BAR Proteins PSTPIP1/2 Regulate Podosome Dynamics and the Resorption Activity of Osteoclasts

Bone resorption in vertebrates relies on the ability of osteoclasts to assemble F-actin-rich podosomes that condense into podosomal belts, forming sealing zones. Sealing zones segregate bone-facing ruffled membranes from other membrane domains, and disassemble when osteoclasts migrate to new areas. How podosome/sealing zone dynamics is regulated remains unknown. We illustrate the essential role of the membrane scaffolding F-BAR-Proline-Serine-Threonine Phosphatase Interacting Proteins (PSTPIP) 1 and 2 in this process. Whereas PSTPIP2 regulates podosome assembly, PSTPIP1 regulates their disassembly. PSTPIP1 recruits, through its F-BAR domain, the protein tyrosine phosphatase non-receptor type 6 (PTPN6) that de-phosphophorylates the phosphatidylinositol 5-phosphatases SHIP1/2 bound to the SH3 domain of PSTPIP1. Depletion of any component of this complex prevents sealing zone disassembly and increases osteoclast activity. Thus, our results illustrate the importance of BAR domain proteins in podosome structure and dynamics, and identify a new PSTPIP1/PTPN6/SHIP1/2-dependent negative feedback mechanism that counterbalances Src and PI(3,4,5)P3 signalling to control osteoclast cell polarity and activity during bone resorption.


Introduction
Bone remodeling is a key process that occurs continuously throughout life, needed during the development, maintenance and repair of the skeleton of vertebrates. It involves the coordinated activity of bone-building osteoblasts and bone-digesting osteoclasts. An unbalanced interaction between these two cell types results in disabling diseases such as osteopetrosis, osteopenia or osteoporosis. Osteoclasts are multinucleated cells arising from hematopoietic, mono-nucleated precursors. Macrophage-stimulating factor (M-CSF) triggers the proliferation of these precursors, and the cytokine receptor-activator of NF-κB ligand (RANKL) induces their differentiation into cells able to fuse with each other to generate multi-nucleated osteoclasts [1]. To digest large bone surface areas, mature osteoclasts create between their bone-facing ruffled membrane and the bone surface an acidic resorption lacuna, into which lysosomal hydrolases are

RNA Interference and Gene Transduction
In vitro osteoclastogenesis was induced by addition of RANKL (3% (v/v)~50ng/ml) to Raw264.7 cells, plated at a density of 1x10 6 cells. At day four, cells were differentiated into osteoclasts. For microscope analysis, differentiated cells were washed 2 times with PBS and then incubated for 15min with PBS to remove undifferentiated mononucleated cells. Then osteoclasts were detached with a cell lifter (Corning Incorporated Costar). Detached cells were centrifuged at 220 g for 5 min, resuspended in DMEM/3% RANKL, and then transferred either to glass cover slips (; 11mm) or to BD BioCoat osteological™ discs (BD Bioscience), Ibidi μm, 35mm, glass bottom dish, hydroxyapatite coated chambers or subjected to electoporation. Soluble recombinant RANKL was produced in Pichia yeast as described previously (Czupalla et al., 2005).
HEK293a cells were seeded on glass cover slips (; 11 mm, Menzel) 24h before transfection. HeLa cells seeded at a density 3x10 4 cells/mL were transfected with a mixture of 1μg plasmid DNA and 3μL cationic polyethylenimine transfection reagent JetPEI™ (Peqlab Biotechnologie GmbH). 24h post transfection cells were processed for subsequent analysis.
Stealth RNAi duplexes were designed according to Invitrogen's BLOCK-iT algorithm and purchased from Invitrogen. After 4 days of differentiation on plastic dishes. Osteoclasts were detached as described, centrifuged for 5min at 220g, and resuspended in Electroporation Isoosmolar Buffer (Eppendorf). 1μM of predesigned stealth RNAi or Negative Control Medium GC stealth RNAi duplexes (Invitrogen) were electroporated into osteoclasts with a single square wave pulse of 2750 V/cm field strength and 0.4ms pulse length using an ECM830 ElectroSquar-ePorator™ (BTX, Harvard Apparatus). Electroporated osteoclasts were resuspended in DMEM/ 3% RANKL and allowed to recover for 46hrs. Osteoclasts were processed for subsequent analysis. At least three stealth RNA I duplexes were tested to silence any given gene. Below are listed the stealth RNAi duplexes used in this study:

Lipofection of Osteoclasts
Raw264.7 cells were seeded on Ibidi μm, 35mm glass bottom dish in RANKL containing media, at density 20x10 4 cells/ml. After 48 h transfection by Lipofectamin 2000 (Invitrogen) was performed according to manufacturers protocol. 48 h post transfection osteoclasts were analysed by TIRF microscopy. Adenovirus production and gene transduction into osteoclasts. Raw264.7 cells were seeded on Ibidi μm, 35 mm glass bottom dish at density 20x10 4 cells/ml or after 4 days of differentiation seeded onto BD BioCoat osteological™ discs as described, in RANKL containing media. After 48 h, transfection by Lipofectamin 2000 (Invitrogen) or Transficient™ DNA Transfection Reagent (MBL) was performed according to manufacturer's protocol. 48 h post transfection osteoclasts were analysed by TIRF microscopy or confocal microscopy. Adenoviral vectors and recombinant adenoviruses were generated using the AdEasy™ system (QBIOgene) developed by He et al. [14]. Target genes were subcloned from various vectors into the transfer vector pShuttle-CMV. For homologous recombination with the pAdEasy-1 plasmid that encodes the Adenovirus-5 genome (E1/E3 deleted), pShuttle-CMV was linearised with PmeI (New England BioLabs 1 ) and 100ng of linearised DNA was electroporated into 33μL of electrocompetent BJ5183-AD1 cells using 2.5 kV, 2mm cuvettes. The recombinant adenoviral construct was then cleaved with PacI and 3μg were transfected into 0.6x10 6 HEK293A cells by using JetPEI™ for production of virus particles. The first virus particles were collected after 10-15 days post transfection. At least two steps of virus amplification were necessary to achieve a good virus titer. Sufficient first viral amplification was harvested from 175 cm 2 of culture, whereas, at least ten 175 cm 2 flasks of HEK293A cells were required for infection with the recombinant adenovirus construct in the second viral amplification and harvested after 2-5 days. Viral particles were released from the cells by 3 quick freeze/thaw/cycles using liquid nitrogen. Viral particles were purified and concentrated using a discontinuous iodixanol gradient (OptiPrep 1 , Axis Shield) adapting the method from Zolotukhin et al (Zolotukhin et al., 1999). The purified virus solution was supplied with 1/3, 3x storage buffer (15mM Tris pH8.0, 150mM NaCl, 0.15% BSA, 50% (v/v) glycerol), and 1/3 of glycerol. The viruses were stored at -20°C. After 4 days of differentiation osteoclasts were transduced with titrated adenovirus and grown for additional 48 h either plated on BD BioCoat osteological™ discs. Then cells were processed for subsequent analysis.

