ARAP1 Bridges Actin Dynamics and AP-3-Dependent Membrane Traffic in Bone-Digesting Osteoclasts.

Bone-resorbing osteoclasts play a central role in bone remodeling and its pathology. To digest bone, osteoclasts re-organize both F-actin, to assemble podosomes/sealing zones, and membrane traffic, to form bone-facing ruffled borders enriched in lysosomal membrane proteins. It remains elusive how these processes are coordinated. Here, we show that ARAP1 (ArfGAP with RhoGAP domain, ankyrin repeat and PH domain-containing protein 1) fulfills this function. At podosomes/sealing zones, ARAP1 is part of a protein complex where its RhoGAP domain regulates actin dynamics. At endosomes, ARAP1 interacts with AP-3 adaptor complexes where its Arf-GAP domain regulates the Arf1-dependent AP-3 binding to membranes and, consequently lysosomal membrane protein transport to ruffled borders. Accordingly, ARAP1 or AP-3 depletion in osteoclasts alters their capacity to digest bone in vitro. and AP-3δ-deficient mocha mice, a model of the Hermansky-Pudlak storage pool syndrome, develop osteoporosis. Thus, ARAP1 bridges F-actin and membrane dynamics in osteoclasts for proper bone homeostasis.

Introduction

Bone remodeling is the continuous process of bone digestion and rebuilding occurring throughout life. It is essential for the proper development, maintenance, and repair of the skeleton of vertebrates and is also central in bone pathology. It involves the coordinated activity of bone-building osteoblasts and bone-digesting osteoclasts to maintain bone mass. Osteoclasts originate from hematopoietic, mononucleated precursors, which upon their activation by the cytokines, macrophage-stimulating factor, and receptor activator of nuclear factor-κB ligand (RANKL), differentiate and fuse with each other to generate multinucleated osteoclasts (Teitelbaum, 2000). To digest the bone matrix, osteoclasts attach onto the bone surface and condense their podosomes, structural units linking cell adhesion molecules and actin meshworks, into podosomal belts to form a sealing zone. This sealing zone segregates the highly convoluted, bone-facing membrane, also known as ruffled border, from other membrane domains. This ruffled border and the bone surface limit the extracellular resorption lacunae, which are acidified to dissolve the inorganic bone matrix and are rich in secreted hydrolytic, lysosomal enzymes to digest the organic bone matrix.

Sealing zones and ruffled borders are highly dynamic structures. They are formed when osteoclasts attach onto bone surfaces and are then disassembled when osteoclasts move to other bone areas to digest. The ruffled border is highly enriched in lysosomal membrane proteins such as LAMPI, LAMPII, V-ATPase, and chloride channels, which are found normally in late endocytic compartments of other cell types (Palokangas et al., 1997, Akamine et al., 1993, Vaananen et al., 1990), and Rab7, a small GTPase, which normally regulates membrane traffic from early to late endocytic compartments (Palokangas et al., 1997, Zhao et al., 2002). The intracellular transport of newly synthesized membrane proteins from the biosynthetic pathway to the endosomal system relies on the hetero-tetrameric, AP-3 adaptor complex, one of the coat components regulating protein sorting in eukaryotes (Dell'Angelica et al., 1997). AP-3 interaction with its target membranes is tightly regulated by the Arf-1 GTPase (Ooi et al., 1998) and its corresponding guanine nucleotide exchange factors (GEFs) or GTPase-activating proteins (GAPs). How lysosomal membrane proteins are targeted to the ruffled border of osteoclasts is unknown, although a current view is that in osteoclasts secretory lysosomes can fuse with the plasma membrane to deliver their membrane components and discharge their hydrolytic, lysosomal enzymes into the resorption lacunae (Mulari et al., 2003, van Meel et al., 2011). However, other mechanisms may exist, for example, the endosomal recycling pathway that contributes to the efficient delivery to the ruffled border (Palokangas et al., 1997, Steenblock et al., 2014).

The formation of sealing zones is instrumental in ruffled border biogenesis. They result from the condensation of individual podosomal units consisting of F-actin cores linked to cell adhesion molecules, thereby forming large protein complexes containing paxillin, gelsolin, talin, filamin A, integrins, and many other proteins (Linder and Aepfelbacher, 2003). Podosome/sealing zone assembly and disassembly is regulated by GTPases of the Rho family (Buccione et al., 2004), in particular Cdc42, RhoA, and Rac (Heckel et al., 2009, Saltel et al., 2004, Chellaiah et al., 2000, Chellaiah, 2005). Their dynamics is also regulated by the Src kinase-dependent phosphorylation and activation of several podosomal components (Soriano et al., 1991), such as the Arf-6 GAP GIT2 (Heckel et al., 2009), the Cdc42 guanine exchange factor FGD6 (Steenblock et al., 2014), or the BAR-domain scaffolding protein proline-serine-threonine phosphatase-interacting protein (PSTPIP) 1 (Sztacho et al., 2016), which, as part of a protein complex with the tyrosine-protein phosphatase non-receptor type 6 and the phosphatidylinositol-3,4,5-trisphosphate 5-phosphatase 1/2, controls the local dephosphorylation of membrane phospholipids and Src-phosphorylated substrates at podosomes, thereby triggering their disassembly (Sztacho et al., 2016, Heckel et al., 2009).

