected using ECL reagent (Amersham).

Localization studies of FasL and PSTPIP using confocal microscopy
293T cells on coverslips were transfected with the appropriate expression constructs and fixed 48 h later with 3% formaldehyde (Roth) in PBS for 10 min, followed by permeabilisation with 0.1% Triton X-100 in PBS for 5 min. Cells were washed twice in PBS/3% BSA, incubated with PBS/3% BSA for a further 20 min to block nonspecific binding of the antibody, and then stained for 1 h in PBS/3% BSA containing 10 µg/ml anti-FLAG antibody M2, 1 µg/ml anti-FasL antibody Nok-1, or a 1:40 dilution of the anti-mPSTPIP rabbit antiserum. After washing twice with PBS/3% BSA, cells were incubated for 60 min with a 1:1000 dilution of Alexa488 anti-mouse or Alexa546 anti-rabbit secondary antibody (Molecular Probes) in PBS/3% BSA. Cells were again washed in PBS/3% BSA and mounted in Vectashield Mounting Medium with DAPI (Vector laboratories). The samples were analyzed by confocal laser scanning microscopy (Leica TCS SL). HeLa cells were grown on coverslips for 24 hours after transfection. For fixation they were washed once with PBS/pH 8, treated with 3% paraformaldehyde (PFA) in PEM buffer (100 mM PIPES, 1 mM MgCl 2 , 1 mM EGTA, pH 6.5) for 5 minutes and with 3 % PFA in borate buffer (100 mM sodium borate, 1mM MgCl 2 , pH 11) for 10 minutes. Cells were then permeabilized with 1% Triton X-100 in PBS/pH 8 for 15 minutes, washed three times with PBS-Triton (0.1 % Triton X-100 in PBS/pH 8) and treated twice with NaBH 4 in PBS/pH 8 (1 mg/ml) for 10 minutes to reduce unreacted aldehydes. Cells were washed three times again with PBS-Triton, blocked with PBS/0.2% BSA for one hour and then incubated with anti-Myc antibody (Cell Signalling) and appropriate fluorochromeconjugated secondary antibodies.
Co-culture killing assay 0.25x10 6 FasL-transfected 293T effector cells were incubated with 2.5x10 6 Fas-sensitive Jurkat A20 target cells in 2 ml medium for 2 or 6 hours at 37 0 C. After incubation, cells were fixed with 70% ethanol and stored at 4 0 C for at least 12 hours. Jurkat cell death was determined by propidium iodide staining and by quantification of cells with hypoploid DNA content (sub G1) using FACS analysis.

Identification of PSTPIP as a FasL-binding protein in a Yeast-Two-Hybrid screen with the intracellular portion of mFasL as the bait
To identify binding partners which interact with the intracellular FasL portion and which are involved in FasL localization and/or reverse signaling, we screened a mouse thymoma cDNA library in a Yeast-Two-Hybrid screen (17) using the mouse intracellular FasL domain as the bait. One of the several clones that we isolated from the 6.5x10 6 transformants screened was the Cterminal part of the proline, serine, threonine phosphatase-interacting protein (PSTPIP) cDNA. PSTPIP was first described as an interaction partner of the phosphatase PTP-HSCF, and its functional role has been linked to the organization of the cytoskeleton (15,18). The human PSTPIP was identified as a CD2-binding protein, hence its name CD2BP1 (16). While the N-terminus of PSTPIP contains an FCH (Fes/CIP-homology) domain and a coiled-coil region for protein interactions, an SH3 domain (binding to proline-rich sequences as found within the intracellular FasL domain) is localized at the PSTPIP C-terminus (19). The PSTPIP deletion mutant isolated in the Yeast Two Hybrid screen codes for aa 130 to 415 and therefore includes the SH3 domain and parts of the coiledcoil region (aa 123-288 in full length PSTPIP; see Fig. 1c).