Immunocytochemistry
For immunocytochemistryanalysis, cells were grown on glass cover slips (; 11 mm) or BioCoat osteological™ discs and fixed with 3% (w/v) paraformaldehyde in PBS for 15min at 37°C, quenched with 50mM NH 4 Cl/PBS for 10 minutes and permeabilised with 0.1% (w/v) Triton X-100 in PBS for 6min. Samples were blocked with 3% (w/v) BSA in PBS for 30 min at room temperature. Then, cover slips were transferred into a humid chamber and incubated with 30μl primary antibody diluted in 3% (w/v) BSA in PBS for 1h at room temperature. Following three PBS washing steps, cover slips were incubated, in the dark, with the appropriate secondary antibody as described above. Staining of the actin cytoskeleton was performed with Phalloidin Alexa 546 or Phalloidin Alexa 633 (Invitrogen). Coverslips were washed five times with PBS, once with water and mounted on glass slides by inverting them onto a droplet of Mowiol containing 10μg/mL DAPI (Invitrogen). BioCoat osteological™ discs were washed five times with PBS and once with water and mounted as following: 1droplet of Mowiol was added to the glass slide, then the BioCoat osteological™ discs were placed with the cells facing the top, 20μL Mowiol+DAPI and a glass slide with nail polish droplets was inverted on the top of the BioCoat osteological™ discs.

Immunocytochemistry with PIP 3 IgM
Cells were fixed by 4% PFA for 30 min at 37°C. Permeabilization and blocking were performed simultaneously by in 0.5% saponine in 3% BSA in PBS on ice for 45 min. From this point work was done on ice. Primary IgM PIP3 antibody (RC6F8) in 0.5% saponine in 3% BSA in PBS buffer was added onto fixed cells and stained for 1 hr (Chen et al., 2002). Cells were washed 3x by ice cold PBS and appropriated secondary antibody was added. After 30 min cells were 4x washed by ice cold PBS. Post fixation was done by 10 min incubation with 2% PFA on ice.

Time-Lapse Videomicroscopy
Raw 264.7 cells derived osteoclasts were transferred to coverslips coated with hydroxyapatite (BD Biosciences, Heidelberg, Germany) and grown in Minimum Essential Eagle Medium (Sigma-Aldrich, St. Louis, USA) supplemented with 10% FCS and 3% soluble recombinant RANKL. Cells were transfected by adenovirus coding RFP-tagged actin binding domain of Ezrin protein and observed with a Zeiss Axiovert 200 M inverted microscope equipped with an automated stage, an Incubator XL3 for temperature maintenance and CO 2 buffering (PeCon). Sequential images were acquired every 1 min for 30 min and processed with the MetaMorph version 6.1 Imaging software (Molecular Devices, Sunnyvale, USA). Sealing zone diameters were assessed using Fiji software and statistical significance was tested using unpaired students t-test (Graph Pad Prism 6).
For monitoring of individual podosomes Raw 264.7 cells were seeded on Ibidi μm, 35mm glass bottom dish in RANKL containing media, at density 20x104 cells/ml. After 48 h transfection by Lipofectamin 2000 (Invitrogen) was performed according to manufacturer's protocol. 48 h post transfection osteoclasts were analysed by Leica AFLX6000 TIRF. Sequential images were acquired every 2 sec for 5 min.

Image Analysis
Images from fluorescence and confocal acquisitions were processed with Adobe Photoshop v7.0 (Adobe Systems). All image processing and analysis were carried out with FIJI software (Schindelin et al., 2012) Immunoprecipitation Cells were washed with PBS, harvested in ice-cold lysis buffer (50 mM Tris pH 7.5, 150 mM NaCl, 1% (v/v) NP-40, 0.1% (w/v) sodium deoxycholate, 1 mM EDTA, 10 mM sodium β-glycerophosphate, 10 mM NaF, 1 mM sodium orthovanadate, and protease inhibitors (Comple-teTM tablets, Roche Diagnostics, Mannheim, Germany), homogenized by resuspension first three times with a 22.5 Gauge needle, second five times with a 27 Gauge needle, then lysed for 15 min on a rotating wheel at 4°C, and centrifuged at 14,000 g at 4°C for 10 min. Protein concentrations of the lysate supernatants were estimated using the DC protein assay including reagents S, A, and B (Bio-Rad, Munich, Germany).
For immunoprecipitation with specific antibodies (PSTPIP1, PSTPIP2), lysates (total protein concentration > 2 mg/ml) were precleared on Protein G-sepharose beads for 1 h at 4°C. Supernatants were incubated with antibodies (5 μg/mg of lysate) for 1 h at 4°C. Then, Protein G-sepharose beads were added for additional 2 h at 4°C. Precipitates were washed two times with lysis buffer, proteins were eluted, and resolved on SDS-PAGE. For immunoblotting, gelseparated proteins were transferred onto nitrocellulose membranes and incubated with the corresponding antibodies. After incubation with secondary antibodies conjugated with HRP (Jackson Immuno Research, Suffolk, UK), bands were detected with enhanced chemiluminescence (ECL) Western Blotting Detection Reagents (GE healthcare, München, Germany).
For immunoprecipitation of tyrosine phosphorylated proteins followed by SILAC mass spectrometry-basedanalysis (see below), lysates of treated and non-treated cells were combined in 1:1 ratio, for immunoprecipitation of tyrosine phosphorylated proteins followed by immunoblotting, lysates were kept separated. Such lysates (total protein concentration > 2 mg/ml) were precleared on Protein A-sepharose beads for 2 h at 4°C. Supernatants were incubated with immobilized anti-phosphotyrosine antibodies (25 μg 4G10/mg of lysate and 10 μl P-Tyr-100/mg of lysate) for 6 h at 4°C. Precipitates were washed four times with lysis buffer, proteins were eluted, and resolved on SDS-PAGE. For immunoblotting, gel-separated proteins were transferred onto nitrocellulose membranes and incubated with the corresponding antibodies.

