CD13 is a critical regulator of cell–cell fusion in osteoclastogenesis

The transmembrane aminopeptidase CD13 is highly expressed in cells of the myeloid lineage, regulates dynamin-dependent receptor endocytosis and recycling and is a necessary component of actin cytoskeletal organization. Here, we show that CD13-deficient mice present a low bone density phenotype with increased numbers of osteoclasts per bone surface, but display a normal distribution of osteoclast progenitor populations in the bone marrow and periphery. In addition, the bone formation and mineral apposition rates are similar between genotypes, indicating a defect in osteoclast-specific function in vivo. Lack of CD13 led to exaggerated in vitro osteoclastogenesis as indicated by significantly enhanced fusion of bone marrow-derived multinucleated osteoclasts in the presence of M-CSF and RANKL, resulting in abnormally large cells containing remarkably high numbers of nuclei. Mechanistically, while expression levels of the fusion-regulatory proteins dynamin and DC-STAMP1 must be downregulated for fusion to proceed, these are aberrantly sustained at high levels even in CD13-deficient mature multi-nucleated osteoclasts. Further, the stability of fusion-promoting proteins is maintained in the absence of CD13, implicating CD13 in protein turnover mechanisms. Together, we conclude that CD13 may regulate cell–cell fusion by controlling the expression and localization of key fusion regulatory proteins that are critical for osteoclast fusion.

absence of CD13, osteoclastogenesis is perturbed without affecting osteoblast function. In addition, the in vitro cytokine-mediated induction of flow-sorted BM-derived CD13-deficient OC progenitors generated increased numbers of OCs that were considerably larger in size, contained many more nuclei and resorbed bone much more efficiently than those from wild type progenitors. Consistent with the primary murine OC data, human osteoclast-like cells generated from CRISPR-engineered CD13 KO U937 cells led to an increase in multinucleated, TRAP+, functional OC compared to the control cells, confirming that CD13 modulates osteoclastogenesis. Furthermore, we demonstrated that while the expression of certain fusion proteins is typically downregulated in mature osteoclasts post-fusion 23 , it is abnormally preserved via post-transcriptional mechanisms in osteoclasts lacking CD13. We have shown that CD13 is a mediator of homotypic cell interaction and a regulator of molecular events defining cell membrane organization, fluidity and movement, all processes critical to cell-cell fusion 14,15,20 . We hypothesize that CD13 is a negative regulator of cell-cell fusion in osteoclastogenesis and potentially, a universal modulator of membrane fusion and thus is a novel target for therapeutic intervention in pathological conditions mediated by defects in cell-cell fusion.