PSTPIP interacts in vitro with the proline rich region of the intracellular FasL domain via its SH3 domain
To confirm the binding of PSTPIP to FasL which we had observed in yeast, we performed glutathione S-transferase (GST) in vitro pulldown experiments. Full length hFasL and different FasL deletion mutants were in vitro translated and labeled with 35 S-methionine. mPSTPIP was expressed and purified as a GSTfusion protein, incubated with the different FasL proteins and tested for binding by GST pull down analysis (Fig. 1A). While full length FasL and a Δ1-39 FasL deletion mutant (which lacks the first N-terminal 39 aa) strongly interacted with PSTPIP, the Δ40-80 deletion mutant (lacking the proline rich region within the intracellular FasL domain) showed only weak in vitro interaction. Residual binding of this mutant could be explained by the presence of a further SH3-binding site within the cytoplasmic FasL tail between aa 22 and 26. In the reverse experiment, in vitro translated mPSTPIP bound to the intracellular domain of mFasL fused to GST (Fig. 1B). To demonstrate that PSTPIP binds to FasL with its C-terminal SH3 domain, we used an in vitro-translated hPSTPIP mutant lacking the SH3 domain (hPSTPIP ΔSH3) in a GST pull-down experiment. While full length hPSTPIP interacted with the intracellular domain of mFasL, the ΔSH3 mutant did not (see Fig.  1B), confirming binding of PSTPIP to FasL via a classical SH3 domain interaction (Fig. 1C). The data also demonstrate that both human and mouse FasL and PSTPIP proteins bind to each other independently of species origin, a result which is explained by the high homology between the human and murine PSTPIP SH3 sequences (93% on the protein level) and the intracellular FasL domains (76% homology).

Co-immunoprecipitation experiments confirm binding of PSTPIP to FasL in vivo
We used different strategies to detect in vivo P S T P I P / F a s L i n t e r a c t i o n s b y c oimmunoprecipitation. In the first experiments, 293T human embryonic kidney cells were transiently co-transfected with mouse PSTPIP and either with FLAG-tagged mouse or human FasL. Immunoprecipitation with an anti-FLAG antibody demonstrated co-precipitation and binding of mPSTPIP to both mouse and human FasL ( Fig. 2A). In a complementary experiment, we transiently expressed FLAG-tagged mouse PSTPIP in 293-005 cells which are stably transfected with human FasL. Again, we could show binding of PSTPIP to FasL by coimmunoprecipitation with anti-FLAG antibody (Fig. 2B). When we used untagged expression constructs we were able to co-precipitate mouse and human PSTPIP with hFasL from 293T cell lysates using anti-FasL antibody (Fig. 2C). As a control to exclude unspecific binding of any SH3 domain-containing protein to the proline rich region within the intracellular FasL domain, we performed a co-immunoprecipitation experiment with the p56 Lck kinase. As shown in Suplementary Fig. 1, mouse Lck could not be co-precipitated with FLAG-tagged mouse FasL, while in the same experiment, mPSTPIP was efficiently precipitated together with mFasL. We also tested the same FasL deletion mutants that we had used before in our GST pull-down analysis in co-immunoprecipitation experiments. When we co-overexpressed either full length, Δ 1-39 or Δ 40-80 FLAG-FasL together with mPSTPIP in 293T cells, we were able to coprecipitate PSTPIP with full length and the Δ1-39 FasL mutant, but not with the Δ40-80 FasL version which lacks the proline rich region within its intracellular domain (Fig. 2D). In the control, PSTPIP was not co-immunoprecipitated with FLAG-HMGB1 or FLAG-TNFα. In some experiments, residual PSTPIP binding with the Δ40-80 FasL mutant was again observed. This may be due to the aforementioned PxxPP sequence motive present between aa 22 and 26 of FasL. We then transfected the Fas receptor-deficient mouse thymoma cell line L12.10 with mouse FasL cDNA. After selection, we used these cells to look for binding of endogenous mPSTPIP to FasL. Because FasL expression was not stable in the L12.10 FasL cells and ceased after some time in culture, we regularly confirmed FasL expression by FACS (fluorescence-activated cell sorter) analysis (data not shown). Before precipitating FasL from the transfected L12.10 FasL cells, we stimulated them for different time lengths with the biotinylated anti-FasL antibody Mfl-3, to test whether the extent of PSTPIP recruitment by FasL is dependent upon FasL stimulation. Fig. 3A shows that endogenous m P S T P I P c o u l d a l r e a d y b e c oimmunoprecipitated with FasL when the cells had not been pretreated with agonistic anti-FasL antibody. Antibody stimulation of the ligand resulted in