1. Medicine
Download icon

The role of extracellular matrix phosphorylation on energy dissipation in bone

  1. Stacyann Bailey  Is a corresponding author
  2. Grazyna E Sroga
  3. Betty Hoac
  4. Orestis L Katsamenis
  5. Zehai Wang
  6. Nikolaos Bouropoulos
  7. Marc D McKee
  8. Esben S Sørensen
  9. Philipp J Thurner
  10. Deepak Vashishth  Is a corresponding author
  1. Department of Biomedical Engineering, Center for Biotechnology and Interdisciplinary Studies, Rensselaer Polytechnic Institute, United States
  2. Faculty of Dentistry, McGill University, Canada
  3. Faculty of Engineering and Physical Sciences, University of Southampton, United Kingdom
  4. Department of Mechanical, Aerospace, and Nuclear Engineering, Rensselaer Polytechnic Institute, United States
  5. Department of Material Science, University of Patras, Greece
  6. Department of Anatomy and Cell Biology, Faculty of Medicine, McGill University, Canada
  7. Department of Molecular Biology and Genetics, Aarhus University, Denmark
  8. Institute of Lightweight Design and Structural Biomechanics, Vienna University of Technology, Austria
Research Article
  • Cited 0
  • Views 679
  • Annotations
Cite this article as: eLife 2020;9:e58184 doi: 10.7554/eLife.58184

Abstract

Protein phosphorylation, critical for cellular regulatory mechanisms, is implicated in various diseases. However, it remains unknown whether heterogeneity in phosphorylation of key structural proteins alters tissue integrity and organ function. Here, osteopontin phosphorylation level declined in hypo- and hyper- phosphatemia mouse models exhibiting skeletal deformities. Phosphorylation increased cohesion between osteopontin polymers, and adhesion of osteopontin to hydroxyapatite, enhancing energy dissipation. Fracture toughness, a measure of bone’s mechanical competence, increased with ex-vivo phosphorylation of wildtype mouse bones and declined with ex-vivo dephosphorylation. In osteopontin-deficient mice, global matrix phosphorylation level was not associated with toughness. Our findings suggest that phosphorylated osteopontin promotes fracture toughness in a dose-dependent manner through increased interfacial bond formation. In the absence of osteopontin, phosphorylation increases electrostatic repulsion, and likely protein alignment and interfilament distance leading to decreased fracture resistance. These mechanisms may be of importance in other connective tissues, and the key to unraveling cell–matrix interactions in diseases.

Introduction

In recent years, the role of extracellular matrix (ECM) proteins and their post-translational modifications (PTMs) in modulating cell activity, cell-matrix interactions, and biomineralization processes has sparked tremendous interest in different connective tissue biological systems. In particular, it has been postulated that different levels of phosphorylation of matrix proteins play a critical role in coordinating calcification processes in normally (Murshed et al., 2004; Clarke, 2008; Murshed and McKee, 2010; Nudelman et al., 2010; Addison et al., 2010; Deng et al., 2013) and pathologically (Boskey, 2013; Bertazzo et al., 2013) mineralized bone tissues either independently and/or in combination with collagen. These phosphoproteins also accumulate at the interfaces found across bone's hierarchical levels (McKee and Nanci, 1996; Thurner et al., 2009; Thurner, 2009), but it remains unclear as to how their phosphorylation levels influence the mechanical properties of bone. We have recently shown the importance of phosphorylation of ECM proteins in regulating bone quality. Global phosphorylation level varied between cortical and trabecular bone (Sroga and Vashishth, 2016), declined with age, and was associated with an increase in age-related skeletal fragility (Sroga and Vashishth, 2018).

Given the importance and incomplete understanding of how total phosphorylation levels, as well as the heterogeneity of phosphorylation observed for different bone matrix proteins, contribute to skeletal fragility, animal models provide a valuable resource to investigate this further. In particular, certain animal models recapitulate key metabolic and skeletal characteristics seen in humans displaying, for example, the phenotypes of major phosphate-handling diseases such as hypophosphatemia (Barros et al., 2013; Liu et al., 2006; Boukpessi et al., 2017), hyperphosphatemia (Yuan et al., 2014), and hypophosphatasia (Harmey et al., 2006; Narisawa et al., 2013; Yadav et al., 2014). Hyp mice – the murine analog of X-linked hypophosphatemia (XLH) – display low-serum phosphate and accumulation of osteopontin (OPN) (Barros et al., 2013), a well-known noncollagenous protein serving as a powerful inhibitor of mineralization, and a key determinant of bone’s resistance to fracture. In contrast to Hyp mice, fibroblast growth factor 23-deficient mice (Fgf23-/- mice) are hyperphosphatemic, but like the Hyp mice also show accumulation of OPN (Yuan et al., 2014). Both of these phosphate-handling disease models exhibit a soft-bone (osteomalacia) phenotype and display decreased cortical area, thickness, and strength (Liu et al., 2016; Murali et al., 2016). The hypophosphatasia mouse model (Alpl−/− mice) displays mineralization deficiencies characterized by rickets/osteomalacia as well as elevated levels of inorganic pyrophosphate (PPi). The Alpl−/− mice also show increased levels of phosphorylated OPN compared to wild type (WT) mice (Narisawa et al., 2013). Interestingly, Opn KO mice also show elevated levels of PPi despite having more mineralized osteoid than wildtype (WT) controls (Harmey et al., 2006). As such, it appears that OPN levels, and possibly its phosphorylation status, contribute to impaired matrix mineralization and may play a role in skeletal integrity in these models.

The degree of OPN phosphorylation has significant effects on its structure and physiological function (Kazanecki et al., 2007). For example, osteoclast adhesion is increased with phosphorylation (Katayama et al., 1998) and correlates with the extent of bone resorption (Razzouk et al., 2002). Hydroxyapatite (HA) crystal formation and growth are inhibited by OPN in a dose-dependent manner (Hunter et al., 1994), and dephosphorylation of OPN abolishes the inhibitory effect of OPN on HA formation by at least 40-fold (Razzouk et al., 2002). In addition to bone resorption and mineralization, OPN has been shown to play a mechanical role in bone, influencing its resistance to fracture (Duvall et al., 2007; Thurner et al., 2010; Poundarik et al., 2012). The negatively charged phosphate groups of serine and threonine residues on OPN bind to multivalent positive ions on hydroxyapatite, and this interaction is part of a bonding/cohesion process that limits separation of mineralized collagen fibrils during mechanical loading (Zappone et al., 2008). Also important in this bone-toughening process are large, covalently crosslinked (by transglutaminase) networks formed between neighboring OPN molecules and between OPN and other bone matrix proteins (Kaartinen et al., 1999; Kaartinen et al., 1997; Kaartinen et al., 2002). Networks of crosslinked OPN polymers are abundantly present in bone, and may reside in the interfibrillar collagenous matrix, at cell-matrix interfaces, and in interfacial cement lines (Kaartinen et al., 2002; Goldsmith et al., 2002) where they may be critical for maintaining the overall strength of bone tissue (Kaartinen et al., 1999; Hoac et al., 2017; Cavelier et al., 2018). Analysis of different tissues revealed that phosphorylation of OPN is highly variable, and typically only some of the potential phosphorylation sites are occupied in vivo (Neame and Butler, 1996). In fact, it is currently unknown how many of all available amino acid residues in mouse OPN are phosphorylated in vivo because the balance between the activities of protein kinases and phosphatases reflects the phosphorylation state of the protein. Importantly, the difference in phosphorylation status results in altered biological and mechanical responses. Considering the functional relevance of OPN phosphorylation, this PTM may be an important determinant of bone matrix quality and fragility.

In this study, we investigate the role of OPN phosphorylation on bone fracture. To execute this, we first demonstrate that in mouse models of impaired phosphate regulation and increased skeletal fragility, the level of OPN phosphorylation declines. Next, we captured the effects of phosphorylated OPN on bone fracture toughness (resistance to crack propagation and fracture) by developing methods to enzymatically phosphorylate and dephosphorylate WT and OPN-deficient mouse bones ex vivo, then measure the resultant change in their mechanical competence. In an effort to gain a better understanding of the various factors that contribute to the mechanical function of phosphorylated OPN in bone matrix, we then conducted atomic force microscopy-force spectroscopy (AFM-FS) experiments demonstrating the effect of pH, ion charges, and phosphorylation levels on the energy dissipation properties of the OPN network using simplified synthetic and physiologically relevant surfaces. Based on these results, we propose that for appropriate mechanical function of bone, the phosphorylation status of OPN promotes fracture toughness up to a beneficial point. Phosphorylation or dephosphorylation alters the interaction between charged groups on OPN, and between OPN and bone mineral leading to increased or decreased energy dissipation.

Results

Evidence of decreased osteopontin phosphorylation in mouse models of impaired phosphate metabolism and decreased mechanical properties

We first investigated whether the phosphorylation state of OPN varied using in vivo mouse models having phosphate disorders and known skeletal pathology linked to soft osteomalacic bones. Mineral-bound proteins were extracted from long bones of WT, Hyp, and Fgf23-/- mice. Total protein was quantified using a colorimetric detection system. From each sample, 2 µg of protein extract was loaded onto a 4–20% gradient SDS-PAGE gel. Since the vast majority of OPN phosphorylation occurs at serine residues, we performed immunoblotting for phosphoserine in the mineral-binding protein extracts. In the bone matrix of both Hyp and Fgf23-/- mice, we found that mineral-bound OPN increased (Figure 1a) but global phosphorylation decreased (Figure 1b) as compared to WT controls. In addition, the post-immunoprecipation results show that despite the accumulation of OPN in Hyp and Fgf23-/- mice (Figure 1c), the proportion of phosphorylated OPN was reduced compared to the bone of WT mice (Figure 1d). Given that these models have opposite levels of serum phosphate deviation from normal (hypophosphatemic vs. hyperphosphatemic), and display a reduction in bone strength (Liu et al., 2016; Murali et al., 2016; Sitara et al., 2004; Camacho et al., 1995) which may be dependent on defective mineralization but driven by the mineral-inhibiting protein OPN, our results suggest that osteopontin phosphorylation may be an important contributor to the fracture resistance of bone.

Pre-immunoprecipation (Pre-IP) of mineral-bound OPN.

(a) and global phosphorylation (b) in protein extracts of long bones from WT, Hyp and Fgf23-/- mice. Post-immunoprecipation (Post-IP) indicates that despite similar levels of OPN (c), phosphorylation of OPN is reduced in these disease models (d).

Phosphorylation status of osteopontin influences bone fracture toughness

To capture the effects of OPN phosphorylation on bone fracture toughness, we performed separate ex-vivo phosphorylation and dephosphorylation of whole femurs from WT and Opn KO mice and subsequent mechanical testing. The global phosphorylation level in bone matrix increased in both genotypes with ex-vivo casein kinase-II (CKII) phosphorylation (WT-phosphorylated vs. WT-nonphosphorylated control, p=0.008; Opn KO-phosphorylated vs. Opn KO-nonphosphorylated control, p=0.007) (Figure 2a). We observed a significant reduction in phosphoproteins with ex-vivo dephosphorylation by alkaline phosphatase (WT-dephosphorylated vs. WT-nondephosphorylated control, p=0.033; Opn KO-dephosphorylated vs. Opn KO-nondephosphorylated control, p=0.006) (Figure 3a). Although the change in ex-vivo phosphorylation between WT and Opn KO (delta-WT vs. delta-Opn KO, Figure 2b) was not statistically significant, we observed a significant difference in dephosphorylation between delta-WT and delta-Opn KO (Figure 3b), indicating that OPN-deficient bone can be modified to a greater extent than WT bone, likely attributable to increased permeability of enzymes into bones lacking OPN.