GST-Tagged Protein Purification
To 250 ml of LB overnight culture were added 750 ml of fresh media with appropriate antibiotics and shake at 190 rpm for 3,5 hrs at 37°C. Protein expression was induce with 0,1 mM IPTG and grown overnight at 15°C. Bacteria were sedimented by centrifugation 4000g for 20 min at 4°C. Pellet was washed with 1x volume LB2 BFR (5 mM Tris, 15 mM NaCl pH 8) and resuspended in LB1 BFR1 (50 mM Tris, 150 NaCl, pH 8, complete protease inhibitors, 5 u Benzonase). Suspension was lysed on French press for 5 min at 10000 kPa. Lysate was centrifuged for 30 min, 10 000g at 4°C. Sepharose beads were washed 2x with 500 ul with water and 2x with LB1 BFR. 50% beads slurry was prepared by adding 250 ul of LB1 BFR to washed beads. Supernatant from bacterial lysate was filtered through 0,45 um and added to washed 50% bead slurry and rotated for 2 hrs at 4°C. After this incubation beads were washed 3x 500 ul with LB1 BFR and 3x 500 ul with fresh binding BFR (20 mM Hepes, 100 mM KCl, 0,05%, 1 mM DTT, complete protease inhibitors). For pull down experiments 50% bead slurry in binding BFR was prepared. GUVs tubulation studies PreScission protease was applied to beads and purified recombinant protein was eluted as described in manufacturers protocol (Ge Healthcare).

GST Pull-Down
Expression of GST-tagged PSTPIP1, PSTPIP1 F-BAR domain only, PSTPIP1 SH3 domain only and PSTPIP2 in E. coli, was induced with 0.1 mM isopropyl-D-thiogalactoside (IPTG) for 20 h at 20°C. Protein was pulled down from bacterial lysate by glutathione-sepharose beads. Briefly: osteoclast lysates were incubated with glutathione-sepharose beads for 1 h at 4°C before they were added to the GST-tagged protein variants on glutathione-sepharose beads for 1 hr incubation at 4°C. Precipitates were washed 4 times with lysis buffer; proteins were eluted, and resolved by SDS-PAGE.

Mass Spectrometry and Data Analysis
Coomassie stained protein bands were excised from the gel, cut into 1mm-cubes and washed twice with ultra-pure water to remove SDS. The gel pieces were then washed twice with 50% (v/v) acetonitrile (ACN) in 25mM ammonium bicarbonate (ABC) for 5min and shrunk by dehydration in ACN. The ACN was removed and the gel pieces were re-hydrated in 50mM ABC. After 5min the same volume of ACN was added for 5min and finally removed completely. The gel pieces were shrunk again in ACN for 5min, ACN was removed and gel pieces were dried in a vacuum centrifuge. Disulfide bonds were reduced by incubation with 10mM DTT in 100 mM ABC for 45min at 56°C. Alkylation was performed by replacing the DTT solution with 55mM iodoacetamide in 100 mM ABC. After 20min at 25°C in the dark, the gel pieces were washed with twice 50% (v/v) ACN in 25 mM ABC, shrunk by dehydration in ACN, and dried in a vacuum centrifuge. The gel pieces were incubated with 100 ng trypsin (sequencing grade, Promega) at 37°C overnight in 20 μl of 25mM ABC. To extract the peptides, 20 μl of 0.5% (v/v) trifluoroacetic acid (TFA) in ACN were added, the samples were sonicated and vortexed for 5min each. The supernatant was transferred into new tubes and the gel pieces were washed, sonicated and vortexed again with 20 μl ACN. The supernatants were combined and dried in a vacuum centrifuge. For mass spectrometric analysis of the peptide mixture, samples were re-dissolved in 5 μl 0.1% (v/v) TFA in water, referred as analyte solution.
Peptides were separated on an UltiMate3000 nanoHPLC system (Dionex, Amsterdam, The Netherlands) equipped with a PepMap C18 nano trap column (3mm, 100Å, 2cm x 75mm i.d.) and a PepMap C18 analytical column (3mm, 100Å, 15cm x 75mm i.d.) directly coupled to the nanoelectrospray source (Proxeon, Odense, Denmark) of a LTQ Orbitrap XL mass spectrometer (Thermo Fisher Scientific, Bremen, Germany). Peptides were eluted with an 80 min linear gradient of 5-45% acetonitrile in 0.1% formic acid at 200 nL/min. Mass spectra were acquired in a data-dependent mode with one MS survey scan (resolution of 60,000) in the Orbitrap and MS/MS scans of the eight most intense precursor ions in the LTQ. Data analysis was done using MaxQuant version 1.2.2.5 (Cox 2008). Peak lists were searched against a database containing 16,339 entries from the UniProt-KB/Swiss-Prot mouse database (release 2011_02) and 255 frequently observed contaminants as well as reversed sequences of all entries and the following search criteria: (i) enzyme specificity, trypsin; (ii) mass accuracy, 6 ppm and 0.5 Da for precursor ion and fragment ion mass tolerance, respectively; (iii) fixed and variable modifications, cysteine carbamidomethylation and methionine oxidation as well as modified arginine and lysine (for SILAC experiments), respectively; (iv) maximum of two missed cleavage sites.
In SILAC experiments cells were cultured in DMEM lacking arginine (PAN Biotech) supplemented with 10% (vol/vol) dialyzed FCS. Arg-6 and Arg-0 SILAC media were prepared by adding l-arginine-U-13 C 6 (Cambridge Isotope Laboratories) or the corresponding nonlabeled amino acid, respectively, and cells were cultured for 5 days in SILAC media. Peptide identifications were accepted based on their posterior error probability until less than 1% reverse hits were retained while protein false discovery rates were < 1%. Proteins were considered if at least two peptides were identified. SILAC-based protein quantification was performed by MaxQuant based on the median SILAC ratios of at least two peptides per sample. Results were only included if the experiment-to-experiment variation of protein ratios were <30% in two independent experiments with SILAC label swapping.
Microscopy chambers Lab-TEK II ((Nunc, Langenselbold, Germany) were blocked 2 hrs with 2% BSA. Washed 3x with 200 μl of BFR (BFR 10 mM Hepes, 150 mM NaCl, 295 mOsm/ Kg). To each chamber 20 μl of BFR was added. 0.5 μl of GUVs were transferred by use of cut pipet tip into isoosmolar BFR and let settle for 5 min at RT. Proteins were added to chambers with GUVs to final concentration 1,5 μM. Osmolarity was increase by addition of 10 ul of 750 mOsm/Kg sucrose to obtain final 400 mOsm/Kg. GUVs were than incubated for 15 min at RT and subjected to inverted LSM 510 META confocal microscope equipped with a 40x, 1.2 numerical aperture water-immersion objective.

