Osteoclast Activated FoxP3+ CD8+ T-Cells Suppress Bone Resorption in vitro

Background Osteoclasts are the body’s sole bone resorbing cells. Cytokines produced by pro-inflammatory effector T-cells (TEFF) increase bone resorption by osteoclasts. Prolonged exposure to the TEFF produced cytokines leads to bone erosion diseases such as osteoporosis and rheumatoid arthritis. The crosstalk between T-cells and osteoclasts has been termed osteoimmunology. We have previously shown that under non-inflammatory conditions, murine osteoclasts can recruit naïve CD8 T-cells and activate these T-cells to induce CD25 and FoxP3 (TcREG). The activation of CD8 T-cells by osteoclasts also induced the cytokines IL-2, IL-6, IL-10 and IFN-γ. Individually, these cytokines can activate or suppress osteoclast resorption. Principal Findings To determine the net effect of TcREG on osteoclast activity we used a number of in vitro assays. We found that TcREG can potently and directly suppress bone resorption by osteoclasts. TcREG could suppress osteoclast differentiation and resorption by mature osteoclasts, but did not affect their survival. Additionally, we showed that TcREG suppress cytoskeletal reorganization in mature osteoclasts. Whereas induction of TcREG by osteoclasts is antigen-dependent, suppression of osteoclasts by TcREG does not require antigen or re-stimulation. We demonstrated that antibody blockade of IL-6, IL-10 or IFN-γ relieved suppression. The suppression did not require direct contact between the TcREG and osteoclasts. Significance We have determined that osteoclast-induced TcREG can suppress osteoclast activity, forming a negative feedback system. As the CD8 T-cells are activated in the absence of inflammatory signals, these observations suggest that this regulatory loop may play a role in regulating skeletal homeostasis. Our results provide the first documentation of suppression of osteoclast activity by CD8 regulatory T-cells and thus, extend the purview of osteoimmunology.


Introduction
The skeletal system is dynamically and constantly remodeled throughout life to maintain bone integrity. There are multiple layers of regulation imposed on the skeletal system homeostasis, including physiological levels of phosphate, calcium, hormones, mechanical loading (e.g. Wolff's law) and energy metabolism (reviewed in [1]). Two cells play a key role in remodeling bone: osteoclasts and osteoblasts. Osteoclasts are large multinucleated cells that are the principal, if not sole, bone resorbing cells in the body. Osteoclasts are derived from the myeloid lineage and therefore may be considered a specialized immune cell. Balancing the function of the osteoclasts are osteoblasts, of mesenchymal origin, which form new bone. Osteoblasts also provide essential signals for, and regulate the differentiation of, the myeloid lineage osteoclast precursors by producing macrophage colony-stimulating factor (M-CSF), receptor activator of NF-kB ligand (RANKL; Tnfsf11) and other co-stimulatory factors in the bone marrow [2,3,4,5].
It has been recognized in the last decade that skeletal homeostasis is dynamically influenced by the immune system. This emerging field, called osteoimmunology [6], arose from observations demonstrating that lymphocyte-derived cytokines, including RANKL, interleukin (IL)-17 and type I and II interferons, are potent mediators of osteoclast function and osteoclastogenesis [7,8,9,10,11,12,13]. Osteoclast activity and numbers are increased by cytokines produced by effector T-cells leading to bone erosion in inflammatory diseases such as rheumatoid arthritis and periodontitis. T-cell produced cytokines also play a critical role in bone cancers, post-menopausal osteoporosis and in Paget's disease [14,15,16,17].
The immune system also maintains two counterbalancing cell types: the effectors (e.g. T H 17), which are dominant during the inflammatory phase, and the regulatory T-cells (T REG ). The transcription factor FoxP3 is a marker of T REG that have the ability to suppress aberrant activation of self-reactive lymphocytes. Loss of FoxP3 function results in fatal autoimmune pathology affecting multiple organs in both humans and mice [18,19,20,21]. Adoptive transfer of T-cells expressing FoxP3 into mice with FoxP3 loss-of-function abolishes the autoimmune pathology [22,23,24,25].
Regulatory T-cells that express FoxP3 also express CD25, the achain of the IL-2 receptor. The transfer of CD4 T-cells depleted of the CD25 + fraction (,10%) from a normal adult mouse into a mouse lacking an intact immune system produces autoimmune disease [26]. Conversely, transfer of the CD25 + CD4 T-cells from normal mice into T-cell-deficient mice suppressed allergy and prevented graft-versus-host disease after bone marrow transplantation [27]. T REG mediate their regulatory function through a number of mechanisms. First, T REG express anti-inflammatory cytokines including IL-10, TGFb and IL-35 [28,29,30,31]. Another mechanism of regulation is by cell-cell contact: cytotoxic Tlymphocyte antigen-4 (CTLA-4) expressed on T REG binds with ,10 fold higher affinity to co-stimulatory B7 molecules on antigen presenting cells (APC) than CD28, and thus prevent APC from activating naïve T-cells [32]. T REG have also been proposed to prevent differentiation of effector T-cells by consuming IL-2, IL-4 and IL-7 required for T-cell activation and polarization [33].
Compared to CD4 T REG , the FoxP3 + CD8 T-cells (referred to as Tc REG here) have not been extensively studied due in part to their low abundance in lymphoid tissue, and the ability of CD4 T REG to regulate activation of both CD4 and CD8 T-cells [34,35,36,37]. While a few recent studies have indicated that Tc REG may also regulate the immune system [38,39,40], their physiological role in immune regulation has not been definitively established. Other regulatory CD8 T-cells that do not express FoxP3 have also been observed and studied [41].
Here we examined the interaction between osteoclasts and Tc REG . We uncovered the osteoclast-CD8 T-cell interaction using a time-series microarray dataset to study osteoclastogenesis. Our data showed that RANKL treatment of bone marrow monocytes coordinately induced a number of chemokines and MHC class I presentation pathway during differentiation [42]. In follow up experiments, we showed that the osteoclast-secreted chemokines preferentially recruited CD8 T-cells. Cross-presentation (i.e. presentation of peptides from extracellular proteins on MHC class I) of antigens to CD8 T-cells induced IFN-c, IL-6 and IL-2 only in the presence of antigen. Antigen presentation by osteoclasts has also been subsequently demonstrated for human osteoclasts derived from peripheral blood mononuclear cells [43]. The priming of naïve CD8 T-cells by osteoclasts induced FoxP3, CD25, CTLA-4, RANKL, and CD122 [44]. The osteoclastprimed FoxP3 + CD8 T-cells were able to suppress proliferation of naïve responder CD8 T-cells by DCs and thus have a potential regulatory T-cell activity [44]. Interestingly, the Tc REG express IFN-c and CTLA-4, which inhibit osteoclast activity [45,46,47,48,49,50], while RANKL increases osteoclast activity. IL-6 directly inhibits osteoclast activity but can induce RANKL expression in osteoblasts to increase osteoclast function [14,51]. Thus, these T-cell produced cytokines could potentially activate or suppress osteoclast activity. We therefore tested if Tc REG could regulate osteoclast resorption activity.

