Vitamin K2 Modulates Vitamin D‐Induced Mechanical Properties of Human 3D Bone Spheroids In Vitro

ABSTRACT Rotational culture promotes primary human osteoblasts (hOBs) to form three‐dimensional (3D) multicellular spheroids with bone tissue‐like structure without any scaffolding material. Cell‐based bone models enable us to investigate the effect of different agents on the mechanical strength of bone. Given that low dietary intake of both vitamin D and K is negatively associated with fracture risk, we aimed to assess the effect of these vitamins in this system. Osteospheres of hOBs were generated with menaquinone‐4 (MK‐4; 10μM) and 25‐hydroxyvitamin D3 [25(OH)D3; 0.01μM], alone and in combination, or without vitamins. The mechanical properties were tested by nanoindentation using a flat‐punch compression method, and the mineralized extracellular bone matrix was characterized by microscopy. The in vitro response of hOBs to MK‐4 and 25(OH)D3 was further evaluated in two‐dimensional (2D) cultures and in the 3D bone constructs applying gene expression analysis and multiplex immunoassays. Mechanical testing revealed that 25(OH)D3 induced a stiffer and MK‐4 a softer or more flexible osteosphere compared with control. Combined vitamin conditions induced the same flexibility as MK‐4 alone. Enhanced levels of periostin (p < 0.001) and altered distribution of collagen type I (COL‐1) were found in osteospheres supplemented with MK‐4. In contrast, 25(OH)D3 reduced COL‐1, both at the mRNA and protein levels, increased alkaline phosphatase, and stimulated mineral deposition in the osteospheres. With the two vitamins in combination, enhanced gene expression of periostin and COL‐1 was seen, as well as extended osteoid formation into the central region and increased mineral deposition all over the area. Moreover, we observed enhanced levels of osteocalcin in 2D and osteopontin in 3D cultures exposed to 25(OH)D3 alone and combined with MK‐4. In conclusion, the two vitamins seem to affect bone mechanical properties differently: vitamin D enhancing stiffness and K2 conveying flexibility to bone. These effects may translate to increased fracture resistance in vivo. © 2020 The Authors. JBMR Plus published by Wiley Periodicals LLC on behalf of American Society for Bone and Mineral Research.