further recruitment and coimmunoprecipitation of PSTPIP with FasL, and the strongest interaction was observed after 8 minutes of FasL stimulation, before recruitment decreased again. This decline could be explained by a general downregulation of the response shortly after FasL activation, but it could also be due to internalization of the FasL/antibody complex or just to antibody degradation in cell culture. Differences in the PSTPIP/FasL interaction resulting from ligand stimulation were not observed with PSTPIP-transfected 293-005 cells which express a large amount of exogenous FasL (data not shown). The interaction of endogenous FasL and PSTPIP was also studied in activated primary mouse Tcells. Freshly isolated mouse splenocytes were stimulated with PMA and ionomycin for 18 hours and were then used for coimmunoprecipitation experiments with different anti-mFasL antibodies. Upregulation of both murine FasL and PSTPIP during T-cell activation has been published before (16,(20)(21)(22)(23), although we could not demonstrate the presence of mFasL protein in the input control of our experiment, due to the lack of a suitable antibody specifically recognizing mouse FasL in Western Blot analysis. However, cell surface FasL expression was confirmed by FACS analysis for the activated mouse T-cells in a parallel experiment (data not shown). The result shown in Fig. 3B reveals that endogenous mPSTPIP could be immunoprecipitated with antibodies against mFasL in activated mouse T-cells, while no PSTPIP/FasL interaction was seen in nonactivated control cells.

Co-localization of FasL and PSTPIP
The interaction of two molecules should be visible by their co-localization in the cell. We therefore overexpressed PSTPIP and FLAG-FasL in in HeLa cells and analyzed localization of both proteins by confocal laser scanning microscopy. Full length human FasL indeed colocalized with hPSTPIP ( Fig. 7 d, e, f). In contrast, the Δ40-80 deletion mutant, lacking the proline rich region within the FasL intracellular domain, did not co-localize with PSTPIP ( Fig. 7  g, h). The same results were obtained in 293T cells, where full length mouse and human FasL co-localized with mPSTPIP, in contrast to the Δ40-80 deletion mutant (Supplementary Fig. 2). These results support the biochemical data suggesting that PSTPIP interacts with the proline rich region of the intracellular FasL domain.

PSTPIP recruits the phosphatase PTP-PEST to FasL
PSTPIP is known to bind to the protein tyrosine phosphatase PTP-PEST. Other PSTPIPinteracting proteins, such as WASP or c-Abl, have been shown to exist in a ternary complex with PSTPIP and PTP-PEST (16,18,(24)(25)(26). In GST pull-down experiments we were able to coprecipitate both in vitro translated PTP-PEST and FasL simultaneously with PSTPIP-GST, while the intracellular FasL domain fused to GST could co-precipitate PSTPIP, but -in the absence of PSTPIP -not PTP-PEST, (data not shown).
We then performed coimmunoprecipitation experiments in COS cells to test whether FasL also binds to PTP-PEST via the adaptor protein PSTPIP in vivo. Fig. 4A demonstrates that both FasL and PTP-PEST bind to the full length PSTPIP protein, while only PTP-PEST interacts with the PSTPIP ΔS H 3 mutant lacking the C-terminal SH3 domain, presumably via the PSTPIP coiled-coil domain, as has been published before (25). When we performed the immunoprecipitation with an anti-PTP-PEST antibody, FasL was co-precipitated only in the presence of PSTPIP (Fig. 4B). These data support a model according to which PSTPIP serves as an adaptor protein, recruiting FasL and PTP-PEST together into one complex (see Fig.  4C). The participation of the phosphatase PTP-PEST in the formation of a ternary FasL/PSTPIP/PTP-PEST complex is likely to influence the phosphorylation status of these proteins. We therefore analyzed whether FasL and PSTPIP are subject to phosphorylation reactions in vivo. For this purpose we treated transfected COS cells with the phosphatase inhibitor pervanadate and used an anti-phospho-tyrosine antibody for Western Blot analysis of anti-PSTPIP immunoprecipitations. As can be seen in Fig.  4 A , blockage of phosphatases unmasked phosphorylation of PSTPIP and PSTPIP ΔSH3. The fact that overexpression of PTP-PEST leads to decreased PSTPIP phosphorylation even in the presence of pervanadate indicates that PTP-PEST indeed binds to and dephosphorylates PSTPIP protein. However, we were not able to demonstrate phosphorylation of FasL in COS or 293T cell overexpression systems. This may be due to the absence of kinases in these cells which specifically phosphorylate FasL, or due to insufficient inhibition of phosphatases (e.g. PTP-PEST).