Mean global protein phosphorylation.

(a) and change in phosphorylation (b) for WT and Opn KO groups. * indicates significance at p<0.05 and error bars represent standard deviation.

Mean global protein phosphorylation.

(a) and change in phosphorylation (b) after removal of phosphate groups (dephosphorylation) for WT and Opn KO groups. * indicates significance at p<0.05 and error bars represent standard deviation.

We observed higher fracture toughness with phosphorylation of WT bones (WT-phosphorylated vs. WT-nonphosphorylated control, p=0.009). In contrast, toughness declined in Opn KO mice following ex-vivo phosphorylation (Opn KO-phosphorylated vs. Opn KO-nonphosphorylated control, p=0.025) indicating that phosphorylation of other bone matrix proteins in the absence of OPN did not improve the fracture resistance of bone (Figure 4). Ex-vivo dephosphorylation caused a decrease in fracture toughness for the WT group (WT-dephosphorylated vs. WT-nondephosphorylated control, p=0.012) (Figure 5) while an increase in fracture toughness was observed for Opn KO mice (Opn KO-dephosphorylated vs. Opn KO-nondephosphorylated control, p=0.037).

Mean fracture toughness (a) and change in fracture toughness (b) due to ex-vivo phosphorylation for WT and Opn KO groups.

* Indicates significance at p<0.05 and error bars represent standard deviation.

Figure 4—source data 1

Fracture toughness of phosphoryled WT and Opn KO mice.

https://cdn.elifesciences.org/articles/58184/elife-58184-fig4-data1-v3.xlsx
Mean fracture toughness (a) and change in fracture toughness (b) attributable to ex-vivo dephosphorylation for WT and Opn KO groups.

* Indicates significance at p<0.05 and error bars represent standard deviation.

Figure 5—source data 1

Fracture toughness of dephosphoryled WT and Opn KO mice.

https://cdn.elifesciences.org/articles/58184/elife-58184-fig5-data1-v3.xlsx

Energy dissipation of the osteopontin network is altered by levels of phosphorylation

We conducted atomic force microscopy-force spectroscopy (AFM-FS) studies using an in-vitro experimental system to demonstrate that the phosphorylation status of OPN can affect bone toughness by altering energy dissipation. At pH 8.5, both OPN and hydroxyapatite (HA) surfaces are negatively charged. Under high Ca2+ concentration, the detachment energy increased more significantly as compared to H2O and Na+ environments. At pH 6.0 however, the protein and HA bear opposite charges. The slightly acidic pH lead to a moderate dissolution of HA over time, and therefore, it is expected that Ca2+ ions are present in solution from the beginning of the experiment. Further addition of Ca2+ ions decreased energy dissipation attributable to the reduction of sacrificial bond formation (Gao et al., 2003) (increase of effectively positively charged sites in OPN) as well as increased repulsion between OPN and HA. These results were also confirmed by the decline in energy dissipation that was observed for dephosphorylated OPN in Ca2+ solution at pH 7.4 as compared to native OPN, as well as dephosphorylated OPN in Na+ solution at pH 7.4, all on mica substrates. Thus, the balance between Ca2+ ions in solution and the availability of negatively charged groups are both important for energy dissipation within the OPN network as well as at the OPN-HA interface. The results from AFM force spectroscopy are summarized in Figure 6.

Figure 6 with 3 supplements see all
Energy dissipation of OPN networks during AFM-FS experiments.

Energies are normalized to dissipation levels in EDTA for OPN deposited on mica and pulled with a pristine AFM tip (pH 7.4) and to dissipation levels in H2O for OPN deposited on HA and pulled with a HA-functionalized tip. All values are significantly different except OPN between HA, pH 8.5 H2O vs. Na+. It should be noted that the relative differences are similar to what is seen for quantitative values, except for EDTA and H2O levels due to normalization. These values are provided in Supplementary files 1 and 2. * indicates significance at p<0.05 and error bars represent standard error (SE) of the mean.

Figure 6—source data 1

Energy dissipation of native (phosphorylated) and dephosphorylated OPN film on mica in EDTA and calcium solution.

https://cdn.elifesciences.org/articles/58184/elife-58184-fig6-data1-v3.xlsx
Figure 6—source data 2

Energy dissipation of native (phosphorylated) OPN film on HA under various pH and ionic conditions.

https://cdn.elifesciences.org/articles/58184/elife-58184-fig6-data2-v3.xlsx

Discussion

Extracellular bone matrix phosphoprotein osteopontin (OPN) has been recently implicated in disease models of hypophosphatemia (Barros et al., 2013; Liu et al., 2006; Boukpessi et al., 2017), hyperphosphatemia (Yuan et al., 2014), and/or hypophosphatasia (Harmey et al., 2006; Narisawa et al., 2013; Yadav et al., 2014). Consistent with these studies, we observed full-length OPN in bone extracts of Hyp and Fgf23-/- mice. We further provide evidence that the phosphorylation level of OPN declined in these mouse models as detected by immunoblotting for OPN’s phosphoserine residues. Given that these models display an aging-like skeletal phenotype (Sroga and Vashishth, 2018) with impaired mineralization and osteomalacia, we considered whether the phosphorylation status of bone matrix proteins including OPN is an important determinant of their skeletal fragility. Our experimental model involved phosphorylation and dephosphorylation of both normal WT bones with OPN, and bones without OPN (Opn KO); thus, by comparing the change in fracture toughness caused by phosphorylation or dephosphorylation of the organic matrix between these two samples (WT-treated minus WT-control, delta-WT; and Opn KO -treated minus Opn KO -control, delta- Opn KO), the contribution of OPN phosphorylation alone can be isolated. Our results suggest that OPN and its phosphorylation level may be one of the dominant phosphoproteins in the determination of global bone matrix phosphorylation level and bone fracture toughness.

Fracture resistance of bone emerges from various mechanisms that exists at multiple length scales across bone hierarchy, and involves in part growth and packing of mineral foci into larger crossfibrillar aggregates such that the ECM becomes highly mineralized. In the context of the present work, at the nanoscale, major contributions to the intrinsic toughness of bone originate from the OPN-crosslinked protein networks (Cavelier et al., 2018; Fantner et al., 2007) and the formation of dilatational bands involving osteocalcin (OC)-osteopontin complexes (Poundarik et al., 2012). The OC-OPN complex has been recently shown to provide high shear toughness and ductility to the interfibrillar interface (Wang et al., 2020). Both the OPN-crosslinked protein networks and the OC-OPN complex presumably work together to control deformation and separation of mineralized collagen fibrils (Gao et al., 2003; Zimmermann et al., 2012). Here, we propose two co-existing mechanisms to elucidate how the addition or removal of phosphate groups on proteins, and particularly OPN, could affect bone mechanical function.

First, cation-mediated crosslinks are formed between two binding regions on one OPN polymer, multiple OPN polymers, and OPN and charged surface ions on HA (e.g., Ca2+, Na+) (Figure 6—figure supplement 3; Fantner et al., 2005). These salt-bridges are weak, but reformable sacrificial bonds that prevent portions of OPN polymers from rupturing (cohesion of the OPN meshwork) and debonding of OPN from HA during repetitive mechanical loading (Zappone et al., 2008; Fantner et al., 2007; Lai et al., 2014). The high affinity of OPN to Ca2+ ions was reinforced in our AFM-FS studies. We used bovine milk OPN as the model protein because of its natural and extensive phosphorylated status (Sørensen et al., 1995). In the presence of Ca2+ ions and when both phosphorylated milk OPN and HA are negatively charged, detachment energy increased significantly. The increase in detachment energy was also observed between OPN and mica substrate (Figure 6). Ca2+-mediated crosslinks were also formed between OPN polymers, which increased cohesion of the OPN meshwork, indicated by higher detachment energy (Figure 6—figure supplement 2), which is generally associated with loading of multiple molecules in parallel (Fantner et al., 2006). Thus, via the effects mentioned above the meshwork is able to stretch more and increase the energy required for complete detachment. Dephosphosphorylation of milk OPN or reversing the charge on HA both resulted in decreased energy dissipation (Figure 6). Similarly, phosphorylation of WT bone specimens ex-vivo under our experimental conditions caused an approximate 18% increase in fracture toughness (Figure 4a) whereas, dephosphorylation decreased toughness by 25% (Figure 5a). These results suggest that phosphorylation is enabling various matrix/mineral interactions, and hence, dissipating energy.

Second, phosphorylation can alter protein network conformation, the mechanical behavior of the organic matrix, and consequently the macroscopic fracture toughness of bone (Thurner et al., 2009; Fantner et al., 2007; Fisher et al., 2001). A recent experimental study (Malka-Gibor et al., 2017) demonstrated that intrinsically disordered proteins (IDPs) and their phosphorylation status can alter neurofilament protein alignment and distance between filaments, resulting in changed energy dissipation of the network (Figure 7). Neurofilaments are a valuable model system for examining phosphorylation-driven interactions of IDPs owing to their high modularity in protein content and phosphorylation levels. Both collagen and neurofilaments are bundled network systems that interact with IDPs (Laser-Azogui et al., 2015; Orgel et al., 2006). For example, non-collagenous proteins in bone matrix such as small integrin-binding ligand, N-linked glycoproteins (SIBLINGs) are IDPs, interact with collagen, and gain more folded features when post-translationally modified (phosphorylation, glycation, acetylation, sulfation, cleavage) (Boskey and Villarreal-Ramirez, 2016). In this regard, the SIBLING proteins (e.g. osteocalcin, osteonectin, fibrillins, etc.) interacting with collagen filaments/fibrils may be considered analogous to neurofilament proteins (Laser-Azogui et al., 2015; Yuan et al., 2017; Khalil et al., 2018). Our AFM-FS studies showed that in the presence of excess Ca2+ ions strong cohesion and excessive crosslinking of the OPN meshwork reduces the stretching ability of the meshwork, leading to shorter pulls, increased repulsion of all positive sites, which likely increased distance in the meshwork, and diminished detachment energy. We postulate that the increase in global phosphorylation of other bone matrix proteins in the absence of OPN ( e.g. other SIBLING matrix proteins) may also potentially result in increased protein alignment and larger interfilament distance between mineralized collagen fibrils, to a detrimental degree that decreases matrix interaction, energy dissipation, and consequently fracture resistance in osteopontin-deficient mice (Figure 4).

Schematic diagram showing differential effects of phosphorylation on conformation of protein systems.

In protein system (A), phosphorylation tends to increase inter- and intrafilament interactions, hence the interfilament distance is reduced. In protein system (B), phosphorylation tends to create interfilament repellant, hence increasing the protein system alignment and inter- filament distance.

