Human osteoclastogenesis in Epstein-Barr virus-induced erosive arthritis in humanized NOD/Shi-scid/IL-2Rγnull mice

Many human viruses, including Epstein-Barr virus (EBV), do not infect mice, which is challenging for biomedical research. We have previously reported that EBV infection induces erosive arthritis, which histologically resembles rheumatoid arthritis, in humanized NOD/Shi-scid/IL-2Rγnull (hu-NOG) mice; however, the underlying mechanisms are not known. Osteoclast-like multinucleated cells were observed during bone erosion in this mouse model, and therefore, we aimed to determine whether the human or mouse immune system activated bone erosion and analyzed the characteristics and origin of the multinucleated cells in hu-NOG mice. Sections of the mice knee joint tissues were immunostained with anti-human antibodies against certain osteoclast markers, including cathepsin K and matrix metalloproteinase-9 (MMP-9). Multinucleated cells observed during bone erosion stained positively for human cathepsin K and MMP-9. These results indicate that human osteoclasts primarily induce erosive arthritis during EBV infections. Human osteoclast development from hematopoietic stem cells transplanted in hu-NOG mice remains unclear. To confirm their differentiation potential into human osteoclasts, we cultured bone marrow cells of EBV-infected hu-NOG mice and analyzed their characteristics. Multinucleated cells cultured from the bone marrow cells stained positive for human cathepsin K and human MMP-9, indicating that bone marrow cells of hu-NOG mice could differentiate from human osteoclast progenitor cells into human osteoclasts. These results indicate that the human immune response to EBV infection may induce human osteoclast activation and cause erosive arthritis in this mouse model. Moreover, this study is the first, to our knowledge, to demonstrate human osteoclastogenesis in humanized mice. We consider that this model is useful for studying associations of EBV infections with rheumatoid arthritis and human bone metabolism.

Introduction the Nihon University Biosafety Committee for Gene Recombination (certification number, 2006-M). 31 mice were prepared for the experiment, health checkups were performed thrice a week, and weight measurements were taken once a week. In order to reduce the pain of the mice, the experiment was planned such that the timing of euthanasia coincided with the time when the mice exhibited anguish symptoms or significant weight loss (20% or more per week) during the course of the experiment. Seven-week-old female NOG mice were transplanted with human cord blood CD34 + HSCs (2C-101, Lonza, Basel, Switzerland) at a rate of 8.0-10 × 10 4 cells/mouse via the tail vein. The reconstituted human immune system components were evaluated and characterized by monitoring the percentages of human CD3 + , CD4 + , CD8 + , CD19 + , and CD45 + cells in mouse PBLs using flow cytometry.

Preparation of EBV and infection of hu-NOG mice with EBV
The protocols for our experiments with EBV were approved by the Biorisk Management and Control Committee of Nihon University School of Medicine (certification number, 20-13-5). For preparing EBV, cells of the EBV-producer line B95-8 (JCRB9123, Japanese Collection of Research Bioresources, Osaka, Japan) were cultured with Roswell Park Memorial Institute (RPMI)-1640 medium (Sigma Aldrich, St. Louis, MO, USA), supplemented with penicillin G (100 U/mL), streptomycin (100 μg/mL), and 10% fetal bovine serum (FBS) at 37˚C in a 5% CO 2 incubator. The culture supernatant of the B95-8 cells was collected 7 days after the final medium change. The supernatant was removed after centrifugation at 400 × g for 5 min at 4˚C. The enriched virus fluid was filtered through a 0.45-μm membrane and stored at −80˚C.
For titrating EBV, donor cord blood samples were obtained with written informed consent of the donor's parents. The protocols for our titration experiments with donor cord blood samples were approved by the Nihon University Itabashi Hospital Clinical Research Judging Committee (certification number, RK-140613-11). Mononuclear cells isolated from cord blood were plated at a density of 2.0 × 10 5 cells/well in 96-well plates and then inoculated with serial 10-fold dilutions of the virus preparation. The number of wells with clumps of proliferating cells was counted 6 weeks after infection, and the titer of the virus in 50% transforming dose (TD 50 ) was determined using the Reed-Muench method [26]. After 3-4 months of human HSC transplantation, the hu-NOG mice were inoculated with EBV at a dose of 1.0-2.0 × 10 1 TD 50 /100 μL/mouse via the tail vein.

