Osteocyte physiology and response to fluid shear stress are impaired following exposure to cobalt and chromium: Implications for bone health following joint replacement

ABSTRACT

The effects of metal ion exposure on osteocytes, the most abundant cell type in bone and responsible for coordinating bone remodeling, remain unclear. However, several studies have previously shown that exposure to cobalt (Co²⁺) and chromium (Cr³⁺), at concentrations equivalent to those found clinically, affect osteoblast and osteoclast survival and function. In this study, we tested the hypothesis that metal ions would similarly impair the normal physiology of osteocytes. The survival, dendritic morphology, and response to fluid shear stress of the mature osteocyte‐like cell‐line MLO‐Y4 following exposure to clinically relevant concentrations and combinations of Co and Cr ions were measured in 2D‐culture. Exposure of MLO‐Y4 cells to metal ions reduced cell number, increased dendrites per cell and increased dendrite length. We found that combinations of metal ions had a greater effect than the individual ions alone, and that Co²⁺ had a predominate effect on changes to cell numbers and dendrites. Combined metal ion exposure blunted the responses of the MLO‐Y4 cells to fluid shear stress, including reducing the intracellular calcium responses and modulation of genes for the osteocyte markers Cx43 and Gp38, and the signaling molecules RANKL and Dkk‐1. Finally, we demonstrated that in the late osteoblasts/early osteocytes cell line MLO‐A5 that Co²⁺ exposure had no effect on mineralization, but Cr³⁺ treatment inhibited mineralization in a dose‐dependent manner, without affecting cell viability. Taken together, these data indicate that metal exposure can directly affect osteocyte physiology, with potential implications for bone health including osseointegration of cementless components, and periprosthetic bone remodeling. © 2016 The Authors. Journal of Orthopaedic Research Published by Wiley Periodicals, Inc. on behalf of Orthopaedic Research Society. J Orthop Res 35:1716–1723, 2017.

Osteocytes account for approximately 95% of the cell population of bone, and arise as terminally differentiated osteoblasts that have become embedded in the mineralizing osteoid matrix. Osteocytes connect to other osteocytes and cells on the bone surface to form an extensive dendritic network that senses mechanical strain and renders the osteocyte as the primary mechanosensory cell within bone.1

The osteocyte network plays a master role in maintaining bone homeostasis through signaling pathways that orchestrate osteoblast and osteoclast activities necessary for normal bone health.2 Osteocyte cell death and alterations in dendritic morphology and connectivity play a role in osteonecrosis of the femoral head, especially following chronic steroid treatment.3 Decreased osteocyte dendritic connectivity is also found in osteoporosis and osteoarthritis, while osteomalacic bone exhibits an increased, but chaotic connectivity,4, 5 highlighting the importance of dendricity in pathophysiological conditions.

Successful osseointegration and survival of prosthetic joints also requires normal bone health. Data from several large national joint replacement registries have shown that hip replacement prostheses that use a large diameter metal‐on‐metal (MOM) bearing made of a cobalt–chromium alloy have a survivorship that is profoundly poorer than prostheses that use non‐metal‐on‐metal bearings.6, 7, 8, 9 This poor survivorship is due to complications including failed prosthesis osseointegration, osteolysis, femoral neck fracture, and hypersensitivity responses, and has resulted in a requirement for the regular clinical monitoring of indwelling prostheses,10 the recall of several prosthesis brands,11 and litigation against their manufacturers. It has also recently become more widely recognized that wear‐accelerated corrosion at taper junctions of all modular prostheses generates metallic wear debris and impairs prosthesis survival.12, 13, 14

Normal physiological concentrations of cobalt (Co) and chromium (Cr) in unexposed individuals are <0.3 μg/L,15, 16 metal‐on‐metal bearings are known to cause an elevation in Co and Cr in both local tissues of the hip joint and systemically. In well‐functioning prostheses at 10‐year follow‐up, median Co and Cr concentrations in the circulation have been reported to be 0.75 μg/L (range: 0.3–50.1 μg/L) and 0.95 μg/L (range: 0.3–58.6 μg/L) respectively.17 Periprosthetic concentrations are several fold higher (median Co, 113 μg/L; median Cr, 54 μg/L),18 while patients with failing prosthesis have reported Co and Cr concentrations as high as 528 μg/L (range: 0–13,000 μg/L) and 1,844 μg/L (range: 0–38,600 μg/L), respectively.19 We recently demonstrated that exposure to Co²⁺ and Cr³⁺ adversely affects both osteoclast and osteoblast survival and function, including the mineralization of prosthesis surfaces in vitro, at concentrations equivalent to those found clinically.20, 21, 22