Results
Bone density is reduced in CD13 KO mice in vivo. Based on the notion that the highly sustained expression of CD13 in all cells of the myeloid lineage reflects its important role in myeloid cell biology, we and others have demonstrated that it contributes to many fundamental cellular processes that impact myeloid cell function in various tissues. These studies prompted our current focus on the myeloid cells of the bone, the osteoclasts. We initially examined the effect of a global loss of CD13 on the phenotype of developing bone. Analysis of bone micro-architecture and function in the cortical and trabecular bone isolated from 8 to 10 week old WT and CD13 KO mice by μ-CT revealed that the femur cortical and trabecular bone density and thickness in CD13 KO mice is reduced compared to WT animals ( Fig. 1a,b). Reduced bone volume/total volume (BV/TV, WT vs. CD13 KO Fig. 1h). In addition, histochemical analysis showed that these osteoclasts were TRAP + (in purple) in CD13 KO femurs compared to their wildtype counterparts (Fig. 1i). Bone remodeling is tightly controlled by the coordinated action of bone forming osteoblasts and bone degrading osteoclasts whereby an increase in bone resorption with no change in formation can lead to loss of bone density. To investigate potential underlying causes of CD13-dependent loss of bone density in CD13 KO mice, we evaluated bone formation in vivo in young mice. Importantly, both the mineral apposition rate (MAR, measurement of the linear rate of new bone deposition), (Fig. 1j) and dynamic bone formation rate (BFR/BS), (Fig. 1k) were unchanged in CD13 deficient mice. These data demonstrated that the overall bone loss in CD13-deficient mice is a result of enhanced osteoclast number and activity while bone formation and osteoblast function is unaltered.
Defects in bone structure are exaggerated in CD13-deficient aged mice. Next, we investigated if the steady state phenotype of lower bone mass observed in CD13 deficient young mice is amplified during aging. Indeed, μCT and histomorphomtery indicated a significant reduction in BV/TV (WT vs. CD13 KO ; 15.0 vs. 7.5; Fig. 2a and WT vs. CD13 KO ; 9.0 vs. 3.4; Fig. 2e), trabecular number (Fig. 2b) and thickness (Fig. 2c), with a concomitant increase in trabecular spacing (Fig. 2d). Finally, the increase in osteoclast numbers (Oc.S/BS, WT vs. CD13 KO ; 12.8 vs. 20.3; Fig. 2f) in CD13-deficient aged mice compared to the wildtype counterpart was sustained. However, over time, while the MAR (Fig. 2g) was unaffected between genotypes in the aged mice, the BFR/BS (Fig. 2h) was significantly diminished in older CD13 KO animals, indicating that CD13 also impacts osteoblast function long-term, which may contribute to the accelerated bone loss in CD13 KO aged mice (Fig. 2a,e). Together, these data strongly suggested a defect in bone structure in the absence of CD13 that is sustained over time, supporting a contribution of CD13 to homeostatic bone remodeling in vivo.
In vitro osteoclastogenesis is exaggerated in CD13 KO cells. The previous data suggested that CD13 may mediate remodeling predominantly at the level of the osteoclasts. Stimulation of flow-sorted monocyte lineage-committed hematopoietic progenitors with two principal BM cytokines, M-CSF and RANKL, triggers the expression of molecules involved in cell-cell fusion and functional bone resorption [6][7][8] and the generation of multinucleated osteoclasts. Flow sorted, BM-derived OCP (CD3 − B220 − NK1.1 − CD11b lo/− CD115 hi Ly6G + , Supplementary Fig. S1) were differentiated to mature OC in the presence of recombinant M-CSF and RANKL and stained with TRAP. Analysis of these cultures revealed significant increases in the number of osteoclasts containing > 3 nuclei per field (2.8-fold) and the average OC size (fourfold) in CD13 KO cells grown on bovine cortical bone slices or plastic (Figs. 3a-c, 4) compared to those generated from WT progenitors. These differences were evident by d5, suggesting that the lack of CD13 accelerates OC fusion and multinucleation but not OCP proliferation rates, as the cell density (total number of nuclei/dish) was not significantly different between genotypes (Fig. 3f). In agreement with amplified osteoclastogenesis in vivo in CD13-deficient mice, in vitro multinucleated OC formation ( Supplementary Fig. S2a-c) was significantly elevated in OCs generated from CD13 KO aged BM. Similarly, spleen-derived CD13 KO cells produced OCs of larger area (threefold), more cells with > 3 nuclei (2.5-fold) and an increased number of nuclei per cell (threefold) compared to WT ( Supplementary Fig. S3). To assess OC bone resorptive capacity, we utilized a 24-well Osteo assay plate, (Corning), coated with synthetic bone mimetic that allows measurement of in vitro osteoclast activity. We plated WT and CD13 KO flow-sorted BM-derived OCP on Osteo assay plates in the presence of recombinant M-CSF and RANKL and allowed them to mature to OC over time. At d10, OCs were removed, individual or multiple resorption pit areas were imaged and www.nature.com/scientificreports/ the area of resorption quantified by ImageJ. As expected, the increase in OC nuclei/cell positively correlates with resorption where CD13 KO OCs showed increased resorption area (threefold) compared to WT, confirming that the elevated fusion in CD13 KO mice translates into exaggerated functional activity in both young (Fig. 3d,e) and aged ( Supplementary Fig. S2d,e) mice. Importantly, WT cells treated with CD13 blocking antibody (SL13; 1 μg/ ml) prior to fusion in the presence of M-CSF and RANKL led to a significant increase (threefold) in multinucleated OC with greater than 3 nuclei per cell which was analogous to the CD13-deficient OCs, illustrating that blocking CD13 as well as its absence led to accelerated fusion (Fig. 4a,b) and confirming the CD13 specificity of osteoclast fusion phenotype.
Osteoclast progenitors with osteoclastogenic potential are similar in WT and CD13 KO bone marrow and periphery. Previously we have shown that the distribution of the hematopoietic population comprised of early hematopoietic progenitors, myelo-erythroid progenitors, and granulocyte macrophage pro-