Co-expression of PSTPIP leads to decreased FasL cell surface localization and reduced FasLmediated target cell killing
In many cell types FasL is constitutively expressed at the cell surface, while in hematopoietic cells the molecule is localized and stored in cytoplasmic secretory lysosomes, from where it is delivered to the cell surface upon activation of the cell (13). The intracellular domain of FasL has been shown to be responsible for this intracellular storage, and we therefore tested whether binding of PSTPIP plays a role in FasL sorting. Epithelial 293T cells do not express detectable levels of PSTPIP ((25) and our own observation), and when we transfected hFasL into these cells we could measure profound FasL expression at the cell surface by FACS analysis. However, cotransfection of human or mouse P S T P I P decreased the amount of surface FasL significantly (Fig. 5A).
To collect further experimental evidence that the interaction of PSTPIP with the cytopolasmic part of FasL regulates cell surface expression of the ligand, we overexpressed a PSTPIP mutant lacking the SH3 domain together with FasL in RBL 2H3 cells which normally retain FasL intracellularly (14). The PSTPIP ΔSH3 mutant is still able to interact with other proteins and structures via its N-terminal protein-protein interaction motives, but it is not capable to bind to FasL. Therefore, PSTPIP ΔSH3 is likely to act as a dominant negative PSTPIP mutant with respect to its FasL-relevant functions(s).  5C). In 14 independent co-culture experiments, we observed a reduction in FasL-mediated target cell killing of 35.2% (+/-17.6%) after PSTPIP co-transfection. This reduction in FasL cell surface expression and FasL-induced target cell death upon PSTPIP co-expression was due to increased intracellular FasL localization and not due to declined total FasL protein expression, as was seen in Western Blot analysis of lysates prepared from the transfected cells (Fig. 5D). To confirm these data and to ensure that cells with equal amounts of total FasL were compared for relative FasL cell surface expression, we used a hFasL construct tagged with G F P at the cytoplasmic N-terminus for transfection experiments. By gating regions of identical GFP fluorescence during FACS analysis, we analyzed cells with similar quantities of total FasL protein (which we again confirmed by Western Blot; data not shown). FACS measurement of cell surface FasL with the anti-FasL antibody Nok-1 once more revealed a decrease in membraneanchored FasL and consequently an increased proportion of intracellular FasL within the PSTPIP-co-transfected cells (Fig. 6A). The reduction in surface FasL upon co-expression of PSTPIP was validated by decreased killing rates in co-culture experiments (Fig. 6B). The specificity of PSTPIP in its ability to relocalize FasL from the cell surface to the cytoplasm was supported by control transfections with XIAP and HMGB1 (High Mobility Group Box 1) cDNAs, which code for cytoplasmic proteins. Their co-expression with FasL did not influence cell surface staining and killing activity of the ligand in co-culture assays (data not shown). Importantly, co-transfection of CD34, another human cell surface molecule, together with h P S T P I P did not change CD34 surface expression when compared to CD34 single transfected cells (data not shown). These findings demonstrate that the FasL-binding protein PSTPIP is involved in the storage of FasL in intracellular compartments, a result which we could support by localization studies with HeLa cells using confocal laser scanning microscopy (Fig. 7). While single transfection of GFP-tagged hFasL led to predominantly cell surface expression of the molecule (Fig. 7a, b), co-transfection of FasL with hPSTPIP resulted in the expected co-localization of both proteins and the retention of FasL in intracellular vesicles (Fig. 7c, d, e, f).

