PSTPIP2, a Protein Associated with Autoinflammatory Disease, Interacts with Inhibitory Enzymes SHIP1 and Csk

The activity of leukocyte signaling pathways must be properly controlled to prevent excessive responses with harmful consequences. This is ensured by a number of negative regulators that limit the duration and magnitude of the signaling. One such regulatory protein is the lipid phosphatase SHIP1, which controls the activity of the PI3K pathway by dephosphorylating its key mediator, phosphatidylinositol-3,4,5-trisphosphate [PI(3,4,5)P₃]. This results in the reduced activity of some of its downstream effectors (1). A prototypical example of such an effector is Akt (also known as PKB), a serine/threonine kinase, which is involved in the regulation of cell activation, proliferation, metabolism, and survival, and which is recruited to the plasma membrane by PI(3,4,5)P₃ for further activation (2). The deletion of SHIP1 results in the increased activity of Akt in the mast cells (3) and in the increased translocation of the Akt PH domain to the plasma membrane in neutrophils, suggesting that in myeloid cells SHIP1 regulates Akt localization and activity in vivo (4). There is also ample evidence that SHIP1 negatively regulates MAPK pathways by various mechanisms, in some cases independently of SHIP1 enzymatic activity (5).

Another critical negative regulator of leukocyte signaling is the protein tyrosine kinase Csk. It is involved in the regulation of Src-family kinases (SFKs), a family of protein tyrosine kinases indispensable to the initiation of signal transduction via ITAM-bearing immunoreceptors, and with additional and important roles in signaling by cytokine, growth factor, and pattern recognition receptors, and many others (6). Csk phosphorylates an inhibitory tyrosine residue at the C terminus of SFKs. This phosphorylated tyrosine then interacts with an SH2 domain in the same molecule, resulting in an autoinhibited conformation (7–9). Csk itself is recruited to the plasma membrane and is activated by binding to the phosphorylated tyrosine motifs of transmembrane proteins such as PAG (10, 11), LIME (12), and SCIMP (13). An additional mechanism of SFK inhibition is dependent upon protein tyrosine phosphatases (PTPs) that dephosphorylate the activation loop tyrosine necessary for SFK catalytic activity. This residue can be dephosphorylated by multiple phosphatases, including the receptor tyrosine phosphatases CD45 and CD148 (14–16), PTPs of the Shp family (17), as well as PTP LYP (PTPN22, also known as PEP in mice). In T cells, LYP/PEP forms a complex with Csk. Together, they simultaneously dephosphorylate the activation loop tyrosine and phosphorylate the inhibitory C-terminal tyrosine of SFKs, exerting a combined inhibitory effect (18, 19). LYP/PEP is a member of a small family of PTPs known as proline-, glutamic acid-, serine- and threonine-rich–family PTPs (PEST-PTPs), consisting of only three members: PEP/LYP, PTP-HSCF (PTPN18), and PTP-PEST (PTPN12) (reviewed in Ref. 20). All three family members bind Csk via proline-rich or tyrosine-containing motifs in their central region (18, 21, 22). In addition, a conserved C-terminal homology domain of PEST-PTPs binds two related proteins, PSTPIP1 and PSTPIP2 (23, 24). These are adaptor proteins involved in the control of inflammation, and their altered function results in the development of autoinflammatory diseases (25–27). In general, these disorders are characterized by sterile inflammation and consequential tissue damage. They are mainly the result of dysregulated activity of the innate immune system with no or limited involvement of adaptive immunity (28). The autoinflammatory disease caused by PSTPIP2 mutations is, in mice, characterized by sterile inflammatory lesions in the bones and various degrees of skin and paw inflammation (26, 27, 29). The disease closely resembles the human disorder known as chronic recurrent multifocal osteomyelitis. However, human patients with genetic alterations in the PSTPIP2 gene have not been identified yet. PSTPIP2 is a member of the F-BAR family of proteins (also known as the pombe cdc15 homology family), characterized by the presence of N-terminal F-bar domain mediating interactions with membrane phospholipids and a C-terminal tail containing various interaction motifs (30). The F-bar domain of PSTPIP2 interacts with PI(4,5)P₂ (31), whereas its C-terminal tail binds PEST-PTPs via the interaction motif, which includes tryptophan 232 (24, 32). The C-terminal tail of PSTPIP2 contains several tyrosines of unknown function that are phosphorylated in macrophages after exposure to M-CSF (33). Two different mouse strains where point mutations in PSTPIP2 result in the autoinflammatory disorder have been established. As a result of these mutations, one of these strains, LUPO, displays approximately a 70% reduction in PSTPIP2 protein expression (27), whereas the other strain, CMO, shows a complete absence of PSTPIP2 (34). The disease is independent of T or B cells, and it was originally attributed to the enhanced activity of macrophages and osteoclasts (27, 32, 34). In addition, more recent work has suggested that neutrophils may also be a critical cell type in disease initiation (35, 36). Similar to several other autoinflammatory syndromes, the disease appears to be, at least in part, caused by the enhanced production of IL-1β (35, 37). The molecular mechanism of how PSTPIP2 prevents autoinflammatory disease development remains largely unknown. The effects of PSTPIP2 binding to PEST-family phosphatases have been tested recently in osteoclasts and megakaryocytes, where it had an effect on their differentiation. However, in these experiments, a profound effect was also seen after the mutation of C-terminal tyrosines, which rendered PSTPIP2 especially in the case of osteoclasts essentially nonfunctional without affecting the interaction with PEST-family phosphatases (32, 38). This suggested that the binding partners of these tyrosines are critical for PSTPIP2 function. However, their identity remained unknown.

In this article, we show that these phosphorylated tyrosine residues bind the lipid phosphatase SHIP1. In addition, we also show that PSTPIP2 interacts with Csk via a mechanism that is, at least in part, independent of PEST-PTPs. Our data also bring evidence for the involvement of these inhibitory enzymes in the PSTPIP2-mediated suppression of the inflammatory response.

