An SNX10-dependent mechanism down-regulates fusion between mature osteoclasts

Homozygosity for the R51Q mutation in sorting nexin 10 (SNX10) inactivates osteoclasts (OCLs) and induces autosomal recessive osteopetrosis in humans and in mice. We show here that fusion of wild-type murine monocytes to form OCLs is highly regulated, and that its extent is limited by blocking fusion between mature OCLs. In contrast, monocytes from homozygous R51Q SNX10 mice fuse uncontrollably, forming giant dysfunctional OCLs that can become 10100 fold larger than their wild-type counterparts. Furthermore, mutant OCLs display reduced endocytotic activity, suggesting that their deregulated fusion is due to alterations in membrane homeostasis caused by loss of SNX10 function. This is supported by the finding that the R51Q SNX10 protein is unstable and exhibits altered lipid-binding properties, and is consistent with a key role for SNX10 in vesicular trafficking. We propose that OCL size and functionality are regulated by a cell-autonomous, SNX10-dependent mechanism that down-regulates fusion between mature OCLs. The R51Q mutation abolishes this regulatory activity, leading to excessive fusion, loss of bone resorption capacity and, consequently, to an osteopetrotic phenotype in vivo. Abbreviations used: ARO Autosomal Recessive Osteopsteosis; IRM-Interference reflection microscopy; OCLOsteoclast; M-CSF-Macrophage colony-stimulating factor; MPC-Mean pair circularity; PI3PPhosphatidylinositol 3-phosphate ; PI(3,5)P2-Phosphatidylinositol 3,5-bisphosphate; PX-Phoxhomology; RANKL-Receptor activator of NFB ligand; RQ/RQ-Homozygous for R51Q SNX10 ; SNX-Sorting nexin; SNX10-Sorting nexin 10; SNX10-KO-SNX10 knockout; SZSealing zone; SZL-Sealing zone-like; WT,+/+-Wild type. Jo ur na l o f C el l S ci en ce • A cc ep te d m an us cr ip t


Introduction
In order to examine the cellular and whole-organism manifestations of the R51Q mutation in SNX10 we recently generated knock-in mice bearing this mutation (Stein et al., 2020).
Homozygous R51Q SNX10 (RQ/RQ) mice exhibit massive, early-onset osteopetrosis and display many other clinical symptoms of the corresponding human ARO patients. It was further shown that OCLs of RQ/RQ mice do not resorb bone, in vivo and ex vivo, due to the absence of ruffled borders and to their inability to secrete protons and acidify the resorption lacunae (Stein et al., 2020). In the current study we address the cell-autonomous manifestations of the homozygous R51Q SNX10 mutation in monocytes throughout their differentiation into OCLs in culture. We show that the mutation results in highly deregulated cell fusion that generates giant cells that fuse continuously and fail to resorb bone, unlike wild-type OCLs whose fusion is regulated by a cellautonomous mechanism. These properties of the mutant OCLs are attributed to loss of function of SNX10 that is caused by the instability of the R51Q SNX10 protein and by its altered functional abilities, which affect membrane trafficking in the mutant cells and lead to loss of their boneresorbing capacity.

Cultured RQ/RQ OCLs are gigantic and unstable
In order to monitor osteoclastogenesis of homozygous R51Q SNX10 (RQ/RQ) OCLs, splenocytes from RQ/RQ and wild-type (+/+) mice were induced to differentiate, in culture, for 5-7 days in the presence of M-CSF and RANKL. Cells were then either fixed and fluorescently labeled for actin, tubulin, and DNA, or were subjected to live-cell video microscopy using phase contrast or interference reflection microscopy (IRM) optics. Splenocytes were used for these experiments since the massive osteopetrosis of RQ/RQ mice radically reduced the bone cavity volume and prevented isolation of sufficient bone marrow cells for routine OCL preparation (Stein et al., 2020).