How sealing zone and ruffled border assemblies are coordinated is largely unknown. Here, we illustrate the functional importance of the PSTPIP1/2 interactor ARAP1, an ArfGAP with RhoGAP domain, ankyrin repeat, and PH domain protein 1, in these processes. In different cell types, ARAP1 associates with various intracellular compartments and actin-related structures such as filopodia (Miura et al., 2002). In NIH 3T3 cells, ARAP1 regulates through Arf1/5 circular dorsal ruffles, which share many characteristics with podosomes (Hasegawa et al., 2012, Buccione et al., 2004). We now show that in osteoclasts ARAP1 regulates sealing zone assembly, most likely through its Rho GAP domain, and Arf1-dependent AP-3 binding to endosomal membranes, most likely through its ArfGAP domain, for delivery of lysosomal membrane proteins to the ruffled border. Thus, our study shows that ARAP1 coordinates sealing zone dynamics and ruffled border biogenesis to modulate bone resorption.

Results

ARAP1 Expression during Osteoclastogenesis

We first assessed by western blotting the expression level of ARAP1 during the RANKL-induced differentiation of RAW264.7 macrophage-like cells into mature osteoclasts. We used an antibody recognizing the C-terminal part of murine ARAP1 conserved in all ARAP1 isoforms (Figure 1A). In RAW264.7 cells, this antibody recognized one major band corresponding to the size of isoform 3 (Figure 1B). Two smaller bands were detectable but only after long exposure times. They could be assigned to ARAP1 isoforms 1 or 4 as they disappeared upon ARAP1 knockdown (Figure S1A). ARAP1 isoform 2 was not detected by western blotting or qPCR using different primers designed for the specific 5′ UTR (data not shown). ARAP1 expression increased during the differentiation of mononucleated RAW264.7 cells with RANKL, reaching an ∼10 fold increase at day 4 after RANKL treatment, when mature, multinucleated osteoclasts were observed (Figure 1C). This suggested that ARAP1 could play an important role in osteoclasts as seen for several other proteins strongly up-regulated during osteoclastogenesis such as the Src tyrosine kinase, the tartrate-resistant acid phosphatase TRAP, or the protease cathepsin K (Czupalla et al., 2005).

Figure 1

Expression of ARAP1 during Osteoclastogenesis

(A) Transcript variants of validated murine ARAP1 isoforms. The different domains are indicated. The target sequence of the stealth siRNA and AB, the epitope recognized by the anti-murine ARAP1 antibody, are situated at the C terminus of all isoforms.

(B) Immunoblotting of ARAP1 in osteoclasts.

(C) ARAP1 expression in osteoclasts and its precursors. RAW264.7 cells were grown on plastic in the absence or presence of RANKL and harvested at the indicated times. Undifferentiated RAW264.7 cells (“Day 0”) were grown without RANKL, and cell extracts were prepared. Protein expression was determined by western blotting and quantified using the Fiji software. GAPDH was used as a control. The figures presented are representative of at least 3 independent experiments (mean ± SD; **p < 0.01, ***p < 0.001).

See also Figure S1.

ARAP1 Interactome

To understand ARAP1 function in osteoclasts, we first identified ARAP1 binding partners. We immunoprecipitated with an anti-ARAP1 antibody the endogenous ARAP1 from an osteoclast lysate and identified its binding partners by semi-quantitative mass spectrometry. We identified 1,434 proteins with a threshold of 95%, of which ∼150 were at least 2-fold enriched after immunoprecipitation (Table 1). These putative ARAP1 interactors first included the CIN85 (Cbl-interacting protein of 85 kDa) protein (also known as SH3 domain-containing kinase-binding protein 1 isoform 1), previously identified as an ARAP1 interactor by the yeast two-hybrid system and liquid chromatography-tandem mass spectrometry (LC-MS/MS) (Havrylov et al., 2009, Yoon et al., 2011). Second, several components of actin-rich podosomes were identified, e.g., subunits of the actin-related protein 2/3 complex, gelsolin, filamin-A, the F-actin capping protein complex, and the BAR-domain containing PSTPIP1/2, two key regulators of podosomes/sealing zones dynamics in osteoclasts (Sztacho et al., 2016). Thus, this analysis confirmed our previous studies showing that PSTPIP1 binds ARAP1 (Sztacho et al., 2016). Finally, the MS analysis of ARAP1 interactors identified several components involved in membrane traffic. This included the vacuolar protein sorting-associated protein 13 (VPS13), whose function is still poorly understood, and the four subunits of the adaptor complex 3, a coat component involved in lysosomal targeting of membrane proteins such as LAMP1 or LAMP2 (Peden et al., 2004). These findings also confirmed our previous studies showing that ARAP1 is part of a protein network selectively recruited together with AP-3 onto membranes (Baust et al., 2008). Collectively, our analysis revealed that ARAP1 could regulate both actin dynamics and AP-3-dependent membrane traffic in osteoclasts.