Cre Recombination of PSTPIP1 in Osteoclasts
Bone marrow cells of 8 to 10 weeks old B6.129-Pstpip1tm1Spg/J +/+ Ctsk-CreE2+/-were sacrificed by CO 2 asphyxiation to isolate long bones (femur and tibia). Femurs and tibias were cleaned of the surrounding soft tissue. Following excision of the ends of the long bones, bone marrow was removed by flushing with aMEM medium. Cells were then differentiated into osteoclasts or macrophages by respective addition of 20 ng/ml Macrophages Colony Stimulating Factor (MCSF, PeproTech, germany) and 50 ng/ml recombinant soluble Receptor activator of nuclear factor kappa-B ligand (RANKL, PeproTech, germany). Cells were grown in aMEM medium supplemented with 10% FCS, 2 mM HEPES, 1% penicillin/streptomycin and 1% Lglutamine in 10% CO2, 95% humidity at 37°C. After differentiation, 4-OHT (10 -3 M) diluted in ethanol was added to the cultured cells.

PSTPIP1 and PSTPIP2 Regulate Podosome and Sealing Zone Dynamics
PSTPIP1/2 amino acid sequences predict that they contain putative F-BAR domains. Such domains have the property to promote membrane tubulation both in vitro and in vivo [10,11].
In agreement with others [9], the incubation of recombinant PSTPIP1/2 with giant unilamellar vesicles led to the formation of membrane tubules (S1A Fig) and the over-expression of GFPtagged PSTPIP1 or PSTPIP2 in HEK cells also led to membrane tubule formation (S1B Fig). Thus, although their crystal structure was not yet established, PSTPIP1/2 exhibited the typical properties of bona fide BAR-domain containing proteins.
To investigate PSTPIP1/2 functions, we first localized them in Raw.267-derived osteoclasts grown on osteological discs (hydroxyapatite-coated surfaces). Both endogenous PSTPIP1 and PSTPIP2 localized to actin-rich sealing zones (Fig 1A and 1B). A mRFP-tagged PSTPIP2 expressed in Raw.267-derived osteoclasts also localized to actin rich sealing zones (S2 Fig). Whereas the siRNA-mediated PSTPIP2 depletion (90% efficiency without affecting PSTPIP1 expression, Fig 1E) led to a complete disappearance of podosomes and sealing zones (Fig 1C), the siRNA-mediated PSTPIP1 depletion (>90% efficiency without affecting PSTPIP2 expression, Fig 1E) did not produce any apparent effects on sealing zones observed in fixed osteoclasts ( Fig 1C). However, the examination of PSTPIP1-depleted osteoclasts using time-lapse video microscopy revealed that their sealing zones, labeled by the mRFP-tagged actin-binding domain of Ezrin (mRFP-ABDE), had significantly lost their dynamic state when compared to control osteoclasts (Fig 1D, 1F and 1G, S1 and S2 Movies). These results are in agreement with the observation that PSTPIP1 depletion in macrophages stabilizes podosomes [15]. Similar to PSTPIP2 depletion, the depletion of both PSTPIP1 and PSTPIP2 abolished podosome and sealing zone formation (Fig 1C). These results suggest that PSTPIP2 regulates podosome and sealing zone assembly, whereas PSTPIP1 regulates their disassembly.
To test this hypothesis, we used time-lapse videomicroscopy to analyze podosome dynamics. Osteoclasts were grown on glass surfaces that allow podosome assembly, but prevent their packing into sealing zones. Podosomes labeled with mRFP-ABDE had a half-life of 2-4 minutes, as previously described [3], and their assembly/disassembly correlated with the synthesis and turnover of PI(3,4,5)P3, detected with the GFP-PH domain of Akt (Fig 2A  and 2B, S3 Movie). PSTPIP1 and PSTPIP2 exhibited different dynamics. PSTPIP2 was always detected on already formed F-actin-rich podosomes, which disassembled while PSTPIP1 was recruited. (Fig 2A and 2B, S4 and S5 Movies). Altogether, these results indicate that PSTPIP1 and PSTPIP2 have opposite roles in podosome dynamics. PSTPIP2 controls Fluorescence intensities across the indicated white lanes are indicated and Pearson's coefficients were calculated (0.65 for PSTPIP1 and 0.25 for PSTPIP2 from datasets of three different experiments N = 3, and n = 68 measurements). C Effect of siRNA-mediated depletion of PSTPIP1 or PSTPIP2 or both PSTPIP1 and PSTPIP2 on sealing zone assembly. Osteoclasts were treated with siRNA targeting the indicated genes and then grown for 48 hours on osteological discs as indicated in Materials and Methods. Cells were then fixed and stained with phalloidin. Scale bars 20 μm. D Sealing zone dynamics in PSTPIP1-depleted osteoclasts. Osteoclasts were treated with siRNAs targeting PSTPIP1 and then plated on osteological discs. After 24 hours, they were infected with a recombinant adenovirus encoding the mRFP-Ezrin actin-binding domain. After 32 hours, osteoclasts were observed by time-lapse videomicroscopy (300 msec. per frame, 1 frame per 1 min., see S1 and S2 Movies 1, 2). E The knockdown efficiencies were determined by western blotting and quantified. The figures presented are representative of at least 3 independent experiments (mean ± SD). F Sealing zone diameter was measured using the Fiji software. The relative sealing zone diameter (biggest sealing zone as reference) was plotted for each sealing zone assessed (n = 3 independent experiments). G The change of relative sealing zone diameter per minute was plotted and tested using students t-test. (mean ± SD * represents p<0.05 and ** p<0.01, *** p< 0.001).