Journal of Cell Science • Accepted manuscript
Mature +/+ OCLs, cultured on either glass or plastic substrates, were typically round with a projected diameter of 150-300 µm, and displayed a single circumferential SZ-like (SZL)  and extended significantly beyond the field of view. As in +/+ OCLs, SZLs were noted in RQ/RQ OCLs mainly at the cell periphery. Similarly large cells were also produced when small numbers of RQ/RQ bone marrow cells were differentiated in culture (Fig. S1E), indicating that production of the giant mutant OCLs is not limited to cells of splenic origin. Large RQ/RQ OCLs often contained "islands" of extracellular space within their projected area, which were surrounded by SZLs and typically contained "trapped" mononucleated cells and small multi-nucleated OCLs ( Fig. 1B and Fig. S1D). The origin of these enclosed areas is discussed below.
Importantly, RQ/RQ splenocytes cultured on a physiological bone substrate also fused into huge, multi-nucleated cells (Fig. 1C). These large RQ/RQ OCLs stained positively for TRAP, but the staining was weak and diffuse compared to +/+ cells. Longer-term studies indicated that the lifespan of RQ/RQ OCLs was significantly shorter than their +/+ counterparts (Fig. 1C, Day 15).
Each RQ/RQ OCL grown on bone displayed a single peripheral SZ, in contrast to +/+ cells that typically contained several much smaller SZs per cell (Fig. 1D).
Examination of individual OCLs indicated that the average area of RQ/RQ OCLs was 14.6 times larger than +/+ OCLs (Fig. 1E, left); this ratio increased to 49.0 when the upper 5 percentiles of the two cell populations were compared (Fig. 1E, right). In order to examine if the large size of RQ/RQ OCLs results from enhanced cell spreading or from increased fusion, we plotted the number of nuclei in individual +/+ and RQ/RQ OCLs as a function of their area. As shown in Fig. 1F the projected area and number of nuclei were highly correlated in both RQ/RQ and +/+ OCLs, including in the exceptionally large RQ/RQ OCLs (e.g. Fig. 1G). We conclude that the large projected area of the giant RQ/RQ OCLs is attributable to the increased fusion of mutant cells.

RQ/RQ OCLs exhibit altered fusion dynamics
The fusion dynamics of +/+ and RQ/RQ OCLs were compared by analyzing phase-contrast live cell movies (Movies 1-4, Fig. 2) in conjunction with immune fluorescence microscopy of fixed cells. As noted previously (Levaot et al., 2015;Soe et al., 2015) RANKL induced +/+ monocytes to fuse, and the resulting bi-nucleated cells grew further by fusing with adjacent mononucleated cells ( Fig. 2A) and, occasionally, with neighboring OCLs. Many of these initially-formed multi-nucleated cells were motile, assumed an irregular shape, and displayed podosomes scattered throughout their ventral membrane (immature cell, Fig. 2B). With time, these cells assumed a rounder and smoother shape, which we refer to as 'mature' morphology.
Concomitantly, these cells became less motile and developed a podosomal belt at their periphery (mature cell, Fig. 2B). Mature +/+ OCLs continued to grow, fusing either with mononucleated cells or with neighboring immature multi-nucleated cells ( Fig. 2A; Movies 1, 3). Fusion between pairs of mature +/+ OCLs was, however, observed very rarely. Most pairs of mature +/+ OCLs remained juxtaposed for long periods of time and then separated (e.g., cells 1 and 2 in Fig. 2C).
In agreement with this observation, the sequence of fusion events of +/+ OCLs often ended when mature cells became completely surrounded by other mature OCLs, with which they did not fuse ( Fig. 2D, Fig. S1F, Movie 3). RQ/RQ OCLs displayed a similar morphological maturation process. Yet, in sharp contrast with +/+ OCLs, mature RQ/RQ OCLs readily fused with each other (dashed arrow in Fig. 2A), forming giant OCLs that continued to fuse and grow ever larger (Movies 2, 4; cells 3, 4, 5 in Fig. 2C, Fig. 2D). Fusion of the mutant OCLs ended only when no fusion partners remained. Fusion between mutant OCLs typically started at several points along the cell-cell interface and proceeded rapidly, often engulfing areas of extracellular space that Journal of Cell Science • Accepted manuscript remained surrounded by podosomal SZLs (Fig. 1B and Fig. S1D). These areas were gradually integrated into the surrounding cell and their SZLs dissolved. Taken together, these observations point to the existence of a regulatory mechanism that blocks fusion between mature +/+ OCLs, and indicates that this mechanism is dependent on SNX10 and that it is disabled in RQ/RQ

OCLs.
Complementary information about differential osteoclastogenesis in RQ/RQ OCLs and their +/+ counterparts was obtained by dynamic live cell IRM, a microscopy-based imaging method that enables direct examination in culture of the gap between the ventral membrane of When examined by IRM, large +/+ OCLs exhibited complex and rapidly-changing shading patterns, indicating that the distance between the cells and the underlying surface is highly dynamic. The peripheral regions of these cells, where SZLs are located, were typically darker and more stable, consistent with presence of stable adhesions in these areas. This region remained dark also in mature juxtaposed +/+ OCLs that did not fuse (Fig. 3A, +/+ cells, t=685'). In contrast, the local IRM intensity of the SZLs of neighboring RQ/RQ OCLs became considerably lighter just prior to and during the fusion event (Fig. 3A, RQ/RQ cells, t=505'), indicating that local adhesion to the substrate is reduced during fusion between mature RQ/RQ OCLs.