Discussion
The Fas Ligand molecule is well known for its capacity to induce apoptosis by activating the corresponding death receptor Fas (27). However, its highly conserved 80 aa long cytoplasmic domain contains several prominent signaling motives, which argues for the existence of a reverse signaling mechanism via FasL into the ligand-bearing cell (1,2,6,12). In addition, the cytoplasmic domain has been implicated in FasL sorting in hematopoietic cells to specialized intracellular organelles, the secretory lysosomes (13). Cellular proteins interacting with the intracellular FasL domain may also influence the capability of the ligand to induce apoptosis by Fas receptor activation, for example by preventing oligomerization of FasL. So far, the molecules which bind to the intracellular FasL domain and which are responsible for such aspects of FasL biology have not been functionally characterized, although different proteins have been suggested as FasL-interacting proteins (28,29). In this study we show by a number of methods that the adaptor protein PSTPIP binds to the intracellular FasL domain. We confirmed our initial Yeast-Two-Hybrid result by in vitro GST pull down, co-immunoprecipitation of overexpressed and endogenous proteins and by co-localization. PSTPIP is a homolog of S . pombe CDC15p, a protein involved in the assembly of the actin ring in the cytokinetic cleavage furrow (15). Mouse PSTPIP has been identified as an interactor of the PEST-type protein tyrosine phosphatase PTP-HSCF (15), and human PSTPIP has been found to interact with the CD2 cytoplasmic domain and to downregulate CD2-stimulated T-cell adhesion through coupling of the protein tyrosine phosphatase PTP-PEST to CD2 (16). PSTPIP belongs to a family of structurally related proteins characterized by a similar modular composition of an N-terminal FCH (Fes/CIP4 homology) domain, followed by a coiled-coil region and a C-terminal SH3 (Src homology 3) domain (FCH/SH3 protein family). Our experiments show that PSTPIP interacts with the N-terminal proline rich domain within FasL via its SH3 domain. We also demonstrate that cotransfection of PSTPIP with FasL decreases cell surface FasL expression, while the amount of intracellular FasL is enlarged, arguing that the interaction of PSTPIP with FasL is important for intracellular sorting of FasL. PSTPIP is able to bind to the phosphatase PTP-PEST via its coiled-coil domain (16,(24)(25)(26), and we could prove the existence of a trimeric complex consisting of FasL, PSTPIP and PTP-PEST by co-immunoprecipitation experiments. Interaction with the PSTPIP SH3 domain has not only been reported for CD2, but also for the Wiskott-Aldrich syndrome protein WASP and for the c-Abl kinase. PSTPIP serves as a scaffold protein between PTP-PEST and WASP, allowing the phosphatase to dephosphorylate WASP and thereby inhibit WASP-driven actin polymerization and formation of the immunological synapse (25,26). PSTPIP also directs PTP-PEST to the c-Abl kinase to mediate Abl dephosphorylation and inactivation (24). The data on PSTPIP interactions with CD2 and WASP and its homology to yeast CDC15p suggest a role for PSTPIP in the organization of the cytoskeleton. Recently, mutations in human PSTPIP that disrupt binding to PTP-PEST have been linked to PAPA syndrome, an autoinflammatory disorder (30). Further interactions of PSTPIP with CD4, CD8, CD54 and CD62L have been published, the functional roles of which remain to be established (20). Our experiments reveal that binding of PSTPIP to the intracellular domain of FasL influences the localization of the ligand within the cell. In Tand NK-cells, FasL is targeted to specialized secretory lysosomes which fuse with the plasma membrane upon recognition of a target cell (13,14). These Golgi-derived vesicles carry lysosomal markers (e.g. lysosome associated membrane proteins (LAMP)) and, apart from FasL, they contain other cytolytic effector molecules including perforins and granzymes. Storage of newly synthesized FasL ensures that the ligand appears on the cell surface only upon activation and precisely at the contact site between effector and target cell (immunological synapse), thereby preventing non-specific