Histochemistry of knee joint tissues
EBV-infected and -uninfected hu-NOG mice were sacrificed via cervical subluxation by trained technicians. After confirmation of death, the knee joints were dissected out from these animals, fixed in 10% formaldehyde solution, and embedded in paraffin. Serial sections were generated along the longitudinal bone axis from the paraffin-embedded samples and subjected to staining for tartrate-resistant acid phosphatase (TRAP) and with hematoxylin-eosin. The sections of EBV-infected and -uninfected hu-NOG mice were stained for human cathepsin K and human matrix metalloproteinase-9 (MMP-9) using immunohistochemistry.

Examination of mouse knee joint using three-dimensional computed tomography
EBV-infected and -uninfected hu-NOG mice were euthanized and three-dimensional images of the knee joints were constructed from multiple tomographic images using a high-definition microfocus X-ray computed tomography scanner (Kureha Special Laboratory Co., Ltd., Fukushima, Japan).

Bone marrow cell culture
Osteoclasts are generally derived from their progenitor cells of the monocyte/macrophage lineage in the bone marrow [27]. Long bones, such as the femur of the EBV-infected and -uninfected mice were dissected and cut off at the epiphysis. The marrow was flushed out with RPMI-1640 medium (Sigma Aldrich) containing 10% heat-inactivated FBS using a 26-gauge needle attached to a 1.0-mL syringe and collected in a 1.5-mL tube. After centrifugation (200 × g for 15 min at 26˚C), bone marrow cells were obtained from EBV-infected mice, suspended in osteoclast growth medium (PT9501, Lonza) containing recombinant human macrophage colony-stimulating factor (M-CSF) (33 ng/mL) and soluble human receptor activator of nuclear factor κB ligand (RANKL) (66 ng/mL), and supplemented with L-glutamine (0.1%), penicillin G (100 U/mL), streptomycin (100 μg/mL), and 10% FBS. Bone marrow cells from EBV-uninfected mice and commercially available mouse osteoclast progenitor cells (OSC14C, Cosmo bio, Tokyo, Japan) were suspended in (human) osteoclast culture medium (OSCMHB, Cosmo bio) and (mouse) osteoclast culture medium (OSCMM, Cosmo bio), respectively. These bone marrow cells and the mouse osteoclast progenitor cells were then plated into chamber slides (Laboratory-Tek 8-well Permanox Slides, Thermo Scientific/Thermo Fisher, Grand Island, NY, USA) at a density of 1.0 × 10 4 cells/well and incubated at 37˚C in a 5% CO 2 incubator. After 10-14 days of culture, the chamber slides were subjected to cytochemistry.
Pit formation assay was performed as follows. Bone marrow cell preparations from EBVinfected hu-NOG mice obtained were placed in osteo assay surface plates coated with a synthetic inorganic bone mimetic calcium phosphate (Corning, Kennebunk, ME, USA), which allows direct assessment of osteoclast activity in vitro [28], and incubated with osteoclast growth medium (Lonza) in the presence of recombinant human M-CSF (33 ng/mL) and soluble human RANKL (66 ng/mL) at 37˚C in a 5% CO 2 incubator. After 10 days of incubation, the plates were stained for TRAP and examined using light microscopy.
Osteoclasts were separated from bone marrow cells using a previously reported method of culturing osteoclasts in vitro as adherent cells [29]. The pelvic bones of the EBV-infected and -uninfected mice were dissected and cut into several pieces. The marrow was flushed out with RPMI-1640 medium (Sigma Aldrich) containing 10% heat-inactivated FBS and collected in a 1.5-mL tube. The bone marrow cells obtained from pelvic bone were filtered using a 40 μm cell strainer and stored at −80˚C using Cell Banker (Nippon Zenyaku Kogyo Co., Ltd, Fukushima, Japan) until use. After thawing, the bone marrow cells from pelvic bones were cultured in α-minimal essential medium (MEM) (Gibco/Thermo Fisher, Grand Island, NY, USA) containing penicillin G (100 U/mL), streptomycin (50 μg/mL), gentamicin (50 μg/mL), and 20% FBS overnight at 37˚C in a 5% CO 2 incubator to remove adherent cells. After 12 hours, the non-adherent cells were collected and seeded on glass-bottom dishes. After 3 days of culture, the non-adherent cells were washed out with fresh media. After 8 days of culture, the adherent cells were subjected to cytochemistry.