Table 1

Putative ARAP1 Interactors Identified by Quantitative LC-MS/MS

Description	Alternative Names	Protein Molecular Weight (Da)	Exclusive Unique Peptide Count	Exclusive Unique Spectrum Count	Percentage Sequence Coverage (%)
SH3 domain-containing kinase-binding protein 1 isoform 1	CIN85	73,330.3	9	10	23.20%
		73,330.3	25	29	57.60%
F-actin-capping protein subunit beta isoform d	F-actin-capping protein	29,296.1	13	17	63.80%
		29,296.1	11	12	47.70%
F-actin-capping protein subunit alpha-1	F-actin-capping protein	32,751.7	8	9	51.00%
		32,751.7	7	8	54.50%
F-actin-capping protein subunit alpha-2	F-actin-capping protein	32,967.2	12	12	46.90%
		32,967.2	8	9	50.30%
mCG21234; isoform CRA_b	Filamin-α	280,463.8	46	48	26.90%
		280,463.8	2	2	1.21%
Gelsolin isoform X2	Gelsolin	88,212.5	3	3	5.24%
		88,212.5	13	13	22.80%
CD2-associated protein	METS-1	70,434.6	4	4	8.16%
		70,434.6	22	23	43.20%
Actin-related protein 3	ARP3	47,357.7	20	25	64.10%
		47,357.7	3	3	88.50%
Actin-related protein 2/3 complex subunit 3	ARP2/3 subunit3	20,525.9	2	2	22.50%
Actin-related protein 2/3 complex subunit 1B	ARP2/3 subunit 1B	41,063.6	12	15	39.80%
		41,063.6	2	2	6.18%
Actin-related protein 2/3 complex subunit 2	ARP2/3 subunit 3	34,357.8	11	14	43.00%
		34,357.8	2	2	7.33%
Proline-serine-threonine phosphatase-interacting protein 1	PSTPIP1	47,576.4	5	5	17.30%
		47,576.4	5	5	16.40%
Pstpip2 protein	PSTPIP2	39,020.7	4	4	22.20%
AP-3 complex subunit mu-1	AP-3μ1	46,937	16	26	53.60%
		46,937	14	16	58.10%
Adaptor-related protein complex 3; sigma 2 subunit	AP-3σ2	22,017.8	5	6	27.50%
		22,017.8	2	2	98.54%
Adaptor-related protein complex 3; beta 1 subunit; isoform CRA_b	AP-3β1	122,872.4	23	26	25.70%
		122,872.4	17	17	20.40%
AP-3 complex subunit delta-1	AP-3ɗ1	135,085.1	11	11	13.50%
		135,085.1	25	28	26.70%
Vacuolar protein sorting-associated protein 13C	VPS13C	417,316.3	50	52	19.60%
		417,316.3	24	24	8.64%
Vacuolar protein sorting-associated protein 13A	Chorein, VPS13A	359,406	3	3	1.52%
		359,406	2	2	0.85%

Anti-ARAP1-IgG-magnetic beads were mixed with an osteoclast lysate to purify endogenous ARAP1 and its interactors. After washing, the bound proteins were identified by LC-MS/MS. In this table putative interactors identified were illustrated.

ARAP1 Localization

Previous studies indicated a widespread localization of ARAP1 on different compartments of the biosynthetic and endocytic pathways as well as circular dorsal ruffles of cultured cells (Daniele et al., 2008, Miura et al., 2002, Hasegawa et al., 2012). In mature osteoclasts grown on glass or apatite-collagen-coated surfaces, endogenous ARAP1 localized to actin-rich podosomes and sealing zones as well as AP-3-coated compartments (Figure 2). The co-localization of ARAP1 with AP-3 was highly specific as indicated by the Pearson correlation of ∼0.6 for both endogenous and overexpressed ARAP1 (Figure 2G). AP-3-positive structures also showed strong co-localization with an early endosome marker (EEA1) or VPS35, a subunit of the retromer complex found on early endosomes, but to a lesser extent with Rab7a, a late endosomal maker, and GM-130, a Golgi marker (Figures 2H and S2) (Dell'Angelica et al., 1997, Le Borgne et al., 1998). Thus, part of ARAP1 is present on AP-3-coated early endosomes, which also contain a large amount of the lysosomal membrane protein LAMP1 in osteoclasts (Figures 2B, 2F, and S2). ARAP1 localization would therefore be consistent with our MS-based proteomic analysis of ARAP1 interactors.