Induction of Tc REG by Mature Osteoclasts is Antigendependent
We first show (Fig. 1A) that mature osteoclasts can induce FoxP3 and CD25 in ovalbumin (OVA)-specific OT-I CD8 T-cells only in the presence of antigen. The induction of FoxP3 in the presence of antigen was assayed by both multicolor flow cytometry (top panels) and by reverse transcription (RT) of RNA isolated from the co-culture, followed by PCR (middle panel). As controls, we isolated RNA from purified GFP positive and negative CD4 and CD8 T-cells from the transgenic FoxP3 eGFP reporter mouse [52,53]. These reporter mice co-express the enhanced green fluorescent protein (eGFP) and transcription factor FoxP3 under control of the endogenous promoter. We show that only mature osteoclasts (incubated in the presence of RANKL for 4 days) have the ability to induce FoxP3 as the osteoclast precursors (bone marrow derived monocytes) could not induce the regulatory phenotype in CD8 T-cells.
TGFb1 can induce regulatory T-cells in the periphery and in vitro [24,54,55,56]. Avian osteoclasts have been reported to express the latent form of TGFb [57]. To assess if murine osteoclasts express TGFb that could induce Tc REG , we measured the expression of TGFb by quantitative real-time PCR (qRT-PCR). The relative expression (2 DDCt with respect to b-actin) of TGFb1 in osteoclasts (+M-CSF, +RANKL) over osteoclast precursors (+MCSF, -RANKL) was determined to be 5.560.8 fold. The relative expression decreased to 0.960.6 fold (about the same as in the precursors) in the presence of Tc REG , indicating that Tc REG suppress TGFb expression by osteoclasts. However, these results do not address the role of TGFb in Tc REG induction. Furthermore, as bone contains significant levels of TGFb that is released by the action of osteoclasts [58,59,60,61], and the effect of TGFb can be dose-dependent, we added increasing amounts of TGFb to the osteoclast and OT-I co-cultures to test its effect on the induction of Tc REG . While adding an antibody that neutralizes all three forms of TGFb at 50 mg/ml efficiently blocked the induction of CD4 T REG , the antibody had no effect, at any dose tested, on the numbers of Tc REG induced (Fig. 1B left). Addition of recombinant murine TGFb1 also had no statistically significant effect on the induction of Tc REG by osteoclasts at any concentration of TGFb tested (Fig. 1B right). Taken together these results indicate that while osteoclasts express TGFb1, the cytokine is not required for induction of Tc REG by osteoclasts; furthermore, in the presence of Tc REG , TGFb1 mRNA levels decreased in osteoclasts.
Consistent with our previous observations [44], the induction of FoxP3 coincides with production of the cytokines IFN-c, IL-6, IL-10, again only in the presence of OVA (Fig. 1C top). All of the cytokines measured by ELISA were found to be expressed by the CD8 T-cells as measured by intracellular staining and flow cytometry and not by osteoclasts (Fig. 1C bottom). These results show that only mature osteoclasts have the ability to induce CD25, FoxP3 and the cytokines IFN-c, IL-6 and IL-10 in T-cells in an antigen-dependent manner. Finally, we did not detect any change in the levels of CD80, MHC class I surface expression, or detectable levels of MHC class II, on the osteoclasts after 48 hour of co-culture with CD8 T-cells or Tc REG (data not shown).