Previous studies have suggested that exposure to Co²⁺ in vitro has a cytotoxic and pro‐inflammatory effect on osteocytes.23, 24, 25, 26 Here, we develop understanding in this area by examining the effects of exposure to both Co²⁺ and Cr³⁺ at clinically relevant concentrations on osteocyte physiology—in particular their survival, dendritic morphology, response to fluid shear stress (FSS), and mineralization.

MATERIALS AND METHODS

Metal Ion Preparation and Treatments

Cobalt (II) hexahydrate (CoCl₂ · 6H₂O) and chromium (III) chloride hexahydrate (CrCl₃· 6H₂O) (Fluka, Gillingham, UK) served as salts for Co²⁺ and Cr³⁺, respectively. Working concentrations of each metal ion were prepared, as previously described.20 The stability of these metal ions in culture media has been confirmed previously using flame‐atomic absorption spectroscopy.20 Control treatment contained equivalent volume of sterile distilled water to maintain conditions, and referred to as 0 µg/L treatments.

Cell Culture

MLO‐A5 cells are thought to represent the post‐osteoblast, pre‐osteocyte cells responsible for triggering mineralization of osteoid.27 MLO‐A5 cells were maintained at 37°C in a humidified atmosphere of 95% air and 5% CO₂ with α‐MEM GlutaMAX™ (ThermoFisher Scientific, Paisley, UK) containing 100 U/ml penicillin, 100 μg/ml streptomycin, and 10% FBS (Life Technologies, Paisley, UK) (referred to as complete α‐MEM).

MLO‐Y4 cells were derived from the same transgenic mice as the MLO‐A5 cells, but exhibit a more mature osteocyte phenotype, with characteristic dendritic morphology complemented with expression of osteocytic genes such as osteocalcin and connexin‐43.28 The cells were grown in rat‐tail type 1 collagen (ThermoFisher Scientific) coated T75 flasks (0.15 mg/ml in 0.02 M acetic acid) and maintained at 37°C in a humidified atmosphere of 95% air and 5% CO₂, with α‐MEM+ containing 100 U/ml penicillin, 100 μg/ml streptomycin, 2.5% FBS, and 2.5% BCS (complete α‐MEM+).

MLO‐Y4 Survival and Dendricity

To assess survival and dendricity, MLO‐Y4 cells were cultured overnight in collagen coated 96‐well plates at 2,500 cells per well in complete α‐MEM+ at 37°C in a humidified atmosphere of 95% air and 5% CO₂. Subsequently, the cells were treated with vehicle (0.5% FCS supplemented α‐MEM +) ± metal ion treatments for 24 h after which they were fixed and stained with 0.1% crystal violet29 (Sigma–Aldrich, Gillingham, UK).

To quantify MLO‐Y4 cell survival, cell numbers were counted from five non‐overlapping random fields of view per treatment with four wells quantified for each treatment. Dendricity was assessed by counting the number of dendrites per cell and the average length of dendrites, for dendrites greater than 5 μm in length. Quantification was done using CellD imaging software (Olympus, Southend‐on‐Sea, UK) on 200× magnified images with Leica DMI4000B microscope.

Intracellular Calcium ([Ca²⁺]_i) Response to Fluid Flow

MLO‐Y4 cells were cultured in complete α‐MEM+ on collagen coated slides at a density of 15 × 10³ and 25 × 10³ cells per slide. Following a 24 h incubation, the cells were treated with vehicle ± 50 or 500 μg/L of Co²⁺ and Cr³⁺ (1:1) for 30 min or 24 h, respectively. Following the treatments, cells were loaded with a 5 μM of Fluo‐4AM (ThermoFisher Scientific), a fluorescent Ca²⁺ indicator for 45 min at 37°C to observe the cellular response to FSS. The cells were then subjected to 20 s of mechanical stimuli using a laminar fluid‐flow chamber to create a physiologically relevant FSS of 16 dynes/cm², previously demonstrated to induce increases in [Ca²⁺]_i in vitro.30, 31, 32 Changes in [Ca²⁺]_i were monitored in real‐time using Leica AF6000 time‐lapse fluorescent microscope. Cellular fluorescence of individual cells was measured using ImageJ (http://imagej.nih.gov/ij/) and the average change in fluorescence for each cell relative to baseline was calculated and plotted. The area under the curve (AUC), a measure of cellular response, and peak intensity were calculated using GraphPad Prism® (GraphPad Software, La Jolla, CA) and expressed relative to control from the same experimental time‐point. All experiments were repeated three times with 40 cells analyzed for each experiment.