killing by T-and NK-cells. In non-hematopoietic cells with conventional lysosomes, however, FasL appears directly on the plasma membrane. FasL is targeted to secretory lysosomes via the proline rich domain in its cytoplasmic tail, between aa 40 and 70, and it has been speculated that an interaction of this domain with an SH3 domaincontaining protein might facilitate direct sorting of FasL to secretory lysosomes (14). We could show that PSTPIP binds to the proline rich domain of FasL via its SH3 domain and that cotransfection of PSTPIP with FasL into nonhematopoietc cells decreases the amount of cell surface FasL and increases its storage in intracellular granules. PSTPIP is a promising candidate for FasL sorting to secretory lysosomes since its expression is limited to hematopoietic cells, namely T-and NK-cells (15,16) which contain the secretory lysosomes used for intracellular FasL storage (13,14). Several members of the FCH/SH3 protein family have been functionally linked to the cytoskeleton and to endocytosis. In particular, the Forminbinding Protein 17 (FBP17) has been shown to interact with Dynamin and to regulate endocytosis by forming vesicotubular structures (31). PSTPIP on the other hand is involved in actin polymerization (32), a process known to be important for endocytosis (33). One could speculate that co-expression of PSTPIP with FasL leads to decreased FasL cell surface expression because of a PSTPIP-induced increase in endocytosis. However, such an explanation is not plausible in the light of our observation that the level of cell surface expression of another receptor, hCD34, is not influenced by co-expression of PSTPIP. Reverse signaling into the ligand-bearing cell has been described for several members of the TNF family. So far, this bi-directional signaling capacity has been associated with 6 out of 16 TNF family members (CD27L, CD30L, CD40L, CD137L, TNFα and FasL) that contain a putative casein kinase I (CKI) substrate motive (for review see (2)), although CKI-dependent serine phosphorylation upon receptor activation has, up until now, only been shown for TNFα (11). Studies with primary T-cells from wildtype mice and animals defective for FasL (gld) or Fas (lpr) have suggested that FasL influences the proliferative behavior of T-cells ((7-9) and our own unpublished data). Furthermore, an increase in Erk1/2 phosphorylation and activation of the cytosolic phospholipase A2 have been reported for the mouse-derived sertoli cell line TM4 after crosslinking of endogenous FasL with antibodies (10). In our own experiments, we observed an increase in Erk1/2 phosphorylation in FasLtransfected 293T cells upon stimulation with the anti-FasL antibody Nok-1, as well as an increase in p38-MAPK phosphorylation after treatment of the same cells with a soluble Fas receptor molecule (Fas-Fc) which binds to surface FasL (data not shown). Such reverse signaling events may also depend on the proline rich domain within the cytoplasmic part of FasL and on recruitment of a SH3 domain-containing protein like PSTPIP. PSTPIP is able to bind to the protein tyrosine phosphatases PTP-HSCF and PTP-PEST as an adaptor protein and to recruit them to other target proteins (15,16). Trimeric complexes of PSTPIP and PTP-PEST with the kinase c-Abl or the WASP protein have so far been described (24)(25)(26). In such complexes, PTP-PEST dephosphorylates both the adaptor protein PSTPIP and the PSTPIP-binding protein, either c-Abl or WASP. While the phosphorylation status of PSTPIP may influence its capability to interact with its binding partners (18), dephosphorylation of c-Abl and WASP by PTP-PEST deactivates these proteins. In overexpression studies, we could demonstrate the existence of FasL/PSTPIP/PTP-PEST complexes in COS cells. While the proline rich domain of FasL interacts with the PSTPIP SH3 domain, PTP-PEST binds independently to the Nterminal PSTPIP coiled-coil stretch via its Cterminal CTH domain (19,25). To which extent the PSTPIP/PTP-PEST/FasL complex formation is influencing the other aggregates of PSTPIP/PTP-PEST with c-Abl or WASP (or whether it is regulated itself by such complexes) will be a matter of future studies. In accordance with published