Figure 2

Localization of Endogenous and GFP-Tagged ARAP1 in Osteoclasts

(A–E) Osteoclasts were grown on glass or apatite-collagen-coated surfaces (ACCs), fixed, and stained for (A, B, D, and E) ARAP1 (green) and AP-3 (magenta) actin (red) and DAPI (blue) and (C) AP-3s1 (blue), actin (magenta), ARAP1 (red), and LAMPI (green). White arrows are pointing towards AP-3 positive endosomes. Outlines of single osteoclasts are depicted by white circles in the actin channel. Scale bar: 10 μm.

(F) Enlargement of rectangle in B. Fluorescence intensities across the indicated white lines (at sealing zones) were plotted. Images were analyzed using the Fiji software. Scale bar: 10 μm.

(G) Enlargement of rectangle in E. Fluorescence intensities across the white lines (at sealing zones) as indicated were plotted. Images were analyzed using the Fiji software. Scale bar: 10 μm.

(H) Descriptive statistics to identify ARAP1/AP-3-positive endosomes. The global Pearson correlation coefficients for endogenous and overexpressed ARAP1, AP-3ɗ, EEA1, Rab7a, LAMPI, GM-130, and VPS35 were determined using the Volocity software with automated thresholding (Costes et al., 2004). (n = 3, mean ± SD).

See also Figure S2.

We also expressed a GFP-tagged ARAP1 isoform 3 in osteoclasts using a recombinant adenovirus. Expressed GFP-ARAP1 isoform 3 also localized to podosomes and sealing zones as well as to AP3-coated structures (Figures 2D and 2E). Interestingly, osteoclasts overexpressing GFP-ARAP1 isoform 3 had remarkably enlarged vesicles coated with AP-3 (Figures 2D, 2E, and S3). This phenomenon could be especially appreciated on glass, where osteoclasts are flat and large (Figure S3). The influence of ARAP1 overexpression on the morphology of AP-3-positive structures suggests that ARAP1 may regulate AP-3-dependent trafficking and/or endosome maturation.

ARAP1 Regulates Sealing Zone Dynamics and the Resorption Activity of Osteoclasts

We first followed by time-lapse video microscopy ARAP1 dynamics at individual podosomes using total internal reflection fluorescence microscopy and sealing zones labeled with the monomeric red fluorescent protein (mRFP)-tagged actin-binding domain of ezrin (mRFP-ABDE) expressed with a recombinant adenovirus. Figure S1 shows that ARAP1 dynamics follows that of actin at podosomal cores (Figures S1B and S1C and Video S1) and sealing zones (Video S2), indicating that ARAP1 remains associated with these actin-rich structures during their lifetime. We then performed small interfering RNA (siRNA)-mediated knockdown of ARAP1 isoforms using siRNAs targeting the C-terminal end of ARAP1 conserved in all murine transcript variants. Upon ARAP1 knockdown (90% efficiency), the podosomal belt normally observed at the periphery of osteoclasts grown on glass was disrupted. Podosomes had an abnormal podosome organization as they appeared more dispersed forming arbitrary clouds and rings (Figures 3A and 3D). When grown on apatite-collagen-coated surfaces, the osteoclasts assemble classical sealing zones, which are dynamic structures as observed by time-lapse video microscopy (Figure 3, Video S3). Upon ARAP1 depletion, siRNA-treated osteoclasts could only assemble very small sealing zone remnants (Figures 4A and 4B) whose dynamics was drastically reduced (Figure 4B and Video S4). These sealing zone remnants remained very static, unable to further assemble or disassemble (Figure 4B). Moreover, ARAP1-depleted osteoclasts were unable to resorb collagen/apatite-coated surfaces (Figure 4C). Remarkably, this phenotype could be rescued upon re-expression of the GFP-ARAP1 isoform 2, which lacks the ArfGAP domain (Figures 3B and 3C). Collectively, these results demonstrate that ARAP1 is a key component of podosomes, which regulates their assembly and dynamics to form sealing zones required for efficient degradation of bone-mimicking surfaces. Furthermore, our results indicate that only the RhoGAP domain of ARAP1 may regulate podosome/sealing dynamics.

Figure 3

ARAP1 Affects Podosomal Belt Organization

(A) Effect of ARAP1 and AP-3 knockdown on podosomes in overview (scale bar: 100 μm) and detail (scale bar: 10 μm). (A) ARAP1 and AP-3 were depleted in osteoclast using stealth siRNA as described. Cells were fixed and stained on glass coverslips for DAPI (blue) and phalloidin (red). The outline of single osteoclasts were marked in white. Scale bar in overview: 100 μm; scale bar in detail: 10 μm.

(B) ARAP1 knockdown cells were subsequently transfected with GFP-tagged ARAP1 isoform 2 (green) using a recombinant adenovirus. Fluorescence intensities across the white lines (at podosomes) as indicated were plotted. Scale bar in overview: 100 μm; scale bar in detail: 10 μm.