Quantification of the morphological "signature" of mature OCLs
The observations described above indicate that mature +/+ and RQ/RQ OCLs differ radically in their capacity to fuse with other mature OCLs. As described above, maturation of OCLs grown on non-degradable surfaces includes three main manifestations: loss of the ability to fuse with other mature OCLs, formation of a single peripheral SZL, and a circular morphology. In order to quantify the morphological manifestations of the mature state in cultured OCLs of both genotypes, we employed comparative morphometry. Specifically, running osteoclastogenesis phase-contrast movies 'backwards', we identified pairs of fusing multi-nucleated OCLs in +/+ and in RQ/RQ cultures and marked their individual boundaries just before they fused. We then calculated the circularity of each of these cells (4 [(cell area)/(cell perimeter) 2 ]), where a value of 1 indicates a perfect circle. Since fusion occurs between cell pairs, we represented each pair of fusing OCLs by the arithmetic mean of their individual circularity values (mean pair circularity, MPC). Low-MPC fusion events between multi-nucleated cells, in which one or both were immature and irregularly-shaped, occurred in both genotypes (Fig. 3B). However, high-MPC fusion events between pairs of mature, round OCLs were significantly more prevalent in RQ/RQ cultures than in +/+ cultures. For example, an MPC value of 0.7 is close to the median of RQ/RQ fusion events, but represents the 90 th percentile of +/+ fusions ( Fig. 3B). High-MPC interactions between +/+ OCLs were common; however, most ended in separation after a relatively long time (393.8±56.2 minutes, Mean±SE, N=59 interactions), indicating that this particular fusion modality is inhibited in this genotype. Collectively, these results agree well with the observation that fusion between pairs of round, morphologically mature OCLs is much more common among RQ/RQ OCLs than among +/+ OCLs.
Further studies revealed that in +/+ cultures, fusion events involving two mononucleated cells or a mononucleated cell and a multi-nucleated cell were rapid, with the cells juxtaposing their membranes and fusing after 30-60 minutes (Fig. 3C, mono-mono and mono-multi modalities, +/+ cells). Fusion between two multi-nucleated cells, which in +/+ cultures refers predominantly to pairs of OCLs with low MPC values, was slower than other fusion modalities ( Fig. 3C, multi-multi +/+ cells) and, as indicated, was a rare event. In contrast, fusion between RQ/RQ cells in all three modalities proceeded faster than in +/+ cells (Fig. 3C, RQ/RQ cells). As indicated, all cases in which multi-nucleated RQ/RQ cells became adjacent to each other led to their rapid fusion, irrespective of their MPC values (Movies 2, 4), highlighting a major functional distinction between OCLs of the two genotypes.

Deregulated fusion of RQ/RQ OCLs is not caused by aberrant RANKL signaling
Cell fusion during osteoclastogenesis is driven by continuous presence of RANKL, hence we examined whether the deregulated fusion of RQ/RQ OCLs might be caused by increased sensitivity to this cytokine. Similar to +/+ cells, RQ/RQ OCL precursor cells neither differentiated nor fused in the absence of RANKL (Fig. 3D, Fig. S4A), indicating that production of RQ/RQ OCLs is strictly RANKL-dependent. Moreover, when +/+ and RQ/RQ cells were cultured separately in the presence of increasing concentrations of RANKL, the relative area of the growth surface covered by multi-nucleated OCLs increased in a dose-dependent manner that was similar in both genotypes (Fig. 3D). Yet, individual RQ/RQ OCLs were fewer and consistently larger than the more numerous +/+ OCLs grown at the same concentration of RANKL ( Fig. S4A), indicating that RQ/RQ OCLs fuse abnormally also at low concentrations of RANKL. qPCR studies showed that mRNAs for RANKL-induced proteins, including the key osteoclastogenic transcription factor NFATc1, TRAP, cathepsin K, the ATP6V1D2 V-ATPase subunit, CLCN7, the RANKL receptor RANK, and the fusion-related proteins DC-STAMP and OC-STAMP, were expressed at similar levels in +/+ and in RQ/RQ OCLs (Fig. S4B). We conclude that +/+ and RQ/RQ OCLs and their precursor cells respond similarly to RANKL, and that the fusion phenotype of the mutant cells is not caused by abnormal sensitivity to this cytokine.