data, we could demonstrate that t y r o s i n e -p h o s p h o r y l a t e d P S T P I P i s dephosphorylated by the interacting PTP-PEST phosphatase (25,34). Nevertheless, we could not show tyrosine phosphorylation of FasL upon phosphatase inhibition in COS cells which may be due to the absence of FasL kinases or to inefficient PTP-PEST inactivation. The latter speculation is supported by the fact that, in our experiments, overexpression of PTP-PEST led to decreased PSTPIP phosphorylation even in the presence of pervanadate. In essence, our data suggest that binding of PSTPIP and PTP-PEST may influence the phosphorylation status of FasL in relevant cell types, such as T-cells, which express endogenous PSTPIP and PTP-PEST proteins (25) and in which reverse signaling via FasL has been shown to influence activationinduced proliferation (7)(8)(9). It is interesting to note that we observed an increase in endogenous PSTPIP recruitment to overexpressed cell surface FasL in mouse thymoma cells upon stimulation of the ligand with soluble Fas receptor. Such recruitment of PSTPIP to stimulated membrane-bound FasL might indicate a further role for PSTPIP (and PTP-PEST) in the regulation of cell surface FasL and/or FasL reverse signaling in T-cells which express the endogenous PSTPIP protein.
FasL is a very dangerous molecule for a cell to display at the cell surface due to its potent ability to induce cell death. Therefore, the protein is expressed in only a few cells and is regulated at different levels. The FasL gene contains binding sites for several transcription factors and has been reported to be transcriptionally upregulated by e.g. NF-kB, c-Myc, RARg, Nur77, c-Jun/FosB (AP-1), Sp1, Egr1/2/3, NF-AT (35) in activated T-cells. Its localization within intracellular organelles in T-and NK-cells, followed by a controlled delivery to the immunological synapse upon activation, ensures that only the recognized target cell is killed (13,14). FasL surface expression is then downregulated by matrix metalloproteases which cleave the extracellular FasL domain (36)(37)(38)(39). However, some cells release apoptosis-inducing full length FasL in microvesicles into their environment (40)(41)(42). Therefore, proteolysis and exocytosis are also involved in FasL regulation. In addition, we have observed FasL localization of FasL in lipid rafts at the cell surface (manuscript in preparation) which further influences localization, stability and the cell death-inducing capacity of the molecule. The sheer number of regulatory levels indicates a complex interdependency and interplay of these processes, which are likely to involve several different interacting proteins. With PSTPIP we have identified a molecule which binds to the intracellular FasL domain and which is involved in at least one of these regulatory processes, namely localization. In addition, by forming a trimeric complex with FasL and the phosphatase PTP-PEST, PSTPIP is likely to be implicated in the regulation of FasL phosphorylation and possibly in an additional reverse signaling role of FasL. Future studies, including analysis of knockout/knockin mouse models in which the wildtype FasL allele has been replaced by a deletion mutant lacking the intracellular domain, will help to elucidate the consequences of PSTPIP binding to FasL and the physiological role of FasL reverse signaling.  wildtype (wt) cells were added to the Mfl-3-coated wells for the indicated times. For subsequent immunoprecipitation cells were lysed, and further biotinylated Mfl-3 antibody was added (1µg/ml). Streptavidin agarose was employed for precipitation. The figure shows some co-immunoprecipitation of endogenous mPSTPIP with FasL without stimulation of FasL, while maximal binding of PSTPIP protein to FasL occurred after 8 minutes of anti-FasL stimulation. The IP lanes display the IgG heavy chain of the Mfl-3 antibody used for precipitation. B: 1.5x10 7 mouse splenocytes were either stimulated with 10ng/ml PMA and 0.2µg/ml ionomycin (+) or left untreated (-). After 18 hours the cells were treated for 30 minutes with the metalloprotease inhibitor phenanthroline (1.5 mM) and then lysed for co-immunprecipitation with different anti-FasL