Osteoclast-like cell formation is enhanced in U937 cells expressing CD13 CRISPR.
To confirm that the effect of loss of CD13 in cell fusion extends across species, we utilized human U937 myeloid cells engineered to delete CD13 or express scrambled wildtype control by CRISPR-based technology (Fig. 5f). Differentiation of U937 cells with PMA 25-27 for 3 days led to formation of equivalent levels of adherent monocytic cells in both genotypes (data not shown). Subsequent stimulation with human recombinant M-CSF and RANKL led to the aggregation of monocyte-like cells followed by differentiation into large multinucleated TRAP + osteoclastlike cells over time (8-10d) which was significantly accelerated in cells lacking CD13 compared to scrambled controls ( Fig. 5a-c). TRAP+, multinucleated CD13 KO CRISPR cells exhibited increased resorptive activity when grown in Osteo assay plates for 17 days as indicated by the relative resorption area (Fig. 5d,e), confirming that the cells generated from U937 are indeed functional osteoclasts and that CD13 is a negative regulator of cell-cell fusion in both mouse primary cells (BM and periphery) and a human monocytic cell line.

Expression of fusion proteins is dysregulated in CD13 KO mature osteoclasts.
A number of the molecular mechanisms mediating osteoclast fusion and multinucleation have been elucidated 28 . In particular, the small GTPase dynamin 2 (the major isoform in OC), the "master fusogen" DC-STAMP (DCST1) and the tetraspanins CD9 and CD81 are common and critical regulators of osteoclast fusion 23,29 . Our recent studies have   www.nature.com/scientificreports/ shown that CD13 is a potent negative regulator of dynamin-dependent endocytosis of a variety of receptors [20][21][22] , suggesting that CD13 may participate in cell fusion by regulating endocytic processes. Indeed, immunofluorescence ( Fig. 6) and immunoblot (Fig. 7a,b) analyses of OC lysates derived from flow-sorted WT BM-OCPs demonstrated high levels of dynamin and DC-STAMP expression by 2d-post differentiation, which was subsequently reduced by 3d when cell fusion and maturation into multinucleated WT osteoclasts is complete, as previously reported 23 . However, while dynamin, DC-STAMP and CD9 are highly expressed in CD13 KO OCP, rather than being downregulated, this strong expression is maintained in mature multinucleated OCs (Figs. 6, 7a,b), suggesting that CD13 may impact fusion by regulating the levels of these key fusion and/or endocytic molecules critical for OC fusion. In addition, flow cytometry of dynamin, DCST1 and CD9 confirmed persistent surface expression of fusion promoting proteins in multinucleated OC lacking CD13 compared to WT OC (Supplementary Fig. S5a-c). Furthermore, immunoblot analysis of cell lysates obtained from WT BM-derived progenitor cells stimulated with M-CSF and RANKL over 3d indicated that CD13 is highly expressed in myeloid progenitor cells but its expression level is unaltered upon stimulation with M-CSF and RANKL over time (d0-3) (Fig. 7c,d), consistent with CD13 regulating fusion mechanisms independent of RANKL signaling. Next, we investigated the overall mechanistic process by which CD13 sustains expression levels of dynamin, DCST1 and tetraspanins. Quantitative RT-PCR analysis demonstrated that transcript levels for DNM2, DCST1 and tetraspanins CD9 and CD81 were similar in BM progenitor cells stimulated with M-CSF + RANKL over 0-5d, indicating that CD13 does not regulate transcription levels of these proteins (Fig. 8a). To explore if CD13 limits the stability of fusion-regulatory proteins in WT cells, BM-derived OCP were grown in the presence of M-CSF + RANKL and treated with cycloheximide (100 μg/ml) to inhibit new protein synthesis. While dynamin and DCST1 protein expression in CD13 KO OC remained stable over 8-12 h, loss of dynamin and DCST1 protein expression occurred by 4-8 h in WT OC, indicating that CD13 controls fusion-regulatory protein turnover and stability (Fig. 8b,c).