TRAP staining
The plates of serial sections from the knee joint samples, the chamber slides of cultured bone marrow cells of long bones, the pit formation assay plates of cultured bone marrow cells of long bones, and glass-bottom dishes of cultured bone marrow cells of pelvic bones were subjected to staining for TRAP using the acid phosphatase leukocyte kit (3864-1KT, Sigma Aldrich). The samples were fixed in a citrate/acetone solution. After rinsing in deionized water, they were incubated with a mixture of acetate solution, naphthol AS-BI phosphoric acid solution, tartrate solution, and Fast Red Violet LB salt solution (F3881, Sigma Aldrich) in a dark room. After washing, serial sections in the plates and cultured bone marrow cells of pelvic bones in glass-bottom dishes were stained with hematoxylin (cultured cells on the chamber slides and pit formation assay plates were not stained).

Statistical analysis
The number of CD34 + cells transplanted between the EBV-infected and EBV-uninfected groups was analyzed using the Mann-Whitney U-test. The incidence rate of bone erosion between the EBV-infected and EBV-uninfected groups was compared using Fisher's exact test. The relationship between the severity of bone erosion and titer of EBV inoculated was analyzed using the Mann-Whitney U-test. These analyses were performed using Excel for Mac 2011 version 14.7.7 (Microsoft, Redmond, WA, USA). A P value < 0.01 was considered to indicate a statistically significant difference.

Results
The 31 NOG mice used in this study were humanized, and 21 of these were infected with EBV. There was no mortality except for the planned euthanasia.

Bone erosion induced by EBV infection
In this study, we initially confirmed our previous findings of rapid increase in the number of CD8 + cells and the reversal of CD4/CD8 ratio in the peripheral blood of humanized mice following EBV infection [24]. The percentages of CD4 + cells and CD8 + cells in the PBLs of EBVinfected (n = 10) and EBV-uninfected mice (n = 5) were sequentially assessed using flow cytometry once per week. In all 5 EBV-uninfected mice, the CD4/CD8 ratio remained above 1.0 during our observation period. In contrast, the CD4/CD8 ratio in EBV-infected mice decreased to less than 1.0 in all 10 EBV-infected mice, reaching minimal values 7-10 weeks after EBV inoculation in most mice. The changes in the percentages of CD4 + , CD8 + , CD19 + and CD45 + cells in PBLs of EBV-infected mice are shown in Fig 1. As the rapid decrease in the CD4/CD8 ratio indicated successful infection with EBV [24], the histology of the knee joint tissues of mice was analyzed. Breakdown of the bone surface was observed in the area close to the knee joint capsule in all 10 EBV-infected mice, whereas no sign of bone destruction was detected in any of the five EBV-uninfected mice (Table 1). Although the number of CD34 + cells transplanted between the EBV-infected and EBV-uninfected groups did not show any correlation (P = 0.052), the incidence rate of bone erosion was significantly high in the EBVinfected group (P = 0.0003). The histological severity of bone erosion around the knee joint was classified into three grades according to the depth of cortical bone breakdown (CBB): CBB � 2/3 rd of the bone thickness, 1/3 rd � CBB < 2/3 rd of the bone thickness, and

PLOS ONE
CBB < 1/3 rd of the bone thickness were defined as grades 3+, 2+, and 1+, respectively (Typical knee joint tissue images are shown in S1 Fig). Table 1 shows the severity of bone erosion, the number of transplanted CD34 + HSCs, and the dose of inoculated EBV in each mouse. Although variation in the severity of bone erosion was observed among individual EBVinfected mice, correlation between the severity of bone erosion and titer of inoculated EBV was not observed (P = 1.0). Three-dimensional computed tomography (3D-CT) images of knee joints revealed that bone erosive changes resembled RA in EBV-infected mice (n = 2) at locations where histological analysis identified CBB, whereas no such changes were observed in uninfected mice (n = 2). The 3D-CT images of the knee joint in representative mice are shown in Fig 2.