Tc REG Suppress Osteoclast Resorption Activity in vitro
To directly test if osteoclast-induced Tc REG affect resorption activity, we co-cultured purified ovalbumin-specific transgenic OT-I Tc REG with osteoclasts (1:1 ratio). The osteoclasts were grown on hydroxyapatite-coated slides or wells to measure resorption activity, rather than bone slices to avoid any confounding effects of TGFb present in the bone. The data is presented as the total area resorbed and the number of pits in the presence of T-cells relative to osteoclast only cultures. The co-cultures were performed with either ''in-situ'' produced Tc REG , or Tc REG induced by osteoclasts plated on tissue-culture plates, then transferred to osteoclasts growing on hydroxyapatitecoated slides.
Our results ( Fig. 2A, B) show that the in situ produced Tc REG suppressed osteoclast resorption activity. We also tested TGFbinduced OT-II (CD4) T REG (iT REG ) in the co-culture assay. Whereas, in the presence of the induced OT-II T REG osteoclast resorption activity modestly increased (1.5-fold in area resorbed), re-stimulation of the T REG with anti-CD3 + anti-CD28 was required to observe suppression. In contrast, the osteoclastinduced Tc REG could suppress in the absence of re-stimulation. We also tested if CD8 T-cells when activated by anti-CD3 and anti-CD28 in the presence of IL-2 could suppress resorption by osteoclasts. While CD8 T-cells activated in this manner produce significant levels of IFN-c, they were only able to partially suppress (,25% suppression) pitting by osteoclasts. Finally, we tested if in vivo generated Tc REG could suppress osteoclast activity. Tc REG were purified from the bone marrow space by cell sorting to .95% purity, then co-cultured with osteoclasts (at a ratio of 3 osteoclasts to 1 GFP + T-cell) on hydroxyapatite-coated plates. The CD8 + GFP + Tc REG population was able to efficiently suppress osteoclast resorption activity, whereas the CD8 + GFP 2 population did not (Fig. 2B).
Our results show that the osteoclast-induced Tc REG , whether produced by co-culture, or isolated from a mouse, have the ability to suppress osteoclast activity. In experiments shown below, we address the mechanism by which the Tc REG mediate this suppression of osteoclasts.