RQ/RQ OCLs exhibit reduced dextran endocytosis
SNX10 participates in vesicle trafficking and in endocytosis, cellular activities that are critical for osteoclastogenesis and whose disruption may lead to major membrane-related phenotypes such as loss of ruffled border structures and failure of acidification and bone resorption by OCLs (Palagano et al., 2018;Pangrazio et al., 2013;Stein et al., 2020). Similar defects cause aberrant cell fusion in other cell types (Smurova and Podbilewicz, 2017). In order to directly examine the effect of R51Q SNX10 on endocytosis in OCLs, we measured dextran internalization by these cells, a process that occurs by both fluid-phase and receptor-mediated endocytosis (Pustylnikov et al., 2014). Cells were cultured in the presence of TRITC-labelled dextran, and their ability to internalize the compound was evaluated by fluorescence microscopy.
Control OCLs from both genotypes incubated at 0 o C exhibited low and similar levels of TRITC signals (Fig. 3E). Internalized dextran was clearly visible in mononucleated and immature polynucleated cells of both genotypes that were incubated at 37ºC (Fig. 3F, TRITC-dextran panels).

The R51Q SNX10 protein is unstable, exhibits aberrant lipid binding properties, and is mis-localized in RQ/RQ OCLs
Next, we examined the expression and functional properties of wild-type and mutant SNX10 in OCLs. qPCR studies indicated that Snx10 mRNA is induced during OCL differentiation, and that total Snx10 mRNA levels are similar in +/+ and RQ/RQ OCLs (Fig. 4A).
cDNA sequencing confirmed that +/+ OCLs express wild-type Snx10 mRNA and that RQ/RQ OCLs express mutant mRNA exclusively (Fig. S4C). Detection of endogenous SNX10 protein in OCLs by protein blotting with commercial anti-SNX10 antibodies yielded inconsistent results, hence we examined SNX10 expression directly using targeted proteomics. Analysis of whole-cell lysates of +/+ OCLs identified two peptides derived from the SNX10 protein. These peptides were either absent or were detected at low levels in RQ/RQ OCLs (Fig. 4B), indicating that the R51Q SNX10 protein is present at low amounts in the mutant OCLs. Poor expression of the R51Q SNX10 protein was also observed when the exogenous mutant protein was expressed in RAW264.7 cells and detected via its FLAG tag (Fig. 4C). Exposing these cells to the proteasome Journal of Cell Science • Accepted manuscript inhibitor MG132 increased detection of R51Q SNX10 protein (Fig. 4C), indicating that its low levels are caused at least in part by its reduced stability and degradation.
Arginine 51 is located within the phosphoinositide-binding PX domain of SNX10, suggesting that the R51Q mutation might affect its lipid-binding specificity and through it the function of the residual mutant protein. In order to examine this possibility we produced wildtype and R51Q SNX10 proteins in bacteria and used equal amounts of each protein to probe a phospholipid array in vitro. As shown in Fig. 4D, wild-type SNX10 bound phosphatidylinositol 3-phosphate (PI3P) and phosphatidylinositol 3,5-bisphosphate (PI(3,5)P 2 ), while R51Q SNX10 did not bind any of the phospholipids present on the array. Since lipid binding by the PX domain enables SNX10 to associate with vesicular membranes, we examined if the R51Q mutation altered the cellular localization pattern of SNX10. To this end we expressed FLAG-tagged cDNAs for +/+ SNX10 and for its R51Q mutant in +/+ and in RQ/RQ OCLs, respectively, followed by immunofluorescence staining with anti-FLAG antibodies. +/+ SNX10 was readily expressed in the cells, and appeared as small, regularly-shaped punctate signals in the cytosol that are consistent with vesicular localization (Fig. 4E). In agreement with its low amounts in OCLs, exogenous R51Q SNX10 protein was expressed at lower levels and in fewer cells, where it gave rise to larger, irregularly-shaped fluorescent signals that may represent protein aggregates (Fig.   4E). Collectively, these data suggest that the RQ/RQ OCL fusion phenotype is caused by loss of function of SNX10 due to a severe reduction in the amount of R51Q SNX10 protein, which is likely caused by the inability of the R51Q SNX10 protein to bind specific phospholipids that leads to its subsequent mis-localization and degradation.

R51Q SNX10 induces OCL hyper-fusion by a loss of function mechanism
In order to examine directly if the RQ/RQ OCL hyper-fusion phenotype is caused by loss of function of SNX10, we knocked out the Snx10 gene in RAW 264.7 cells using CRISPR resorbing activity when seeded on bovine bone (Fig. 5F). Adding as little as 12.5% +/+ cells resulted in detectable bone-resorbing activity, and the average size of the resulting multinucleated cells was reduced ten-fold. Increasing the fraction of +/+ cells to 25% resulted in close to normal bone-resorbing activity and OCL size, and cultures containing 50% or more +/+ cells were indistinguishable in size and activity from pure +/+ OCL cultures (Fig. 5F). Growth of +/+ or RQ/RQ OCLs in medium conditioned by prior growth of OCLs from the other genotype did not alter the fusion characteristics of either genotype (not shown), indicating that the RQ/RQ phenotype is not induced by differential secretion of soluble factors from the OCLs. Fusion of even small amounts of +/+ cells with RQ/RQ cells can therefore provide sufficient wild-type SNX10 protein to rescue the fusion and activity phenotypes of the mutant OCLs.