Abs to the following Ags were used in this study: FLAG (M2) and GAPDH (from Sigma-Aldrich, St. Louis, MO); phospho-Erk (T202/Y204), phospho-Akt (S473, T308), and Myc (9B11) (Cell Signaling Technology, Danvers, MA); Csk (C-20) (Santa Cruz Biotechnology, Santa Cruz, CA); CD18 (YTS 213.1) (Serotec, Kidlington, U.K.); B220-biotin (RA3-6B2), TER119-biotin, Gr-1–PE (RB6-8C5), c-Kit–biotin, CD3ε-biotin, CD11b-FITC, Ly6C-PE-Cy7 (Biolegend, San Diego, CA); B220-FITC, DX5-biotin, F4/80-biotin, F4/80-PE, and F4/80-Alexa700 (eBioscience, San Diego, CA); Thy1.2-FITC, CD271-FITC, anti-FITC, and anti-biotin MicroBeads (Miltenyi Biotec, Bergisch Gladbach, Germany); HRP-conjugated goat anti-mouse L chain–specific Ab, HRP-mouse anti-rabbit L chain–specific, purified mouse IgG-whole molecule, goat anti-mouse F(ab)₂, goat anti-rat F(ab)₂ (Jackson Immunoresearch, West Grove, PA), Fc Bloc (2.4G2) (BD Biosciences, San Jose, CA). The rabbit antisera against SHIP1, PEP, PTP-PEST, and PTP-HSCF were a gift from Dr. A. Veillete (Institut de Recherches Cliniques de Montreal, University of Montreal). The Ab against human SHIP1 was from Exbio (Vestec, Czech Republic).

The mouse mAb that recognizes murine and human PSTPIP2 was generated by the immunization of mice (F1 hybrids of BALB/c × B10A) with a full-length recombinant murine PSTPIP2 produced in Escherichia coli. Splenocytes from immunized mice were fused with Sp2/0 myeloma cells and cloned by limiting dilution. Ab production was tested by ELISA and Western blotting.

To prepare mouse or human aggregated Ig, we first purified IgG from mouse serum (Sigma-Aldrich) or human AB serum (Invitrogen, Carlsbad, CA) on protein A-Sepharose (GE Healthcare, Uppsala, Sweden), transferred them to PBS, and concentrated them to 30 mg/ml on an Amicon Ultracel–30K unit (Millipore, Merck, Darmstadt, Germany). The aggregation was induced by heating to 63°C for 30 min.

CMO mouse strain (Cg-Pstpip2cmo/J) carrying the c.293T→C mutation in the Pstpip2 gene (26, 29) on the BALB/c genetic background, resulting in an L98P change in the PSTPIP2 protein, and BALB/cByJ, were obtained from The Jackson Laboratory (Bar Harbor, ME). The BALB/c and F1 hybrids of BALB/c × B10A were from the animal facility of Institute of Molecular Genetics, Academy of Sciences of the Czech Republic (Prague, Czech Republic). All the experiments in this work that were conducted on animals were approved by the Animal Care and Use Committee of the Institute of Molecular Genetics and were in agreement with local legal requirements and ethical guidelines.

All the primary cells and cell lines were cultured at 37°C with 5% CO₂ in the following media supplemented with 10% FCS and antibiotics: WEHI-231 cell line (ATCC) in RPMI 1640, HEK293FT cells (Invitrogen), Phoenix Eco cells (Origene, Rockville, MD), J774.2 (ATCC) cells, and immortalized macrophage progenitors in DMEM. IMDM was used for immortalized granulocyte progenitors and primary granulocytes.

Bone marrow (BM) was isolated from CMO and BALB/cByJ or BALB/c mice sacrificed by cervical dislocation. Full BM was cultured in DMEM conditioned with 10% L929 culture supernatant containing M-CSF to differentiate BM-derived macrophages (BMDMs). Osteoclasts were differentiated in DMEM supplemented with 20 ng/ml M-CSF and 100 ng/ml RANKL (Peprotech, Rocky Hill, NJ). Murine granulocytes were isolated from BM by negative selection using B220, F4/80, DX5, c-Kit, CD3ε, and Ter119 biotinylated Abs, and anti-biotin MicroBeads, on an AutoMACS magnetic cell sorter (Miltenyi Biotec), and the purity (>90%) was determined by flow cytometry. Human granulocytes were purified from buffy coats (purchased on a commercial basis from the Blood Bank of Thomayer Hospital, Prague, Czech Republic) by sedimentation in 2% Dextran T500 (Pharmacosmos, Holbaek, Denmark) followed by Ficoll centrifugation (Ficoll-Paque PLUS, GE Health Care) and erythrocyte lysis in an ACK buffer [150 mM NH₄Cl, 0.1 mM EDTA (disodium salt), 1 mM KHCO₃]. The generation of immortalized macrophage progenitors has been described previously (39); in brief, BM was centrifuged over Ficoll-Paque PLUS gradient, and mononuclear cells were directly infected with ER-Hoxb8 retrovirus by spinoculation. The cells were cultivated in a medium conditioned with 1% LUTZ culture supernatant as a source of GM-CSF and 1 μM β-estradiol (estrogen; Sigma-Aldrich). The cells could be differentiated into macrophages by the withdrawal of estrogen from the media within a week. To prepare immortalized granulocyte progenitors, we used a modified version of this protocol (39). The progenitors were first enriched by the depletion of Mac-1⁺, B220⁺, and Thy1.2⁺ from mouse BM cells and cultured in the presence of IL-3, IL-6, and SCF (supplied as culture supernatants from HEK293 cells transduced with constructs coding for respective cytokines) for 2 d. Next, they were transduced with the same ER-HoxB8 construct described earlier. The transduced cells were enriched for the GMP progenitor population by FACS (Lin⁻, Sca-1⁻, c-Kit⁺, FcγR⁺, CD34⁺) and propagated in a media containing 1 μM β-estradiol and 1% SCF-containing supernatant. Granulocyte differentiation was induced by β-estradiol withdrawal and the addition of 50 ng/ml G-CSF.