RT‐PCR

Immediately following mechanical stimuli, RNA was extracted from MLO‐Y4 cells using ReliaPrep™ RNA Cell Miniprep System (Promega, Southampton, UK) and reverse transcribed to cDNA using the High Capacity RNA‐to‐cDNA Kit (ThermoFisher Scientific). Quantitative RT‐PCR was performed with 7900HT Fast Real‐time PCR system using TaqMan® Gene expression assays (ThermoFisher Scientific) for RANKL, Gp38, Dkk‐1, and Cx43 and gene expression was expressed relative to Gapdh and controls with no exposure to metal ions and FSS using the ΔΔCT method. The expression of Gapdh did not vary across samples with respect to FSS and metal ion treatments (data not shown).

Mineralization Assay

MLO‐A5 cells were seeded in 48‐well plates at a density of 2 × 10⁴ cells per well in 0.5 ml complete α‐MEM till they reached confluence (usually day 3). The media was then replaced with vehicle ± metal ion treatments supplemented with 10 nM dexamethasone and 50 µg/ml l‐ascorbic acid (Sigma–Aldrich) (referred to as osteogenic media) to promote osteoblast differentiation.20, 22, 33 Vehicle ± metal ion treatments in osteogenic media was replenished every 2–3 days until 2 days prior to the end of experiment, when 5 mM inorganic phosphate was added to the osteogenic media to promote mineralization. On day 7 the cells were fixed overnight in 100% ethanol and stained with 40 mM Alizarin Red S (pH 4.2, Sigma–Aldrich). The plates were washed extensively with 95% ethanol and air‐dried prior to scanning on a flatbed scanner. The percentage area of mineralization per well was quantified using ImageJ (http://imagej.nih.gov/ij/) and expressed as percentage response to vehicle.

Statistical Analysis

All experiments were conducted on three separate occasions with three minimum repeats per occasion. The average of the individual repeat experiments were then used for statistical analysis. All values are reported as the mean ± standard deviation (SD) or mean ± 95% confidence interval (CI) and were analyzed using GraphPad Prism® (GraphPad Software). Data were tested for normality using the D'Agostino and Pearson omnibus normality test and analyzed by one‐way analysis of variance (ANOVA) with post‐test for linear trend or t‐test to address specific questions. If a significant effect was found, specific interactions were examined by performing two‐way analysis of variance with Sidak's multiple comparisons test. All tests were performed two‐tailed using a critical p‐value of 0.05.

RESULTS

Metal Ions Reduce MLO‐Y4 Osteocyte Cell Number In Vitro

Co²⁺ ion treatment significantly decreased the number of MLO‐Y4 cells over the full concentration range, while Cr³⁺ treatment had no effect on cell number (Fig. 1A). Combined metal ion treatment showed a significant interaction (p = 0.0396), reducing cell viability over the full concentration range (Fig. 1A and B), with 500 μg/L Co²⁺:Cr³⁺ combined treatment having a greater effect over Co²⁺ treatment alone.

Figure 1.

Effect of metal ions on MLO‐Y4 cell number. (A) Graph depicts the changes in MLO‐Y4 cell numbers following 24 h treatment with clinically relevant concentrations of Co²⁺, Cr³⁺, or Co²⁺:Cr³⁺. Cell numbers were counted from five non‐overlapping random fields of view per well with four wells quantified for each treatment. All data are expressed as mean ± SD, relative to vehicle. Vertical dotted line separates the upper limit of previously reported ion concentrations equivalent to those in the patient serum and the hip aspirate. Significant linear trends across the concentration range analyzed using One‐way ANOVA are represented by arrows in the corresponding direction of effect and p‐values alongside the respective treatment condition. Asterisks are used to denote differences between Co²⁺ and combined Co²⁺:Cr³⁺ treatments (*p < 0.05). (B) MLO‐Y4 cells stained with crystal violet—typical fields of view for vehicle (0 μg/L), Co²⁺ (5,000 μg/L), Cr³⁺ (5,000 μg/L), and Co²⁺:Cr³⁺ (5,000 μg/L each), scale bar = 50 μm.