We observed a non-linear dose response relationship between the level of global matrix phosphorylation and bone fracture toughness in WT mice (Figure 8a). Phosphorylation explained ~36% of the variance in fracture toughness and this relationship was not observed in the absence of OPN (Figure 8b). Taken together, this data supports the previously mentioned mechanism involving increased interaction energy and sacrificial bond formation between OPN and HA as well as between OPN polymers. The AFM-FS studies show that adhesion is not only dependent on the charge of OPN and HA under a certain environment but also the availability of free Ca2+ ions. The remaining variance in fracture toughness may be associated with the formation of OC-OPN complexes or enzymatic OPN-crosslinked protein networks. It has been previously shown that crosslinking of OPN by transglutaminase-2 enzyme (TG2) increases interfacial adhesion and toughness (Cavelier et al., 2018). However, OC inhibits TG2 crosslinking activity most likely by competing for the binding site on OPN (Kaartinen et al., 1997). As such, there is insufficient evidence at present that TG2 crosslinking of OPN and phosphorylation of OPN are independent. Although ex-vivo phosphorylation of Opn KO mice bone decreased fracture toughness (Figure 4a), and dephosphorylation increased toughness compared to the respective untreated Opn KO mice bone (Figure 5a), unlike WT, we did not observe an association between the level of global matrix phosphorylation and fracture toughness in these mice (Figure 8b). As noted above, excessive crosslinking can be detrimental to protein networks by increasing repulsion, interfilament distance, and stretching ability. This data suggests that although global matrix level of phosphorylation affects fracture toughness, the contribution of phosphorylated OPN may be critical in the determination of bone toughness.

Schematic of the relationship between global protein phosphorylation and fracture toughness of wild-type (a) and Opn KO (b) mice.

By continuing the increase in phosphorylation of WT bone, fracture toughness improves exponentially. There is no significant relationship between global phosphorylation and fracture toughness in Opn KO mice following ex-vivo phosphorylation and dephosphorylation.

Our current study is not devoid of limitations and we acknowledge other phosphorylation interactions that may potentially influence the outcomes. The gross skeletal phenotype of Opn KO mice is normal compared to WT mice (Rittling et al., 1998; Yoshitake et al., 1999). However, increased mineralization was found in some areas of cortical bone (Boskey et al., 2002), and the bones are mechanically weaker. The collagen structure in Opn KO mice was also shown to be highly disorganized which further causes disorganization of mineral (Depalle et al., 2020). OPN in bone resides at its surfaces (including lining the lacuno-canalicular system) in the thin structure known as the lamina limitans (McKee and Nanci, 1996), and throughout bulk bone. Thus, its alterations in vivo may affect many processes including mineral-binding (Addison et al., 2010), cell attachment as part of the bone remodeling cycle, cell signaling that may affect mechanosensation, and the structural integrity of bone. Our ex-vivo experiments were conducted under physiological conditions to alter the organic matrix with buffer solutions containing magnesium chloride, calcium, and EDTA to prevent any alterations in mineral. The AFM measurements are not fully quantitative, but the potential lies in examining relative differences, as was done in this study. Also, by using the same cantilever, the measurements are very accurate and reproducible. Bovine milk OPN contains approximately 28 phosphorylation sites and all but a few residues in this motif are phosphorylated. The higher phosphorylation levels essentially allow for demonstration of the principal effects seen in whole-bone fracture toughness tests following increased phosphorylation. Attempts to over-phosphorylate bovine milk OPN (our source OPN) would likely be unsuccessful as the nonphosphorylated serine residues in bovine OPN are not located in recognition sequences of any specific kinase. The lack of experiments on OPN with a varying range of phosphorylation levels may be seen as a limitation, but nevertheless we provide data points for the most extreme cases, and, different from the physiological system, we control the concentration of Ca2+ ions.

Despite the fact that the maximum-load method for measuring fracture toughness demonstrates the least variability compared to other methods (Ritchie et al., 2008), there is inherent variability in fracture toughness tests. For example, we have shown with a larger sample size that Opn KO mice have lower fracture toughness compared to WT mice (Thurner et al., 2010; Poundarik et al., 2012). The data in Figures 4 and 5 are from a different set of control bones and fracture toughness values vary across bones from the same batch of mice attributable to inherent differences between animals, and because of variations in any mechanical testing method (including fracture toughness testing). Accordingly, we have minimized variations between animals by conducting pairwise comparison i.e. WT-dephosphorylated vs WT-controls. Such comparison, as noted above, also allows us to determine the contribution of OPN phosphorylation and dephosphorylation while accounting for compounding effects of other changes in the bone matrix.

In conclusion, this study shows for the first time that osteopontin and its phosphorylation level promotes fracture toughness of bone. The heterogeneity in osteopontin phosphorylation, alters interfacial adhesion and cohesion of the OPN meshwork leading to increased or decreased energy dissipation. In the absence of osteopontin, phosphorylation and de-phosphorylation of other bone matrix proteins impact bone toughness in a binary stepwise manner. We expect that our study holds the potential to begin understanding the need for regulation of global matrix phosphorylation and heterogeneity in phosphorylation for different proteins with respect to maintaining skeletal health and whose alterations influence bone fragility in diseases.

Materials and methods

Key resources table
Reagent type
(species) or resource
DesignationSource or referenceIdentifiersAdditional
information
Genetic reagent (M. musculus)C57BL/6NCrlCharles RiverRRID:IMSR_CRL:27
Genetic reagent (M. musculus)B6.Cg-PhexHyp/JJackson LaboratoryCat#: 000528
RRID:IMSR_JAX:000528
Animals maintained in Dr M Mckee lab.
Genetic reagent (M. musculus)Fgf23-/-PMID:15579309Animals were a gift from Dr. B. Lanske
Genetic reagent (M. musculus)B6.129S6(Cg)-Spp1tm1Blh/JPMID:9661074Animals were a gift from Dr S. Rittling.
Genetic Reagent (B. taurus)Milk protein (Mammary gland)PMID:8320368Provided by Dr ES Sorensen
Chemical compound, drugSynthetic hydroxyapatiteAndriotis et al., 2010. Crystal Research and TechnologyProduced by Dr N. Bouropoulos
Commercial assay or kitpIMAGO-biotin HRP DetectionTymora AnalyticalCat# 900–100
Antibodyanti-OPN (goat polyclonal)R and D SystemsCat# AF808, RRID:AB_2194992(1:100,000 µL)
Antibodyanti-phosphoserine (rabbit polyclonal)Thermo Fisher ScientificCat# 61–8100, RRID:AB_2533940(1:2500 µL)

Immunoprecipitation and immunoblotting for OPN in mouse models of phosphate disorder

Request a detailed protocol

Long bones from 6-week-old Hyp and Fgf23-/- mice (and WT age-, strain-, and sex-matched controls, n = 3) were collected and bone proteins extracted as described previously (Goldberg and Sodek, 1994). In brief, cleaned frozen bone samples were pulverized, cooled in liquid nitrogen, and bone protein extracted from this powder twice at 4°C for 24 hr with 4 M guanidium‐HCl in 50 mM Tris‐HCl, pH 7.4 containing protease and phosphatase inhibitors (0.1 mM phenylmethylsulfonyl fluoride (PMSF), 100 µg/mL of benzamidine, 5 µg/mL leupeptin, 1 mM sodium pyrophosphate, 1 mM β-glycerophosphate, 1 mM sodium orthovanadate, and 5 mM sodium fluoride). Mineral-bound proteins were then extracted twice at 4°C for 24 hr with 0.5 M EDTA, 50 mM Tris‐HCl, pH 7.4 containing protease and phosphatase inhibitors. The mineral-binding protein fraction was then concentrated and washed in 5 mM sodium bicarbonate, then quantified using the bicinchoninic acid protein assay (Pierce, Rockford, IL, USA).

For each sample, 10 µg of total mineral-bound bone protein extract was mixed with 300 µL of 100 mM sodium acetate, pH 5.5 containing 1 mM PMSF and 0.1 mM leupeptin and incubated on ice for 3 min, and then gently mixed with rotation at 4°C for 10 min. Next, 10 µL of 0.2 mg/mL goat anti-mouse osteopontin antibody (R and D Systems, Cat# AF808-CF, Minneapolis, MN, USA) was added and samples were gently rotated at 4°C for 1 hr, followed by the addition of 50 µL of Protein A/G PLUS-Agarose beads (Santa Cruz, SC-2003, Dallas, TX, USA) and gentle rotation at 4°C for 1 hr. Samples were spun at 2000 × g for 1 min, and supernatants were removed. Beads were then washed in cold 100 mM sodium acetate buffer, pH 5.5 three times, and immunoprecipitated proteins were eluted in 2 × Laemmli protein loading buffer. Samples were resolved on a 4–20% gradient SDS-PAGE gel, transferred onto PVDF membranes and immunodetected using anti-mouse osteopontin (R and D Systems, Minneapolis, MN, USA) and anti-phosphoserine (Invitrogen, Cat# 61–8100, Carlsbad, CA, USA) antibodies. Two technical replicates were performed for Fgf23-/- mice and corresponding WT littermates experiments while four technical replicates were performed for the Hyp and corresponding WT littermates experiments.

In-vitro phosphorylation and dephosphorylation of whole mouse bone

Sample preparation

Request a detailed protocol

Previously frozen femora were dissected from twenty-eight six-month-old male Opn KO (n = 14) and C57BL/6NCrl wild-type (WT, n = 14) mice. The sample size reflects the number of independent biological replicates and were based on results from previous pilot studies and publications from the laboratory (Sroga and Vashishth, 2016; Poundarik et al., 2012). Bones were cleaned of soft tissue and femoral head and condyle removed for experimental uniformed treatment throughout the bone. A notch was created on the anterior side in the mid-shaft of all samples using a slow speed diamond blade saw and sharpened using a razor blade (IsoMet Low Speed Saw, Buehler) This method produces a sharp notch with a root radius of ~10 µm (Ritchie et al., 2008). The crack length is defined in terms of the half crack angle and fracture toughness testing is accurate for half crack angles between 0–110 degrees (Ritchie et al., 2008). A specimen was considered an outlier and removed if crack angles were larger than two standard deviations from the mean, and if notches were off-centered or extended greater than 1/3 of the cortex. Consistent with physiological loading, the anterior side was chosen so that the notch experiences tension during bending test. The notch represents a pre-existing crack that will initiate and propagate into a large-scale catastrophic fracture. The bones were then rinsed with 1 x phosphate buffered saline (PBS) and stored in saline soaked gauze at −80°C until use.

In-vitro phosphorylation and dephosphorylation

Request a detailed protocol

One limb of each animal (left or right) was randomly selected for phosphorylation and the contralateral limb served as the non-phosphorylated control [Opn KO (n = 7) and WT (n = 7)]. Phosphorylation was conducted by incubating the samples for 48 hr at 30°C with casein kinase-II (CK2) and the reaction buffer (New England BioLabs, Ipswich, MA). Adenosine triphosphate (ATP) (2 mM) was added to the buffer as the phosphoryl donor for CK2. The incubating solution also contained protease and phosphatase inhibitors (final concentration 2 x, Pierce Biotechnology, Rockford, IL), and antibiotics [ampicillin (100 μg/μL) and kanamycin (50 μg/μL)]. ATP, CK2, and antibiotics were also added second time to the reaction at the 20 hr of incubation. The non-phosphorylated samples (i.e. controls) were placed in a similar solution without added enzymes for the same time period and temperature.