The CLG construct (generated by the PCR-mediated joining of individual overlapping fragments and subcloned to MSCV-IRES-Thy1.1; Clontech, Mountain View, CA) was composed of the SH2 and SH3 domains of human Csk (aa 1–171) followed by the kinase domain from human Lck (aa 225–509) with Y505 mutated to phenylalanine and then by a Myc tag sequence and GST from Schistosoma japonicum lacking methionine 1 (Fig. 1A). Murine PSTPIP2 was cloned from cDNA prepared by the reverse transcription (RevertAid, Fermentas, Thermo Fisher Scientific, Waltham, MA) of RNA from WEHI-231 cells. Wild type (WT) PSTPIP2 or its various mutants generated by PCR mutagenesis were cloned into the MSCV-IRES-LNGFR vector (Clontech) with an N-terminal FLAG tag. Human PSTPIP2 was cloned from cDNA prepared from peripheral blood leukocytes. A Csk-Myc construct in a pMSCV retroviral vector and its variants with inactivating point mutations in the SH2 and SH3 domains (R107K and W47A, respectively) were prepared by PCR from the analogous Csk-Myc-OFP constructs described earlier (40). Lipofectamine 2000 (Invitrogen) was used according to the manufacturer’s protocol for the transfection of Phoenix Eco cells to produce viral particles. Retrovirus-containing supernatants were then harvested, supplemented with polybrene (10 μg/ml; Sigma-Aldrich), and added to the cells. The cells were then centrifuged at 1250 × g for 90 min at 30°C. Infected cells were sorted based on the expression of reporter markers (Thy1.1 or LNGFR) by a BD Influx FACS (BD Biosciences).

For Western blotting, the cells were washed and resuspended in serum-free conditioned IMDM or DMEM at a concentration of 1–4 × 10⁷ cells/ml. Subsequently, the cells were stimulated as indicated at 37°C. The activation of cells was stopped by the addition of an equal volume of a 2× concentrated SDS-PAGE sample buffer containing 0.1 mM diisopropyl fluorophosphate, followed by the sonication and heating of the samples (96°C for 2 min). The samples were analyzed by SDS-PAGE followed by Western blotting. For pervanadate-induced activation, the cells were incubated with 0.1 mM pervanadate for 10 min. To inhibit SHIP1 during the activation, we treated the cells with 10 μM 3-α-aminocholestane (3AC; Echelon Biosciences, Salt Lake City, UT).

For immunoprecipitation, the cells were lysed in the lysis buffer (1% lauryl maltoside [Calbiochem, Merck], 20 mM Tris [pH 7.5], 100 mM NaCl, 5 mM iodoacetamide, 50 mM NaF, 1 mM Na₃VO₄, 2 mM EDTA, 1 mM Pefabloc [Sigma-Aldrich] or 100× diluted Protease Inhibitor Cocktail set III [Calbiochem]; for neutrophil granulocyte lysis, 1 mM PMSF [Sigma-Aldrich] and 0.1 mM diisopropyl fluorophosphate [Fluka; Sigma-Aldrich] were also added) at 5 × 10⁶/ml for 30 min on ice. Postnuclear supernatants were then incubated for 1 h with a primary Ab (2 μg/ml), followed by 1–2 h of incubation with Protein A/G Plus agarose beads (Santa Cruz Biotechnology) or Protein G magnetic beads (Millipore, Billerica, MA) at 4°C. After washing on chromatography columns or a magnetic stand, immunoprecipitates were eluted with a 70 μl SDS-PAGE sample buffer.

For the in vitro kinase assay, cells were lysed in the lysis buffer where NaF, Na₃VO₄, and EDTA were omitted. The immunoprecipitation was carried out in 96-well plates precoated with 0.1 mg/ml goat anti-mouse IgG (Sigma-Aldrich) and Myc Ab (see earlier). The immunoprecipitates were incubated with a kinase buffer (25 mM HEPES pH 7.4, 5 mM MgCl₂, 5 mM MnCl₂, 0.1% lauryl maltoside, and 0.1 μCi γ[³²P]-ATP per well) for 20 min at room temperature. After washing, the immunoprecipitates were eluted with a 50 μl SDS-PAGE sample buffer and subjected to SDS-PAGE and autoradiography.

Protein bands were cut from the gel, then further cut into small pieces and decolorized in a sonic bath at 60°C several times with 0.1 M 4-ethylmorpholine acetate (pH 8.1) in 50% acetonitrile (ACN). After complete destaining, proteins were reduced by 50 mM TCEP in 0.1 M 4-ethylmorpholine acetate (pH 8.1) for 5 min at 80°C and alkylated using 50 mM iodoacetamide in 0.1M 4-ethylmorpholine acetate (pH 8.1) in the dark at room temperature. Next, the gel was washed with water and ACN and was partly dried using a SpeedVac concentrator (Savant, Holbrook, NY). Finally, the gel was reconstituted with a cleavage buffer containing 0.01% 2-ME, 0.05M 4-ethylmorpholine acetate (pH 8.1), 10% ACN, and sequencing-grade trypsin (10 ng/μl; Promega). Digestion was carried out overnight at 37°C, and the resulting peptides were extracted with 30% ACN/0.1% TFA and subjected to mass spectrometric analysis. Mass spectra were acquired using the positive ion mode on a MALDI-FTMS APEX-Ultra (Bruker Daltonics, Bremen, Germany) equipped with a 9.4 T superconducting magnet and a SmartBeam laser. The acquisition mass range was 700–3500 m/z, and 512k data points were collected. The instrument was externally calibrated using the PepMix II peptide standard (Bruker Daltonics, Bremen, Germany). This results in a typical mass accuracy <2 ppm. A saturated solution of α-cyano-4-hydroxy-cinnamic acid in 50% ACN/0.2% TFA was used as a MALDI matrix. One microliter of the matrix solution was mixed with 1 μl of the sample on the target and the droplet was allowed to dry at ambient temperature. After the analysis, the spectra were apodized using square sine apodization with one zero fill. The interpretation of the mass spectra was done using the DataAnalysis version 3.4 and BioTools 3.2 software packages (Bruker Daltonics, Billerica, MA). The proteins were identified by peptide mass fingerprinting using the search algorithm MASCOT (Matrix Science).