Metal Ions Increase the Dendricity of MLO‐Y4 Cells In Vitro

Co²⁺ ion treatment increased the number of dendrites per cell over the full concentration range (Fig. 2A), while Cr³⁺ treatment had no significant effect. Combined metal ion treatment showed a significant interaction (p = 0.0427), with concentrations of 5,000 μg/L Co²⁺:Cr⁺³ reducing the increase in dendrite number observed with Co²⁺ treatment alone. An increase in average length of MLO‐Y4 dendrites was also observed with metal ion treatments. Co²⁺ treatment increased the number of dendrites over the full concentration range (Fig. 2B), while Cr³⁺ treatment had no significant effect. Co²⁺:Cr³⁺ combined treatment similarly showed an increase in the number of dendrites over the full the concentration range (Fig. 2B) but no significant interaction was observed (p = 0.6861). The combined Co²⁺:Cr³⁺ treatment did not have any further effect over that of Co²⁺ alone.

Figure 2.

Metal ions increase the dendricity of MLO‐Y4 cells in vitro. Graphs depict the changes in (A) dendrite number and (B) dendrite length per cell following 24 h treatment with clinically relevant concentrations of Co²⁺, Cr³⁺, or Co²⁺:Cr³⁺. Dendrite (>5 μm) numbers were counted and lengths measured for cells in five non‐overlapping random fields of view per well with four wells quantified for each treatment. All data are expressed as mean ± SD, relative to vehicle. Vertical dotted line separates the upper limit of previously reported ion concentrations equivalent to those in the patient serum and the hip aspirate. Significant linear trends across the concentration range analyzed using One‐way ANOVA are represented by arrows in the corresponding direction of effect and p‐values alongside the respective treatment condition. Asterisks are used to denote differences between Co²⁺ and combined Co²⁺:Cr³⁺ treatments (***p < 0.0001).

Metal Ions Reduce the Intracellular Ca²⁺ Response in MLO‐Y4 Following Mechanical Stimuli

The response of MLO‐Y4 osteocyte cells to mechanical stimuli in the form of FSS was assessed by measuring intracellular Ca²⁺ changes in real time. Following 30 min and 24 h exposure to Co²⁺:Cr³⁺, a reduction in cellular response (AUC) to mechanical stimuli was observed for 50 and 500 μg/L compared to the vehicle (Fig. 3A and B). A reduction in the peak response was also observed for both 50 and 500 μg/L Co²⁺:Cr³⁺compared to the vehicle, for both time‐points. The effects were dose‐dependent with 500 μg/L Co²⁺:Cr³⁺ showing reduced cellular and peak responses compared to 50 μg/L following 30 min and 24 h exposures.

Figure 3.

Metal ions reduce the intracellular Ca²⁺ response in MLO‐Y4 following mechanical stimuli. The graphs illustrate changes in intracellular Ca²⁺ mediated fluorescence due to FSS (16 dynes/cm²) following exposure to 50 and 500 μg/L of Co²⁺ and Cr³⁺ for (A) 30 min and (B) 24 h. Data are from three separate experiments with 40 cells analyzed for each experiment and are represented as mean ± 95%CI. The inset graphs illustrate the changes in area under curve (AUC) and peak intensity at both time points (***p < 0.0001).

MLO‐Y4 Gene Expression in Response to FSS Is Altered With Metal Ion Exposure

The MLO‐Y4 cells robustly expressed the osteocyte markers Cx43 and Gp38, and the signaling molecules RANKL and Dkk‐1, with the rank order of gene expression Gp38 ≥ Cx43> RANKL>Dkk‐1 (data not shown). FSS resulted in increased expression of RANKL (p = 0.0272), Gp38 (p = 0.0017), and Cx43 (p = 0.01148), and a reduction in Dkk‐1 expression (p = 0.0245) compared to unstimulated cells (Fig. 4A). These changes in gene expression following FSS were also observed for the 24 h experimental group‐increased RANKL (p = 0.0415), Gp38 (p = 0.0128), and Cx43 (p = 0.0218), and reduced Dkk‐1 (p = 0.0245) (Fig. 4B). Cells exposed to Co²⁺:Cr³⁺ for 30 min and 24 h prior to FSS had blunted responses, with the osteocyte markers Gp38 and Cx43 showing reduced expression in significant linear trend over the concentration range (Fig. 4B, p = 0.0338 and p = 0.0107, respectively).