Properties of multinucleated cells present in bone erosion sites
To determine the characteristics of the multinucleated cells, serial sections of knee joint tissues from EBV-infected (n = 2) and EBV-uninfected mice (n = 2) were stained for TRAP, human cathepsin K, and human MMP-9. Bone erosion lesions were observed in both EBV-infected

Differentiation of osteoclasts from the bone marrow cell culture
Human osteoclasts were identified in the bone erosion sites of EBV-infected hu-NOG mice, as mentioned above. However, whether human osteoclasts can develop in HSC-transplanted humanized mice, such as hu-NOG mice, was not known. Therefore, we used a procedure known to induce in vitro osteoclast differentiation on bone marrow cells from long bones of EBV-infected hu-NOG mice (n = 2) and EBV-uninfected hu-NOG mice (n = 2).  Next, using the pit formation assay, we investigated whether these human osteoclasts derived from the bone marrow of EBV-infected hu-NOG mice showed the functional characteristics of osteoclasts (n = 3). The results showed that the TRAP-positive multinucleated cells derived from the bone marrow of these EBV-infected mice formed many resorption pits on the surface of plates coated with synthetic bone mimetics (Fig 4).
According to the manufacturer's specifications, the antibodies specific to the human MMP-9 and human mitochondrial protein used in this experiment do not react with mouse proteins. For confirmation, we examined the species specificity of these antibodies. The images of the multinucleated cells cultured from mouse osteoclast progenitor cells are shown in Fig 5. These cultured multinucleated cells were stained positively with antibodies that react with mouse mitochondria and mouse MMP-9 proteins. However, those were not stained with the antibodies specific for the human MMP-9 and human mitochondrial protein.
Furthermore, to determine whether the human osteoclasts developed from bone marrow cells without human RANKL and human M-CSF, the bone marrow cells from pelvic bones of EBV-infected mice (n = 3) were cultured in α-MEM for 8 days. The cultured adherent cells were stained for TRAP and human MMP-9. The images of the cultured representative multinucleated cells are shown in Fig 6. Several multinucleated cells were observed in the cultured bone marrow cells. These multinucleated cells stained positively for TRAP and with the antihuman MMP-9 antibody. In contrast, adherent cells could not be cultured from the bone marrow of EBV-uninfected mice (S2 Fig).

Discussion
In the present study, we established experimental procedures to reproducibly induce erosive arthritis in hu-NOG mice using EBV inoculation. Using these procedures, we confirmed that EBV indeed induces erosive arthritis, which histologically resembles RA in hu-NOG mice. Furthermore, 3D-CT definitively revealed bone destruction in the knee joints of EBV-infected mice, providing diagnostic imaging evidence similar to that observed in patients with RA. Histological analysis of 10 EBV-infected mice revealed inter-individual variation in the level of bone erosion ranging from 1+ to 3+. Although these data were obtained for other EBVinfected hu-NOG mice under different experimental conditions, we determined the bone mineral density (BMD) of trabecular bone at the distal portion of the femur. Comparison of BMD between mice with severe erosion and those with mild erosion indicated that the former had significantly lower BMD grades than the latter (S3 Fig), suggesting that the severity of bone erosion was affected by the activity level of the bone-resorbing osteoclasts.