The initial goal of this study was the identification of novel Csk-interacting proteins. To achieve this, we designed a construct (hereafter termed CLG) consisting of the SH2 and SH3 domains of Csk and the kinase domain from the SFK Lck, followed by the Myc tag and GST (Fig. 1A). The construct was expressed in the murine B cell line WEHI-231 via retroviral infection. Because many Csk-binding proteins are substrates of SFKs, we expected that the Lck kinase domain would phosphorylate proteins bound to the Csk SH2/SH3 module of the construct, and we thus label them for relatively easy detection in the in vitro kinase assay or else after phosphorylation in vivo. This would allow for the subsequent optimization and scaling up of the procedure to identify these proteins by mass spectrometry. The Myc-GST module was intended for the tandem purification of CLG using glutathione Sepharose followed by the second round of affinity purification with the anti-Myc Ab. However, this approach was eventually abandoned because of the major technical difficulties caused, presumably, by the competition between the construct and the endogenous murine GST present in the cell lysates. Hence we eventually used this construct in only single-step Myc immunoprecipitations, without engaging its GST moiety. In this setup, we immunoprecipitated the CLG fusion protein from WEHI-231 lysates and incubated the immunoprecipitate with γ[³²P]-ATP. The proteins radioactively labeled by the CLG Lck kinase domain were then subjected to SDS-PAGE and autoradiography. This procedure revealed several proteins with molecular masses of 75, 51–55, 37, and 16 kDa in the immunoprecipitates (Fig. 1B). Whereas the 75-kDa band most likely represented the autophosphorylated CLG construct, the identity of the other proteins remained unknown. To identify these proteins, we performed a large-scale immunoprecipitation from WEHI-231 cell lysates (from 4 × 10⁸ cells) infected with the CLG construct followed by SDS-PAGE and Coomassie blue detection. In this way, we were able to detect a protein of 37 kDa, which likely corresponded to a 37-kDa band from the in vitro kinase assay (Fig. 1C). Mass spectrometry identification revealed that this protein was PSTPIP2 (Fig. 1D, Supplemental Fig. 1). Although it was known to interact with PEST-PTPs (24, 32), interaction of PSTPIP2 with Csk has not yet been described.

Because of the unusual composition of the CLG construct, we wanted to confirm that nonaltered full-length endogenous Csk interacts with PSTPIP2. To achieve this, we infected a murine WEHI-231 cell line with a FLAG-tagged PSTPIP2 construct and then immunoprecipitated this protein from the lysates of infected cells. We could readily detect endogenous Csk in FLAG immunoprecipitates from PSTPIP2-FLAG–infected cells, but not from noninfected cells, demonstrating that full-length endogenous Csk interacts with PSTPIP2 (Fig. 2A). To find out which of the binding domains present in Csk is responsible for PSTPIP2 binding, we infected WEHI-231 cells with Csk constructs containing inactivating mutations in their SH2 and SH3 domains. Intriguingly, the immunoprecipitations of these mutated Csk proteins from the cell lysates revealed that the mutation of any of these domains substantially reduced or abolished the binding of PSTPIP2 to Csk, suggesting that the binding is likely cooperative, involving both the SH2 and the SH3 domains of Csk (Fig. 2B).

Because Csk is known to interact with PEST-PTPs (18, 21, 22), we wanted to find out whether the Csk–PSTPIP2 interaction is dependent on these enzymes. For this purpose, we again used WEHI-231 cells transfected with WT PSTPIP2 and its W232A mutant, which is unable to bind PEST-family phosphatases. Coprecipitation experiments revealed that the W232A mutation diminished Csk binding. However, a substantial fraction of Csk still interacted with PSTPIP2 W232A, whereas at the same time the binding of all the PEST-family phosphatases was completely abolished (Fig. 2C). These data suggested that a fraction of Csk binds to PSTPIP2 independently of PEST-PTPs, most likely via the direct interaction with PSTPIP2. In this experiment, we also demonstrated that PSTPIP2 interacts with PEP in a similar manner (requiring the presence of W232) to other PEST-family phosphatases (Fig. 2C). PEP is the last PEST-family member for which this interaction has not yet been demonstrated.