PSTPIP1/2 Interactomes
To better understand PSTPIP1/2 function, we then identified their interactors. For this, we first performed pull-down experiments using recombinant PSTPIP1, its F-BAR, or SH3 domains fused to GST or recombinant GST-PSTPIP2, as baits, and osteoclast lysates as a source of proteins. Pulled-down proteins were identified using semi-quantitative mass spectrometry based on MS2 spectrum counting. We identified %150 putative PSTPIP1 interactors binding either to its F-BAR or SH3 domain, including several known PSTPIP1 interactors and podosomal components (Table 1). These were classified into several functional groups. The first group comprised phospho-tyrosine protein phosphatases of the PEST family such as PTPN12, PTPN18, PTPN22 and PTPN6 that bound to the F-BAR domain. The second group comprised tyrosine protein kinases such as Syk and BTK binding to the F-BAR domain, and Abl1/2 binding to the SH3 domain. A third group, mostly interacting with the PSTPIP1 SH3 domain, comprised the GTPase dynamin (dynamin2), several ARF or Rho GTPase activating proteins (GAPs) including GIT1, ARAP1 and ASAP1, Rho-GAPs (RHG17), and the Rho  Guanine exchange factors DOCK 5 and 8. A fourth group comprised of several actin-nucleation promoting factors such as WASP, VASP, WASL and IQGAP1, and the actin motors Myosin-9 and Myosin-II. Finally, the phosphoinositol phosphate 5-phosphatases SHIP1/2 and synaptojanin bound to the PSTPIP1 SH3 domain. We also established the PSTPIP2 interactome (S1 Table). Compared to PSTPIP1, fewer PSTPIP2 interactors were identified, as expected from the fact that PSTPIP2 lacks a SH3 domain. Interestingly, several proteins (PTPN12, PTPN22 and IQGAP1) interacted with both PSTPIP1 and PSTPIP2. Other proteins exclusively bound the F-BAR domain of either PSTPIP1 (e.g. PTPN6) or PSTPIP2 (e.g. Talin-1). The same interactors were found by labelfree quantitative proteomics comparing GST-tagged PSTPIP1 and GST-tagged PSTPIP2 interactomes (data not shown). We confirmed several of these in vitro interactions by immunoprecipitating endogenous PSTPIP1/2 from osteoclast lysates with specific antibodies, followed by western blotting using specific antibodies against selected interactors (e.g. PTPN6 and SHIP1/ 2) (S3 Fig). In particular, we confirmed that PSTPIP1, PTPN6 and SHIP1/2 formed a complex in osteoclasts.

PTPN Phosphatases and Podosome/Sealing Zone Dynamics
To investigate the biological significance of this PSTPIP1, PTPN6 and SHIP1/2 complex, we first examined PTPN6 localization and function. PTPN6 localized to sealing zones ( Fig 3A). Its localization was lost upon PSTPIP1 depletion (Fig 3B), as expected for a specific PSTPIP1 interactor. To establish its functional importance, we examined by time-lapse video microscopy the dynamics of mRFP-ABDE-positive sealing zones in PTPN6-depleted osteoclasts (90% knockdown efficiency without affecting PTPN12 or PTPN22 expression, Fig 3D). Fig 3C (S6 and S7 Movies) shows that PTPN6-depleted osteoclasts grown on osteological discs significantly lost sealing zone dynamics (Fig 3E and 3F), thus fully recapitulating PSTPIP1 depletion.
We also examined the localization and function of PTPN12 and PTPN22. Although they were detected at sealing zones, their localization was not affected by PSTPIP1 depletion, as expected from proteins also interacting with PSTPIP2 (S4A Fig). We therefore determined how PTPN12-and PTPN22-depletion could affect sealing zone dynamics using time-lapse video microscopy in osteoclasts expressing the mRFP-tagged actin-binding domain of Ezrin. Their depletion (95-90% efficiency, S4C Fig) gave rise to phenotypes resembling PSTPIP2 depletion. Only rare and small, quickly collapsing, sealing zones could be observed in PTPN12-or PTPN22-depleted osteoclasts (S4D and S4E Fig, S8-S10 Movies). This indicated that the main function of these PSTPIP2 interactors was to regulate podosome and sealing zone assembly, as did PSTPIP2.