Discussion
The fusion of plasma membranes is essential to and indispensable for many physiologic processes such as fertilization through sperm/egg fusion 30 , muscular development through myoblast fusion 31 , skeletal development and maintenance of skeletal integrity through formation of osteoclasts and control of certain viral infections and spreading through the formation of macrophage giant cells (MGC) 32 . Thus, this important biological process directly defines the course of many pathological processes including infertility, skeletal defects (osteoporosis and osteopetrosis), failure of skeletal repair, failure to maintain prosthetic implants as well as fusion of host and viral membranes in viral diseases. Clearly, potential common regulators and mechanisms would be attractive therapeutic targets in these disorders. Two of the cell types that fuse, osteoclasts and multinucleated giant cells, are derived from a common progenitor, are rendered fusion competent by common molecular mediators and ultimately regulate specialized functions in specific microenvironments. While OC can undergo fusion in both normal or pathological states such as Paget's disease 33 , macrophages fuse to form MGC primarily under inflammatory conditions such as chronic granulomatous disease or the foreign body response 34 . However, the fact that the fusion of OC and MGCs are governed by common signaling mechanisms again makes these and their component molecules attractive targets for therapeutic intervention. Interestingly, we have previously shown that in response to ischemic injury, CD13 KO skeletal muscle satellite cells fused more readily than WT cells to form multinucleated myoblasts, suggesting that CD13 may also participate in fusion of other cell types, thus affecting processes such as skeletal muscle repair 16 .
Osteoclastogenesis comprises many steps from the commitment and survival of osteoclast progenitor cells, their differentiation into mononuclear pre-osteoclasts that fuse to generate multinucleated mature osteoclasts and finally, activation of osteoclasts for bone resorption. Among the different steps, osteoclast fusion is thought to be the critical step in this phenomenon. Our data clearly indicate that osteoclast progenitor survival, differentiation and proliferation is not dependent on CD13 expression, suggesting that CD13 may be specifically involved in the fusion mechanism to generate multinucleated osteoclasts.
Defective osteoblastic bone forming activity can also contribute to osteolysis. Previously, we have shown that CD13 expression does not affect mesenchymal stem cell formation or their survival 16 , confirming that elevated levels of osteoprogenitors in the in vitro osteoblastogenesis analysis is not due to differences in mesenchymal stem cell formation. We showed that despite an increase in the osteoblast progenitor population in the absence of CD13, bone formation rate and mineral apposition rate remain unaltered between genotypes, indicating that impaired skeletal mass is not due to a defect in mature osteoblast function. While dynamic bone formation is equivalent between genotypes of young mice, CD13 KO aged mice have a significantly reduced bone formation rate compared to their wildtype counterparts, perhaps implicating a CD13 contribution of osteoblast function in bone remodeling during aging.
Studies have implicated various membrane-associated processes as critical to osteoclastogenesis such as clustering of membrane tetraspanins 35 , endocytosis of surface molecules 36 as well as the initial increases in    20 and Src activation 42 , each of which have been shown to participate in OC fusion. In the current study, fusion protein expression in CD13 KO OCs is aberrantly sustained by a posttranscriptional mechanism. Alternatively, organizers of actin-based protrusions are also pivotal in myeloid cell as well as gamete fusion 43 , which depends on step-wise reorganization of the actin cytoskeleton, initiated by formation of "podosome-like" membrane protrusions in myeloid cells [44][45][46][47][48] . Importantly, overexpression of DC-STAMP generates cells with numerous cell protrusions and increased fusogenic capacity. Similarly, the abundant filopodia in OC precursors and OC in active fusion are significantly downregulated as OCs mature 39,47,49 . Recently, we have reported that CD13 is a critical signaling platform that links the plasma membrane to dynamic mediators of actin cytoskeletal assembly and rearrangement 20 . We propose that CD13 may regulate the expression of these endocytic and fusion regulatory proteins, perhaps by mediating their internalization, endocytic trafficking, recycling and/or In conclusion, in the present study we demonstrate that CD13 expression controls osteoclastogenesis specifically at the level of cell-cell fusion. Further investigation into the relationship between CD13 and dynamin, DCST1 and CD9 and other regulators of osteoclast fusion such as OC-STAMP and the osteoclast receptor αvβ3 integrin 28 will clarify mechanisms regulating CD13-mediated cell fusion in osteoclasts as well as in fusion of other cell lineages including foreign body giant cells and satellite stem cells. Considering the diversity and importance of pathologies that are influenced by cell-cell fusion, identification of CD13-dependent molecular mechanisms and signaling that regulate myeloid fusion will provide novel therapeutic approaches in fusion pathologies.