antibodies (H11 and 3C82 (Alexis), Alf2 (Ancell)). All three antibodies were able to co-immunoprecipitate endogenous mPSTPIP with mFasL in stimulated, but not in unstimulated, T-cells, while the M2 anti-FLAG antibody did not precipitate detectable amounts of mPSTPIP. Efficiency of FasL precipitation could not be controlled due to the lack of anti-mouse FasL antibodies suitable for Western Blot analysis. Moreover, this figure illustrates that PSTPIP is tyrosine-phosphorylated, since the antiphosphotyrosine antibody 4G10 only detects a specific signal after tyrosine phosphatase-inhibition with pervanadate (anti-P-Tyr 4G10 Western Blot; the band marked with an asterisk is non-specific). Pervanadate treatment also resulted in a small shift of the PTP-PEST signal towards higher molecular weight (compare for example lane 1 with lane 2, 3 with 4 and so on in the anti-PTP-PEST Western Blot), indicating that PTP-PEST itself is regulated by phosphorylation, as has been published before (15,24). B: When immunoprecipitation was performed with anti-PTP-PEST in transfected COS cells, co-precipitation of FasL was only observed in the presence of PSTPIP. No differences were observed between pervanadate-treated and untreated cells. These data confirm that FasL and PTP-PEST form a ternary complex in vivo by binding to different regions of the adaptor protein PSTPIP, as shown schematically in C. In this complex PSTPIP is dephosphorylated, most likely by PTP-PEST which may also influence FasL phosphorylation. A : Human FasL (pCR33-FLAG-hFasL) was transfected either alone or together with hPSTPIP (pcDNA3.1-hPSTPIP) into 1x10 6 293T cells. 48 hours later cells were stained with anti-FasL antibody (Nok-1) plus anti-mouse secondary antibody coupled to FITC for subsequent FACS analysis. Average FasL cell surface expression was measured as an arbitrary geometrical mean FITC fluorescence. Empty vector (pcDNA3.1) transfection and staining only with secondary anti-mouse antibody (2 nd Ab) served as negative controls. B: RBL 2H3 cells were electroporated with full length hFasL together with either empty vector (thick line) or with PSTPIPΔSH3 (lacking the PSTPIP SH3 domain; thin line with grey filling). FasL cell surface expression was quantified 24 hours later with Nok-1 anti-FasL antibody and APC-coupled anti-mouse secondary antibody by FACS analysis. The increase in FasL cell surface expression upon co-transfection of PSTPIPΔSH3 was measured as an arbitrary geometrical mean fluorescence shift (ΔGeo Mean). C: Single and double transfected 293T cells from A were used 48 hours after transfection for a co-culture killing assay with Fas-sensitive A20 Jurkat target cells. D: Transfected 293T cells used in A and C were analyzed by Western Blot for total FasL protein expression. G247 antibody was used to detect FasL, and equal loading of lysates was confirmed by coomassie staining of the gel. Several (glycosylated) FasL bands were detected between 28 and 37 kD, and co-expression of hPSTPIP did not decrease the overall amount of FasL in the cells. The anti-hPSTPIP serum detected the target protein above an unspecific 50 kD signal. Fig. 6: GFP-FasL cell surface expression and its killing capacity are decreased upon PSTPIP cotransfection. A: hFasL with an N-terminal GFP-tag was transfected alone or with hPSTPIP. 48 hours later, cells with similar GFP fluorescence, and therefore equal amounts of GFP-FasL, were gated during FACS analysis in the FL1 channel and analyzed for FasL surface expression by staining with the anti-FasL antibody Nok-1 and anti-mouse secondary antibody coupled to APC (2 nd Ab). Staining with antimouse antibody alone served as a negative control. The top histogram shows an overlay of cells transfected with GFP vector control (GFP) and cells transfected with GFP-FasL alone (GFP-FasL) or in combination with PSTPIP (GFP-FasL+PSTPIP). The histograms below were used to quantify the average FasL surface expression as the arbitrary geometrical fluorescence of GFP-FasL-transfected cells (left) and of cells transfected with GFP-FasL plus PSTPIP (right). B: A co-culture assay was performed using the transfected target cells from A with Fas sensitive A20 Jurkat target cells.