In a different set of animals, one limb was randomly selected for de-phosphorylation and the contralateral limb served as the nondephosphorylated control [Opn KO (n = 7) and WT (n = 7)]. De-phosphorylation was conducted by incubating the samples for 48 hr at 37°C with calf intestinal alkaline phosphatase (CIP) and the CIP reaction buffer (New England BioLabs, Ipswich, MA). In pilot and published studies (Sroga and Vashishth, 2016) we did not observe increase in either phosphorylation or dephosphorylation of bone samples after 48 hr. The incubation solution also contained protease inhibitor and antibiotics as previously described. CIP enzyme was also added second time to the reaction at the 20 hr of incubation. The non-dephosphorylated samples (i.e. controls) were placed in a similar solution without added enzymes.

Mechanical testing

Request a detailed protocol

All femora were scanned using micro-computed tomography (μCT) at 70 kVp, 114 mA, 200 ms integration time and at high resolution 10.5 µm voxel size (vivaCT 40, Scanco Medical AG, Bassersdorf, Switzerland) for measuring bone geometry. Following in-vitro phosphorylation and dephosphorylation treatment, samples were loaded in three-point bending until failure at a loading rate of 0.001 mm/s (Elf Enduratec 3200). The resulting load displacement curve was used to calculate a single-valued fracture toughness Kc at maximum load for each sample (Ritchie et al., 2008). Toughness measured here is dependent on the material reflecting the changes due to phosphorylation or dephosphorylation.

Protein extraction, quantification, and phosphoprotein detection

Request a detailed protocol

After mechanical testing, all bones were defatted, lyophilized (freeze-dried), and weighed (approximately 20–40 mg). Samples were then placed in eppendorf tubes with 600 μL of extraction buffer consisting of 0.05 M EDTA, 4 M guanidine chloride, and 30 mM Tris-HCl. The bones were subsequently homogenized, centrifuged, and the supernatant collected (Omni Inc, Kennesaw, GA). The supernatant from each sample was placed into a micro-dialysis vial and underwent simultaneous protein isolation and demineralization over two days at 4°C, pH 7.4, against several changes of 1 x PBS and 5 mM EDTA.

The amount of protein in the samples was quantified using the Coomassie Plus (Bradford) Assay. The measurement of phosphorylated proteins was done using the pIMAGO-biotin Phosphoprotein Detection assay kit (Sroga and Vashishth, 2016) (Tymora Analytical, West Lafayette, IN). Samples were tested in triplicates for each assay. Briefly, protein mixtures were bound to the wells by overnight incubation at 4°C. After a series of washing and blocking, the wells were incubated with pIMAGO reagent for attachment of the nanopolymer to phosphate groups on proteins. The wells were washed again, incubated with avidin-HRP followed by the provided colorimetric-based detection system. The absorbance was read at 415 nm using a micro-plate reader (Infinite M200, Tecan). The amount of global protein phosphorylation was calculated as absorbance/ng of protein. Assays for protein concentration and phosphoprotein detection were ran in triplicates.

Data analysis for global phosphorylation

Request a detailed protocol

The mean and standard deviation were calculated for total protein phosphorylation amount and fracture toughness. Paired samples t-test was used to compare differences between the groups (WT-phosphorylated vs. WT-nonphosphorylated; Opn KO-phosphorylated vs. Opn KO -non phosphorylated). Because phosphorylation modifies the organic matrix including OPN, we compared the change in fracture toughness caused by phosphorylation of the organic matrix with (WT-treated minus WT-control, delta-WT) and without osteopontin (Opn KO-treated minus Opn KO -control, delta- Opn KO) by independent samples t-test. The same analysis was done for dephosphorylated samples and nondephosphorylated controls conducted on separate animals. All analyses were conducted using IBM SPSS 21 and two-tailed significance threshold set at 0.05 for both paired and independent samples t-test.

Atomic force microscopy – force spectroscopy studies

Chemicals

All chemicals were purchased from Sigma-Aldrich (Sigma-Aldrich Company Ltd., Gillingham, Dorset, UK) unless otherwise stated.

Preparation and characterization of hydroxyapatite (HA) powder

Request a detailed protocol

Synthetic HA was produced for the functionalization of the AFM cantilever in order to simulate the mineralized fiber – NCP – mineralized fiber interaction. The preparation of the synthetic HA crystals was performed by the simultaneous addition of 250 mL aqueous solution of H3PO4 (0.3 M) and 250 mL aqueous solution of CaCl2·2H2O (0.5 M) to 500 mL ultrapure boiling water. To avoid temperature fluctuation, both reactants were added at a rate of approximately 2.5 mL per minute under continuous stirring. Prior and during the addition of the reactants, nitrogen gas was bubbled through the solution in order to remove the dissolved CO2. At all times, the pH was kept between 9.0 and 10.0 by the addition of concentrated NH4OH solution. Upon the completion of the addition, the solution was kept under stirring for 24 hr at 80°C before cooling to room temperature. To retrieve the HA crystals, the suspension was filtered through a 0.22 μm membrane filters (Whatman, Maidstone England). Finally, the crystals were dried and ‘matured’ at 150°C for 24 hr and stored in a desiccator. The end product was characterized by means of X-Ray Diffraction (XRD), Fourier Transform Infrared Spectroscopy (FTIR) and Scanning Electron Microscopy equipped with Energy-dispersive X-ray analyser (SEM/EDX; Zeiss Supra 35VP). XRD analysis was performed using a standard powder diffractometer (Siemens D8) with Ni-filtered CuKa1 radiation (λ = 0.154059 nm) and the acquired diffraction spectra were matched against JCPDS reference data using the EVA XRD software. The FTIR spectra were acquired using an Excalibur spectrophotometer (Digilab, Japan) at a resolution of 2 cm−1 using the KBr pellet method.

Preparation and characterization of hydroxyapatite (HA) surfaces

Request a detailed protocol

HA surfaces were prepared through the ‘maturation’ of CaP cements in Ringer solution as described previously (Knychala et al., 2013; Andriotis et al., 2010). In brief, the cements were made by mixing alpha-tricalcium phosphate (a-TCP) powder with 4.0 % w/v disodium hydrogen phosphate (Na2HPO4) solution at liquid (mL)/powder (g) ratio of 0.32, homogenized by a spatula for 1 min in agate mortar and then spread carefully on Silastic M RTV Silicone Rubber moulds. The specimens were kept in 100% humidity for 12 hr and then placed in 60 mL of Ringer’s solution at 37°C for 7 days to harden. During the maturation period, the a-TCP is transformed into calcium-deficient HA following the hydrolysis of the a-TCP according to the reaction 3Ca3(PO4)2 + H2O → Ca9(HPO4)(PO4)5OH (Ginebra et al., 2004).

OPN purification and dephosphorylation

Request a detailed protocol

In bovine milk, OPN is subjected to proteolytic processing by proteinases such as thrombin (Grassinger et al., 2009), plasmin, cathepsin D or matrix metalloproteinases (Christensen et al., 2010). In this work isolated OPN from bovine milk as essentially described in Sørensen and Petersen, 1993. The principal components are N-terminal OPN fragments ending between residues 145 and 153 of the mature protein as well as the mature full-length protein (Christensen and Sørensen, 2014). After isolation, OPN was stored in a desiccator at room temperature until use. Dephosphorylated milk OPN was prepared as described in Boskey et al., 2012. Briefly, OPN was incubated with bovine alkaline phosphatase (ALP) (20 mU ALP/μg protein) in 10 mM ammonium bicarbonate (pH 8.5) overnight at 37°C and subsequently analyzed by MALDI-TOF MS to verify dephosphorylation.

Buffer solutions

Request a detailed protocol

The buffers used were the same as previous studies (Katayama et al., 1998; Fantner et al., 2005; Lai et al., 2014). More specifically, Na Buffer (150 mM NaCl, 10 mM HEPES), Ca Buffer (40 mM CaCl2, 110 mM NaCl, 10 mM HEPES), and ultra pure water (H2O). Each solution was divided into separate vials and ph adjusted for each experiment using either HCl or concentrated NaOH solution.

Adsorption of OPN on model surfaces

Request a detailed protocol

The lyophilized OPN was dissolved in ultrapure water (concentration 2 μg/µL) and absorption of OPN film on the model surfaces (HA or mica) was accomplished using the ‘drying droplet’ method. During this process, a small drop (4 μL) of OPN solution was deposited onto a freshly cleaned and dried HA or mica surface which was previously glued on the bottom of the fluid cell using 5 minute-setting epoxy. The droplet was then left to dry inside the AFM hood forming a thin protein film on the model surface, and then rehydrated with the appropriate solution.

AFM cantilevers for force spectroscopy measurements

Request a detailed protocol

One aggregate of synthetic HA crystals was glued to a tipless monolithic silicon AFM probe (AIO-TL, Budget Sensors) using epoxy glue (Araldixe, Huntsman, The Woodlands, Texas, USA). For this, a few micrograms of the synthetic HA crystals were added in 5 mL of ethanol and stirred vigorously to produce a dispersion. At this stage, 500 μL of this dispersion were deposited onto a glass slide and left to dry. A droplet of epoxy was placed by the side of the dry crystals and the glass slide was placed into the AFM. The AFM probe was then engaged carefully onto the epoxy, pulled back, and engaged again on the aggregate of choice. After two minutes in contact, the probe was withdrawn and left in the AFM for an additional 30 min to ensure complete setting of the epoxy. An example of the end result is presented in Figure 6—figure supplement 1.

Force spectroscopy experiments

Request a detailed protocol

Force spectroscopy measurements of the adhesive properties of the OPN film under various ionic environments were conducted by means of an atomic force microscope (MFP3D, Asylum Research, Santa Barbara, CA, USA) using an open fluid cell setup. Following Fantner et al., 2005 protocol, all experiments were performed subsequently and at the same location. Exchange of solution, for altering the ionic environment and the pH, was carried using a syringe-pump inlet/outlet system without moving the head. In each environment, 50–80 pulls were collected and analyzed using a custom made Matlab script (version 7.10.0.4999, The MathWorks Inc, Natick, Massachusetts, USA). For each force curve, the cantilever was positioned 3 μm away of the surface, driven in full contact with it, and after a dwell time of 10 s was retracted back to the starting position. During these cycles, the approach and retraction velocities were set to 2.0 μm/sec and 5.0 μm/sec, respectively. Full contact was defined as the tip-sample repulsive force reaching a threshold value of 15 nN. The spring constant, k, of the cantilever probe was measured prior to the functionalization using the thermal noise method (Ritchie et al., 2008), and followed by the Inverse Optical Lever Sensitivity (InvOLS) of the system. The later was determined by acquiring ten (10) force curves on a nominally infinitely stiff surface (i.e. the glass slide). A line was then fitted on the loading part of each force curve and the slope of the fitted line was used as the InvOLS. The mean InvOLS value of all ten curves was then used as the InvOLS of the cantilever. In the case of the HA-functionalized cantilevers the spring constant was reassessed using the thermal method post-functionalization and the resulting value was used for the analysis. Force spectroscopy measurements of phosphorylated/dephosphorylated OPN on mica surfaces were conducted using Olympus BL-RC150VB-C1 Bio-levers (Olympus Optical Co., Ltd., Tokyo, Japan); spring constant 6 pN/nm (0.006 N/m), while stiffer (c. 0.18 N/m) cantilevers were used for the HA experiments. Maximum force from force spectroscopy experiments are reported in Supplementary files 3 and 4.