Figure 4.

MLO‐Y4 gene expression in response to fluid shear‐stress is altered with metal ion exposure. The effect of FSS on osteocyte gene expression was assessed in the absence or presence of 50 and 500 μg/L of Co²⁺ and Cr³⁺ exposure for (A) 30 min and (B) 24 h. Changes in MLO‐Y4 gene expression are depicted as 2^−ΔΔCt relative to Gapdh and untreated cells without application of FSS. All data are mean ± 95%CI from three separate experiments with three technical repeats per experiment. Comparisons between untreated cells and cells following FSS are represented by asterisks (*p < 0.05; **p < 0.01). Significant linear trends across the concentration range analyzed using one‐way ANOVA with p < 0.05.are denoted as “f.”

Metal Ions Reduce Mineralization by Late Osteoblast/Early Osteocyte MLO‐A5 Cells

Using MLO‐A5 cells, we found no effect of individual Co²⁺ ion treatment on mineralization. However, Cr³⁺ reduced the mineralization of the MLO‐A5 cells over the concentration range equivalent to those observed in serum of patients with accelerated MOM bearing wear (Fig. 5), without affecting cell viability (data not shown). Combined metal ion treatment similarly showed a significant reduction in mineralization over the same concentration range; however, no interaction was seen (p = 0.8257) or any further effect of Co²⁺:Cr⁺³ combined treatment over the effect of Cr³⁺ treatment alone.

Figure 5.

Metal ions reduce mineralization by late osteoblast/early osteocyte MLO‐A5 cells. Percentage area mineralized by MLO‐A5 cells following exposure to clinically relevant concentrations of Co²⁺ and Cr³⁺. All data are expressed as mean ± SD, relative to vehicle. Vertical dotted line separates the upper limit of previously reported ion concentrations equivalent to those in the patient serum and the hip aspirate. Significant linear trends across the concentration range analyzed using one‐way ANOVA are represented by arrows in the corresponding direction of effect and p‐values alongside the respective treatment condition.

DISCUSSION

The osteocyte is the most abundant type of bone cell and plays a critical role in mechanosensing and in the orchestration of bone remodeling. However, little is known about the effects of clinically relevant metal exposure on this bone cell population. Here, we show that exposure to increased Co²⁺ and Cr³⁺ concentrations is detrimental to osteocyte physiology. We find that osteocyte cell survival and dendritic morphology are sensitive to metal ions at concentrations seen clinically in patients with prostheses. We also show for the first time that the typical osteocytic response to FSS, including intracellular calcium signaling and gene expression of osteocyte markers and bone remodeling cytokines, is impaired in the presence of these metal ions.

Reduction in MLO‐Y4 osteocyte cell viability after 24 h exposure to Co²⁺ occurred in a dose‐dependent manner, consistent with and expanding on previous reports demonstrating a cytotoxic effect at high concentrations of Co²⁺ (0.05 mM; ∼3,000 μg/L) following 48 h exposure.25 The lack of any effect of Cr³⁺ alone is consistent with the observations of Kanaji and coworkers and the previously held dogma of the relative inability of Cr³⁺ to cross cell‐membranes.34 In contrast to this dogma, we have previously shown the intracellular localization of Cr³⁺ in osteoblasts,21 as well as an additive effect of Co²⁺ and Cr³⁺ treatment.22 Thus the increased effect of combined ions at 500 μg/L over the effect of Co²⁺ alone observed in this study suggests for the first time potential facilitation of Cr³⁺ entry into osteocytes, as well as common downstream signaling. The fact that we don't observe any increased effect with combined metal ion treatment for concentrations at the higher end equivalent to those observed in the patient hip aspirates suggests that Co²⁺ is having the dominate effect.

As osteoblasts transition to osteocytes and mature, they form dendritic processes and establish highly oriented intercellular communication between the vascular space and the bone surface.1 Reduced osteocyte dendricity has a negative impact on osteocyte function and on the mechanical properties of bone.35 Several diseased states have been associated with alterations in osteocyte dendricity. A decrease in interconnectivity has been observed in osteoporotic and osteoarthritic bone, while osteomalacic bone exhibits an increased interconnectivity which is chaotic.4, 5 We detected an increase in the number and length of dendrites in MLO‐Y4 osteocytes cells treated with increasing concentrations of metal ions. One possible explanation for this observation is that it represents an attempt by the residual cell population to maintain or re‐establish inter‐cellular communication.