PIP3, 5 Phosphatases SHIP1/2 and Podosome/Sealing Zone Dynamics
We then reasoned that the protein-tyrosine phosphatase PTPN6 bound to the F-BAR domain of PSTPIP1 could regulate the phosphorylation state and the activity of proteins bound to the SH3 domain of PSTPIP1. To identify PTPN6 substrates, we performed SILAC (Stable Isotope Labeling with Amino acid in cell Culture) experiments to evaluate the phosphotyrosine-containing proteomes of osteoclasts treated with phenylarsine oxide (PAO), a membrane permeable PTPN inhibitor [16]. Lysates of osteoclasts treated or not with this inhibitor were mixed (equal volumes) and immunoprecipitated with anti-phosphotyrosine antibodies. The immununoprecipitates were then analyzed by quantitative mass spectrometry. Among the %350 proteins identified, %150 proteins exhibited changes in their tyrosine phosphorylation state ( Table 2). Among these we found the PSTPIP1 SH3 domain interactors SHIP1, ABL1, GIT1, VASP, as well as other proteins critical for podosome/sealing zone assembly, such as the protein-tyrosine kinase Pyk2, the focal adhesion kinase FAK1 and paxillin. We confirmed that PTPN6 or PSTPIP1 knockdown results in a higher phosphorylation state of tyrosine residues in SHIP1/2 (S5 Fig). In osteoclasts grown on osteological discs, the PIP3, 5 phosphatases SHIP1 and SHIP2 were detected at sealing zones and, to a small extent, at the ruffled border ( Fig 4A). Their localization at sealing zones was reduced upon PSTPIP1 knockdown (Fig 4A). We then examined, using time-lapse videomicroscopy, the phenotype of osteoclasts expressing the GFP-tagged ezrin actin-binding domain and depleted of either SHIP1 or SHIP2. Fig 4C,  4D and 4F (S11-S13 Movies) shows that the siRNA-mediated depletion of SHIP1 or SHIP2 led to the formation of sealing zones with significantly reduced dynamic states, thus phenocopying PSTPIP1 or PTPN6 depletion. This illustrates the complementary functional importance of  : 20μm). B PSTPIP1-depleted osteoclasts were treated similarly (scale bars: 20μm) Images were analyzed using the Fiji software. Fluorescence intensities across the indicated white lanes are indicated and Pearson's coefficients were calculated (0.35 for PTPN6 from datasets of three different experiments N = 3, and n = 37 measurements). C Sealing zone dynamics in PTPN6-depleted osteoclasts. Osteoclasts were treated with siRNAs targeting PTPN6 and then plated on osteological discs. After 24 hours, they were infected with a recombinant adenovirus encoding the mRFP-Ezrin actin-binding domain. After 32 hours, osteoclasts were observed by time-lapse videomicroscopy (100 msec. per frame, 1 frame per 1 min., see S6 and S7 Movies). D The knockdown efficiencies were determined by western blotting and quantified. The figures presented are representative of at least 3 independent experiments. Scale bars 20 μm (mean ± SD). E Sealing zone diameter was measured using the Fiji software. The relative sealing zone diameter (biggest sealing zone as reference) was plotted for each sealing zone assessed. F The change of relative sealing zone diameter per minute was plotted and tested using students t-test. (mean ± SD * represents p<0.05 and ** p<0.01, *** p< 0.001).

Podosome/Sealing Zone Dynamics and Bone Degradation
In osteoclasts, the inability to assemble podosomes and sealing zones impairs bone degradation. However, it is unknown how changes in podosome and sealing zone dynamics affect osteoclast activity in bone degradation. To address this question, we examined the capability of osteoclasts depleted of PSTPIP1, PTPN6 or SHIP1/2, which reduced sealing zone dynamics (see Fig 5F), to digest osteological discs mimicking bone surfaces. Fig 5A shows that osteoclasts depleted of any of these components exhibited a 2-3 fold higher capacity of digesting the surface of osteological discs. In contrast, osteoclasts depleted of PSTPIP2, PTPN12, PTPN22 or Src, that could not assemble podosomes and sealing zones were also unable to digest such  surfaces. We also evaluated the ability of mouse primary PSTPIP1-/-osteoclasts to digest osteological discs. For this, mice carrying a floxed PSTPIP1 gene were crossed with mice expressing Cre-ERT2 under the control of the Cathepsin K promoter that allows performing tamoxifeninduced conditional knockouts in osteoclasts [17]. Cre-ERT2-positive and negative mice with a floxed PSTPIP1 gene were generated, and primary osteoclasts were obtained after the differentiation of their bone marrow precursors with M-CSF and RANKL. The treatment of these primary osteoclasts with tamoxifen resulted in an efficient knockout of PSTPIP1, without affecting PSTPIP2 expression (Fig 5C). Tamoxifen treatment of Cre-ERT2+/+, PSTPIP1 +/+ and Cre-ERT2-/-, PSTPIP1+/+ osteoclasts did not affect their ability to form sealing zones ( Fig 5B). However, it increased (%3 fold) the capacity of Cre-ERT2+/+, PSTPIP1+/+ osteoclasts to digest the surface of osteological discs ( Fig 5D). Altogether, these results indicate that the dynamic instability of podosomes and sealing zones controlled by PSTPIP1/PTPN6/SHIP1/2 complex is key in modulating the activity of osteoclasts in bone digestion.