Data processing and analysis

Request a detailed protocol

All force curves were exported in ASCII (plain text files) and processed in Matlab. Each force curve was split into its approaching and retraction parts (Figure 6—figure supplement 2). Energy dissipation was defined as the area enclosed by the retraction curve and the X-axis from the point of contact (X = 0, Y = 0) to the ‘Pulling Length’; where the latter was defined as the length from contact to the maximum distance at which the adhesion is smaller than the 1.0% of the Maximum Force (maximum adhesion force registered during retraction, i.e the Y-minimum of the retraction curve). Statistical analysis was performed in Origin (OriginPro version 9.0.0; OriginLab Corporation, Northampton, MA, USA). The normality of the distributions was assessed by means of a Kolmogorov-Smirnov test. Differences in Energy between the different environments were assessed by means of two-sided unpaired Student’s t-test (significance threshold p=0.05).

Data availability

All data generated or analyzed during this study are included in the manuscript and supporting files.

References

    1. Clarke B
    (2008) Normal bone anatomy and physiology
    Clinical Journal of the American Society of Nephrology 3 Suppl 3:S131–S139.
    https://doi.org/10.2215/CJN.04151206
    1. Liu S
    2. Zhou J
    3. Tang W
    4. Jiang X
    5. Rowe DW
    6. Quarles LD
    (2006) Pathogenic role of Fgf23 in hyp mice
    American Journal of Physiology. Endocrinology and Metabolism 291:E38–E49.
    https://doi.org/10.1152/ajpendo.00008.2006
    1. Thurner PJ
    (2009) Atomic force microscopy and indentation force measurement of bone
    Wiley Interdisciplinary Reviews: Nanomedicine and Nanobiotechnology 1:624–649.
    https://doi.org/10.1002/wnan.56

Decision letter

  1. Clifford J Rosen
    Senior Editor; Maine Medical Center Research Institute, United States
  2. Cheryl Ackert-Bicknell
    Reviewing Editor; University of Colorado, United States
  3. Virginia Ferguson
    Reviewer

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

Increasing attention in the bone field is focused on understanding the nuances of the composition of extracellular matrices and how post translational modifications of constituent proteins alters cellular interactions, tissue integrity and function. In an elegant series of experiments, the authors show that the phosphorylation status of osteopontin can be modulated, altering its energy dissipation properties thereby promoting fracture toughness, but that there is a ceiling to this effect. The mechanisms by which phosphorylation alters facture toughness are explored in detail and these results may be informative for understanding the properties of the extracellular matrix in other connective tissues.

Decision letter after peer review:

Thank you for submitting your article "The Role of Extracellular Matrix Phosphorylation on Energy Dissipation in Bone" for consideration by eLife. Your article has been reviewed Clifford Rosen as the Senior Editor, a Reviewing Editor, and three reviewers. The following individual involved in review of your submission has agreed to reveal their identity: Virginia Ferguson (Reviewer #1).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

We would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). Specifically, when editors judge that a submitted work as a whole belongs in eLife but that some conclusions require a modest amount of additional new data and or revision, as they do with your paper, we are asking that the manuscript be revised to either limit claims to those supported by data in hand, or to explicitly state that the relevant conclusions require additional supporting data.

Our expectation is that the authors will eventually carry out the additional experiments and report on how they affect the relevant conclusions either in a preprint on bioRxiv or medRxiv, or if appropriate, as a Research Advance in eLife, either of which would be linked to the original paper.

Summary:

There is a great deal of interest in the field in determining how the non-collagenous aspects of the bone matrix contribute to the overall resistance of bone tissue to fracture. In this manuscript, the authors seek to investigate the importance of phosphorylation status of osteopontin to bone toughness and how this non-collagenous extracellular matrix protein impacts fracture toughness. Of particular interest in this paper is how the contribution of osteopontin phosphorylation to whole bone toughness is explored at the nano-scale, protein-mineral interactions level. They determine that the phosphorylation of OPN increases electrostatic interactions and cation-mediated crosslinks, allowing for increased energy dissipation under loading. Overall, this was considered to be a well written and interesting paper comprised of well thought out experiments.

During the review many requests for additional information and clarification were made by the reviewers but few experiments. The reviewers wondered if alternate mechanisms could be competing with OPN towards affecting fracture toughness. In other instances, it was wondered if alternate explanations might not explain these data. These concerns are summarized below.

Essential revisions:

1) The limitations of this study were glossed over in terms of the limits of the OPN-/- mouse model (e.g., what else is changing in the OPN deficient bone during development that could influence experimental outcomes?), and what sources of error may exist and be generated from specific methods (e.g., AFM testing or the intrinsic variability of notch fracture testing of whole mouse bones).

2) Could there be additional interactions that were potentially overlooked where phosphorylation status changes the way that OPN interacts with other matrix molecules in bone? What about the pH effects on other proteins, GAGs, etc. in bone – could these have complicated or contributed to the results that were observed in this manuscript in any meaningful way? The sole focus of this manuscript on OPN is beneficial, but it also excludes many other concomitant effects of phosphorylation/dephosphorylation and pH that could compound the experimental outcomes that were observed. The authors are asked to briefly justify (and consider including in their manuscript) a justification of why these effects on other matrix molecules were not the dominant outcome – in other words that OPN is the sole or predominant contributor to the endpoint measurements of interest (e.g., fracture toughness).

3) Figure 4, Figure 5 include bars for statistical comparisons (marked with *) that are sometimes not located in a manner to compare two groups. It is thus difficult to see which groups are intended to be marked in a comparison. Figure 5 also seems to have lost the -/- for OPN-/- group within the figure legend, and Figure 4B is missing "C" in the "Change in.…" title. To provide readers a transparent interpretation of the results, consider overlaying dot plots on the bar charts. That is, show the actual data points. This will also convey the sample size more clearly.

4) Data from control mice in Figure 4 and Figure 5 don't seem to match. What could this mean when the differences in means are comparable to those observed in some of your experimental comparisons? Does this cast doubt on the overall results of fracture toughness testing? How can we believe the results you present if the WT groups are not producing repeatable results?

5) Why was 48 hours chosen for phosphorylation and dephosphorylation? Is the (de)phosphorylation achieved with this protocol physiologically-relevant?

6) The position and mass of the HA cluster on the cantilever would be expected to affect the spring constant of the cantilever. How was this accounted for? It is only mentioned that the cantilevers were calibrated before functionalization.

7) For AFM testing, how was the effect of different shaped crystal clusters on electrostatic interaction forces accounted for? Numbers of replicates/group for AFM testing and some other assays were unclear. This is essentially a contact mechanics question, since the surface area of contact could be quite different for different clusters. More detail about the normalization procedure could help here (was the normalization for the same functionalized HA geometry?). How was the full contact threshold force decided? For experiments with changing buffers, was the order of buffer change randomized? Otherwise, buffer effects may be confounded with washing away part of the OPN or the crystal.

8) Did mineralization or collagen crosslinking change with phosphorylation or dephosphorylation treatments of femurs? The authors state that "loss of OPN resulted in reduced fracture toughness without accompanied changes in calcium variability", but it is not clear if it is assumed that mineralization did not change with (de)phosphorylation processes.

9) Statistics: In the case of a significant interaction, which the authors have found, simple effects can be tested, but family-wise error should be considered and accounted for to correct for the greater likelihood of Type I error. It’s not correct to t-test first and then use ANOVA to look for an interaction. It is better to fit your data to a two-factor ANOVA, and in the case of an interaction, perform post-hoc simple effects testing that is corrected for these multiple comparisons. Further, details ensuring model fit (normality and homoscedascity of residuals) would be appreciated. It is suggested to present data as a mean instead of an average. "Average" can be defined as mean, median, mode… etc.

10) The sentence “Treatment affected fracture toughness of WT bone…” is confusing. There is a significant interaction, and that p-value should be reported, but instead, p-values are reported for phosphorylation and dephosphorylation.

11) For quantification of phosphorylation in Hyp and FGF23-/- mice, the mice were very young (6 weeks old). The mice were much older (6 months) for fracture toughness testing. Phosphorylation changes with aging; how much less phosphorylation would these mouse models have at 6-mo of age, and how would this relate to the WT mice?

12) The fracture toughness of bone originates from nanoscale energy dissipation at the protein-mineral interface. While this toughening mechanism is important, it is not the only important toughening mechanism in bone. Toughening mechanisms span length-scales and involve mineral, collagen, and their interactions. A slight modification to language to contextualize this toughening mechanism as an important one of several would be helpful.

13) The authors' finding that there is a “sweet spot” of phosphorylation was considered to be very interesting. However, this concept of sweet spot confused the reviewers in the manner that it was presented. The authors did not provide a fracture toughness vs. phosphorylation relationship. That is, by continuing the increase in phosphorylation of WT bone, does the improvement in fracture toughness stop and decline? With the loss of OPN, phosphorylation had a negative effect. This was interpreted to mean that phosphorylation levels returned to normal, but really, the finding implicates the phosphorylation of OPN as being required to improve toughness. It is not clear what subset of phosphorylation values within the studied range of phosphorylation here is relevant in disease. The “sweet spot” mechanism likely needs to be removed or qualified. Either way, its description is rather difficult to follow.

14) The in vitro (de)phosphorylation had some variability. Did amount of (de)phosphorylation correlate (within groups) with fracture toughness?

15) The authors postulate that TG2 and phosphorylation may separately act as toughening mechanism. This is supported by FGF23 and Hyp mice having increased TG2, which should toughen bone, but also phosphorylation. But what if TG2 also has a nonlinear relationship with bone mechanics? The reviewers were concerned that there was not enough evidence in the paper to say that TG2 and phosphorylation are independent, but felt this is a very interesting discussion point. Absolutes should be tempered here.

16) A reviewer asked if one could over-phosphorylate the bovine OPN to potentially observe the quadratic relationship with energy dissipation for confirmation of the relationship observed for whole-bone testing? Some discussion on this point would be appreciated.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "The Role of Extracellular Matrix Phosphorylation on Energy Dissipation in Bone" for further consideration by eLife. Your revised article has been evaluated by Clifford Rosen (Senior Editor) and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed before formal acceptance, as outlined below. Please note that these requested clarifications were agreed as necessary by all reviewers during the post review correspondence.

Reviewer #1:

This manuscript is well written and is on a topic that is compelling and important to the field. The authors have given sufficient attention to the concerns expressed by the reviewers, including myself. I feel satisfied that the responses were diligent and thorough and the manuscript is improved as a result.

Upon revisiting Figure 8A, I agree that a quadratic fit is not necessarily the best choice and that an exponential function should be considered. Concerningly, if the very most highly phosphorylated point is excluded, the data look very much like a plateau. This would challenge the “sweet spot” idea as applied to just WT bones. As reviewer 3 notes it “seems like the study showed that phosphorylation promotes fracture toughness (to a point) if OPN is present”. I agree with this interpretation based on the presented data, which is consistent with more of a asymptote than a sweet spot in Figure 8A.