Introduction
T hree-dimendional (3D) bone spheroids, also referred to as osteospheres, represent new in vitro models to study the molecular mechanisms of bone remodeling, (1) as well as the pathophysiology of bone diseases and healing. (2) Slow horizontal clinorotation promotes aggregation and differentiation of bone cells into bone tissue-like structures without the inclusion of any scaffold material. (1,(3)(4)(5) Under these culture conditions, primary human osteoblasts (hOBs) form a self-assembled mineralized extracellular matrix within the 3D bone spheroids. (1) We have previously shown that these spheroids represent a suitable model for assessment of the effect of various stimuli on the biomechanical properties of bone. (6) Vitamin D stimulates the absorption of calcium and phosphate from the intestine. Low serum vitamin D levels induce secondary hyperparathyroidism, leading to increased bone resorption, decreased BMD, and a higher fracture incidence. (7) Vitamin D is mainly synthesized in the skin after exposure to sunlight, but is also obtained from dietary sources. Vitamin D is metabolized to 25-hydroxyvitamin D 3 [25(OH)D 3 ] in the liver, and to 1,25-dihydroxyvitamin D 3 [1,25(OH) 2 D 3 ] in the kidneys by the enzyme 1α-hydroxylase. (8) 1,25(OH) 2 D 3 is the biologically active form, (9) whereas 25(OH)D 3 levels are used as a measure of vitamin D status. (10) 1α-hydroxylase, as well as the vitamin D receptor, is also expressed in osteoblasts. (11)(12)(13)(14)(15) For studies of the effect of vitamin D on osteoblasts in vitro, 25(OH)D 3 is preferred over 1,25(OH) 2 D 3 because of its longer half-life time. (16) Menaquinones, referred to as vitamin K2, are a family of molecules consisting of a 2-methyl-1,4-naphthoquinone structure with a variable number of 3 0 -substituted isoprene units. (17) The main dietary menaquinones are MK-4 to MK-10, which are found in fermented food and animal products. (18,19) Vitamin Kdependent proteins have been isolated in bone, cartilage, kidney, and vascular and soft tissues. (20) These proteins include, among others, osteocalcin (OC) and periostin. (21) OC gene expression is regulated by 1,25(OH) 2 D 3, (22) whereas the protein's capability to bind to calcium relies on the vitamin K-dependent gamma-carboxylation of three glutamic acid residues in the molecule. (23) Periostin is a matricellular protein involved in the regulation of collagen fibril diameter and cross-linking. (24) Vitamin K2 also exerts direct effects on bone cells, stimulating osteoblastogenesis (25)(26)(27) and inhibiting the osteoclast differentiation. (25,27) Vitamin K2 has been reported to bind to the steroid and xenobiotic receptor (SXR), resulting in enhanced expression of several components of the bone matrix. (26) Low vitamin K intake, as well as high levels of undercarboxylated OC (unOC), is associated with an increased risk of bone fragility concomitant with hip fractures in elderly patients. (28)(29)(30) The vitamin K2 synthetic form MK-4 is approved in antiosteoporosis therapy in Japan and is frequently used in combination with bisphosphonates. (31) However, the effect of MK-4 on BMD and fracture risk remains a controversy. (32) Combined administration of vitamin D and K is suggested to have synergistic positive effects on calcium homeostasis and bone and cardiovascular health. (33) Vitamin D enhances vitamin K-dependent bone protein production. (34,35) Both vitamin D and K have been demonstrated to be cofactors in the gamma-carboxylation of OC. (36,37) An increasing number of randomized controlled trials have also evaluated the combined treatment of vitamin K2 and D with different outcomes. (38)(39)(40) Both vitamins D and K play important roles in bone health; however, their combined effects on mechanical properties of 3D bone spheroids have, to our knowledge, not been studied before. Therefore, we wanted to investigate the in vitro effects of vitamin D and K, alone and in combination on the biomechanical properties of 3D bone spheroids of primary hOBs. To elucidate the molecular mechanisms, we aimed at identifying the effect of these vitamins on the gene expression and secretion of proteins and cytokines involved in the biological and mechanical functions of bone in both 2D cell cultures of primary hOBs and in 3D bone constructs.

2D Cell cultures
Commercially available primary hOBs (NHOst cell system; Lonza, Walkersville, MD, USA) were grown in osteoblast growth medium (OGM; Lonza) at 37 C in a humidified atmosphere of 95% air and 5% CO 2 . The medium was changed three times weekly, and the cells were subcultured and seeded in 24-well-plates. At confluence, synthetic vitamin K2, MK-4 (at 1μM and 10μM; gift from Kappa Biosciences, Oslo, Norway), and 25(OH)D 3 (0.01μM; Calcifediol CRS; European Pharacopoeia Reference Standard, EDQM, Strasbourg, France) were added alone or in combination to the culture medium. Cells cultured with regular OGM were used as control. Cell culture media were harvested after 1, 7, 14, and 20 days of incubation.

Mechanical testing of osteospheres
The semispheres were thawed overnight and dried for 24 to 48 hours at room temperature in air. The main global geometry, ie, the surface at the equatorial plane and the height of the samples, was established with a microscope. μCT scanning was not applicable because of the low density of the immature bone tissue. Based on the size of a pixel in the microscope image, the size of the surface was transformed into real size. Assuming an elliptical cross section, a section area was determined and used to calculate the equivalent circular cross section with an equivalent radius. The average cross-section radius and the height of the samples were applied in finding stress and strain measures from the measured global force and displacement in the mechanical testing of the semiosteospheres. The mechanical response of the osteospheres at room temperature was characterized by nanoindentation using a Hysitron TI950 TriboIndenter (Hysitron, Minneapolis, MN, USA). Because of the irregular geometry of the samples, conventional nanoindentation was not applicable. Instead, a so-called flat-punch method for a compression test of the particle-like materials was used. (41) The semispheres were placed on a silicon chip and compressed with a diamond flat punch with a diameter of 1.08 mm, comparable with sample size, as previously illustrated in Haugen and colleagues. (6) A sketch of the compression test set-up is given in Fig. 1. The predefined loading function consisted of one cycle with a small load sequence of maximum 50 mN with a 2-s hold time at load peak. Then, a 10-cycle sequence leading up to a 50-mN maximum load, and finally a 10-cycle sequence of increasing load up to 200 mN were applied. The cyclic loaddisplacement response was done stepwise with the load protocol increasing in 10 steps to 200 mN with partial unloadings, as a viscous effect evolves when the peak load is held constant. A nominal measure of tangential stiffness can be estimated by connecting the 10 points corresponding to each load increase. This leads to the response curves, as shown in Fig. 2A. To remove some of the geometrical influences of the semispheres on the response, the curves in Fig. 2A are mapped into nominal stress and strain. The global load was divided by the equivalent semicircular equatorial cross-section area to get a stress measure (ie, stress = punch force/πr 2 , where r is the radius of the semicircular equatorial cross section). The resultant global displacement was divided by the height of the sample to obtain a strain measure (ie, strain = global displacement/height of the semisphere).