Tc REG Suppress Osteoclast Differentiation but not Survival
IFN-c blocks osteoclast differentiation, while RANKL promotes differentiation and survival of the osteoclasts. As the pitting assays described above require 7 to 10 days, one mechanism by which Tc REG could mediate a loss of pitting would be to suppress osteoclast differentiation. To test for the effect of Tc REG on osteoclastogenesis, we cultured bone marrow cells with M-CSF for three days, and then removed non-adherent cells to enrich for osteoclast precursors; RANKL was then added with Tc REG (generated on independent wells with mature osteoclasts). After four days of co-culturing we visualized and enumerated the osteoclasts using the fluorescent substrate ELF-97 for tartrate resistant acid phosphatase (TRAP), a marker of osteoclasts [62]. The results show that no TRAP positive cells were induced by RANKL (top middle panel Fig. 3A) in the presence of Tc REG . In contrast, many large TRAP positive osteoclasts were observed in the absence of T-cells (top left), and in the presence of naïve CD8 Figure 1. Osteoclast-induced TcREG produce IL-10, IL-6, and IFN-c. Mature osteoclasts (day 4) or osteoclast precursors were cultured with no Ag or OVA protein and then used to prime CD8 T-cells from an OT-I mouse. A. T-cells were collected at 48 h following initiation of co-culture. T-cells were stained for CD44, CD25 and FoxP3 and then analyzed by flow cytometry. T-cells co-cultured with osteoclasts in the absence of antigen do not express FoxP3 + or CD25 (top); FoxP3 and CD25 were induced in the presence of antigen as shown in the representative flow plot. The expression of FoxP3 was confirmed by reverse-transcription of RNA isolated from the co-culture and subsequent PCR of cDNA. GFP sorted cells from FoxP3 eGFP reporter mice were used as controls. Only mature (day 4) osteoclasts supported the generation of Tc REG (right panel). B. Anti-mouse TGFb was added to co-cultures at the dose indicated (left). Addition of recombinant murine TGFb1 to co-cultures of CD8 T-cells and osteoclasts at concentration indicated (right). The percent of input T-cells converted to FoxP3 + are plotted in both panels. No statistically significant effect was observed on Tc REG induction with either the addition of neutralizing antibody or recombinant TGFb. C. Media was collected and cytokine quantitated by multiplexed ELISA. After 48 h of co-culture, cells were treated with Golgi stop and PMA plus ionomycin for 6 h. The cells were permeabilized, stained and evaluated for cytokine production by flow cytometry. While the CD11b + osteoclasts were negative for all cytokines, the CD8 + T-cells stained triple positive for IL-10, IL-6, and IFN-c. Statistical significance was assessed by non-parametric paired T test: *: P,0.05, **: P,0.01, ***: P,0.001 and ****: P,0.0001. doi:10.1371/journal.pone.0038199.g001 Figure 2. TcREG inhibit osteoclast resorption. A. Osteoclasts (day 3) lifted and plated on hydroxyapatite-coated plates. OT-I Tc REG or OT-II TGFbinduced FoxP3 + CD4 T-cells (iT REG ) generated in separate dishes were added next day. The co-cultures were re-fed every two days. After seven days, the wells were treated with bleach, photographed, and total pit area was quantified (results in panel B). Representative images from five replicates are shown in panel A. No pitting was observed in the presence of Tc REG (top right). Larger pits were observed in the presence of iT REG (bottom left), but Tc REG were dominant suppressors (bottom right). B. Tc REG suppressed pitting on hydroxyapatite plates without re-stimulation. In contrast, iT REG could only suppress after re-stimulation (see methods for details). IFN-c producing OT-I T-cells activated by anti-CD3 and anti-CD28 in the presence of IL-2 could partially suppress pitting by osteoclasts. Activated GFP + CD8 T-cells purified by cell sorting from FoxP3 eGFP reporter mice could also suppress osteoclast pitting, while conventional (GFP 2 ) CD8 had no affect on pitting. Statistical significance of area resorbed was assessed by Wilcoxon test: *: P,0.05, **: P,0.01 and ***: P,0.001 relative to osteoclast alone wells. doi:10.1371/journal.pone.0038199.g002 T-cells (upper right). Sato et al. [8] have previously shown that TGFb-induced T REG cannot suppress osteoclastogenesis, where as Zaiss et al. [49] obtained the opposite results. Our results (Fig. 3A) show that iT REG cannot suppress osteoclastogenesis unless they are restimulated with anti-CD3 and anti-CD28. To test if IFN-c is solely responsible for suppression of osteoclast differentiation, we used bone marrow from IFN-c receptor knockout (IFNcR1 2/2 ) mice. We found that Tc REG could efficiently suppress osteoclast differentiation of IFNcR1 2/2 precursors, indicating that additional cytokines play a role in mediating suppression of osteoclastogenesis. Finally, we confirmed the suppression using quantitative real-time PCR of three osteoclast markers (Fig. 3A bottom right). These marker genes were chosen because they are expressed in mature osteoclasts and absent in T-cells.
Next, we considered if Tc REG could control osteoclast numbers, and hence function, by regulating their survival or lifespan. For instance, IFN-c can induce Fas ligand expression in osteoclasts, in the presence of TNFa to induce apoptosis [63]. In other cell types, TGFb and IL-18 have been shown to induce Fas ligand [64,65].

Tc REG Suppress Mature Osteoclasts by Preventing Cytoskeletal Reorganization
Osteoclasts must adhere to the bone surface and migrate along it to resorb [66]; cytoskeletal reorganization is critical for this process [67,68]. Therefore, we stained for actin rings in osteoclasts in the presence and absence of Tc REG . Tc REG were cultured with mature osteoclasts seeded on bone, or on plastic for 24 h. T-cells were then removed and the actin rings were visualized using flourconjugated phalloidin. In the presence of Tc REG , the actin ring was either eliminated on bone ( Fig. 4A and B), or (the actin belt) was reduced in size on plastic (Fig. 4C).
These results indicate that Tc REG can act directly on mature osteoclasts and prevent the formation of sealing zones. Our data indicates that Tc REG and restimulated iT REG can efficiently block osteoclastogenesis, but not their survival. As osteoclasts express MHC class I but not class II [44] it is possible that co-culturing Tc REG in the pitting assay with the osteoclasts re-stimulates the Tc REG but not the iT REG . Therefore, in the next set of experiments we tested if suppression of osteoclast activity requires antigen presentation by osteoclasts. We also tested if direct cell contact between Tc REG and osteoclasts is needed to mediate suppression.