Reviewer 3 asks “why doesn't the increase in fiber network modulus and strength continue to increase in phosphorylation”? I read this differently, with the interpretation that modulus and strength could indeed increase with phosphorylation but to a detrimental degree that decreases strain. Clarification is required as the reader will likely be confused on this point as well.

It would be helpful to show all data (not just four means) for Figure 8B and to indicate which data correspond with each group. This would alleviate confusion surrounding this figure.

The authors should clarify in the Abstract that the phosphorylation and dephosphorylation experiments were conducted ex vivo.

Reviewer 3 comments that the abstract only mentions one mechanism of the three in the discussion. My bigger issue is that mechanisms 1 and 3 are co-dependent as presented (they are described as "three different, co-existing mechanisms". The authors argue that mineral-matrix interactions are facilitated by OPN phosphorylation, but this would be the real mechanism and the “sweet spot” or “benefit to a point” would be the relationship of this mechanism to dose.

The authors should add a little bit of clarifying statistical information (1 vs 2 tailed testing, critical α). These are stated for the AFM experiments but not for the fracture toughness and phosphorylation experiments.

Reviewer #2:

The revised manuscript is considerably improved. The detailed Response to reviewers is appreciated, as well as the updates to figures and text. The revised Figure 8 is particularly appreciated, notwithstanding the comments made by other reviewers.

I agree that their findings are exiting and that the explanations need to be improved so that the readers do not need to make inferences that may be off track.

Reviewer #3:

The study is intriguing because it shows how phosphorylation promotes fracture toughness of bone if osteopontin is present. However, this reviewer still struggles with the authors' interpretation of the mechanisms by which phosphorylation does this (they propose 3). A number of questions come up after reading the Abstract and Discussion. I like the concept that modulus and strength increases with phosphorylation but at a cost of decreasing strain after some threshold. The authors just need to do a better job describing each mechanism and how they are inter-related so that readers don't have to interpret their interpretation.

How did the authors decide on fitting a quadratic equation to fracture toughness vs. phosphorylation levels of WT bone? To this reviewer, it looks like fracture toughness plateaued after 1.5 in Figure 8A. What about fitting an exponential function, exp(1/phos)?

Discussion: "However, further increase in the degree of most favorable interaction (i.e. larger than the "sweet spot" level) can cause a rapid decrease in ultimate strain, and thus, energy dissipation." What is the physical mechanism that supports this assertion? In other words, why doesn't the increase fiber network modulus and strength continue to increase with an increase in phosphorylation?

Another thing that is challenging to the interpretation of the results is that fact OPN bridges mineral crystals, presumably more bridges with more phosphorylation. Correct? This is the first mechanism described. In the OPN-/- bone, this mechanism is gone. Is the relationship between fracture toughness of OPN-/- bone and phosphorylation levels (from the 2 enzyme treatments) non-linear? Sorry, Figure 8B is not too convincing with only 4 data points.

https://doi.org/10.7554/eLife.58184.sa1

Author response

Essential revisions:

1) The limitations of this study were glossed over in terms of the limits of the OPN -/- mouse model (e.g., what else is changing in the OPN deficient bone during development that could influence experimental outcomes?), and what sources of error may exist and be generated from specific methods (e.g., AFM testing or the intrinsic variability of notch fracture testing of whole mouse bones).

The limitations of the study are now reported in the Discussion.

2) Could there be additional interactions that were potentially overlooked where phosphorylation status changes the way that OPN interacts with other matrix molecules in bone? What about the pH effects on other proteins, GAGs, etc. in bone – could these have complicated or contributed to the results that were observed in this manuscript in any meaningful way? The sole focus of this manuscript on OPN is beneficial, but it also excludes many other concomitant effects of phosphorylation/dephosphorylation and pH that could compound the experimental outcomes that were observed. The authors are asked to briefly justify (and consider including in their manuscript) a justification of why these effects on other matrix molecules were not the dominant outcome – in other words that OPN is the sole or predominant contributor to the endpoint measurements of interest (e.g., fracture toughness).

Our experiments were conducted at pH 7.4 under physiological conditions, and any instability in pH would affect proteins in both WT and Opn-/- groups. By using Opn-/- mice we can determine the contribution of all other matrix proteins in the absence of OPN. Likewise, by comparing the change in fracture toughness caused by phosphorylation or dephosphorylation of the organic matrix with osteopontin (WT-treated minus WT-control, δ-WT) and without osteopontin (Opn-/- -treated minus Opn-/- -control, δ-Opn-/-), we can isolate the contribution from OPN alone.

OPN in bone resides at its surfaces (including lining the lacuno-canalicular system) in the thin structure known as the lamina limitans (McKee and Nanci, 1996), and throughout bulk bone, so alterations in vivo may affect many processes where phosphorylation is important. These would seem to be predominantly involved in mineral-binding and possibly modifying TG2 crosslinking sites, and in cell signaling that may affect mechanosensation and bone remodeling. Dephosphorylation of OPN by the enzymatic activity of TRAP also affects osteoclast integrin binding in vitro, and thus altered bone remodeling over time affecting bone's mechanical properties. In addition to its interaction with mineral, the phosphorylation status of OPN is likely to influence its interaction with collagen and osteocalcin. Reflecting this, new text has been added to the Discussion section.

3) Figure 4, Figure 5 include bars for statistical comparisons (marked with *) that are sometimes not located in a manner to compare two groups. It is thus difficult to see which groups are intended to be marked in a comparison. Figure 5 also seems to have lost the -/- for OPN -/- group within the figure legend, and Figure 4B is missing "C" in the "Change in.…" title. To provide readers a transparent interpretation of the results, consider overlaying dot plots on the bar charts. That is, show the actual data points. This will also convey the sample size more clearly.

As suggested, these figures have been corrected. Thank you.

4) Data from control mice in Figure 4 and Figure 5 don't seem to match. What could this mean when the differences in means are comparable to those observed in some of your experimental comparisons? Does this cast doubt on the overall results of fracture toughness testing? How can we believe the results you present if the WT groups are not producing repeatable results?

There is inherent variability in fracture toughness testing despite the fact that the maximum load method for measuring fracture toughness of mice bone used in this study has the least variability compared to other methods (Ritchie et al., 2008). We and others have shown with a larger sample size that Opn-/- mice have lower fracture toughness compared to WT mice (Thurner et al., 2010; Poundarik et al., 2012). This study was not powered to detect differences between controls of different genotypes, but rather the experimental design was to examine differences between paired groups i.e. WT-dephosphorylated vs WT-controls. The data in Figure 4 and Figure 5 are from a different set of control bones and fracture toughness values vary across bones from the same batch of mice due to inherent differences between animals, and due to variations in any mechanical testing method including fracture toughness testing. In this case, we have minimized variations between animals by conducting pairwise comparison (where one bone is kept as control and the other modified in vitro under carefully controlled conditions, thus only one factor i.e. phosphorylation/dephosphosphorylation is changed). This has been noted in the Discussion section.

5) Why was 48 hours chosen for phosphorylation and dephosphorylation? Is the (de)phosphorylation achieved with this protocol physiologically-relevant?

Our in vitro studies were performed using enzymes which require physiological conditions for the reaction to be successful. The physiological reaction time for a given enzyme, both in vivo and in vitro, is its lifetime. All enzymes stop to function (i.e. “die”) due to, for example, enzyme unfolding, damage to the active center, adsorption on the tube wall, etc. In pilot and published studies (Sroga and Vashishth, 2016) we did not observe an increase in either phosphorylation or dephosphorylation of bone samples after 48 hours. This has been noted in methods section in vitro Phosphorylation and Dephosphorylation.

6) The position and mass of the HA cluster on the cantilever would be expected to affect the spring constant of the cantilever. How was this accounted for? It is only mentioned that the cantilevers were calibrated before functionalization.

Thank you for raising this point. The text has been modified in subsection “Force spectroscopy experiments” to include the requested information. In the case of the HA-functionalized cantilevers, the calibration of the cantilever’s spring constant, done prior to functionalization, was used to calibrate the Inverse Optical Lever Sensitivity (InvOLS), which required force curves to be acquired on a nominally infinitely stiff surface. (The Inverse Optical Lever Sensitivity (InvOLS) was determined by acquiring ten (10) force curves on a nominally infinitely stiff surface (i.e. the glass slide). A line was then fitted on the loading part of each force curve and the slope of the fitted line was used as the InvOLS. The mean InvOLS value of all ten curves was then used as the InvOLS of the cantilever.) The reason we avoided calibrating the InvOLS post-functionalization was to reduce the chance of mechanical damage to the functionalized end of the cantilever, which could affect both the adhesive (epoxy) and the integrity of the HA aggregate.

Following InvOLS calibration and cantilever functionalization, the spring constant was re-assessed using the thermal noise method and the new value was used for the analysis. However, we agree that the presence of a foreign body on the cantilever can influence its behavior. To ensure comparable results, care was taken so that the selected HA aggregate was always of similar shape and size. This was further confirmed post-calibration by assessing the amount of cantilever deflection. It is worth noting that small variations in spring constant calibration would mainly affect the reading of maximum force and, in this study, we are presenting and discussing the relative energy dissipation change between the OPN groups, which is primarily affected by the pulling Length.

7) For AFM testing, how was the effect of different shaped crystal clusters on electrostatic interaction forces accounted for? Numbers of replicates/group for AFM testing and some other assays were unclear. This is essentially a contact mechanics question, since the surface area of contact could be quite different for different clusters. More detail about the normalization procedure could help here (was the normalization for the same functionalized HA geometry?). How was the full contact threshold force decided? For experiments with changing buffers, was the order of buffer change randomized? Otherwise, buffer effects may be confounded with washing away part of the OPN or the crystal.

Our experiments were conducted in relative terms; thus, correcting for variations of crystal cluster shape was not applicable. We normalized our measurements to the dissipation levels in EDTA (for OPN on mica) and to dissipation levels in H2O (for OPN on HA). For each condition, we collected 50 < n < 60 force curves for the OPN on HA experiments and 80 < n < 100 for the OPN on mica experiment. Full contact was ensured by pressing the cantilever against the coated surface to the point where the force curve slope approached infinity, condition which was achieved for contact forces > 5 nN. The order of the buffer change was not randomized. It followed the order shown in Figure 6 (H2O, Na, Ca).

8) Did mineralization or collagen crosslinking change with phosphorylation or dephosphorylation treatments of femurs? The authors state that "loss of OPN resulted in reduced fracture toughness without accompanied changes in calcium variability", but it is not clear if it is assumed that mineralization did not change with (de)phosphorylation processes.

The lack of changes in calcium variability was reported in Thurner et al., 2010. Crosslinking between collagen remains unchanged. Dephosphorylation reverses the inhibitory effect of OPN on HA formation in vitro and thus mineralization may be affected. In this study, the experiments were conducted under physiological conditions with buffer solutions containing magnesium chloride, calcium, and EDTA to prevent any alterations in mineral. Thus, our conditions do not induce crystallization. This has been noted in the Discussion section.