Microscopy analysis of osteospheres
Osteospheres were washed in sterile PBS, fixed with 4% paraformaldehyde, embedded in OCT frozen sectioning medium (VWR International BVBA, Leuven, Belgium), and sectioned at a thickness of 10 μm using a CryoStar NX70 cryostat (Thermo Fisher Scientific, Waltham, MA, USA). For immunofluorescence characterization, sections were stained with a modified version of Goldner's trichrome method. (42) Weigert's hematoxylin solution, chromotrop 2R, fuchsine acid, orange G, tungstophosphoric acid, and fast green powder, as well as Entellan mounting medium were purchased from Merck KGaA (Merck, Darmstadt, Germany). In brief, sections were incubated in Bouin's solution (Sigma-Aldrich) for 1 hour at 50 C, washed in tap water, stained with Weigert's hematoxylin for 5 min, and washed again. After incubation with chromotrope 2R/fuchsine acid for 15 min, sections were washed in 1% acetic acid, stained with orange G for 7 min, washed in 1% acetic acid, stained with fast green for 10 min, and washed in 1% acetic acid again. After dehydration, the sections were mounted with entellan and imaged with a Leica DM RBE microscope (Leica, Wetzlar, Germany) with a digital camera. Prior to confocal microscopy, the sections were immunostained with primary antibodies against periostin and collagen type I (COL-1). Antigen retrieval was performed in 10mM citrate buffer, pH 6.0, with 0.05% Tween 20 at 60 C overnight. Sections were cooled to room temperature, washed with PBS, permeabilized with 0.1% triton X-100 for 10 min, washed with PBS, and blocked in 10% normal goat serum (NGS; Abcam, Cambridge, UK) for 1 hour at room temperature. Sections were then incubated with rabbit antiperiostin (ab14041; Abcam) and mouse COL-1 (ab90395; Abcam) antibody at a 1:300 dilution in 2% NGS, overnight at 4 C, and washed three times with PBS. Alexa Fluor 488 goat anti-rabbit (Thermo Fisher Scientific) and Alexa Fluor 568 goat anti-mouse (Thermo Fisher Scientific) secondary antibodies were used at a 1:500 dilution in 4% NGS for 1 hour at room temperature, sections were washed three times with PBS, counterstained with Hoechst 33342 (1 μg/mL; Sigma-Aldrich) for 30 min and mounted. Sections were imaged with Leica SP8 confocal microscope (Leica Microsystems CMS GmbH, Mannheim, Germany) using 405-, 488-, and 552-nm excitation, and 420-to 480-nm, 500-to 550-nm, and 580-to 630-nm emission filters for Hoechst 33342, Alexa Fluor 488, and Alexa Fluor 568, respectively. Confocal images were processed with ImageJ  software (NIH, Bethesda, MD, USA; https://imagej.nih.gov/ij/). For each image, random ROIs (n = 5) on each section were selected and quantified for their mean intensity. Five ROIs were also selected in the image areas containing no section (background), quantified for their mean intensity, averaged, and subtracted from the section mean intensities. 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium (Sigma-Aldrich) was applied for detection of alkaline phosphatase (ALP) in frozen sections of osteospheres as previously described by Brauer and colleagues. (43) Alkaline phosphatase activity assay ALP activity in the cell culture media of the 3D osteospheres after 1, 3, 7, and 14 days of culture was determined by measuring the hydrolysis of p-nitrophenyl phosphate (pNPP) (Sigma-Aldrich) into the yellow end-product p-nitrophenol, which absorbs at 405 nm. Prior to analysis, aliquots of the cell culture media were concentrated fivefold using MicrosepTM centrifugal tubes with 3 KDa cut-off from Pall Life Science (Ann Arbor, MI, USA). There was 25 μL of each concentrated sample incubated with 100-μL pNPP for 30 min in the dark at room temperature; then, the reaction was stopped by adding 50 μL of 3M NaOH. The absorbance was measured at 405 nm in a plate reader (ELX800; BioTek, Winooski, VT, USA) and the ALP activity was quantified using a standard curve based on calf intestinal ALP (Promega, Madison, WI, USA).