Suppression of Osteoclasts Activity is Mediated by Tc REG Secreted Cytokines
To test if the regulation of osteoclast activity was antigendependent, OT-I FoxP3 + T-cells were cultured with osteoclasts in the presence or absence of OVA peptide. As shown in Fig. 5A, Tc REG could suppress osteoclasts to similar levels in the presence or absence of antigen presentation. Therefore, we next tested if direct contact between osteoclasts and Tc REG was required to mediate suppression. Tc REG were able to suppress osteoclast activity to a significant level when separated by a membrane with a 0.45 m pore (Fig. 5B). These results indicated that the secreted cytokines were, to a large extent, responsible for the suppressive activity.
To identify the specific cytokines responsible for osteoclast suppression, pre-differentiated Tc REG were incubated with mature osteoclasts in the presence of neutralizing antibodies. Blockade of IFN-c showed the most significant reduction in osteoclast inhibition while neutralizing IL-10 and IL-6 also relieved suppression (Fig. 5C). CTLA-4 blockade showed no effect confirming that the secreted cytokines IL-6, IL-10, and IFN-c all contribute to osteoclast suppression. Osteoclasts express F C receptors (FcR), and there could be a concern about confounding effects from the added antibody. Although, we used rat anti-mouse antibodies to minimize the likelihood of mouse FcR from binding to the rat antibodies, a possibility remains. Second, the active moiety of some of the cytokines, notably IFN-c is a dimer [69] it is possible that immune complex formation could confound the results. To directly address these issues we added duck Hepatitis B virus (dHBV) polymerase to our co-culture, and titrated mouseanti dHBV polymerase antibody (20 to 70 mg/ml). We observed no effect on suppression (data not shown), indicating that the antibodies, and potential immune complex formation did not confound the assay.
To confirm the results of antibody neutralization we generated osteoclasts from IFNcR1 knockout mice. We found that Tc REG generated from OT-I mice could suppress IFNcR1 2/2 osteoclast pitting to a similar extent as wild-type osteoclasts (Fig. 5D). Taken together, our data document that Tc REG , once generated can suppress osteoclast resorption directly through secreted cytokines. Altering the concentration by neutralizing antibody against all secreted cytokines expressed by Tc REG relieved suppression, while blockade of Tc REG expressed CTLA-4 had no measurable effect of suppression.