(C) The number of osteoclasts in A and B with an intact and altered (not intact) podosomal belt were counted on an 11-mm glass coverslip and plotted. The intactness of the podosomal belt was judged based on the presence of a clear belt on the periphery of an osteoclast. If osteoclasts had dispersed podosomes or arbitrary rings that were phalloidin-positive within the cell, they were judged as non-intact. (n = 3, mean ± SD).

(D) The knockdown efficiencies were assessed by immunoblotting (ARAP1) or by qPCR (AP-3) and plotted. (mean ± SD).

Figure 4

ARAP1 and AP-3 Depletion Affects Osteoclast Resorption

(A) Effect of siRNA-mediated ARAP1 depletion on the podosomal belt and sealing zones. Osteoclasts were treated with siRNA targeting the ARAP1 genes or mock siRNA and then grown for 48 hr on osteological disks as indicated in Methods. Cells were then fixed and stained with phalloidin (red) and DAPI (blue). (Scale bar: 10 μm). The knockdown efficiency was determined by western blot. The number of osteoclasts with intact and altered (not intact) podosomal belts was counted on an 11-mm glass coverslip and plotted.

(B) Effect of siRNA-mediated ARAP1 and AP-3μ1 depletion on sealing zone dynamics. Osteoclasts were treated with siRNAs targeting the corresponding genes or mock siRNA and then plated on apatite-collagen-coated (ACC) multiwell dishes. After 24 hr, they were infected with a recombinant adenovirus encoding the mRFP-ezrin actin-binding domain. After 32 hr, osteoclasts were imaged by time-lapse microscopy (200 ms per frame, 1 frame per minute for 30 min, scale bar: 10 μm). See also Videos S3, S4, and S5. Sealing zone diameter was measured using the Fiji software. The relative sealing zone diameter was plotted for each sealing zone assessed. The change of the sealing zone diameter per minute was plotted and tested using Student's t test. (n = 3, mean ± SD; ****p < 0.0001) The size of the sealing zones above 20 μm in control and siARAP1- and siAP-3-treated osteoclasts was counted on an 11-mm ACC-coated glass coverslip (n = 3, mean ± SD; **p < 0.01). The knockdown efficacy of AP-3μ1 was determined with qPCR.

(C) Effect of siRNA-mediated ARAP1 and AP-3μ1 depletion on resorption. The resorptive capacity of osteoclasts treated with siARAP1 or mock siRNA was measured using Corning Osteo Assay Surfaces. The resorbed area from siARAP1 and siAP-3μ1 osteoclasts on 24-well Osteo Assay dishes was plotted using Student's t test. (n = 3, mean ± SD; ****p < 0.0001).

Video S1. ARAP1 at Individual Podosomes, Related to Figure 2

Dynamics of ARAP1 at podosomes visualized with the GFP-ARAP1 and with the mRFP-ezrin actin-binding domain in osteoclasts grown on glass coverslips shown in Figure S1. (100 ms per frame, 1 frame per 2 s).

Video S2. ARAP1 at Dynamic Sealing Zones, Related to Figure 2

Dynamics of ARAP1 at sealing zones visualized with the GFP-ARAP1 and with the mRFP-ezrin actin-binding domain in osteoclasts grown on apatite-collagen-coated (ACC) surfaces shown in Figure S1 (200 ms per frame, 1 frame per minute for 30 min).

Video S3. Sealing Zone Dynamics of Control Osteoclasts, Related to Figure 4

Video corresponding to Figure 4B. The dynamics of sealing zones visualized with mRFP-ezrin actin-binding domain in osteoclasts grown on apatite-collagen-coated (ACC) surfaces.

Video S4. Effect of ARAP1 Depletion on Sealing Zones, Related to Figure 4

Effect of siRNA-mediated ARAP1 depletion on sealing zone dynamics illustrated by time-lapse video microscopy shown in Figure 4B. The dynamics of sealing zones visualized with mRFP-ezrin actin-binding domain in osteoclasts grown on apatite-collagen-coated (ACC) surfaces.

Video S5. Effect of AP-3 Depletion on Sealing Zones, Related to Figure 4

Effect of siRNA-mediated AP-3μ1 depletion on sealing zone dynamics illustrated by time-lapse video microscopy shown in Figure 4B. The dynamics of sealing zones visualized with mRFP-ezrin actin-binding domain in osteoclasts grown on apatite-collagen-coated (ACC) surfaces.

ARAP1 Regulates AP-3 Coat Binding to Membranes and Trafficking of Lysosomal Membrane Proteins to the Ruffled Border

Our study shows that ARAP1 interacts with the AP-3 complex, a coat component required for the proper targeting of selected lysosomal membrane proteins to the endocytic system of every cell type. The Arf1 GTPase regulates AP-3 binding to membranes of early endosomes (Ooi et al., 1998). Therefore, it could be postulated that ARAP1 regulates Arf1 binding to membranes and consequently AP-3 recruitment onto membranes and the trafficking of lysosomal membrane proteins in osteoclasts. To test this hypothesis, we used an siRNA-based approach to inactivate ARAP1 and monitored the interaction of AP-3 with early endosomal membranes and the targeting of lysosomal membrane proteins in mature osteoclasts.