9) Statistics: In the case of a significant interaction, which the authors have found, simple effects can be tested, but family-wise error should be considered and accounted for to correct for the greater likelihood of Type I error. It’s not correct to t-test first and then use ANOVA to look for an interaction. It is better to fit your data to a two-factor ANOVA, and in the case of an interaction, perform post-hoc simple effects testing that is corrected for these multiple comparisons. Further, details ensuring model fit (normality and homoscedascity of residuals) would be appreciated. It is suggested to present data as a mean instead of an average. "Average" can be defined as mean, median, mode… etc.

Kindly note that, as mentioned above and in the manuscript (Discussion section, Materials and methods secrtion), this study is motivated from results showing differences between Opn-/- and WT mouse bone, and the goal was to test whether modification of OPN by phosphorylation (and not all matrix proteins) affect the outcome. Thus, given that in vivo phosphorylation and dephosphorylation of WT and Opn-/- mice may cause changes other than in the level of phosphorylated OPN, we designed our experimental method where WT and Opn-/- mouse bones were paired and altered in vitro under carefully controlled conditions to modify a selected aspect. Pairwise comparisons where both limbs are from the same animal are valid (for example, comparing WT-phosphorylated vs WT-non-phosphorylated control where both limbs are from the same animal; likewise, comparing Opn-/- phosphorylated vs Opn-/- non-phosphorylated controls). Both Kolmogorov-Smirnov and Shapiro-Wilk Tests of Normality revealed that phosphorylation levels and fracture toughness were normally distributed (p>0.05). Next, to determine whether phosphorylation of osteopontin contributes to fracture toughness, we compared δ-WT (the change between WT-phosphorylated and WT-non-phosphorylated control), and δ-Opn-/- (the change between Opn-/- phosphorylated vs Opn-/- non-phosphorylated controls). By comparing δ-WT and δ-Opn-/- the difference detected would be attributed to the presence of osteopontin since all other proteins that were phosphorylated would be subtracted as background. Two-way ANOVA tests are not designed to evaluate the difference between WT control and WT phosphorylated and Opn-/- control and Opn-/- phosphorylated but between the mean values of these four groups and do not allow us to test whether phosphorylation of OPN causes a difference.

10) The sentence “Treatment affected fracture toughness of WT bone…” is confusing. There is a significant interaction, and that p-value should be reported, but instead, p-values are reported for phosphorylation and dephosphorylation.

Please see the response to the above comment where we include the rationale of not performing two-way ANOVA tests. To avoid any confusion; this sentence has been removed.

11) For quantification of phosphorylation in Hyp and FGF23-/- mice, the mice were very young (6 weeks old). The mice were much older (6 months) for fracture toughness testing. Phosphorylation changes with aging; how much less phosphorylation would these mouse models have at 6-mo of age, and how would this relate to the WT mice?

This is a very interesting question. Our assay for the quantification of global protein phosphorylation was recently developed (Sroga and Vashishth, 2016) and to date no study has quantified the amount of phosphorylation in normal and diseased bone during development. We have previously reported that total phosphorylation of bone matrix proteins including OPN declines with age (Sroga and Vashishth, 2018). Despite our looking at only one mouse age so far, our results here are the first to show that global phosphorylation, as well as OPN phosphorylation, declines in osteomalacic Hyp and Fgf23-/- mice, and in our view, this makes an important initial contribution that we chose to mention here. Quantification of the amount of phosphorylation in these models, and at different ages, will be done in future studies as an additional full-scale project. Fracture toughness testing on the 6-week old mice used here is very difficult to perform, and less reliable, but older mice will be incorporated into our future planned work on these mutant mice.

12) The fracture toughness of bone originates from nanoscale energy dissipation at the protein-mineral interface. While this toughening mechanism is important, it is not the only important toughening mechanism in bone. Toughening mechanisms span length-scales and involve mineral, collagen, and their interactions. A slight modification to language to contextualize this toughening mechanism as an important one of several would be helpful.

The statement has been clarified to “Fracture resistance of bone emerges from various mechanisms that exists at multiple length scales across bone hierarchy, which involves in part growth and packing of mineral foci into larger crossfibrillar aggregates such that the extracellular matrix becomes highly mineralized[…]Both the OPN-crosslinked protein networks and the OC-OPN complex presumably work together to control deformation and separation of mineralized collagen fibrils” please see Discussion section.

13) The authors' finding that there is a “sweet spot” of phosphorylation was considered to be very interesting. However, this concept of sweet spot confused the reviewers in the manner that it was presented. The authors did not provide a fracture toughness vs. phosphorylation relationship. That is, by continuing the increase in phosphorylation of WT bone, does the improvement in fracture toughness stop and decline? With the loss of OPN, phosphorylation had a negative effect. This was interpreted to mean that phosphorylation levels returned to normal, but really, the finding implicates the phosphorylation of OPN as being required to improve toughness. It is not clear what subset of phosphorylation values within the studied range of phosphorylation here is relevant in disease. The “sweet spot” mechanism likely needs to be removed or qualified. Either way, its description is rather difficult to follow.

Thank you for the suggestion. We have revised Figure 8 and provided a fracture toughness vs phosphorylation graph using the values from the study. By continuing the increase in phosphorylation of WT bone, fracture toughness does improve and then declines. This relationship is modeled by following equation y = 2.086+ 2.574x- 0.6415x2 (R2=0.49, F=7.72, p=0.00450).

14) The in vitro (de)phosphorylation had some variability. Did amount of (de)phosphorylation correlate (within groups) with fracture toughness?

Due to the variability, the amount of dephosphorylation did not correlate with fracture toughness. In addition, the correlation coefficient will detect a linear relationship, but our data shows that global phosphorylation levels in WT bone follows a non-linear relationship as shown in comment #12.

15) The authors postulate that TG2 and phosphorylation may separately act as toughening mechanism. This is supported by FGF23 and Hyp mice having increased TG2, which should toughen bone, but also phosphorylation. But what if TG2 also has a nonlinear relationship with bone mechanics? The reviewers were concerned that there was not enough evidence in the paper to say that TG2 and phosphorylation are independent, but felt this is a very interesting discussion point. Absolutes should be tempered here.

We agree that there is not enough evidence and have tempered down the independence of TG2 and phosphorylation. Please see the Discussion section.

16) A reviewer asked if one could over-phosphorylate the bovine OPN to potentially observe the quadratic relationship with energy dissipation for confirmation of the relationship observed for whole-bone testing? Some discussion on this point would be appreciated.

Thank you for this suggestion. We have added relevant discussion on this topic (please see the Discussion section). In particular, OPN is phosphorylated by the FAM20C kinase. The primary motif for phosphorylation is S/T-X-E/S(P)/D. Bovine milk OPN contains approximately 28 phosphorylation sites and all but a few residues in this motif are phosphorylated. If more sites should be phosphorylated, a kinase that phosphorylates serines/threonines in another motif should be used. However, in this context, we do not think that this approach would be successful as the non-phosphorylated serines in bovine OPN are not located in recognition sequences of any specific kinase. Therefore, milk OPN is already close to the over-phosphorylated OPN animal model. In hindsight, it may be possible to perform measurements on partially de-phosphorylated OPN and, in this sense, assess the whole curve shown. However, we note that several assays would be necessary to characterize the modified protein. Importantly, the consequence for mechanical properties of OPN networks is also influenced by the abundance and concentration of Ca2+ ions. As shown in the AFM experiments, situations that lead to high Ca2+ ion concentration result in shielding of the phosphorylated sites, charge reversal and repulsion. While the lack of experiments on OPN with varying phosphorylation-levels may be seen as a limitation, we nonetheless provide data points for the most extreme cases, and, different to the physiological system, control the concentration of Ca2+ ions. Taken together, our experiments do reproduce the effects seen in the animal models. We are planning to do further studies with variably and reproducibly phosphorylated OPN, as has recently been achieved in co-author Sorensen's lab.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Reviewer #1:

This manuscript is well written and is on a topic that is compelling and important to the field. The authors have given sufficient attention to the concerns expressed by the reviewers, including myself. I feel satisfied that the responses were diligent and thorough and the manuscript is improved as a result.

Upon revisiting Figure 8A, I agree that a quadratic fit is not necessarily the best choice and that an exponential function should be considered. Concerningly, if the very most highly phosphorylated point is excluded, the data look very much like a plateau. This would challenge the “sweet spot” idea as applied to just WT bones. As reviewer 3 notes it 'seems like the study showed that phosphorylation promotes fracture toughness (to a point) if OPN is present. I agree with this interpretation based on the presented data, which is consistent with more of a asymptote than a sweet spot in Figure 8A.

Thank you for raising this point. Figure 8A has been revised with an exponential function which is also significant (p<0.001). Consistent with suggestion from reviewer #3, our data shows that fracture toughness increases with phosphorylation only when OPN is present. Figure 8B has been added to reinforce this point.

Reviewer 3 asks “why doesn't the increase in fiber network modulus and strength continue to increase in phosphorylation”? I read this differently, with the interpretation that modulus and strength could indeed increase with phosphorylation but to a detrimental degree that decreases strain. Clarification is required as the reader will likely be confused on this point as well.

Based on the revised Figure 8A, these previous statements have been removed and we provide the following information to explain the increase of fracture toughness with phosphorylation for wildtype samples (Discussion section):

“We observed a non-linear dose response relationship between the level of global matrix phosphorylation and bone fracture toughness in wildtype mice (Figure 8A). Phosphorylation explained ~36% of the variance in fracture toughness and this relationship was not observed in the absence of OPN (Figure 8B). Taken together, this data supports the previously mentioned mechanism involving increased interaction energy and sacrificial bone formation between OPN and HA as well as between OPN polymers. The AFM-FS studies show that adhesion is not only dependent on the charge of OPN and HA under a certain environment but also the availability of free Ca2+ ions”

It would be helpful to show all data (not just four means) for Figure 8B and to indicate which data correspond with each group. This would alleviate confusion surrounding this figure.

As described above in response #1 the previous Figure 8A and the related sweet spot hypothesis have been removed. Thank you for the suggestion, all the data points are now included in Figures 8A and 8B.

The authors should clarify in the Abstract that the phosphorylation and dephosphorylation experiments were conducted ex vivo.

This has been clarified. We now state that “Fracture toughness, a measure of bone’s mechanical competence, increased with ex-vivo phosphorylation of wildtype mouse bones and declined with ex-vivo dephosphorylation. In osteopontin-deficient mice, global matrix phosphorylation level was not associated with toughness.”

Reviewer 3 comments that the abstract only mentions one mechanism of the three in the discussion. My bigger issue is that mechanisms 1 and 3 are co-dependent as presented (they are described as "three different, co-existing mechanisms". The authors argue that mineral-matrix interactions are facilitated by OPN phosphorylation, but this would be the real mechanism and the “sweet spot” or “benefit to a point” would be the relationship of this mechanism to dose.

Thank you for the suggestion. We have now clarified that mineral-matrix interactions facilitated by OPN phosphorylation, is the mechanism for the non-linear exponential increase in toughness with phosphorylation to a point. The second mechanism explains how phosphorylation of bone matrix in the absence of Osteopontin affects fracture toughness. The explanation has been added in Discussion section and in the Abstract.

The authors should add a little bit of clarifying statistical information (1 vs 2 tailed testing, critical α). These are stated for the AFM experiments but not for the fracture toughness and phosphorylation experiments.