Although we identified the interaction between Csk and PSTPIP2 in the B cell line WEHI-231, macrophages were initially thought to be the critical cell type driving inflammatory disease in CMO mice (27, 34). Moreover, PSTPIP2 phosphorylation can be observed in macrophage cell line after stimulation by M-CSF (33), and the absence of PSTPIP2 resulted in an increased macrophage proliferation in response to the same stimulation (34). Thus, to study PSTPIP2 tyrosine phosphorylation and its effects on Csk binding in a more relevant cell type, we used the murine macrophage-like cell line J774. We generated constructs with phenylalanine mutations in Y323, Y329, and Y333, located in the PSTPIP2 C-terminal tail (Y3F). Y323 and Y333 have previously been reported to be responsible for the majority of PSTPIP2 tyrosine phosphorylation (24, 32), and Y329 also seemed a good candidate phosphorylation site (see 11Discussion for more details on the selection of these tyrosines). Next, we expressed these constructs, as well as the W232A construct, in J774 cells stimulated (or not) with M-CSF to induce the tyrosine phosphorylation of PSTPIP2. We observed an increase in the tyrosine phosphorylation of PSTPIP2 after M-CSF treatment, which was not affected by W232A mutation (Fig. 2D). In contrast, the mutation of all three C-terminal tyrosines (Y3F) almost completely abolished PSTPIP2 tyrosine phosphorylation (Fig. 2D), which is in agreement with previously published data (24, 32). In the next experiment, we analyzed the Csk binding to these mutants. For this purpose, we also included a mutation of Y249, which, of all PSTPIP2 tyrosines, most closely resembled the Csk-binding site (12). Similar to the previous experiment, in resting cells we observed Csk binding to WT PSTPIP2 (Fig. 2E). Moreover, upon M-CSF stimulation, Csk binding to PSTPIP2 substantially increased. However, none of the tyrosine mutations were able to reduce basal or M-CSF–induced Csk binding to PSTPIP2. In addition, when we analyzed the W232A mutant of PSTPIP2, which was unable to interact with PEST-family phosphatases, we found that in nonstimulated cells Csk binding was not altered by the W232A mutation. However, the increase in Csk binding after M-CSF stimulation was lost (Fig. 2E). We also tested a number of other mutations (Supplemental Table I), as well as their combinations with W232A, but we did not observe any additional effect on Csk binding (data not shown). Unfortunately, some of the candidate binding sites could not be analyzed because their mutations resulted in instability and degradation of the mutant protein that, as a result, could not be detected in the lysates of the transfected cells. Thus, we can conclude that, in J774 cells, PSTPIP2 constitutively binds Csk and this binding is independent of PEST-PTPs. Moreover, M-CSF stimulation results in enhanced binding, which is likely mediated by PEST-PTPs.

To better understand the function of PSTPIP2 in macrophages and to find relevant situations where the signaling pathways could be regulated by PSTPIP2, we analyzed its expression and phosphorylation in this cell type. To be able to follow the endogenous protein expression, we developed a mouse mAb specific for both human and mouse PSTPIP2. The Ab recognizes transfected (Fig. 3A) and endogenous PSTPIP2 (Fig. 3B), and works well in both Western blot analysis and immunoprecipitation experiments.

It has been previously published that PSTPIP2 expression in macrophages is increased after LPS treatment, suggesting that PSTPIP2 expression may be enhanced during inflammation. Thus, we treated BMDMs with LPS or various mostly proinflammatory cytokines and followed the PSTPIP2 expression. Interestingly, PSTPIP2 was strongly upregulated, not only by LPS but also by treatment with several proinflammatory cytokines, including IFN-γ, TNF-α, and GM-CSF, whereas treatment with IL-4 (which promotes their alternative M2 differentiation) did not change the PSTPIP2 expression (Fig. 3B). These data suggest that an increase in PSTPIP2 expression is a part of a proinflammatory differentiation program in macrophages.

To obtain the more complete information about the involvement of PSTPIP2 in the regulation of inflammatory processes in macrophages, we next tested several additional receptors involved in the macrophage inflammatory response for their ability to regulate the tyrosine phosphorylation of PSTPIP2. In this experiment, we stimulated J774 cells expressing FLAG-PSTPIP2 with M-CSF, an anti-Fc receptor (anti-FcR) Ab or an Ab against an integrin β2 subunit, as well as LPS, GM-CSF, and IFN-γ. In resting cells, we observed a low level of constitutive PSTPIP2 phosphorylation (Fig. 4A). Similar to the previously published data, M-CSF treatment resulted in increased phosphorylation. Importantly, PSTPIP2 tyrosine phosphorylation was also substantially increased after FcR cross-linking with more rapid kinetics (Fig. 4A). This phosphorylation peaked at 1 min and then gradually declined (Fig. 4B). It was accompanied by increased Csk binding, which was also more sustained. The binding of the PEST phosphatases PTP-PEST and PTP-HSCF did not seem to be influenced by FcR stimulation. However, the remaining family member PTP-PEP displayed a comparatively low level of binding in the resting cells, which increased substantially after FcR stimulation (Fig. 4B). Similar phosphorylation was also observed after integrin stimulation, although in this case the involvement of the FcRs could not be excluded because full-length Abs were used to cross-link the integrins in these experiments (Fig. 4A). We did not see any changes in PSTPIP2 phosphorylation after stimulation with LPS, GM-CSF, or INF-γ (data not shown).

The C-terminal tyrosines of PSTPIP2 were identified as major phosphorylation sites in PSTPIP2, and in some experiments they also appeared to be among the functionally most important residues in the PSTPIP2 protein (32, 38). In our experiments, Csk binding did not depend on the presence of these tyrosines. Therefore, we assumed that there must be an additional, unknown binding partner that interacts with these residues, and we attempted its identification. For this purpose, we used immortalized macrophage progenitors (see 2Materials and Methods). These cells are much more closely related to primary cells than tumor cell lines, and thus they are better suited for the detection of the most relevant interactions. Moreover, increased number of macrophage progenitors has been observed in CMO mice, suggesting PSTPIP2 involvement in the regulation of this population (34). Finally, these cells can be propagated in a cell culture and obtained in sufficient numbers for large-scale experiments. We infected these cells with WT PSTPIP2 and PSTPIP2 with mutated C-terminal tyrosines (Y3F) and then performed large-scale PSTPIP2 immunoprecipitations. Mass-spectrometry analysis of the proteins specifically present in the immunoprecipitates of WT, but not mutant, PSTPIP2 resulted in the identification of the lipid phosphatase SHIP1 as a novel PSTPIP2 binding partner (Fig. 5A, Supplemental Fig. 2). To confirm these data under more physiological conditions, we immunoprecipitated PSTPIP2 from the lysates of WT and CMO BMDMs stimulated with M-CSF or via FcR, and we observed a low level of constitutive SHIP1 binding that was substantially increased after both M-CSF and FcR stimulation (Fig. 5B, 5C). To verify that SHIP1 binds to the C-terminal tyrosines of PSTPIP2, we expressed WT and Y3F mutants of PSTPIP2 in CMO BMDMs and stimulated these cells with pervanadate. Indeed, we could readily observe SHIP1 binding to WT, but not to the mutant protein, confirming that functionally important C-terminal tyrosines are required for SHIP1 binding (Fig. 5D).