We first determined the amount of AP-3 bound to membranes by simply measuring the fluorescence intensities after labeling osteoclasts with anti-AP-3 antibodies. Remarkably, the knockdown of ARAP1 induced a 3-fold increase in the fluorescence intensity of membrane-bound AP-3 when compared with mock-treated osteoclasts (Figures 5A, 5B, and 5G). Moreover, the lysosomal membrane protein LAMPI and Rab7a, which are markers of the ruffled border of osteoclasts grown on collagen-apatite-coated surfaces (Figures 5C, 5D, and S5) were barely detectable at the inner part of the sealing zone remnants after ARAP1 knockdown, indicating the absence of a ruffled border. These results suggest that in the absence of ARAP1 more AP-3 binds to early endosomal membranes and retains lysosomal membrane proteins in the endocytic system.

Figure 5

ARAP1 Regulates the Association of AP-3 with Membranes and LAMPI Trafficking to the Ruffled Border

(A–F) Effect of siRNA-mediated depletion of ARAP1 or AP-3μ1 on AP-3 and LAMPI fluorescent signals. Osteoclasts treated with siRNAs targeting ARAP1 or AP-3μ1 or mock siRNA were grown on glass or apatite-collagen-coated (ACC) coverslips, fixed, and stained with DAPI and phalloidin and with anti-ARAP1 or anti-AP-3s1 antibodies or anti-LAMP1 antibodies. (scale bar: 10 μm).

(G) AP-3s1 fluorescence intensities in siARAP1-treated osteoclasts were measured from individual stacks and relative values were plotted (n = 3, mean ± SD; ****p < 0.0001).

(H) The relative LAMPI fluorescence intensity at the ruffled border in AP-3μ1 knockdown (KD) was measured and plotted (n = 3, mean ± SD; Student's t test, ****p < 0.0001). The large error bars can be explained by variations in the KD efficiency in individual osteoclasts.

(I) Effect of siRNA-mediated depletion of ARAP1 and AP-3μ1 depletion on steady-state LAMPI. Osteoclasts treated with siARAP1, siAP-3μ1, or mock siRNA were grown on plastic, harvested, and immunoblotted with anti-LAMPI, ARAP1, and GAPDH antibodies. AP-3 knockdown efficiency was determined by qPCR. The intensity of LAMPI in immunoblotting was analyzed using the Fiji software and plotted.

Our previous studies have shown that AP-3-dependent cargos like the lysosomal membrane protein LAMP1 are directed to the cell surface in the absence of AP-3, which results in a longer half-life of these proteins and their apparent accumulation in cells (Baust et al., 2008, Traikov et al., 2014). Therefore, we depleted osteoclasts from AP-3 using siRNAs targeting its μ subunit. When compared with control osteoclasts, AP-3μ-depleted osteoclasts were strongly enriched in LAMPI and LAMPII (a 4- and 2-fold increase, respectively) at the ruffled border (Figures 5H and S4). In addition, we found by western blotting more LAMPI in osteoclasts depleted of AP-3μ or ARAP1 than in control osteoclasts (Figure 5I). Thus, the ruffled border of AP-3-depleted osteoclasts is enriched in AP-3-dependent cargos, although AP-3μ knockdown had per se no effect on the organization of podosomes and sealing zones (Figures 4A and 4B). In addition, AP-3-depleted osteoclasts grown on collagen-apatite-coated surfaces exhibited a higher capacity to resorb the bone-mimicking surfaces that control osteoclasts (Figure 4D) Thus, AP-3, like ARAP1, controls the membrane composition of the ruffled border of osteoclasts and their capacity of digesting bone.

AP-3-Deficient Mice Are Osteoporotic

The results described above strongly suggest that AP-3-deficient mice should develop osteoporosis due to a higher activity of their osteoclasts to digest bone. To test this hypothesis, we examine the bone phenotype of mocha mice, a mouse model for the human Hermansky-Pudlak storage pool syndrome, which lack a functional AP-3 coat due to mutations in their δ subunit (Kantheti et al., 1998). Homozygous AP-3-depleted mice are extremely difficult to obtain because they die during early development and also during embryogenesis. Therefore, we decided to examine wild-type, heterozygous and homozygous animals from the same litters. We could collect 4 homozygous, 8 wild-type, and 11 heterozygous animals aged between 4 and 6 weeks. Tibias from these animals were collected and analyzed by micro X-ray computed tomography for trabecular bone volume, trabecular separation, trabecular number, and thickness. All these bone parameters were significantly lower in heterozygous and homozygous animals when compared with the wild-type, whereas trabecular separation significantly increased (Figures 6A–6D). This osteoporotic phenotype is in agreement with our in vitro bone digestion studies (Figure 4C) and further illustrates the importance of AP-3 in osteoclast function (Figure 6E).