Discussion
Our study illustrates the essential function of PSTPIP1 and PSTPIP2 in podosome and sealing zone dynamics in osteoclasts. Remarkably, these F-BAR-domain proteins exhibit opposite activities. PSTPIP2, acting as a membrane scaffold, is essential for podosome and sealing zone assembly. PSTPIP1, substitutes PSTPIP2 on mature podosomes, and regulates podosome and sealing zone disassembly. PSTPIP1 recruits through its F-BAR domain the protein tyrosine phosphatase PTPN6 that can dephosphorylate and regulate the activity of podosome components bound to its SH3 domain, as illustrated for the PI(3,4,5)P3 5-phosphatases SHIP1/2. These results provide a mechanism by which the PSTPIP1/PTPN6/SHIP1/2 complex regulates the dynamic instability of podosomes assembled upon Src-dependent phosphorylation and PI (3,4,5)P3 signalling. In addition, it shows that this dynamic assembly/disassembly of podosomes and sealing zones is key for osteoclast activity in bone degradation (Fig 6). PSTPIP1 and PSTPIP2 behave as typical F-BAR domain containing proteins sensing positive membrane curvature able to generate membrane tubules in vitro [10,18,19]. Previous studies have proposed that podosomes and invadopodia of cancer cells are protrusive structures of the plasma membrane [6,20,21]. Some others have proposed that podosomes of Rous sarcoma virus (RSV) transformed cells contain an invaginated tubular membrane in their core [22,23] that contains actin surrounded by 250-500 nm rings containing two podosomal components, vinculin and talin [4]. Our data showing that PSTPIP2, a curvature-sensing protein, is essential for podosome assembly in osteoclasts would be consistent with a structural model in which podosomes contain an invaginated membrane tubule in their core. This structure would also explain the role that dynamin, a SRC-dependent GTPase that assembles on tubulated membranes during clathrin-mediated endocytosis, plays in podosome dynamics [24].
Our study illustrates the critical role of PSTPIP2 in podosome formation in mature Raw cell-derived osteoclasts since its knockdown prevents the formation of these structures. This finding would drastically differ from other studies showing that PSTPIP2 is a negative regulator of actin polymerization in undifferentiated Raw cells [25]. Whereas additional studies depleted osteoclasts. Osteoclasts were treated with siRNAs targeting SHIP1 or SHIP2 and then plated on osteological discs. After 24 hours, they were infected with a recombinant adenovirus encoding the mRFP-Ezrin actin-binding domain. After 32 hours, osteoclasts were observed by timelapse videomicroscopy (100 msec. per frame, 1 frame 1 min., see S11-S13 Movies). D Sealing zone diameter was measured using the Fiji software. The relative sealing zone diameter (biggest sealing zone as reference) was plotted for each sealing zone assessed. E The knockdown efficiencies were determined and quantified by western blotting. The figures presented are representative of at least 3 independent experiments (mean ± SD). Scale bars 20 μm. F The change of relative sealing zone diameter per minute was plotted and tested using students t-test. (mean ± SD * represents p<0.05 and ** p<0.01, *** p< 0.001) doi:10.1371/journal.pone.0164829.g004