Quantification of proteins secreted in the cell culture medium
Multianalyte profiling of protein levels in the culture media of the 2D cultures and of the osteospheres was performed on the Luminex 200 system employing xMAP technology (Luminex Corp., Austin, TX, USA). Acquired fluorescence data were analyzed by the xPONENT 3.1 software (Luminex). Prior to analysis, aliquots of the cell culture media from the 2D experiment were concentrated 10-fold using MicrosepTM centrifugal tubes (Pall Life Science) with 3 KDa cut-off . Analyses were performed using the Milliplex Human Bone Panel kit (EMD Millipore, Billerica, MA, USA). For the 2D cultures, the effect of MK-4 and 25(OH)D 3 , alone and in combination, on the secretion of cytokines and proteins (IL-1b, IL-6, osteoprotegerin [OPG], OC, leptin, osteopontin [OPN], PTH, TNF-α, adrenocorticotropic hormone, adiponectin, and insulin) to the culture medium after 1, 7, 14, and 20 days were measured. The secretion of OC, OPG, OPN, dickkopf-related protein 1, FGF23, IL-6, and sclerostin to the culture medium of the osteospheres was assessed after 1, 3, 7, and 14 days of vitamin treatment. Furthermore, in the 3D experiment, the level of angiogenic markers was determined using the Milliplex Human Angiogenesis / Growth Factor Panel kit (granulocyte-colony stimulating factor, leptin, VEGF-A, VEGF-C, and VEGF-D). All analyses were performed according to the manufacturer's protocols.

RNA isolation and RT-PCR analysis
Total mRNA from 3D osteospheres and 2D cultures was isolated using the Dynabeads mRNA DIRECT kit (Thermo Fisher Scientific ) with some modifications to the manufacturer's protocol. Briefly, the cells were lysed in lysis/binding buffer (100mM Tris-HCl, pH 7.5, 500mM LiCl, 10mM EDTA, pH 8.0, 1% lithium dodecyl sulfate, 5mM dithiothreitol), the lysate was sonicated (UP50H; Hielscher Ultrasonics GmbH, Teltow, Germany) for 10 s and centrifuged for 5 min at 4 C; then the supernatant was collected. mRNA was isolated using magnetic beads [oligo (dT) 25 ] as described by the manufacturer. Beads containing mRNA were suspended in 10mM Tris-HCl, pH 7.5, and stored at −80 C until use. Two-step RT-PCR was performed using technical triplicates of total mRNA for the first cDNA Strand Synthesis kit 1612 according to the manufacturer's protocol (Thermo Fisher Scientific). The second step, real-time PCR was carried out in a Bio-Rad CFX 384 (Bio-Rad Laboratories, Hercules, CA, USA), using SYBR green-based assay iQ SYBR supermix (Bio-Rad Laboratories). RT-PCR data were analyzed using the 2 − ΔΔCt method 2 [−Delta Delta C(T)]. (44) Each treatment was compared with the respective control and normalized against β-actin. The primer sequences are listed in Table 1.

Statistical analysis
Statistical analysis was performed using SigmaPlot software version 14.0 (Systat Software, San Jose, CA, USA). Data obtained by Luminex analysis and RT-PCR (ΔΔCt values) were compared between the groups by t test or Mann-Whitney U test, depending on their normal distribution. Data are presented as percentage of untreated cells (= 100%) at each time point of observation. Mean intensities from the confocal image analysis (n = 5 per sample) were compared between the groups by t test. A probability of ≤0.05 was considered significant.