Figure 6

Bone Phenotype of Mocha Mice

(A–E) Mocha mice aged 4–6 weeks were analyzed for trabecular bone parameters at their right tibias. (A–D) Bone volume, trabecular thickness (Tb.Th.), trabecular separation (Tb.Sp.), and trabecular number were analyzed with the CTan and CTvol software after microCT. Statistical analysis was done in the GraphPad Prism 7 Software using ordinary one-way ANOVA (n = 8 [WT], 11 [HET], and 4 [HOM]; mean ± SD; *p < 0.05, **p < 0.01, ***p < 0.001). (E) 3D projections of the trabecular bone of mocha mice.

Discussion

Our study illustrates the functional importance of ARAP1 as a bridging factor coordinating podosome/sealing zone assembly and ruffled border biogenesis in osteoclasts. The ARAP1 Rho GAP domain may control Rho signaling for actin and sealing zone dynamics. ARAP1, most likely through its ArfGAP domain, controls the Arf-1-dependent association of AP-3 with membranes and consequently the delivery of lysosomal membrane proteins to the ruffled border. Alteration of any of these two coordinated processes affects the capacity of osteoclasts to digest bone.

ARAP1 is present at podosomes and sealing zones where it regulates their dynamics. How is ARAP1 recruited at this location? First, ARAP1 contains five PH domains able to bind to several phosphatidylinositol phosphates (PtdIns) (Campa et al., 2009). The first and third PH domains fit the consensus for binding phosphatidylinositol-3,4,5-trisphosphate [PtdIns(3,4,5)P3] (Daniele et al., 2008, Miura et al., 2002), a phospholipid whose synthesis is activated at the plasma membrane upon integrin binding to extracellular, RGD motif-containing proteins (Yu et al., 2013). In addition, PtdIns(3,4,5)P3 binding to the PH domain stimulates ARAP1 GAP activity (Campa et al., 2009). However, these PH domains exhibit a low affinity for PtdIns(3,4,5)P3. Therefore ARAP1 at podosomes must be stabilized by interactions with other podosomal components. Our quantitative proteomics analysis of the ARAP1 interactome clearly shows that ARAP1 interacts with the scaffolding BAR-domain PSTPIP1/2, which regulates the assembly/disassembly of podosomes and sealing zones. This correlates well with our previous studies showing that ARAP1 is found in the PSTPIP1/2 interactome (Sztacho et al., 2016). PSTPIP1 interacts with the phosphotyrosine protein phosphatase PTPN6 through its BAR-domain and the phosphatidylinositol 5-phosphatase SHIP1/2 (Sztacho et al., 2016) through its SH3 domain. PTPN6 may also dephosphorylate other proteins that are in proximity to the membrane, such as Syk (Maeda et al., 1999). SHIP1 is specific for hematopoietic cells and keeps the levels of phosphatidylinositol (3,4,5)-trisphosphate (PI(3,4,5)P3) low, a regulation that is involved in podosome disassembly (Muraille et al., 1999). Although not identified in the proteomic approach for ARAP1 interactors represented here, ARAP1 may also bind to SHIP1/2 via its SAM domain, similar to the ARAP3 isoform SAM domain, which binds to SHIP2 (Raaijmakers et al., 2007). We also identify as an ARAP1 interactor CIN85, also known as a SH3KBP1 (SH3 domain kinase-binding protein 1); Ruk (regulator of ubiquitous kinase); and SETA (SH3 domain-containing gene expressed in tumorigenic astrocytes) (Narita et al., 2001, Gout et al., 2000, Bogler et al., 2000, Borinstein et al., 2000, Dikic, 2002). CIN85 also interacts with SHIP1 in B cells (Havrylov et al., 2009, Buchse et al., 2011). We therefore propose that ARAP1 is part of a large protein complex composed of PSTPIP1, PTPN6, SHIP1/2, and CIN85. This complex functions by combining local dephosphorylation of PI(3,4,5)P3 lipids through SHIP1/2, dephosphorylation of Src-dependent podosomal substrates through PTPN6, and down-regulation of Rho signaling through ARAP1 to control podosome and sealing zone dynamics.