Discussion
Many studies have now established that under inflammatory conditions, T-cell produced cytokines increase osteoclast activity leading to bone erosion [70,71,72]. This close relationship between the skeletal and immune systems is illustrated in diseases such as autoimmune rheumatoid arthritis and periodontitis. The crosstalk between the immune and skeletal system has been termed osteoimmunology. Osteoimmunology emerged from the early pioneering work of Suda and colleagues (reviewed in [72]) and the demonstration that RANKL, a critical mediator of osteoclasts, is expressed on T-cells [6]. Here we examined the regulation of osteoclasts by FoxP3 + CD8 T-cells (Tc REG ). Although Tc REG have been documented in humans and mice [18,22,39,73,74,75,76,77,78,79,80,81,82,83,84,85], they have not been studied extensively, in part due to their low abundance (0.2 to 2% of CD8 T-cells) in lymphoid organs. In comparison, the wellstudied CD4 regulatory T-cells, T REG , comprise 5-12% of CD4 T-cell in the spleen. The Tc REG and the T REG have overlapping and distinct functions. Both cells express CD25 and the transcription factor, FoxP3 a marker of the regulatory T-cells [34,86,87].
Based on the mediator cytokines produced by the Tc REG , in this work we examined if the osteoclast induce Tc REG could control osteoclast activity. Our results show that the osteoclastinduced Tc REG secreted cytokines (Fig. 1) that could suppress resorption by osteoclasts (Fig. 2). Induction of Tc REG by osteoclasts required antigen (Fig. 1), through cross presentation by osteoclasts [44]. Once activated, Tc REG , unlike the T REG , do not require restimulation to suppress osteoclasts (Fig. 2). To examine the mechanisms by which suppression is mediated, we To test if suppression of osteoclastogenesis was mediated solely by IFN-c, bone marrow cells from IFNcR1 2/2 mice were used. The results of TRAP staining were confirmed by quantitative realtime PCR using primers for Acp5 (TRAP), Cathepsin K (CTSK) and matrix metalloprotease-9 (MMP-9). b-actin expression was used for normalization. B. Osteoclasts, cultured in the absence or presence of Tc REG for five days were counted after TRAP staining with ELF-97. To test for increased apoptosis adherent cells were stained with Annexin V. Each data point is average of three wells per experiment. Statistical significance was assessed by nonparametric paired T test: *: P,0.05, **: P,0.01 and ***: P,0.001 by comparison to untreated (Untx) osteoclast wells. doi:10.1371/journal.pone.0038199.g003 measured the effect of the Tc REG on differentiation of osteoclasts. Our data shows that Tc REG and iT REG block osteoclastogenesis (Fig. 3A). The Tc REG did not affect the survival of osteoclasts (Fig. 3B), but Tc REG could directly act on mature osteoclasts to suppress actin ring formation (and actin belt formation; Fig. 4). To elaborate on the requirement that Tc REG do not require restimulation, we showed that Tc REG could suppress osteoclast activity in the absence of antigen (Fig. 5A) and when separated from the osteoclasts by a membrane with a 0.45 m pore (Fig. 5B) indicating the secreted cytokines mediate the suppression. To indentify the cytokine(s), we used neutralizing antibodies against the secreted cytokines and found that altering the levels of any of the three cytokine relieved the suppression of osteoclast activity (Fig. 5C). Indeed, Tc REG could effectively suppress pitting of IFNcR1 2/2 osteoclasts (Fig. 5D). Although the role of Tc REG in controlling the immune system is still being defined [88], our results provide a new physiological role for the Tc REG : the inhibition of osteoclast activity under homeostatic conditions.
Regulatory T-cells use a number of mechanisms to suppress the immune system as noted above, however only a subset of the mechanisms may be expressed in a given situation. In this context, we note that the osteoclast-induced Tc REG produce IL-6 and IFNc, molecules that would normally be considered pro-inflammatory, but previous studies have shown they can affect osteoclast activity [89]. Based on a number of recent studies that demonstrate a remarkable plasticity in T REG responses [90,91], we speculate that this may be a general phenomenon for regulatory T-cells: the method of activation and their location may allow them to express specific effector functions, to regulate cells at hand. This phenomenon may underlie the apparent multiple types of T REG observed [34,87].
Summarizing, we demonstrate that FoxP3 + CD8 T-cells can suppress osteoclast resorptive function and thus provide a novel control function for regulatory T-cells beyond regulation of the immune system. The ability of osteoclasts to induce Tc REG and the ability of Tc REG to subsequently regulate osteoclast function establishes a bi-directional regulatory loop between these two cells in the bone marrow. Notably, the regulatory loop does not require the presence, in vitro, of pro-inflammatory cytokines. Indeed, our ability to isolate functional Tc REG from mice, in the absence of any inflammatory disease (Fig. 2B), indicates that these cells have a role in maintaining skeletal homeostasis in vivo. In contrast to CD4 T REG , the Tc REG may have a more specialized and local function. The ability of osteoclasts to induce Tc REG and their ability to suppress without restimulation may provide an explanation for the low levels of Tc REG found in vivo: the system is rapidly inducible, so a large reservoir of Tc REG is not needed. The induction of the Tc REG by osteoclasts that suppress osteoclasts would be selflimiting and lead to small number of Tc REG . Osteoclasts remodel bone as part of basic multicellular unit (BMU) that includes osteoblasts and other cells [92,93]. The rapid induction of Tc REG also suggests that the local concentration of the Tc REG -produced cytokines is likely to be high, thus requiring a small number of Tc REG . This notion is consistent with our observation that Tc REG suppressed osteoclast slightly less well when separated in a transwell. Finally, in contrast to CD4 T REG cells, which require MHC class II for activation, regulatory CD8 T-cells can be activated by any MHC class I expressing cells. All cells (except red blood cells) express MHC class I and therefore could activate Tc REG , an abundance of Tc REG could suppress host response to pathogens.
Why does the immune system regulate osteoclasts? We suggest two possibilities: First, regulatory T-cells have evolved to suppress the immune system. As osteoclasts are derived from myeloid cells, they retain the ability to respond to immune signals. Just as cytokines produced by effector T-cells activate osteoclast activity in inflammatory bone erosion diseases, the cytokines produced by Tc REG suppress osteoclasts. In addition to ontogeny, a functional linkage may also exist. The bone marrow is the primary site of hematopoiesis. Stromal cells line, provide essential support and form a specialized sealed compartment (''niche'') for the hematopoetic stem cells (HSC) [94,95,96,97]. It has been documented that osteoclast activity modulates the egress of the hematopoetic precursors (HPC) from the niches [98,99]. We hypothesize that the immune system may increase osteoclast activity through production of effector T-cells (T EFF ). The T EFF secreted cytokines increase osteoclast activity during inflammation to replenish lost immune cells and thus increase circulating hematopoetic precursors (HPC). To maintain balance or restore homeostasis after inflammation, Tc REG may be used to suppress osteoclast activity. For example, it is conceivable that the Tc REG evolved to provide an elegant sensor for T-cell lymphopenia. A Figure 5. Tc REG inhibit osteoclast activity in an antigen-and contact-independent manner by secreted cytokines. A. Osteoclasts were seeded on hydroxyapatite-coated plates and allowed to adhere overnight. The osteoclasts were pulsed with SIINFEKL peptide (+Antigen) or a control FLAG peptide (-Antigen). OT-I Tc REG or naïve CD8 T-cells were added and co-cultured for 7 days. Cells were re-fed with medium containing M-CSF and RANKL every 2 to 3 days. The plates were then treated with bleach solution, washed, dried and photographed. Representative photomicrographs are shown on the left. Quantitation from four experiments of three wells each is shown on the left. B. Tc REG were added to top-insert of transwell (0.45-m membrane) separated from the osteoclast plated on 24-well Corning Osteo-Assay plate. Osteoclast activity was determined by quantifying total pit area resorbed following 10 days of co-incubation. C. In a culture of both osteoclast and Tc REG on Corning Osteo-Assay plates, neutralizing antibodies against IL-10 (25 mg/ml), IFN-c (50 mg/ml), IL-6 (20 mg/ml) or CTLA-4 (10 mg/ml) were added to determine their impact on osteoclast activity. The cells were co-cultured for 10 additional days, and re-fed with media containing M-CSF, RANKL every three days. Re-feed media with antibodies were added on days 3 and 6. D. Pitting assay as described in Panel C were conducted in parallel using osteoclasts from either wild type (WT) or IFNcR1 2/2 mice. Statistical significance of area resorbed was assessed by non-parametric paired T test: *: P,0.05, **: P,0.01 and ***: P,0.001 in comparison to osteoclast alone wells. doi:10.1371/journal.pone.0038199.g005 reduction in Tc REG numbers may lead to an increase in bone resorption, and the subsequent increase in HPC mobilization. More studies are needed to explore the consequences of this bidirectional regulation for both the bone and autoimmune regulation, and to identify the sensors that mediate this regulation.