Discussion
The dynamic process whereby OCL precursor cells fuse to form multi-nucleated, boneresorbing OCLs has been extensively investigated in recent years (Levaot et al., 2015;Soe et al., 2015). In wild-type cultures, this process yields OCLs with wide, yet predictable, ranges of size and number of nuclei that, in mice, rarely exceed 2x10 5 µm 2 and 100 nuclei, respectively (Fig.   1F). This rather common observation, per se, indicates that osteoclastogenic cell fusion is a tightly regulated process that is blocked when the heterokaryons reach a "mature stage". The existence of an upper limit to OCL size is broadly accepted, yet the nature of the regulatory mechanism that underlies the signal that stops fusion is unknown. The current study sheds new light on this process and shows that the membrane trafficking-associated protein SNX10 actively participates in down-regulating fusion between mature OCLs (Fig. 6). In particular, the finding that a defined point mutation in a single protein induces cell-autonomous deregulated fusion indicates that OCL size is actively regulated by a cellular-genetic mechanism. The data suggest that this mechanism functions by limiting fusion between pairs of mature OCLs, that SNX10 is essential for this process, and, surprisingly, that no backup mechanism halts fusion between mature OCLs once this mechanism is disrupted.
How is OCL maturation manifested? As OCLs differentiate and fuse, these cells undergo morphological and functional changes that enable them to effectively degrade bone. We show here that among these changes, cultured OCLs undergo "morphological maturation" and assume a more circular shape, and concomitantly lose their ability to fuse with other morphologicallymature, multi-nucleated OCLs. Interestingly, while mature OCLs cannot fuse with each other, they readily fuse with less mature cells. This observation highlights an intriguing asymmetry, whereby mature OCLs cannot initiate fusion but can fuse with immature cells that apparently initiate this process. While the underlying mechanism is unclear, this asymmetry is reminiscent of the initial stages of osteoclastogenesis, in which long-term priming by RANKL differentiates a sub-population of monocytes into "founder cells", which fuse with "follower" monocytes that are fusion competent but cannot initiate the fusion process (Levaot et al., 2015). Similar asymmetry during cell-cell fusion has been described also during myoblast-myotube fusion in Drosophila (Dworak and Sink, 2002), although it is not a general aspect of cell fusion (reviewed in (Hernandez and Podbilewicz, 2017;Schejter, 2016)).
The correlation between the shape of an OCL and its ability to fuse was further examined in a quantitative manner by calculating the cell's circularity, a morphological parameter that is independent of object size. Since cell fusion occurs between pairs of interacting cells, the decision whether to fuse is likely determined pairwise by the properties of these cells. We Collectively, our results indicate that SNX10 is essential for the physiological arrest of mature OCL fusion and that the R51Q mutation disrupts this regulatory function.
Results presented in this study collectively indicate that, at the cellular level, the hyper-fusion phenotype of the RQ/RQ OCLs is due to loss of SNX10 function. This conclusion is supported by the significantly reduced levels of R51Q SNX10 in the mutant cells, by the ability of relatively small amounts of wild-type SNX10 protein to block formation of giant OCLs, and by induction of hyper-fusion in RAW 264.7 cells by knocking out Snx10. The R51Q SNX10 protein itself cannot bind PI3P or PI(3,5)P 2 and is mis-localized in OCLs. Moreover, expression of exogenous R51Q SNX10 in RQ/RQ OCLs does not rescue their hyper-fusion phenotype, indicating that the mutant SNX10 does not function like its wild-type counterpart even when its amounts are increased in the cells. However, while these results suggest that at the molecular level, loss of function of R51Q SNX10 is caused by a combination of its severely reduced levels and by inactivity of the residual protein present in RQ/RQ OCLs, we cannot exclude at present the possibility that the R51Q SNX10 protein possesses a dominant-negative activity. Further studies are required to address this issue.
While the exact role of SNX10 in regulating OCL fusion is still unclear, nearly all the manifestations of the R51Q mutation in this cell type, in addition to regulation of cell fusion, are related to aberrant membrane properties. These include inability to bind specific phospholipids, OCLs compared to wild-type controls. A similar drop was noted in mononucleated RQ/RQ cells; its effect might be more limited in these less-developed cells due to their overall higher levels of uptake. It is noteworthy that inhibiting endocytosis in epidermal cells of C. elegans by targeting dynamin or RAB-5 results in sustained retention of the EFF-1 fusogen at the apical membrane of these cells, and promotes their excessive fusion (Smurova and Podbilewicz, 2016; Smurova and Podbilewicz, 2017). Conceptually, we can envisage a similar process in mature wild-type OCLs, whereby removal of putative essential fusogens from their plasma membrane through endocytosis Journal of Cell Science • Accepted manuscript down-regulates their fusion capacity. Defective endocytosis in mature R51Q SNX10 OCLs could result in sustained retention of such molecules in the plasma membrane and drive deregulated cell fusion. The capacity of endocytosis to affect fusion in OCLs was previously shown by Shin and colleagues, who reported that clathrin-mediated endocytosis is required for early-stage fusion of osteoclast precursor cells (Shin et al., 2014). The current study focuses on late-stage osteoclastogenesis, whose main characteristic is cessation of cell-cell fusion between mature OCLs. It is therefore possible that the role of endocytosis evolves to become fusion-inhibitory as OCLs mature.
Although OCL activity increases with size and nuclear number (Boissy et al., 2002;Lees and Heersche, 1999;Makris and Saffar, 1982), the present study suggests that optimal OCL survival and activity are associated with a defined size range that has an upper limit, and which is actively regulated by the cell itself. The mechanistic links between the large size of RQ/RQ OCLs, and their lack of ruffled border and inactivity are not clear at the present time, although they could both arise from the membranal abnormalities of these cells. Studies in which OCL hyper-fusion is induced by mechanisms that do not involve SNX10 should shed further light on this issue.
Nonetheless, our data suggest that promotion of deregulated fusion that renders OCLs excessively large and fragile might enable targeting these cells in disease, in a manner distinct from existing strategies that inhibit formation of OCLs or induce their death.