Our study strongly suggests that a down-regulation of Rho signaling is essential for maintaining the highly dynamic nature of podosomes and sealing zones. ARAP1 knockdown resulting in a higher Rho signaling disrupts podosomal belt organization and reduces the size and dynamics of sealing zones, thereby resulting in osteoclasts unable to resorb bone-mimicking surfaces. Re-expression of an ARAP1 isoform containing a Rho GAP domain but lacking the Arf domain is sufficient to restore podosome and sealing zone dynamics. We could not directly demonstrate that the Rho GAP activity is necessary for this function. This would have required the re-expression of an ARAP1 mutant impaired in its Rho GAP activity in osteoclasts depleted of endogenous ARAP1. Such experiments, as described here with ARAP1 isoform 2 (Figures 3B and 3C), are technically challenging due to the adenovirus-based re-expression system, which renders difficult the re-expression of large size proteins like ARAP1. Potential targets of the ARAP1 GAP activity are RhoA, Rac1, or Cdc42 (Miura et al., 2002, Destaing et al., 2003). RhoA signaling is required for osteoclast podosome organization, motility, and bone osteoclasts by activating integrin-dependent activation of phosphoinositide synthesis (Chellaiah et al., 2000). Cdc42 signaling is also necessary since its GEF FGD6 is required for podosome formation through the assembly of complexes comprising the Cdc42 interactor IQGAP1, the Rho GAP ARHGAP10, and the integrin interactors talin-1/2 or filamin A (Steenblock et al., 2014). Similarly, the Rac1 guanine exchange factor FARP2 is required for localized activation of GTP-bound Rac1 into podosome-ring-like structures, where it controls integrin β3 activity during osteoclastogenesis (Takegahara et al., 2010). Therefore, more work will be needed to identify the Rho GTPase down-regulated by ARAP1 and understand its precise role in controlling Rho signaling at podosomes and sealing zones.

ARAP1 is also present on early endosomes where it regulates the AP-3-dependent traffic of lysosomal membrane proteins to the ruffled border. ARAP1 interaction with early endosomes is first due to the presence of the PH 1, 2, 4, or 5 domain, which binds PtdIns(3)P (Daniele et al., 2008), a phospholipid enriched at the cytoplasmic surface of early endosomes (Peden et al., 2004), and second due to its interaction with the AP-3 complex, which is mostly present on this compartment. In vitro, the ARAP1 ArfGAP domain preferentially hydrolyzes Arf1-GTP (Miura et al., 2002, Ooi et al., 1998), which is required for efficient AP-3 binding to PtdIns(3)P-rich membranes (Nie et al., 2003). Accordingly, ARAP1--depleted osteoclasts exhibit a higher amount of AP-3 bound to their membranes, and the overexpression of ARAP1 in osteoclasts resulted in enlarged AP-3-coated endosomes, suggesting that endosome maturation is impaired. Thus, ARAP1, like AGAP1 binding to AP-3 δ and σ subunit (Nie et al., 2003), is one of the ArfGAPs regulating the Arf-1-dependent binding of AP-3 to early endosomes. We could not directly demonstrate that ArfGAP activity of ARAP1 is essential for this function for the same reasons as explained above for its Rho GAP activity. We have previously shown that in HeLa cells AP-3 mediates the intracellular targeting of lysosomal membrane proteins to late endocytic compartments and that in its absence lysosomal membrane proteins are allowed to have access to the cell surface (Traikov et al., 2014, Le Borgne et al., 1998). In osteoclasts, the ruffled border is a highly convoluted membrane enriched in lysosomal membrane proteins. We show here that lysosomal membrane proteins such as LAMP1 and LAMPII are more enriched at the ruffled border and early endosomes after AP-3 depletion. In contrast, we found no change in the secretion of β-hexosaminidase in ARAP1-depleted osteoclasts, a classical lysosomal enzyme whose targeting is AP-3 independent (data not shown). It has been proposed that the ruffled border of osteoclasts results from the fusion of intracellular acidic vesicles, most likely lysosomes, with the ruffled border (Mulari et al., 2003). We propose that the regulation of AP-3 binding onto early endosomes by ARAP1 would modulate the access of newly synthesized lysosomal membrane proteins to the ruffled border of osteoclasts.

What is the consequence of modulating the protein content of the ruffled border? Clearly, our study shows that the activity of osteoclasts in bone resorption is affected. ARAP1-depleted osteoclasts are impaired in bone-mimicking surface digestion, most likely due to the lack of formation of a proper sealing zone and ruffled border. Strikingly, AP-3-depleted osteoclasts, which can still form normal sealing zones, are more active in surface resorption that controls osteoclasts. The most likely explanation is that the ruffled border is, as we show here, enriched in lysosomal membrane proteins and probably other proteins involved in bone digestion like the V-ATPase and chloride channels, making the resorption lacuna more efficient in degradation. Accordingly, AP-3δ-deficient mocha mice develop severe osteoporosis already at a relatively early age. Mocha mice provide a model of Hermansky-Pudlack storage pool syndrome. Hence, these mice are characterized by defective platelets, coat, and eye color as a consequence of defective sorting of AP-3-dependent cargo proteins to lysosomes and lysosome-related organelles of specialized cells such as melanosomes, platelet-dense granules, and neurotransmitter vesicles (Kantheti et al., 1998). Thus, our results would strongly suggest that patients with Hermansky-Pudlack storage pool syndrome also suffer from osteoporosis.

ARAP1 KO mice generated as part of the International Mouse Phenotyping Consortium show a bone phenotype as they have abnormally short tibias, thus indicating a key function of ARAP1 in bone biology (www.impc.org).

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.