Fig 5. Sealing zone dynamics and osteoclast activity in digestion.
A Osteoclasts were treated with siRNAs targeting the indicated genes and then plated onto osteological discs. After 48 hours, the digested areas of osteological discs (seen in white whereas the surface appears in grey) were visualized by microscopy. B Bone marrow osteoclast precursors isolated from long bones of Cre-ERT2+/+,PSTPIP1+/+ and Cre-ERT2-/-,PSTPIP1 +/+ mice were treated with M-CSF and RANKL and the resulting osteoclasts were further treated with tamoxifen as indicated in Materials and Methods. Sealing zones of osteoclasts plated on osteological discs were stained with phalloidin and observed by confocal microscopy. C PSTPIP1 and PSTPIP2 expression was determined by western blotting. D The activity of these osteoclasts in resorption was determined as indicated above and as described in Materials and Methods. Scale bars 50 μm. The figures presented are representative of at least 3 independent experiments. E Quantification of resorption pit assays were perfomed as indicated in materials and methods. Quantifications of areas from 3 different experiments are plotted in chart (mean ± SD * represents p<0.05 and ** p<0.01, significance was calculated using t-test). Knockdown efficiencies (>90%) determined by western blotting were as presented in previous figures. F The change in relative sealing zone diameter per minute was plotted for each condition; controls of individual experiments were taken together. Statistical significance was tested using students t-test for each KD. (Mean ± SD, **** p< 0.0001). would be required to clarify this apparent discrepancy, it is possible that Src-dependent tyrosine phosphorylation of PSTPIP2 in mature osteoclasts modify the properties of this BARdomain containing protein. Our study also shows that PSTPIP1 can substitute PSTPIP2 on mature podosomes, a phenomenon that has never been described for BAR-domain proteins. The mechanism underlying this switch remains unknown. Src-dependent phosphorylation may regulate this switch. Our study demonstrates that PSTPIP1 and PSTPIP2 play essential but opposite roles in podosome dynamics. Whereas PSTPIP2 regulates podosome and sealing zone assembly, PSTPIP1 regulates their disassembly. This dynamic instability of podosomes is important for sealing zone dynamics in osteoclasts. The arrangement of podosomes in higher ordered structures depends on the surface onto which cells adhere [3,5,6]. Podosomes assemble when osteoclasts are grown on glass, but condense into podosomal sealing zones when they are grown on bone [24] or on bone mimicking surfaces, such as osteological discs [3]. Sealing zone expansion and shrinking is allowed by a continuous assembly of podosomes at its outer rim, and a disassembly of podosomes in its inner rim [3]. Whereas this dynamic instability of  (3,4,5)P3 synthesis and Src-dependent phosphorylation of podosomal components. B PSTPIP2, acting as a membrane scaffolding protein, initiates membrane tubulation. PSTPIP2 recruit podosomal components such as Talin, which connects integrins with actin cytoskeleton. This step also involves PTPN12 and PTPN22. C PSTPIP1, also sensing membrane curvature, substitutes PSTPIP2 by a yet unknown mechanism. PSTPIP1 recruits interactors, in particular PTPN6 and SHIP1/2. D PTPN6 can dephosphorylate several podosomal components and regulate their activity, in particular SHIP1/2 to turnover PI (3,4,5)P3 into PI(3,4)P2. PI (3,4,5)P3 turnover combined with dephorylation of Src substrates would destabilize the overall podosome architecture. Podosome and sealing dynamics would therefore rely on mechanisms coordinating PI(3,4,5)P3 synthesis/turnover, membrane scaffolding properties of the F-BAR PSTPIP1/2 and the Src-and PTPN6-dependent phosphorylation state of podosomal components.
doi:10.1371/journal.pone.0164829.g006 podosomes and sealing zones regulate osteoclast activity in bone degradation, it may also play a role during other phases of bone digestion, i.e. the subsequent steps of cell adhesion and cell migration.
Besides acting as membrane scaffolds, PSTPIP1/2 are also used as docking platforms that recruit podosome components in osteoclasts. PSTPIP2 binds talin1, which connects αvβ3 integrins and F-actin, and is essential for bone degradation [26]. IQGAP1, an actin nucleation promoting factor essential for podosome assembly [27] and the Rac1 Guanine exchange factor DOCK5, which is essential for sealing zone assembly and bone degradation [28]. PSTPIP2 is also used as a docking platform by the tyrosine-protein phosphatases PTPN12 and PTPN22, which, together with PSTPIP2, contribute to stabilize podosomes and sealing zones in osteoclasts.
PSTPIP1 activity in podosome and sealing zone dissassembly is mediated by proteins interacting with its BAR or SH3 domain. Interestingly, PTPN6 specifically interacts with the PSTPIP1 BAR domain, although this exhibits a 60% homology with the BAR domain of PSTPIP2. In the future, a detailed analysis to map the PTPN6 binding sites on the PSTPIP1 BAR domain could help to better understand this interaction. Nevertheless, PTPN6 localization at sealing zones was strikingly dependent upon the presence of PSTPIP1. PTPN6 controls osteoclast resorption. In other cell types, such as macrophages [29] and neutrophils [30], PTPN6 regulates integrin-mediated adhesion. When PTPN6 function is impaired, these cells are hyper-adhesive and respond to integrin engagement more strongly than wild-type cells. PTPN6 contains two SH2 domains and a phosphatase domain. The N-terminal SH2 domain functions as an auto-inhibitory domain that blocks the catalytic domain in the ligand-free close conformation, in a phosphorylation-dependent manner [31]. In its open conformation, the PTPN6 phosphatase domain might regulate the phosphorylation state and the activity of proteins binding either to the SH2 domain of PTPN6 or to PSTPIP1. Some of these proteins, such as BTK, SYK or myosin-9 and myosin-I, which bind the BAR domain of PSTPIP1, might interact with the PTPN6 SH2 domains. In B-cells, PTPN6 dephosphorylates BTK and SYK thereby decreasing their kinase activity [32]. We have also identified several proteins belonging to the interactome of the PSTPIP1 SH3 domain, such as the tyrosine-protein kinases ABL1/2, the actin nucleation promoting factors like WASP, VASP, WIPF1 or WASL (N-WASP), the E3 ubiquitin protein ligases Cbl,. PTPNs guided by PSTPIP can dephosphorylate WASP [33].
The dynamic instability of sealing zones affects osteoclast function. Our results show that osteoclasts unable to assemble podosomes and sealing zones, as seen after Src [7], PSTPIP2, PTPN12 or PTPN22 depletion, are also unable to digest osteological discs. In contrast, osteoclasts with a reduced ability to disassemble sealing zones, as seen after PSTPIP1, PTPN6, SHIP1 or SHIP2 depletion, exhibit a higher capability to digest such surfaces. This higher capacity in digestion most likely reflects the fact that osteoclasts with reduced sealing zone dynamics remain longer at every given place, which is more efficiently digested. Thus, the dynamic instability of podosomes and sealing zones has a strong impact in osteoclast activity in digestion. In agreement with this, SHIP1 deficient mice are severely osteoporotic due to increased numbers of hyper-resorptive osteoclasts [35]. Mutations that abrogate PSTPIP2 expression in Lupo mice lead to auto-inflammatory disease involving extra-medullary hematopoiesis, skin and bone lesions [36]. In addition, osteoclast precursors purified from these mice exhibit increased osteoclastogenesis. This finding would contradict our prediction that PSTPIP2 deficient mice would exhibit osteopetrosis due a lack of function of osteoclasts in degradation. However, Lupo mice have been created by random chemical mutation [12]. Therefore, conditional knockouts in mouse osteoclasts would help understanding the precise function of PSTPIP2 in bone physiology. We generated mice with a conditionally deleted allele of PSTPIP1 induced by tamoxifen treatment. PSTPIP1-/-primary osteoclasts exhibit a higher resorption activity than control osteoclasts.
In conclusion, we illustrate the functional importance of PSTPIP1/2 in the structural organization of podosomes and sealing zone dynamics in osteoclasts. While providing a wealth of information about PSTPIP1/2 interactors, our studies illustrate the importance of a protein complex comprising PSTPIP1, PTPN6 and SHIP1/2 that links changes in membrane shape and the dynamics of F-actin-rich structures, with a negative feedback mechanism controlling Src and PI(3,4,5)P3 signaling to regulate osteoclast activity. These findings contribute to understanding determinant cell biological aspects of bone physiology that have been elusive so far.  Fig. PTPN12 and PTPN22 localization and effect of their depletion on sealing zone dynamics. A Non treated osteoclasts or osteoclasts treated with siRNAs targeting PSTPIP1 were grown on osteological discs, then fixed and stained with anti PTPN12 or anti PTPN22 antibodies (green) and phalloidin (red) (scale bars: 50μm). Images were analyzed using the Fiji software. Fluorescence intensities across the indicated white lanes are indicated and Pearson's coefficients were calculated (0.56 for PTPN22, 0.49 for PTPN12). B, C Sealing zone dynamics in PTPN12-or PTPN22-depleted osteoclasts. Osteoclasts were treated with siRNAs targeting PTPN12 or PTPN22 and then plated on osteological discs. After 24 hours, they were infected with a recombinant adenovirus encoding the mRFP-Ezrin actin-binding domain. After 32 hours, osteoclasts were observed by time-lapse videomicroscopy (100 msec. per frame, 1 frame per 1 min., see S8-S10 Movies). The knockdown efficacies were determined and quantified by western blotting. The figures presented are representative of at least 3 independent experiments (mean ± SD). (TIF) S5 Fig. Effect of PSTPIP1 and PTPN6 depletion on SHIP1 and SHIP2 tyrosine phosphorylation. Phosphotyrosine specific immunoprecipitation of osteoclast lysates from control and PSTPIP1 or PTPN6 knockdown was done and analyzed by Western blot. Magnitude of these changes was spectrometrically measured in 3 independent experiments. Statistical significance of relative values was tested using students t-test. (mean ± SD, Ã p< 0.05; ÃÃ p<0.01).  Table. PSTPIP2 interactors. Recombinant GST-PSTPIP2 was incubated with osteoclast lysates. Bound proteins were isolated on Glutathione beads, resolved by SDS-PAGE and identified by semi-quantitative mass spectrometry analysis based on MS2 spectral counting. Protein names, gene names, and accession numbers are indicated. MS2 spectral counts are represented in the last column. (DOC)