R51Q SNX10 knock-in mice:
Mice carrying the R51Q mutation in SNX10 were constructed by CRISPR at The Weizmann Institute, and genotyped and maintained as described (Stein et al., 2020). Mice were in the C57Black/6 genetic background and were housed in a barrier facility kept at 22±2 o C on a light/dark cycle of 12h:12h, with food and water provided ad libitum. Colonies were maintained by intercrossing heterozygous RQ/+ mice to produce +/+, RQ/+, and RQ/RQ littermate mice.
Primary OCLs were prepared from mice of either sex, aged 4-8 weeks. All experiments were approved by the Weizmann Institute IACUC and were conducted in accordance with Israeli law.

Cell culture
Culture of primary mouse osteoclasts from spleens: Spleens from mice aged 4-8 weeks were dissociated into un-supplemented α-Minimal Eagle's Medium (α-MEM; Gibco-Thermo Fisher Scientific, Waltham, MA, or Sigma-Aldrich, St. Louis, MO). Following erythrocyte lysis, cells were seeded at a density of 5x10 6 cells/well (+/+) and 2.5x10 6 cells/well (RQ/RQ) in 6-well plates, or 2x10 6 cells/well (+/+) and 1x10 6 cells/well (RQ/RQ) in 24-well plates. Cells were cultured in complete OCL medium (α-MEM supplemented with 10% fetal calf serum (FCS), 2 mM glutamine, 50 units/ml penicillin, and 50 g/ml streptomycin, as well as the cytokines M-CSF (20 ng/ml, Peprotech, Rehovot, Israel) and RANKL (20 ng/ml, R&D Systems, Minneapolis, MN)). Cells were grown at 37 °C in 5% CO 2 for 5-7 days with daily changes of medium. In some studies, glass coverslips were inserted into wells of 24-well plates and processed as above. For growth on bone, cells were prepared and seeded as above in wells of 24 well plates each containing 2-3 small fragments of bovine cortical bone. Cells grown on bone were cultured in complete OCL medium that included cytokines as above, with changes of medium every 48-72 hours for up to 15 days. Some bone cultures were processed for SEM as described (Stein et al., 2020).
Culture and manipulation of RAW264.7 cells: RAW264.7 cells, obtained from the ATCC, were grown in Dulbecco's Modified Eagle's Medium (DMEM, Sigma), supplemented with 10% fetal calf serum, 2 mM glutamine, 50 units/ml Journal of Cell Science • Accepted manuscript penicillin, and 50 g/ml streptomycin. Cells were grown on plastic tissue culture plates and induced to differentiate with M-CSF and RANKL as above.
For construction of RAW LifeAct-EGFP cells, LifeAct-EGFP was isolated by PCR from pLifeAct-EGFP (Riedl et al., 2008) and cloned into the retroviral vector pBABE/Puro (Morgenstern and Land, 1990). RAW cells were transduced with retroviral particles prepared in 293 HEK cells from pBabe/Puro-LifeAct-EGFP, and selected with 5 g/ml puromycin. For CRISPR-mediated targeting of the Snx10 gene, candidate sgRNA sequences were selected from