Statistical clarification for the phosphorylation and fracture toughness experiments are now included in subsection “Data analysis for global phosphorylation”. We now state “All analyses were conducted using IBM SPSS 21 and two-tailed significance threshold set at 0.05 for both paired and independent samples t-test.”

Reviewer #2:

The revised manuscript is considerably improved. The detailed Response to reviewers is appreciated, as well as the updates to figures and text. The revised Figure 8 is particularly appreciated, notwithstanding the comments made by other reviewers.

I agree that their findings are exiting and that the explanations need to be improved so that the readers do not need to make inferences that may be off track.

Thank you for the feedback. Figure 8 has been revised and the Discussion has been reorganized to limit inferences from the reader.

Reviewer #3:

The study is intriguing because it shows how phosphorylation promotes fracture toughness of bone if osteopontin is present. However, this reviewer still struggles with the authors' interpretation of the mechanisms by which phosphorylation does this (they propose 3). A number of questions come up after reading the Abstract and Discussion. I like the concept that modulus and strength increases with phosphorylation but at a cost of decreasing strain after some threshold. The authors just need to do a better job describing each mechanism and how they are inter-related so that readers don't have to interpret their interpretation.

How did the authors decide on fitting a quadratic equation to fracture toughness vs. phosphorylation levels of WT bone? To this reviewer, it looks like fracture toughness plateaued after 1.5 in Figure 8A. What about fitting an exponential function, exp(1/phos)?

Thank you for the suggestion. Kindly note that as mentioned above, an exponential function rather than a quadratic function is now used to fit the relationship between phosphorylation levels and fracture toughness of WT bone subjected to ex vivo phosphorylation/dephosphorylation (Figure 8A).

Discussion: "However, further increase in the degree of most favorable interaction (i.e. larger than the "sweet spot" level) can cause a rapid decrease in ultimate strain, and thus, energy dissipation." What is the physical mechanism that supports this assertion? In other words, why doesn't the increase fiber network modulus and strength continue to increase with an increase in phosphorylation?

It is possible that fiber realignment and altering the net effective charge on the protein network can have detrimental effects on bulk tissue strain and thus fracture toughness. However, in light of the updated Figures 8A and 8B, these previous statements have been removed.

Another thing that is challenging to the interpretation of the results is that fact OPN bridges mineral crystals, presumably more bridges with more phosphorylation. Correct? This is the first mechanism described. In the OPN-/- bone, this mechanism is gone. Is the relationship between fracture toughness of OPN-/- bone and phosphorylation levels (from the 2 enzyme treatments) non-linear? Sorry, Figure 8B is not too convincing with only 4 data points.

The interpretation of the result that OPN bridges mineral crystals and more bridges are formed with more phosphorylation is correct. This mechanism explains the continuous relationship between increased phosphorylation levels and increased fracture toughness with the presence of osteopontin as shown in Figure 8A. However, in the absence of OPN there seems to be an alternative mechanism where the phosphorylation and dephosphorylation of bone matrix increases and decreases toughness in a binary/step-wise fashion as seen from the results in Figure 4 and Figure 5. In particular, we postulate that the increase in global phosphorylation of other bone matrix proteins in the absence of OPN (e.g., other SIBLING matrix proteins) may potentially result in increased protein alignment and larger interfilament distance between mineralized collagen fibrils to a detrimental degree that decreases matrix interaction, energy dissipation, and consequently fracture resistance. Taken together, these results show that phosphorylation affects bone toughness through two mechanisms. First involving increased bridge formation between OPN and mineral crystals (adhesion) and between OPN molecules (cohesion) with increasing phosphorylation, and the second involving other SIBLING proteins where increased phosphorylation results in increased protein alignment and larger interfilament distance leading to decreased matrix interaction and fracture resistance. Recent work has also shown that the characteristic fibrillar structure of collagen is lost and the collagen network is disorganized when Osteopontin is absent (Depalle et al., 2020) which is likely to affect mechanical function of OPN KO mice. Please refer to the Discussion section.

https://doi.org/10.7554/eLife.58184.sa2

Article and author information

Author details

  1. Stacyann Bailey

    Department of Biomedical Engineering, Center for Biotechnology and Interdisciplinary Studies, Rensselaer Polytechnic Institute, Troy, United States
    Contribution
    Data curation, Formal analysis, Validation, Investigation, Methodology, Writing - original draft, Writing - review and editing
    For correspondence
    stacyann.bailey@mountsinai.org
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-9013-2469
  2. Grazyna E Sroga

    Department of Biomedical Engineering, Center for Biotechnology and Interdisciplinary Studies, Rensselaer Polytechnic Institute, Troy, United States
    Contribution
    Conceptualization, Supervision, Validation, Methodology, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
  3. Betty Hoac

    Faculty of Dentistry, McGill University, Montreal, Canada
    Contribution
    Data curation, Software, Formal analysis, Validation, Investigation, Methodology, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
  4. Orestis L Katsamenis

    Faculty of Engineering and Physical Sciences, University of Southampton, Southampton, United Kingdom
    Contribution
    Data curation, Software, Formal analysis, Validation, Investigation, Methodology, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
  5. Zehai Wang

    Department of Mechanical, Aerospace, and Nuclear Engineering, Rensselaer Polytechnic Institute, Troy, United States
    Contribution
    Software, Formal analysis, Validation, Investigation, Methodology, Writing - original draft
    Competing interests
    No competing interests declared
  6. Nikolaos Bouropoulos

    Department of Material Science, University of Patras, Patras, Greece
    Contribution
    Resources, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  7. Marc D McKee

    1. Faculty of Dentistry, McGill University, Montreal, Canada
    2. Department of Anatomy and Cell Biology, Faculty of Medicine, McGill University, Montreal, Canada
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Investigation, Methodology, Writing - review and editing
    Competing interests
    MDM is a member of the FRQS Network for Oral and Bone Health Research, and he holds the Canada Research Chair in Biomineralization as part of the Canada Research Chairs program which contributed to the funding of this work.
  8. Esben S Sørensen

    Department of Molecular Biology and Genetics, Aarhus University, Aarhus, Denmark
    Contribution
    Resources, Investigation, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-7050-3354
  9. Philipp J Thurner

    Institute of Lightweight Design and Structural Biomechanics, Vienna University of Technology, Vienna, Austria
    Contribution
    Conceptualization, Resources, Software, Supervision, Funding acquisition, Methodology, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
  10. Deepak Vashishth

    Department of Biomedical Engineering, Center for Biotechnology and Interdisciplinary Studies, Rensselaer Polytechnic Institute, Troy, United States
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Methodology, Writing - review and editing
    For correspondence
    vashid@rpi.edu
    Competing interests
    No competing interests declared

Funding

National Institutes of Health (AR 49635)

  • Stacyann Bailey
  • Grazyna E Sroga
  • Zehai Wang
  • Deepak Vashishth

Canadian Institutes of Health Research

  • Betty Hoac
  • Marc D McKee

University of Southampton (Doctoral Prize Fellowship)

  • Orestis L Katsamenis

Canada Research Chairs

  • Marc D McKee

EPSRC (Doctoral Prize Fellowship)

  • Orestis L Katsamenis

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

This work was supported by the following funding agencies: The National Institutes of Health AR 49635, Doctoral Prize Fellowship from Engineering and Physical Sciences Research Council (EPSRC), UK, the University of Southampton, UK, and the Canadian Institutes of Health Research. MDM is a member of the FRQS Network for Oral and Bone Health Research, and he holds the Canada Research Chair in Biomineralization as part of the Canada Research Chairs program which contributed to the funding of this work. We thank Beate Lanske for the generous dontation of the FGF23 knockout mice and to the Center for Biotechnology and Interdisciplinary Studies Imaging Core at Rensselaer Polytechnic Institute for providing access to micro-computed tomography.

Ethics

Animal experimentation: This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All of the animals were handled according to approved institutional animal care and use committee (IACUC) protocols (VAS-001-14) of Rensselaer Polytechnic Institute.

Senior Editor

  1. Clifford J Rosen, Maine Medical Center Research Institute, United States

Reviewing Editor

  1. Cheryl Ackert-Bicknell, University of Colorado, United States

Reviewer

  1. Virginia Ferguson

Publication history

  1. Received: April 23, 2020
  2. Accepted: December 7, 2020
  3. Accepted Manuscript published: December 9, 2020 (version 1)
  4. Accepted Manuscript updated: December 16, 2020 (version 2)
  5. Version of Record published: December 17, 2020 (version 3)

Copyright

© 2020, Bailey et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 679
    Page views
  • 75
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Medicine
    Julia Stokes et al.
    Research Article Updated

    Volatile anesthetics (VAs) are widely used in medicine, but the mechanisms underlying their effects remain ill-defined. Though routine anesthesia is safe in healthy individuals, instances of sensitivity are well documented, and there has been significant concern regarding the impact of VAs on neonatal brain development. Evidence indicates that VAs have multiple targets, with anesthetic and non-anesthetic effects mediated by neuroreceptors, ion channels, and the mitochondrial electron transport chain. Here, we characterize an unexpected metabolic effect of VAs in neonatal mice. Neonatal blood β-hydroxybutarate (β-HB) is rapidly depleted by VAs at concentrations well below those necessary for anesthesia. β-HB in adults, including animals in dietary ketosis, is unaffected. Depletion of β-HB is mediated by citrate accumulation, malonyl-CoA production by acetyl-CoA carboxylase, and inhibition of fatty acid oxidation. Adults show similar significant changes to citrate and malonyl-CoA, but are insensitive to malonyl-CoA, displaying reduced metabolic flexibility compared to younger animals.

    1. Medicine
    Fan Xue et al.
    Research Article Updated

    Imbalances in bone formation and resorption cause osteoporosis. Mounting evidence supports that brain-derived neurotrophic factor (BDNF) implicates in this process. 7,8-Dihydroxyflavone (7,8-DHF), a plant-derived small molecular TrkB agonist, mimics the functions of BDNF. We show that both BDNF and 7,8-DHF promoted the proliferation, osteogenic differentiation, and mineralization of MC3T3-E1 cells. These effects might be attributed to the activation of the Wnt/β-catenin signaling pathway as the expression of cyclin D1, phosphorylated-glycogen synthase kinase-3β (p-GSK3β), β-catenin, Runx2, Osterix, and osteoprotegerin (OPG) was all significantly up-regulated. Knockdown of β-catenin restrained the up-regulation of Runx2 and Osterix stimulated by 7,8-DHF. In particular, blocking TrkB by its specific inhibitor K252a suppressed 7,8-DHF-induced osteoblastic proliferation, differentiation, and expression of osteoblastogenic genes. Moreover, BDNF and 7,8-DHF repressed osteoclastic differentiation of RAW264.7 cells. The transcription factor c-fos and osteoclastic genes such as tartrate-resistant acid phosphatase (TRAP), matrix metalloprotein-9 (MMP-9), Adamts5 were inhibited by 7,8-DHF. More importantly, 7,8-DHF attenuated bone loss, improved trabecular microarchitecture, tibial biomechanical properties, and bone biochemical indexes in an ovariectomy (OVX) rat model. The current work highlights the dual regulatory effects that 7,8-DHF exerts on bone remodeling.