Materials and methods
The authors confirm that all experiments were carried out in accordance with relevant guidelines and regulations. The ethical approval for all animal care and procedures were carried out in accordance with relevant guidelines and regulations by the UConn Health Institutional Animal Care and Use Committee.

Animals.
Global young (8-10 weeks) or aged (18-25 weeks) wildtype and CD13 KO (C57BL/6J) male and female mice were generated and housed at the Gene Targeting and Transgenic Facility at University of Connecticut School of Medicine 12 . All procedures were performed in accordance with the guidelines and regulations approved by the UCONN Health Institutional Animal Care and Use Committee. UConn Health is fully accredited by Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) International and Public Health Service (PHS) assurance number is A3471-01 (D16-00295) and USDA Registration Number is 16-R-0025.
In vivo analysis of WT and CD13 KO mice. Micro-computed tomography and histomorphometry. Samples were scanned in a density-calibrated µCT40 (Scanco Medical, Bassersdorf, Switzerland) in PBS at 8 µm 3 resolution with the following settings: 55 kV, 145 µA, 300 ms integration, 1000 projections/rotation with Gaussian filtering. Analysis was performed following standard guidelines 51 . Briefly, femoral trabeculae were auto-contoured in a 120-slice region 1 mm proximal to the distal condyles with a lower threshold of 2485 Hounsfield units (HU), and femoral cortex was auto-contoured in an 80-slice region just distal to the third trochanter with a lower threshold of 4932HU 50,52 . All samples were scanned, reconstructed, and analyzed in a Scanco µCT40 running Evaluation Program V6.6 (http:// www. scanco. ch/ en/ syste ms-solut ions/ softw are. html).

Isolation of hematopoietic progenitor population from bone marrow and spleen. BM cells
were obtained by flushing femur and tibia from WT or CD13 KO mice with 10 ml 1 × PBS and 2% heat inactivated FBS, followed by RBC lysis and filtering through 40 μm cell strainer (BD Biosciences). Total live cells counted with Countess Automated cell counter (Thermo Fisher Scientific) were stained with antibody cocktail at 4 °C. Cells from mouse spleen was obtained by gentle crushing the organ between frosted microscopic slides in cold 10 ml 1 × PBS and 2% heat inactivated FBS.
Bone resorption assay 52 . Flow-sorted osteoclast progenitors derived from BM were seeded on Osteo Assay plate (Corning) at a density of 50,000 cells/well in α-MEM containing 10%FBS, 1% Penicillin-Streptomycin, 30 ng/ml M-CSF and 30 ng/ml RANKL for d10. Surface pit formation was measured by removing cells with 100 μl of 10% bleach solution at RT for 5 min. Wells were washed with deionized water and allowed to dry. Cluster of pits formed was imaged using a light microscope (Olympus Scientific), using Olympus cellSens Dimension V0118 software (Olympus Scientific) (https:// www. olymp us-lifes cience. com/ en/ softw are/ cells ens/) and the area of resorption was measured by ImageJ software (https:// imagej. nih. gov/ ij/).