Construction and use of adenoviruses and lentiviruses to transduce OCLs:
Adenoviruses: C-terminally FLAG-tagged wild-type and R51Q SNX10 (mouse) cDNAs were cloned into the pShuttle-CMV plasmid (Stratagene, Agilent Technologies, Inc., Santa Clara, CA) and used to produce Adenoviruses using the AdEasy XL adenoviral vector system (Stratagene, Agilent Technologies). Viruses were amplified using HEK293AD cells. For infection, monocytes were grown in the presence of M-CSF for 3 days, and then in M-CSF and RANKL for two Journal of Cell Science • Accepted manuscript additional days. Then, one volume of medium from HEK293AD cells that contained adenovirus particles was added to two volumes of complete OCL growth medium volume (containing M-CSF and RANKL). Cells were incubated for 24 hours, followed by exchange to fresh complete OCL medium and harvesting 24 hours after that.
Lentiviruses: Wild-type and R51Q SNX10 tagged at their C-terminus with FLAG-V5 were cloned into the pUltra-hot (a gift from Malcom Moore, Addgene plasmid 24130), which also directs peptide 2A(P2A)-mediated independent expression of mCherry protein. Packaging was performed in HEK-293 cells following transfection of the pUltra-hot plasmid and the ViraPower lentiviral packaging mix (Thermo Fisher Scientific, mix of pLP/VSVG,pLP1, and pLP2 plasmids). Medium containing lentiviral particles was collected after 48 and 72 hours. For infection of RAW264.7 cells with lentiviruses, the cells were incubated with crude lentiviralcontaining medium for 16 hours, passaged, and then sorted by FACS for mCherry positive cells using FACSAriaIII Instrument (BD Biosciences, San Jose, CA). Cells were seeded in a 6 well plate (2x10 5 cells per well). In some cases, 16 hours later fresh medium was added in the presence or absence of 20nM MG132 (Merck-Millipore, Burlington, MA) for 4 hours, after which cells were lysed with RIPA buffer (50 mM Tris (pH 8), 150 mM NaCl, 1% Nonidet P-40, 0.5% Deoxycholate, 0.1% SDS) and analyzed by protein blotting.

Cell analyses:
OCL staining and immunofluorescence: Bone: OCLs grown on bone slices were stained for TRAP using a Leukocyte Acid Phosphatase kit (Sigma). For actin staining, cells were fixed and permeabilized by incubating the slices in 0.5% Triton X-100/3% paraformaldehyde (PFA) for 3 minutes, followed 3% PFA for 20 minutes and three washes in PBS. Cells were then exposed to TRITC-phalloidin (Sigma, P1951) for 1 hour at room temperature. DNA was visualized by incubating the slices with Hoechst 33342 (Molecular Probes, Eugene, OR, H-3570) for 3 minutes.
Live cell imaging of OCLs: Cells were cultured for three days in complete OCL medium supplemented with 20 ng/ml M-CSF. Cells were detached with Trypsin-EDTA and seeded at 1x10 5 cells/well on glass coverslips placed inside wells of a 24 well plate, or on µ-Slide 4-well imaging chambers with a coverslip polymer bottom (Ibidi, Martinsried, Germany). Cells were cultured in complete OCL medium supplemented with 20 ng/ml M-CSF and 20 ng/ml RANKL, with daily medium changes for 3-4 days prior to imaging. Time-lapse images were acquired with an automated inverted microscope (DeltaVision Elite system IX71 with Resolve3D software modulus; Applied Precision, Inc., GE Healthcare, Issaquah, WA) using a 10×/0.30 air objective (Olympus, Tokyo, Japan). The microscope is equipped with an environmental box kept at 37 °C with a 5% CO 2 humidified atmosphere. Images were acquired every 5 min, for up to 42 h. IRM (Verschueren, 1985) timelapse imaging was carried out using a 20x/0.7 objective at 5 minutes interval between frames.