Image of cultured bone marrow cells from EBV-uninfected hu-NOG mice in the absence of differentiation signals in glass-bottom dish
We showed that the multinucleated cells observed in the bone erosion sites of EBV-infected mice were TRAP-positive osteoclasts expressing human cathepsin K and human MMP-9. As osteoclasts are critical effectors of bone destruction, this result strongly suggested that human osteoclasts play a major role in bone erosion of EBV-infected hu-NOG mice. To further confirm this observation, we investigated the presence of human osteoclast progenitor cells in hu-NOG mice and analyzed whether the bone marrow cells of NOG mice possess the potential to differentiate into human osteoclasts. Upon in vitro culture in the presence of human M-CSF and soluble human RANKL, bone marrow cells obtained from EBV-infected mice differentiated into multinucleated cells expressing TRAP, human cathepsin K, and human mitochondrial protein. As these cells showed bone absorption ability in the pit formation assay, we concluded that human mature osteoclasts can differentiate from bone marrow cells of EBV-infected hu-NOG mice. To our knowledge, this is the first demonstration of human osteoclast differentiation in humanized mice engrafted with human CD34 + HSCs. This showed that human osteoclast progenitor cells were present in EBV-infected hu-NOG mice. Based on these results, we concluded that EBV, which do not infect mouse cells, infected human cells differentiated from HSCs transplanted in NOG mice, and EBV infections induced human osteoclast differentiation from human osteoclast progenitor cells and resulted in bone erosion. Upon in vitro culture in the presence of human M-CSF and soluble human RANKL, bone marrow cells obtained from EBV-uninfected mice differentiated into multinucleated cells expressing human cathepsin K, human MMP-9, and human mitochondrial protein. This result indicates that similar human osteoclast progenitor cells are present in the bone morrow cells of EBV-uninfected mice. In the in vitro bone marrow cell culture of EBV-infected hu-NOG mice lacking the human M-CSF and soluble human RANKL signaling molecules, the multinucleated cells expressed TRAP and human MMP-9. In contrast, adherent cells, which tended to develop into osteoclasts in in vitro cell culture, could not be cultured from the bone marrow cells of EBV-uninfected mice. This indicated that the bone marrow of EBV-infected hu-NOG mice provided an environment in which human osteoclast progenitor cells differentiated into human osteoclasts and induced their excessive differentiation to human osteoclasts and this aberrant activation resulted in bone erosion in hu-NOG mice. Osteoclast differentiation from their progenitors requires M-CSF binding to its receptor CSF1R (CD115) and RANKL binding to RANK. In normal bone remodeling, osteoblasts provide RANKL and M-CSF for osteoclastogenesis, and bone homeostasis is maintained by the balance between bone-resorbing osteoclasts and bone-synthesizing osteoblasts [34]. Therefore, excessive M-CSF and/or RANKL expression possibly results in excessive osteoclastogenesis. M-CSF expression in RA has been reported to increase in the synovial tissue [35]. Furthermore, some reports state that synovial fibroblasts and/or activated T lymphocytes produce RANKL, which triggers aberrant differentiation and activation of osteoclasts, causing bone destruction [36,37].
What is the mechanism underlying human osteoclast differentiation and activation in EBV-infected hu-NOG mice? Regarding CSF1R, reports have shown that along with M-CSF, IL-34 also functions as an agonist. In addition, mouse M-CSF does not react with human CSF1R, although the amino acid sequences of mouse IL-34 is highly homologous to that of human IL-34 [38]. Therefore, human osteoclast progenitor cells possibly receive the differentiation signal from human M-CSF, human IL-34, and/or mouse IL-34 in EBV-infected hu-NOG mice. The tissue distribution of M-CSF and IL-34 differs. Reports have shown that osteoclasts in the osseous tissue are differentiated and maintained by M-CSF, whereas those in the spleen are maintained by M-CSF and/or IL-34 [39]. Thus, the mechanism underlying osteoclast differentiation involving CSF1R, which exists in the human osteoclast progenitor cells of EBV-infected hu-NOG mice, might be complex. In the absence of evidence showing that mouse RANKL binds to human RANK, it appears unlikely that cells of mouse origin, including fibroblasts and stromal cells, may participate in human osteoclast differentiation via mouse RANKL-human RANK interaction.
EBV primarily infects human B cells and transforms them into lymphoblastoid cells of the activated B-cell phenotype [23]. EBV-specific T-cell responses have been demonstrated in EBV-infected hu-NOG mice [24]. EBV-infected lymphoblastoid cells and/or activated T cells responding to the virus might have triggered the aberrant differentiation and activation of human osteoclasts in hu-NOG mice. In our previous study, we identified numerous EBVinfected cells, as well as both CD4 + and CD8 + human T cells, in the edematous bone marrow adjacent to the affected joints, whereas only a few EBV-infected cells were present in the synovial tissue of these joints [25]. Hence, we suggested that in the bone marrow, EBV-infected cells and/or responding human T cells may induce differentiation and activation of osteoclasts, leading to bone erosion. Further investigations are required to elucidate the mechanism of differentiation and excessive activation of human osteoclasts in EBV-infected hu-NOG mice.
The present results suggest that EBV infections promote human osteoclast differentiation in the present mouse model, resulting in erosive arthritis, and the EBV-infected hu-NOG mice have the reproducibility of erosive arthritis resembling RA, which can be established in a mouse model. However, the mechanisms underlying human osteoclastogenesis remain unclear. Further studies are required to determine the sources of human osteoclast differentiation signals including RANKL and M-CSF. Furthermore, this model can help confirm the therapeutic effects of anti-human RANKL antibody or cathepsin K inhibitor. However, a mouse model displaying activated human osteoclasts is extremely valuable and is potentially applicable in studies on the association between EBV infections and RA and with human bone metabolism using mouse models such as those of osteoporosis. Bone mineral density was compared between EBVinfected mice having 3+ or 2+ bone erosion (n = 5) and EBV-infected mice having 1+ bone erosion (n = 5) or EBV-uninfected mice having no bone erosion (n = 5). Severe bone erosion group had significantly lower BMD grades than those of mild erosion group. Values represent mean ± standard deviation. Statistical analysis was performed using student's t-test ( � P value = 0.0439). (TIFF)