SHIP1 dephosphorylates the PI3K product PI(3,4,5)P₃, thus regulating a number of important pathways, whereas Csk inhibits SFKs, which are also expected to have a strong impact on macrophage signaling. Unexpectedly, we could not find any differences in the activity of M-CSF, LPS, or FcR-driven signaling pathways when comparing the BMDMs from WT and CMO mice (data not shown). This included the analysis of Akt and Erk phosphorylation, which are known to be regulated by SHIP as well as SFK phosphorylation at the inhibitory tyrosine (or in the activation loop) regulated by Csk. We also analyzed a number of other pathways in the macrophages (Jak-Stat, NF-κB, JNK, p38) (data not shown), but we did not see any defects. However, recently published work has suggested neutrophil granulocytes and IL-1β as probable driving forces behind inflammatory disease in CMO mice (35, 36). Thus, as a next step, we analyzed PSTPIP2 expression and function in granulocytes. Surprisingly, we found that in these cells the expression of PSTPIP2 is higher than in macrophages or osteoclasts (Fig. 6A). Moreover, in contrast with our observations in the macrophages, CMO neutrophils exhibited increased FcR-triggered phosphorylation of the Erk and p38 MAPKs, as well as higher phosphorylation of Akt (Fig. 6B) and more efficient processing of IL-1β (Fig. 6C), compared with WT neutrophils. The additional analysis of signaling pathways in granulocytes isolated from CMO mice revealed that Akt phosphorylation was also increased after the relatively long (90-min) treatment of neutrophils with LPS (Fig. 6D). However, pro–IL-1β production after LPS priming did not appear to be altered in CMO neutrophils (Fig. 6C).

Finally, we also tested the effects of silica particles, a typical inflammasome activator, on CMO neutrophil signaling and IL-1β cleavage. The silica exposure in these cells resulted in the enhanced activation of MAPKs Erk1/2 (Fig. 6E), whereas no reproducible differences in Akt phosphorylation were observed in CMO mice (data not shown). To analyze IL-1β processing, we induced pro–IL-1β production in neutrophils by priming with LPS and then triggered its cleavage by subsequent incubation with silica. Although the LPS-induced pro–IL-1β production was not markedly affected in CMO neutrophils, silica exposure led to the enhanced production of mature IL-1β of 21 and 17 kDa (Fig. 6F), which are the sizes characteristic for cleavage by neutrophil serine proteases and inflammasome-dependent caspases, respectively (41). This is in agreement with the previously published data showing the enhanced secretion of IL-1β by neutrophils into the culture media under these circumstances (35, 36).

Because recent reports showed that IL-1β and inflammasome-activated caspases are critical for the disease development in CMO mice (35–37), we decided to analyze the silica-triggered IL-1β production in more detail, especially with respect to the interactions among PSTPIP2, SHIP1, and Csk described earlier. To address the question of PSTPIP2 phosphorylation under these conditions, we performed the immunoprecipitation of endogenous PSTPIP2 from unstimulated or LPS-primed and silica-stimulated BM cells. Of the cell types that express PSTPIP2 in the BM, neutrophils comprise the vast majority. Thus, we used nonseparated BM cells for this experiment to increase the yield and to prevent unnecessary manipulations of these sensitive cells. Under these conditions, we observed a slight increase in PSTPIP2 phosphorylation after LPS priming followed by a much more substantial increase after silica treatment (Fig. 7A).

To study the effect of the mutations of SHIP1 binding sites in PSTPIP2 on IL-1β processing in neutrophils, we generated immortalized granulocyte progenitors by the infection of BM progenitors with the ER-HoxB8 fusion construct that, in the presence of estrogen, is maintained in the cell nucleus and blocks further differentiation. Such cells can be propagated in the cell culture and infected with retroviral constructs. The replacement of estrogen with G-CSF results in the differentiation of these progenitors into mature granulocytes. Similar to primary granulocytes, these cells are short-lived and cannot be maintained in the culture for >1 d after reaching maturity. We prepared immortalized granulocyte progenitors from CMO mice and transduced these cells with WT, Y3F, and W232A constructs of PSTPIP2. Next, we tested granulocytes differentiated from these progenitors for the processing of IL-1β. The processing of IL-1β after treatment with silica was more efficient in the cells infected with W232A and 3YF mutants when compared with WT PSTPIP2, demonstrating that despite a slightly higher expression level, the mutant proteins were unable to inhibit IL-1β cleavage as efficiently as WT PSTPIP2 (Fig. 7B). This result shows that SHIP1 and PEST phosphatase-binding sites are important for the PSTPIP2-mediated inhibition of IL-1β processing. We excluded empty-vector transfected CMO progenitors from this experiment because of their increased speed of maturation, which was not observed in PSTPIP2-infected cells (irrespective of the mutation present in PSTPIP2; data not shown). Because vector-infected cells were already progressing to apoptosis at the time when the other cell lines just reached the maturity, they could not be used as a control.

Next, to directly analyze the role of SHIP1 in IL-1β processing, we treated the WT and CMO primary granulocytes from the BM with the specific SHIP1 inhibitor 3AC. In agreement with the proposed role of SHIP1 in PSTPIP2 function, this treatment resulted in enhanced IL-1β cleavage after exposure to silica in WT neutrophils, whereas only a very mild effect was observed in their CMO counterparts (Fig. 7C). Altogether, these data support the conclusion that interaction with SHIP1 is a part of the mechanism of how PSTPIP2 suppresses the IL-1β–mediated inflammatory response in granulocytes.