25(OH)D 3 increases and MK-4 reduces the stiffness of osteospheres
The nominal stress-strain response, illustrating potential effects of the vitamin treatment on the mechanical properties of the irregularly shaped osteospheres, is shown in Fig. 2B
Primary hOBs in frozen sections of 21-day-old mineralized 3D osteospheres expressed periostin and produced COL-1. Interestingly, in the untreated osteospheres, COL-1 was expressed as a stripe-like area in the outer regions of the semiconstructs. Osteospheres treated with MK-4 showed a significant stronger expression of periostin than the control (p < 0.01). Additionally, in these osteospheres, COL-1 was expressed in small amounts over the whole area of the semispheres. Combined administration of MK-4 and 25(OH)D 3 did not induce significant changes in COL-1 and periostin expression or the COL-1 expression pattern compared with the control. COL-1 in osteospheres treated with 25(OH)D 3 alone was expressed at a reduced level (p < 0.001), and also all over the area of the semiconstructs compared with the control (Fig. 3C).

25(OH)D 3 increases the secretion of ALP from 3D osteospheres and enhances ALPL expression in 2D cultures
Exposure of 3D osteospheres to 25(OH)D 3 reduced the mRNA expression of ALPL twofold (p < 0.001) relative to control on day 14, whereas no significant differences were observed after exposure to MK-4 or the vitamins in combination (Fig. 4A). Conversely, incubation of 2D cultures with 25(OH)D 3 increased ALPL expression more than 11-fold (p < 0.01) on day 3 and eightfold (p < 0.01) on day 20. In addition, relative ALPL expression was enhanced sixfold (p < 0.05) by MK-4 and 25(OH)D 3 together on day 3, and more than threefold by both MK-4 alone and the combination on day 20 (p < 0.01 for both; Fig. 4B).
The levels of membrane-bound ALP in frozen sections of 21-day-old mineralized 3D osteospheres were not affected by any of the vitamins compared with control (data not shown). However, the secretion of ALP to the culture medium from 3D osteospheres was decreased to 70 AE 1.3% (p < 0.05) of control

25(OH)D 3 enhances the deposition of mineral in osteospheres
Frozen sections of untreated 21-day-old mineralized bone spheroids showed large osteoid formation in the outer region of the semiconstructs, whereas little deposition of mineral was  Relative mRNA expression levels were normalized to reference gene ACTB and presented as fold-change relative to unexposed controls. Values represent the mean AE SD. Significant different from control at *p < 0.05, **p < 0.01, and ***p < 0.001. detected within the spheres (Fig. 5A,E). MK-4 supplementation did not affect mineralization, but the osteoid appeared to be much more condensed compared with the control (Fig. 5B,F). Osteospheres treated with a combination of MK-4 and 25(OH) D 3 showed extended osteoid formation into the central region of the constructs and increased mineral deposition over the whole area compared with the control (Fig. 5C,G). In osteospheres treated with 25(OH)D 3 alone, increased mineralization organized as a stripe-like area over the semiconstruct was observed (Fig. 5D,H).