Materials and Methods
Mice: Five-week-old male C57BL/6 mice were purchased from Jackson Labs or used from in-house breeding colony. A breeder trio of FoxP3eGFP reporter mice on a C57BL/6 background were purchased from Jackson Labs, and bred in-house for these experiments. OT-I/Rag 2/2 and OT-II/Rag 2/2 mice were purchased from Taconic. All animals were maintained in the Department of Comparative Medicine, Saint Louis University School of Medicine in accordance with institutional and Public Health Service Guidelines. Saint Louis University School of Medicine Institutional Animal Care and Use Committee approved all procedures performed on mice (Protocol 2072). Purchased animals were allowed to acclimate for at least one week, prior to use.

Generation of Osteoclasts
Osteoclast precursors were isolated as previously described [42,44]. Briefly, the mice were sacrificed by CO 2 asphyxiation and the long bones harvested. The caps of the bones were removed and bone marrow cells were flushed with 0.05% collagenase (Worthington) in a-minimum essential medium (aMEM, Invitrogen). The cell population was filtered through a 40 m cell strainer, pelleted, resuspended and maintained in aMEM growth medium (aMEM supplemented with 10% heat-inactivated fetal bovine serum

Isolation of Splenocytes
Single cell suspensions of spleens were prepared in PBS +1% FBS by grinding with a sterile syringe plunger and dispersed by pipetting, then filtering through a 40 m cell strainer. For co-culture experiments, OT-II CD4 or OT-I CD8 T-cells were prepared by first enriching for T-cells using Pan-T-cell beads then purified by negative selection using appropriate magnetic beads (Miltenyi). The resulting T-cells were routinely .97% pure when stained with anti-CD3, anti-CD4 and anti-CD8 antibody.

Generation of Tc REG
Day 3 osteoclasts cultured in 20 ng/ml M-CSF and 50 ng/ml GST-RANKL, were seeded at 5610 5 cells/ml/well in the presence of 5 mM OVA (A-5503; Sigma-Aldrich) in 24-well tissue culture-treated plates (Corning). After overnight incubation, medium was removed and (adherent) cells were washed with pre-warmed (37uC) medium. 2.5610 5 freshly harvested splenic OT-I transgenic T cells purified by negative selection were added in 2 mls of complete T-cell media (RPMI, 10% heat-inactivated FBS (DFBS; HyClone), penicillin-streptomycin-glutamine (Hy-Clone), non-essential amino acids, sodium pyruvate, HEPES, and 55 mM b-mercaptoethanol. Following 48 h co-culture, T-cell aliquots were removed and stained intracellularly to assay for FoxP3 expression. TGFb -induced T REG were generated as previously described [24]. Briefly, CD4 T-cells were purified from an OT-II spleen by magnetic beads, and stimulated with plate-bound anti-CD3 (1 mg/ml) and anti-CD28 (2 mg/ml) in the presence of IL-2 and rhTGF-b1 (5 ng/ml). After 48 h of stimulation, the cells were expanded in IL-2 (100 U/ml) containing media for 48 hours. The iT REG were used directly, or in some experiments, restimulated with plate bound anti-CD3/anti-CD28 (as above) for an additional 48 hours in the absence of rhTGFb.

Cytokine Profiling by Multiplexed ELISA
Cytokine quantitation (IL-6, IL10, and IFN-c) was performed using multiplexed ELISA (Millipore/Linco Research) per the vendor's protocol on a Luminex-100. The data was analyzed using three-point logistic fitting in Microsoft Excel.

Cytokine Staining
For osteoclast staining, Golgi Stop (BD Biosciences) was added to T-cell/osteoclast co-cultures at 4 ml/(6610 6 ) cells and incubated for 6 h. Osteoclasts were collected using ice-cold PBS and vigorous pipetting. For T-cell staining, 50 ng/ml phorbol 12-myristate 13acetate and 1 mg/ml ionomycin were added to T-cell/osteoclast cultures for initial 1 h incubation. Golgi Stop (Invitrogen) was then added, and cultured for 3 h. T-cells were then removed from the plate using pre-warmed PBS. Cells were pelleted (5 min at 4006g) and resuspended in PBS plus 1% DFBS. Cells were blocked with rat anti-mouse FccIII/IIR mAb (BD Biosciences Pharmingen) in 1% DFBS in PBS and stained for 45 min at 4uC with fluorophoreconjugated Ab, washed, fixed with Fix Buffer (4% Paraformaldehyde (Electron Microscopy Sciences), 0.01% Tween-20 (Sigma) in PBS), washed, permeabilized with Perm Buffer (0.5% BSA (Sigma), 0.1% Triton-X100 (Sigma) in PBS), washed, stained overnight at RT, washed and then analyzed using an LSR II instrument with CellQuest (BD Biosciences) software. Analysis was performed using FlowJo software (version 8.73; Tree Star).