Dextran internalization:
Splenocytes from +/+ or RQ/RQ mice were seeded on glass coverslips and differentiated into OCLs. On the day of the experiment, growth medium was replaced with OCL medium that contained 50 M TRITC-dextran (40 kDa; Sigma-Aldrich), M-CSF and RANKL, but lacked serum. The cells were incubated for 30 minutes at 37 degrees, washed in ice-cold PBS twice, and immediately fixed in PBS containing 3.5% PFA and 2% sucrose for 15 minutes at room temperature. Nuclei were stained with Hoechst 33342. Cells analyzed at 0 degrees were initially placed on ice for 10 minutes prior to adding ice-cold medium containing TRITC-dextran, Journal of Cell Science • Accepted manuscript incubated for 30 minutes on ice, and processed as above. Following mounting, the slides were analyzed using a DeltaVision Elite system IX71 (Applied Precision) using a 60X 1.42 oil objective (Olympus). For each experiment, exposure times were set to visualize +/+ OCLs incubated at 37 degrees, and remained unchanged throughout the entire process of data collection from that experiment. The OCL SZLs, visible by actin staining, were used to define the boundary of each cell; the area of each cell and the total intensity of the TRITC signal within this area were measured using Image J. Care was taken to analyze mature, round OCLs of roughly similar sizes in both genotypes.

Image display and analysis:
Were performed using Fiji (Schindelin et al., 2012). For image montages, 3X3 adjacent fields with 10% overlap were stitched using the stitching plugin (Preibisch et al., 2009). Some images were corrected for background and shading variations using the BaSic imaging tool (Peng et al., 2017). The areas and circularities of marked cells were calculated semi-automatically using the Fiji software. For multi-nucleated cells, only cells that included 3 or more nuclei and whose entire area was within the image were considered. For determination of time between contact and fusion, live cell videos were run backwards and the cells followed from fusion to the time when they first made contact. Individual mature multi-nucleated OCLs were followed backwards to the initial fusion event between two mononucleated cells that gave rise to them, and all intermediate fusion events were scored and analyzed. Time between first contact and separation without fusion was measured similarly starting from cells that had just separated.
Counting of nuclei in cell images: To count the number of nuclei in each cell we manually marked the cell outlines and saved them in the RoiManager. The nuclei were automatically detected from the DNA channel by enhancement using Laplacian of Gaussian filter (2um smoothing scale), applying fixed threshold (< -0.5) and separating neighboring nuclei using watershed. To enable counting, each nucleus was shrunk to a single point. The number of nuclei in each cell was measured by counting the number of positive pixels in the single-point image within the cell border using a Fiji macro that can be obtained from the Authors.
Pit resorption studies: Cells were seeded on bone fragments as described above, and grown and analyzed as described (Stein et al., 2020).
Cell mixing studies: Splenocytes from +/+ and RQ/RQ mice were grown separately in M-CSF-containing medium for 3 days with daily medium changes. Cells were detached with trypsin/EDTA, counted, and mixed in the desired proportions while maintaining the same total number of cells in each culture (1x10 5 cells /well in 24-well plates). Cells seeded on plastic were stained for TRAP and their area measured as described above.

Data and Materials Availability
All data supporting the results of this manuscript are either included in the manuscript or are available from the corresponding author (AE) upon request. R51Q SNX10 knock-in mice are available under a Weizmann Institute materials transfer agreement. N=219 RQ/RQ OCLs, R 2 =0.9951. The large group of cells at lower left is the same one shown in (F). In panels D-G, cells were analyzed at their prime (day 8 for RQ/RQ OCLs, day 12 for +/+ OCLs, due to the faster development of the mutant cells).

G A T C G A T A G A A G A T C G A T A G A A A G A T C A A T A G A A
A B C Figure S4: (A) Increased fusion of RQ/RQ OCLs is not due to hypersensitivity to RANKL.
Representative images of TRAP-stained splenocytes from +/+ and RQ/RQ mice after 5 days growth in different concentrations of RANKL as indicated. Bars: 250 mm. See also Figure 3D. Data is from one experiment, representative of 3 performed. (B) mRNA levels of key osteoclast-related genes in +/+ and RQ/RQ OCLs. mRNA samples from +/+ and RQ/RQ splenocytes that were grown with M-CSF alone  Figure S6: +/+ and RQ/RQ OCLs can fuse with each other to form genetically hybrid heterokaryons. Cells used in this study were primary splenocytes from +/+ mT/mG mice, described in Figure S5F, which express Tomato and switch to EGFP in the presence of Cre, and from RQ/RQ mice carrying the Cathepsin-K Cre transgene, which are colorless. Fusion between both genotypes in culture induces the Tomato-EGFP switch. (A) Scheme showing the possible outcomes of fusion in this system. (B) Cells were grown separately for three days in the presence of M-CSF, counted, mixed in the indicated proportions, and grown in the presence of M-CSF and RANKL for 5 additional days. Note that the EGFP signal is present only in the mixed cultures. Bars: 100 µm. Images are representative of two experiments.