Discussion
We demonstrate the differential effects of vitamin D and K2 on the mechanical properties of human 3D bone spheroids in vitro Ä exposure to 25(OH)D 3 -induced increased stiffness, whereas the synthetic vitamin K2, MK-4, induced softer or more flexible osteospheres compared with untreated spheroids. Osteospheres treated with a combination of 25(OH)D 3 and K2 had the same flexibility as those treated with K2 alone. To the best of our knowledge, this is the first study to show that vitamin K2 modulates vitamin D-induced mechanical properties in a 3D bone model based on hOBs.
Cell-based in vitro models, previously generated by a rotational coculture approach of hOBs and osteoclasts without any exogenous scaffolding material, (1,2,6) enable us to study the bone microenvironment. In contrast to these two cell systems, we produced 3D mineralized tissue constructs from single cultures of primary hOBs. Osteoblasts in our 3D bone spheroids formed mineralized bone matrix similar to Clark and colleagues and Haugen and colleagues, (1,6) and secreted the bone matrix protein OPN as reported by Penolazzi and colleagues. (2) In addition to 3D spheroids, we applied 2D cultures of primary hOBs to assess the effect of MK-4 and 25(OH)D 3 on proteins and cytokines involved in the mechanical and biological function of bone. In 3D cultures, cellular behavior reflects in vivo tissue functionality more accurately than in monolayer cultures. 3D cultures are thus better suited for the evaluation of cellular responses to various compounds or drugs. (45) The strength of bone and its ability to resist fracture are dependent on its mass and geometry, but also on the bone material properties, (46) which are determined by, among others, the quality, amount, and orientation of collagen fibers (47) ; and degree of mineralization. (48) The mineral phase of bone influences the ability to resist deformation and provides stiffness and strength to the bone structure, (47,48) whereas collagen is associated with its flexibility (toughness), giving resistance to impact load. (47,49) We observed an increase in ALP secreted to the culture medium from osteospheres exposed to 25(OH)D 3 and in line with this, enhanced mineral deposition, which may, in addition to the reduced expression of COL-1, account for their higher bone stiffness. On the other hand, administration of MK-4 alone reduced the ALP activity in the medium and did not stimulate mineralization, but induced enhanced expression of periostin and altered distribution of COL-1. This was reflected in reduced bone stiffness and higher flexibility in the osteospheres. In agreement, we found significantly upregulated expression of POSTN and COL1A1 in 2D cultures exposed to MK-4. However, mRNA expression in the osteospheres was not altered. Periostin is a vitamin K-dependent protein primarily produced and secreted by osteoblasts and their precursor cells. (21) It is an important mediator of the biomechanical properties of collagen-rich tissues by regulating collagen fibril diameter and cross-linking. (24) In the present study, increased flexibility of the osteospheres was observed after exposure to the two vitamins despite enhanced mineralization. The improvement of flexibility could be attributed to increased synthesis of periostin and COL-1. Accordingly, POSTN and COL1A1 gene expression levels in these osteospheres were significantly enhanced; however, no evident alterations in the protein levels of periostin and COL-1 were revealed. Still, it is reasonable that MK-4 may have facilitated the formation of more collagen with proper physiological function in the osteospheres. Vitamin K2 has been suggested to promote collagen accumulation in osteoblastic cells via the SXRsignaling pathway. (50) Enhanced collagen mRNA expression has also been reported in 2D cultures of osteogenically differentiated human mesenchymal stem cells from amniotic fluid treated with MK-4; however, protein levels in 3D spheroid cultures were not affected. (51) Vitamins D and K2, as well as the combination of the two, have previously been described to enhance mineralization of osteoblasts in vitro. (52) In clinical studies, combined administration of vitamins D and K is suggested to improve bone quality and lower the risk of fractures. (33) Moreover, a higher gain in BMD has been reported in postmenopausal women with osteoporosis treated with a combination of the vitamins compared with each vitamin alone or calcium. (38,39) It is worth noting that the generated osteospheres in our study may comprise osteoblasts in various differentiation stages, similar to the in vivo situation. As previously reported, spheroids with a diameter of 500 μm are made-up of a heterogenic population of cells, depending upon the location within the layer-like structure of the sphere. (53,54) In the outer rim of a sphere, cells are surrounded by media and have the space to proliferate, whereas cells in the inner area have cell-to-cell contact and are dependent on nutrient transport from neighboring cells. (54,55) In contrast to the even periostin staining in our study, immunostaining of unexposed osteospheres revealed COL-1 expressing cells in the outer region of the semispheres. This may indicate that these cells are less differentiated, producing higher amounts of COL-1. (56) The absence of COL-1 expression in the inner region of our osteospheres suggests that these cells are of later osteoblast or early osteocyte differentiation stage, (57) as COL-1 is downregulated when osteoblasts begin to develop into osteocytes in vitro. (56,58) As recently suggested by Kim and Adachi, the cell condensation within spheroids triggers the differentiation of osteoblast-precursor cells to osteocyte-like cells. (59) The uneven differentiation of osteoblasts in 3D cultures has been previously reported by others. (57,60) Alterations in this differentiation pattern within the osteospheres, induced by the vitamins, are reflected in the immunostaining and gene expression analysis.