Differentiation Assay
Freshly harvested osteoclast precursor cells were cultured with 20 ng/ml M-CSF for 3 days on a 15 cm Petri dish (Corning). Cells were removed with Versene and seeded at 8.5610 4 cells/well on a 96 well plate with M-CSF to allow BM precursor to adhere. The following day (day 0), GST-RANKL was added to the osteoclast precursor cell culture in the presence or absence of pre-activated Tc REG . T-cells were washed off (day 5) and the adherent cells were assayed for tartrate-resistant acid phosphatase activity using a fluorescent substrate, ELF-97 (Invitrogen), in accordance with the manufacturer's protocol.

Survival Assay
On day 0, mature osteoclast were plated at 8.5610 4 cells/well on a 96 well plate alone or with pre-differentiated Tc REG . Day 5, non-adherent cells were removed and remaining cells were counted and stained with FITC-conjugated Annexin-V (Biotium).

Resorption Assays
For in-situ differentiated Tc REG , on day 0, mature osteoclast (5610 5 ) were seeded on 24-well hydroxyapatite coated plates (Corning or BD Biosciences). Day 1, osteoclast were pulsed with SIINFEKL peptide for 4 h, washed, and then naïve OT-I T-cells were added. M-CSF and GST-RANKL were added every 48 h. Day 10, cells were removed with 10% bleach and pit area was photographed and quantified using NIH ImageJ.
For pre-differentiated Tc REG , day 0, osteoclasts were seeded as above. Day 1 osteoclasts were pulsed with SIINFEKL peptide or no antigen for 4 h, washed, and then pre-differentiated Tc REG were added. M-CSF and GST-RANKL were added every 48 h. On day 7, cells were removed with 10% bleach and pit area was photographed and quantified using NIH ImageJ.
For transwell assays, osteoclasts were plated as described above. Cells were allowed to adhere overnight, and media was replaced with 700 ml of complete T-cell media containing M-CSF and GST-RANKL in the bottom well. 5610 5 Tc REG were added to top insert in 300 ml. Cells were re-fed every 48 h with M-CSF and RANKL. Following 10 days of incubation, bottom wells were treated with bleach to remove cells, and pit area was quantified, as above.

Purification of Tc REG from FoxP3 eGFP Reporter Mice
To purify activated Tc REG we activated osteoclasts in vivo by injecting the FoxP3 eGFP mice with RANKL (1 mg/kg intraperitoneally). Two doses of RANKL were given intraperitoneally on consecutive days; this protocol is modified from [101], where three doses were administered. The mice were sacrificed 50 hours post first RANKL dose and bone marrow cells harvested as described in osteoclast generation section above. Magnetic beads (Miltenyi) were used to positively select CD8 T-cells. The eluted cells were stained with V450-conjugated CD3 (clone 17A2; eBioscience) and PE-conjugated CD8a antibodies and further purified by cell sorting on Aria instrument with FACS Diva software (BD Biosciences).

Antibody Neutralization Assay
Immature day two osteoclasts were lifted and 8.5610 4 cells were seeded on 96-well hydroxyapatite coated plates (Corning). After overnight incubation, pre-differentiated Tc REG (8.5610 4 ) were added with neutralizing antibodies. M-CSF and GST-RANKL were added every 72 h. After additional six days, cells were removed with 10% hypochlorite, photographed and pit area was quantified using NIH ImageJ. The lowest concentration of antibody (see legend Fig. 5C) that gave maximum number of pits was determined from an antibody titration.

Actin Ring Assay
Mature osteoclast were plated at 3610 3 cell/well on bovine bone slides or on plastic in a 48 well plate. Following overnight adherence, Tc REG or naïve OT-I T-cells were added at a 1:1 Tcell to osteoclast ratio. After a 24 h co-incubation, the nonadherent cells were removed; adherent cells were fixed (4% Paraformaldehyde (EMS), 0.2% Triton-X100 (Sigma) in PBS) for 10 minutes, washed thrice with PBS and then stained with phalloidin-Texas Red (Biotium) for 15 min. The cells were photographed and actin ring size was quantified using NIH ImageJ software.

RT-PCR and Quantitative Real-time PCR
Total RNA was purified from cell lysates following manufacturers instructions (Agilent technologies). For FoxP3, cDNA was generated (Invitrogen) and subsequently PCR amplified for 45 cycles. The reaction contained 2.5 mM MgCl2, 0.2 mM dNTPs. qRT-PCR for CTSK, TRAP and MMP-9 transcripts were performed on ABI7500 instrument using a one-step (Invitrogen) SYBR green kit for 40 cycles using input 1 ng total RNA. The data is expressed as 2 DDCt (relative quantity or RQ) by normalizing to b-actin expression between samples. The primer sequences used for analyses of RNA expression span at least one intron, are shown in Table 1.

Statistical Analysis
All statistical analyses were performed with GraphPad Prism (version 5.0c). Wilcoxon or Student's T test were used for determining statistical significance for osteoclast alone and osteoclast plus treatment as indicated in the Figure legends or in the methods described above.