Generation of CD13 KO U937 cells by CRISPR-Cas9 gene editing.
Oligos containing the guide RNA sequence for human CD13: 5′-CAG TGC GAT GAT TGT GCA CA-3′; guide RNA for scrambled control: 5′-CAG TCG GGC GTC ATC ATG AT-3′ were cloned into lentiCRISPR v2 (addgene plasmid #52961). Packaging plasmids psPAX2 and pMD2.G (addgene plasmid #12260 and addgene plasmid #12259) were utilized to generate the lentivirus as described 20 in presence of 2 ng/µl puromycin to select for lentiCRISPR integration. Multiple CD13 KO and scrambled control clones confirmed by immunoblot, IF and flow cytometry analysis were employed for OC generation.
Isolation of TRAP + osteoclast-like cells from human monocytic cell line U937. U937 cells expressing CD13 CRISPR or scrambled control 20 were grown at a density of 100,000 cells/well in 4-well dish with RPMI 1640 containing 10% heat inactivated fetal bovine serum, 1 mM l-glutamine, 1 mM sodium pyruvate, 1% penicillin/ streptomycin and 0.1 µg/ml PMA for 3 days. Non-adherent cells were removed and adherent cells were stimulated with M-CSF (60 ng/ml) and RANKL (100 ng/ml) in RPMI medium without PMA for an additional 10 days. Cells were fixed in 2.5% glutaraldehyde and osteoclast-like cells identified by TRAP staining.
Immunofluorescence and microscopy. Flow-sorted osteoclast progenitors were grown on glass coverslips that were previously coated with 5 µg/ml fibronectin for indicated time period. Cells were fixed in 4% paraformaldehyde (Electron Microscopy Sciences) at RT for 30 min, permeabilized with 0.1% Triton-X-100 in PBS at RT for 5 min. Cells were blocked with blocking buffer containing 5% goat or donkey serum/5% BSA/1 × PBS at RT for 1 h followed by incubation with primary Ab in blocking buffer at 4 °C for overnight. Cells were washed and treated with secondary Ab (1:1200) and DAPI (nuclear stain) in blocking buffer at RT for 1 h. Coverslips were mounted with ProLong Gold antifade mounting medium (Life Technologies), visualized at excitation www.nature.com/scientificreports/ wavelength of 488 nm (Alexa 488), 543 nm (Alexa 594 or TRITC) and 405 nm (DAPI) and imaged by Zeiss LSM 880 confocal fluorescence microscope and analyzed by using Zeiss Zen 2.0 Pro blue edition software (https:// www. zeiss. com/ conte nt/ dam/ Micro scopy/ Downl oads/ Pdf/ FAQs/ zen2-blue-editi on_ insta llati on-guide. pdf).

Immunoblot analysis. Flow-sorted BM osteoclast progenitors grown in presence of M-CSF and RANKL
for 48 h were LPC (lysophosphatidylcholine)-synchronized by treating with reversible fusion inhibitor LPC (100 µM) for 12 h followed by washing and growing cells in LPC-free medium for fusion to proceed for 0-72 h. Cell lysates were harvested in 1% NP40 lysis buffer containing 1× complete Protease Inhibitor cocktail (Roche). Samples were separated by SDS-PAGE and transferred to nitrocellulose membrane, blocked in 1XTBST containing 5% bovine serum albumin, treated with primary Ab followed by appropriate secondary Ab and imaged by ChemiDoc Imaging system version 3.0.1 (https:// www. bio-rad. com/ en-us/ categ ory/ chemi doc-imagi ng-syste ms? ID= NINJ0 Z15) (Biorad). β actin or GAPDH were used as loading controls. Gels/blots were cropped and indicated by dividing lines.

Statistical analysis.
Statistical analysis was performed using unpaired, two-tailed Student's t test using GraphPad Prism software and results are representative of mean ± SD or ± SEM as indicated. Differences at p ≤ 0.05 were considered significant.

Disclosures.
All animal experiments in this study were reviewed and approved by Animal Care Committee