OC and OPN are major noncollagenous proteins (NCPs) involved in bone matrix organization and deposition, and have been shown to influence bone morphology and mechanical properties. (61) Both proteins interact with collagen and mineral. (23,62) It has been recently suggested that their spatial arrangement in the bone matrix enhances bone toughness. (63) Among these NCPs, OPN has been proposed to act as a glue that counteracts the separation of the mineralized collagen fibers upon mechanical loading of bone. In this structure, energy may be dissipated through the formation and reformation of intramolecular bonds between OPN and divalent Ca 2+ , which increases the total energy to fracture bone. (64,65) Moreover, NCPs influence the mechanical properties of bone through dilatational band formation as suggested by Poundarik and colleagues. (63) Dilatational bands are ellipsoidal voids that result from the disassembly of noncollagenous protein complexes, like OC-OPN complexes, which are integrated in the mineralized matrix of bone when a load is applied. Formation of these microcracks within bone allows for the dissipation of large amounts of energy, which reduces the bone's propensity to fracture. (63) The enhanced BGLAP and SPP1 expressions in osteospheres exposed to MK-4 alone and combined with 25(OH)D 3 , as well as the acute increase in OPN secretion, suggest that the reduced stiffness seen in these osteospheres may be partially mediated by these mechanisms.
The carboxylated form of OC facilitates deposition of calcium into the bone matrix. (66) Both vitamins D and K stimulate synthesis of OC and are also cofactors in the carboxylation, (37) thereby contributing to mineralization. (67)(68)(69) In line with this, we observed a rise in OC levels in the 2D cell cultures after 1 and 7 days of vitamin D administration, and after 7 days of exposure to vitamin K, no further enhancement occurred when combining the two vitamins.
OPN release in 2D cell cultures was promoted by both vitamins after one day of exposure, but only by the higher concentration of MK-4 (10μM) after 7 days. Vitamin D alone and in combination with MK-4 also induced a transient increase in OPN in 3D osteospheres.
Based on our findings, it is reasonable that the effects of vitamins D and K are partly mediated by these proteins. It is worth noting that we were not able to detect OC in the culture medium of the osteospheres. This could be attributed to the fact that OC is expressed late in the osteoblast maturation process, (70) and therefore not detectable in the medium after a culture period of 14 days.
The rate of bone turnover is another determinant of bone quality. Thus, we assessed the impact of the two vitamins on substances regulating bone metabolism. In 2D cultures, MK-4 administration alone and combined with 25(OH)D 3 induced a decline in OPG. In contrast, a sustained increase in OPG by exposure to 25(OH)D 3 , as well as a decreased RANKL/OPG ratio in 3D spheroids by 25(OH)D 3 and combined vitamins was seen. These findings may translate to suppression of bone resorption in vivo. In the 3D spheroids, both vitamins induced a rise in IL6. However, data on the effect of IL-6 on bone metabolism are diverging. (71) Moreover, both vitamins induced an increase in DKK1, an inhibitor of bone formation, (72) 25(OH)D 3 at several time points, MK-4 and the combination of the vitamins only after 14 days. Finally, G-CSF levels were enhanced by treatment with 25(OH)D 3 alone, and to a lesser degree by the combined vitamins compared with control. In summary, the two vitamins induced a rise both in factors stimulating and inhibiting bone resorption, as well as factors favoring and inhibiting bone formation. How this translates to in vivo conditions is, however, impeded by the fact that the osteospheres only contained osteoblasts. We observed enhancement of osteoblast differentiation by 25(OH)D 3 and MK-4 alone and in combination, as reflected in increased OC levels. Given the interplay between osteoblasts and osteoclasts, the presence of both cells would have given a more complete picture. Still, based on our results, it can be hypothesized that combined administration of K2 and 25(OH)D 3 could contribute to stronger bone also in vivo. This should be tested in 3D osteospheres containing both osteoblasts and osteoclasts, as well as in rodents and humans.

Disclosures
MS, EAR, JH, BHS, AML, US, and JER state that they have no conflicts of interest. MEM is a shareholder in Axial Vita AS, which sells vitamin K2. JER is a member of Cost Action 16119 CellFit.