Osteocyte cell death plays a key role in the recruitment of osteoclasts and in bone resorption,36 while dendrites are found increased in and orientated to regions of potential bone remodeling.35 The reduction in osteocyte survival and increase in dendricity that we have observed in vitro at clinically relevant concentrations of metal ions may alter periprosthetic bone remodeling and may compound the previously reported direct effects observed on osteoblasts and osteoclasts.20

A change in intracellular calcium concentration is the immediate cellular response to mechanical stimuli and influences various downstream signaling events.37 Consistent with this, fluid flow‐induced increases in calcium signaling in MLO‐Y4 cells were significantly reduced by combined metal ion treatment in a dose dependent manner, thus confirming a detrimental effect on osteocyte function. Furthermore, the mechano‐responsive genes that are known to be upregulated in response to mechanical loading are increased in our system with fluid flow.35, 38, 39, 40, 41 Exposure to metal ions suppressed the increased expression of osteocyte markers Gp38 and Cx43 after 24 h, but had no effect on the fluid flow‐induced effects on RANKL or DKK1 expression. While Gp38 is an early response gene responsible for dendrite elongation,35 Cx43 is involved cell–cell communication by forming gap‐junctions and “hemi‐channels.”42 It has been shown that prostaglandin E2 (PGE2) signaling from osteocytes occurs via Cx43 following mechanical strain42 and mediates its osteogenic effect. The lack of upregulation of Cx43 with FSS following metal exposure may impede this autocrine/paracrine regulation of periprosthetic adaptive bone remodeling in patients with elevated metal levels. We appreciate that the selection of candidate genes in this study does not provide a comprehensive overview of the effects osteocytes might have on periprosthetic bone remodeling, but it does suggest an aberrance in regulation following FSS.

Finally, using the late osteoblasts/early osteocytes cell line MLO‐A5 we found that Co²⁺ exposure had no effect on mineralization, but Cr³⁺ treatment inhibited mineralization in a dose dependent manner, while the MLO‐A5 cells remained viable and produced a matrix that was left unmineralized. It is known that mineralization by osteoid‐osteocyte cells occurs via the release of mineralized spheres that nucleate collagen fibrils and propagate calcification,43 and there is evidence that Cr³⁺ can form complexes with amino acids in collagen distorting its helical backbone.44 Therefore, Cr³⁺‐related distortion of the collagen structure may perturb matrix nucleation and eventual mineralization by MLO‐A5 cells, as seen in this study.

We acknowledge that our study has limitation due to use of the osteocyte‐like cell lines. Although recent studies have described methods for extraction of primary osteocytes from mouse long bones,45 the yields are generally low and would not have been sufficient for our experimental purposes. Therefore, we have used the MLO‐Y4 cell line which serve as a good osteocyte model as they respond to mechanical stimulation by releasing prostaglandin‐E₂,46 ATP,47 and nitric oxide,48 integral to osteocytes’ orchestration of adaptive bone remodeling. We also used MLO‐A5 cells with their higher expression of ALP and osteocalcin represent cells undergoing osteoblast–osteocyte transition which have the ability to mineralize the osteoid matrix they are embedded in.27 Used together, we believe that these models are a good representation of osteocyte behavior with which to test our hypothesis, although we appreciate that some of the key regulators of bone homeostasis such as SOST/sclerostin and FGF23 are not expressed by these cell lines.49, 50

In conclusion, our current data suggest that metal ions directly affect osteocyte cell number and dendricity. In addition, we have shown that the functional responses and gene expression following FSS is blunted in osteocytes exposed to metal ions with implications for adaptive bone remodeling in patients with prosthesis tribocorrosion.

AUTHORS’ CONTRIBUTIONS

KMS and PO: substantial contributions to research design, or the acquisition, analysis, or interpretation of data and drafting the paper or revising it critically. NM: substantial contributions to research design, or the acquisition, analysis, or interpretation of data. JMW and AG: substantial contributions to research design, or the acquisition, analysis, or interpretation of data; drafting the paper or revising it critically; and approval of the submitted version.