Finally, we wanted to test the expression and interactions of PSTPIP2 in primary human granulocytes. Because these cells can be obtained in relatively large numbers, they also allowed us to test the PSTPIP2 interactions. We isolated human granulocytes from fresh buffy coats and stimulated these cells with silica or human aggregated IgG (as an FcR ligand). Surprisingly, these granulocytes did not appear to interact with silica particles or to phagocytose these particles, and this treatment did not result in any PSTPIP2 tyrosine phosphorylation (data not shown). However, they vigorously responded to the FcR stimulation. This treatment resulted in the increased tyrosine phosphorylation of PSTPIP2 accompanied by the binding of Csk and SHIP1 (Fig. 8). The 100-kDa form of SHIP1 prevailing in these immunoprecipitates is likely the result of proteolytic cleavage and has been described previously (42). These results confirm that PSTPIP2 is also expressed in human granulocytes and that it participates in similar interactions as its murine counterpart.

PSTPIP2 is known for its principal role in the development of the inflammatory disorder described as chronic multifocal osteomyelitis in mice, where its reduced expression or its complete absence is the main cause of the disease (26, 27, 34). However, the mechanism for how PSTPIP2 prevents the disease development has remained unclear. As an adaptor protein, it is likely to function via the recruitment of inhibitory enzymes and other negative regulators. The only molecules of this class previously known to bind to PSTPIP2 were the PEST-PTPs (24, 32). All three family members have been expected to bind to PSTPIP2, although the interaction has formally been proved only for PTP-HSCF and PTP-PEST (24, 32). We filled this gap by showing that PEP/LYP also interacts with PSTPIP2 in a similar manner, dependent on the residue W232 of PSTPIP2 (Fig. 2C). Mice deficient in PEP or PTP-PEST do not develop typical signs of autoinflammatory disorder (43–46), and thus it seems that the major functions of these PTPs are not related to their binding to PSTPIP2. However, if there is a redundancy among the PEST-PTPs that are associated with PSTPIP2, single family member deficiencies would probably not reveal any relevant phenotype. Moreover, additional evidence supports the importance of PEST-PTPs for the functionality of PSTPIP2. The W232A mutation in PSTPIP2, which prevented its binding to PEST-family phosphatases, diminished the ability of PSTPIP2 to inhibit the differentiation of osteoclasts and megakaryocytes (32, 38). In addition, we show in this article that this tryptophan is also important for the suppression of IL-1β processing in neutrophils. The substrate of PSTPIP2-associated PEST-PTPs remains unknown. By analogy with PSTPIP1 (47), one might speculate that their substrate would be PSTPIP2 itself, but we have not observed any increase in the phosphorylation of PSTPIP2 mutated at W232 (Fig. 2D). Thus, it seems more likely that these phosphatases dephosphorylate some of its binding partners, such as SHIP1 and Lyn or other unknown proteins in the vicinity of PSTPIP2.

Another possible role of PEST-PTPs includes the recruitment of Csk to PSTPIP2, which is further enhanced by the independent binding of Csk directly to PSTPIP2. In such a case, SFKs would be the most probable target. We did not observe any effect of PSTPIP2 deficiency on the global phosphorylation of SFKs, whether in macrophages or granulocytes. However, it is possible that PEST-PTPs together with Csk specifically inhibit the pool of SFKs associated with PSTPIP2. The interaction between PSTPIP2 and SFK Lyn was described previously in megakaryocytes (38), and we detected a similar interaction in macrophages (data not shown). The effect of Csk binding could not be fully analyzed because we were unable to identify its binding site in PSTPIP2. We suspect that the binding site is within the F-bar domain or other sequences that could not be evaluated because of the lack of mutated protein expression. Notably, the sequence integrity seems to be essential for PSTPIP2 protein stability, because other mutations, including Lupo and CMO, also result in defective protein expression (27, 34). In addition, there was a fraction of Csk that, in J774 cells, interacted with PSTPIP2 inducibly after M-CSF or FcR stimulation. The binding of the inducible Csk fraction was dependent on the interaction of PSTPIP2 with PEST-family phosphatases. The most likely candidate for bringing this additional Csk to the complex is PTP-PEP, which interacted with PSTPIP2 in an inducible manner. However, another option is that PTP-HSCF, which, unlike PTP-PEST or PTP-PEP, binds Csk via phosphorylated tyrosines in its C-terminal tail (22), is responsible for this inducible binding.

Another critical region in PSTPIP2 comprises the C-terminal tyrosines. Their mutation completely abolished the ability of PSTPIP2 to suppress the differentiation of osteoclasts and reduced its ability to inhibit megakaryocyte differentiation (32, 38). The most likely explanation was that these tyrosines recruit a negative regulatory molecule responsible for these effects. In this work, we have shown that these tyrosines are responsible for binding to the lipid phosphatase SHIP1. Indeed, Y329 and Y323 very well match the consensus binding site for the SHIP1 SH2 domain (48, 49), whereas Y333 bears a certain resemblance to the ITIM motif [hydrophobic amino acid (L) in position −2 with respect to the tyrosine], which is also known to bind the SHIP1 SH2 domain (50). All the previous work on PSTPIP2 tyrosine phosphorylation has analyzed only Y323 and Y333. It was based on the assumption that these tyrosines are analogous to the phosphorylated tyrosine residues found in the related and more thoroughly studied protein PSTPIP1 (Y344 and Y367, respectively) (24). However, our sequence analysis did not reveal any relationship between these tyrosines. Y367 of PSTPIP1 is located inside its SH3 domain, which is not present in PSTPIP2. Moreover, it is a highly conserved residue forming a part of the SH3 domain ligand-binding pocket that is found within a number of these domains in various proteins where it is thought to regulate their interactions with respective ligands (51). The sequence surrounding Y344 of PSTPIP1 (LVYASI) is also somewhat different from Y323 of PSTPIP2 (PDYSVV), and therefore we concluded that the C-terminal tyrosines are a unique feature of PSTPIP2 not found in PSTPIP1. Notably, in the previous work, the mutation of Y323 and 333 severely reduced, but did not completely abolish, the phosphorylation of PSTPIP2, suggesting that there is an additional tyrosine phosphorylation site in PSTPIP2 (24). Y329 was shown to be phosphorylated in mast cells (52), and thus we assumed that Y329 is the most likely candidate. Therefore, we decided to mutate all three C-terminal tyrosines in PSTPIP2 at the same time for our analysis of SHIP1 binding and function. This analysis showed that these three tyrosines are required for SHIP1 binding and that the PSTPIP2 phosphorylation in their absence is negligible. Moreover, we also showed that these tyrosines are required for the maximum inhibition of IL-1β processing in neutrophils. This, together with our data demonstrating increased IL-1β cleavage in WT neutrophils in the presence of the SHIP1 inhibitor, strongly supports the idea that some of the key functions of PSTPIP2 are, at least in part, mediated by SHIP1. Additional support for this idea is also derived from the fact that the Akt and Erk pathways, which are both known to be regulated by SHIP1, are affected by PSTPIP2 absence.

When considering the role of SHIP1 in PSTPIP2 function, it is also interesting to compare the phenotypes of SHIP1-deficient mice with the phenotype caused by PSTPIP2 deficiency. Similar to CMO, SHIP1^−/− mice develop a disease with an autoinflammatory component where inflammatory myeloid infiltrate into multiple organs and enhanced serum IL-6 levels can be observed (3, 53–55). Both strains display splenomegaly with extramedullary hematopoiesis and an increased number and proliferation of myeloid progenitor cells (3, 34, 53). In addition, they also share a phenotype of increased osteoclastogenesis and osteoclast activity, resulting in osteopenia and osteoporosis (32, 56). These features shared between SHIP1^−/− and CMO mice suggest that they are components of a common signaling pathway. In contrast, the reduced survival of SHIP^−/− mice because of massive myeloid cell infiltration in the lungs is not observed in CMO mice (53), and PSTPIP2 does not seem to be involved in endotoxin tolerance. This phenomenon is mainly caused by the dramatic increase in SHIP1 expression after LPS treatment (57). Despite the fact that the expression of PSTPIP2 is also increased, we did not observe any significant difference in the signaling of endotoxin-tolerized WT and CMO BMDMs, suggesting that PSTPIP2 plays no role in this process (data not shown). The more serious phenotype of SHIP1^−/− mice is, however, not surprising, because PSTPIP2 is not the only binding partner of SHIP1, and this enzyme interacts with multiple other proteins and has an effect on other pathways.

The initial characterization of CMO and LUPO mice led to the conclusion that macrophages and osteoclasts are the major cell types responsible for the disease development (27, 32, 34). However, recent work has suggested that, in fact, granulocytes are the most important cell type triggering the disease via the increased production of IL-1β (35, 36). Indeed, we had difficulty detecting any differences in the signaling pathways between the WT and CMO macrophages. In contrast, in CMO neutrophils, we observed enhanced activity of multiple signaling pathways after LPS or silica exposure or after FcR stimulation. The CMO disease development was recently shown to require the activity of inflammasome-associated caspases (36). Indeed, we have observed increased CMO neutrophil responses to a typical inflammasome stimulator silica, which was also used in the previous studies (35, 36). Moreover, we also observed enhanced responses to FcR cross-linking, and to some extent also to TLR stimulation by LPS. This suggests that PSTPIP2 has relatively broad effects on neutrophil sensitivity to various stimuli, and that it is likely that other receptors and pathways that have not been tested yet are influenced by the absence of PSTPIP2.

Which of the receptors regulated by PSTPIP2 are responsible for disease initiation in vivo remains unclear. One possibility is that a receptor for environmental silica or other related compounds triggers the disease in living animals. Our data, as well as previous reports, show an exaggerated neutrophil response to silica treatment. However, to our knowledge, the silica receptor on granulocytes has not been identified yet, and so how silica particles initiate phagocytosis and signaling remains unclear. Fc receptors are also good candidates for triggering the disease, because they not only recognize Abs but they also bind pentraxins and pentraxin-opsonized particles (58). Pentraxins, such as C-reactive protein and serum amyloid P, are a family of acute-phase proteins involved in the recognition and opsonization of bacteria, and could well represent a link described by Lukens et al. (36) between altered bacterial microflora in the gut and inflammatory disease development in CMO mice. Moreover, FcRs use a relatively universal ITAM-dependent signaling pathway, which is also used by a number of other innate immune receptors (e.g., the C-type lectin family of pattern recognition receptors and the TREM family receptors), as well as integrins (59), and our results thus imply that signaling by these receptors will also potentially be affected. Clearly, additional research is needed to address all these possibilities, but at present our data suggest that pathways downstream of multiple receptors are affected by PSTPIP2 deficiency and, in fact, they all may contribute in concert to the CMO disease development. Our data provide evidence that SHIP1 and Csk are part of the PSTPIP2-dependent inhibitory network that regulates these pathways and prevents the development of autoinflammatory disease. Deciphering the mechanisms that govern the assembly and activity of this network will help us to understand the cause of autoinflammatory diseases, including chronic recurrent multifocal osteomyelitis and other related disorders, and will facilitate the development of successful treatments.

We thank Dr. Andre Veillette for providing Abs to SHIP1 and PEST-PTPs, Dr. Mark Kamps for DNA vector encoding ER-HoxB8, and Zdenek Cimburek for technical assistance.

The authors have no financial conflicts of interest.