2016). Differential effects of altered patterns of movement and strain on joint cell behaviour and skeletal morphogenesis. Osteoarthritis and Cartilage, 24(11), 1940-1950

Objective: There is increasing evidence that joint shape is a potent predictor of osteoarthritis (OA) risk; yet the cellular events underpinning joint morphogenesis remain unclear. We sought to develop a genetically tractable animal model to study the events controlling joint morphogenesis. Design: Zebra ﬁ sh larvae were subjected to periods of ﬂ accid paralysis, rigid paralysis or hyperactivity. Immunohistochemistry and transgenic reporters were used to monitor changes to muscle and cartilage. Finite Element Models were generated to investigate the mechanical conditions of rigid paralysis. Principal component analysis was used to test variations in skeletal morphology and metrics for shape, orientation and size were applied to describe cell behaviour. Results: We show that ﬂ accid and rigid paralysis and hypermobility affect cartilage element and joint shape. We describe differences between ﬂ accid and rigid paralysis in regions showing high principal strain upon muscle contraction. We identify that altered shape and high strain occur in regions of cell differentiation and we show statistically signi ﬁ cant changes to cell maturity occur in these regions in paralysed and hypermobile zebra ﬁ sh. Conclusion: While ﬂ accid and rigid paralysis and hypermobility affect skeletal morphogenesis they do so in subtly different ways. We show that some cartilage regions are unaffected in conditions such as rigid paralysis where static force is applied, whereas joint morphogenesis is perturbed by both ﬂ accid and rigid paralysis; suggesting that joints require dynamic movement for accurate morphogenesis. A better understanding of how biomechanics impacts skeletal cell behaviour will improve our understanding of how foetal mechanics shape the developing joint


Introduction
We now understand that the mechanical environment experienced during early development is important for normal skeletal development. There are multiple conditions for which abnormal or reduced movement are causal; including developmental dysplasia of the hip (DDH), which affects 1.3 per 1000 births 1,2 , arthrogryposis which affects around 1:4000 births 3 and fetal akinesia deformation sequence (FADS) which affects 1:15,000 births 4,5 . Additionally, there is evidence that early changes to joint shape lead to osteoarthritis (OA) later in life 6 . This can arise if conditions such as DDH are uncorrected 7 , but also subtle changes to hip shape have been identified as conferring increased risk of OA 8,9 . Despite the clinical significance of joint shape, relatively little is known about the underlying cellular events underpinning joint morphogenesis 10,11 .
Many studies have investigated the effect of temporal paralysis on joint formation. The majority have been undertaken in developing chick and mouse limbs and have shown that paralysis caused flattening of articular surfaces and a failure of joint cavitation, resulting in fusion of opposing elements 12e15 . By contrast, few studies have focused on the role of biomechanics in craniofacial development, though craniofacial morphogenesis is also affected by paralysis, with different joints differentially affected in chicks 13,16,17 . It is less clear what the effect of more sustained hyperactivity of the system will be and whether this would be beneficial to the skeletal system; for example, in chick, the effects of hypermobility have been described to increase joint cavity size 18 .
Previously it was believed that joint morphology developed after cavitation, however, recent studies of chick knee and hip joints have revealed that morphogenesis precedes cavitation, with most anatomical features present prior to element separation 19,20 , lineage tracing in mouse also reveals morphogenesis prior to separation 21 .
We still know relatively little about the cellular events that underpin morphogenesis; though recent work has started to address this question. In chick knee development, patterns of mechanical strain co-localise with regions of increased cellular proliferation, giving clues that in vivo cellular behaviour is altered mechanically 22 . In zebrafish, movement is required for normal chondrocyte intercalation 23 and correct cell orientation at the joint 24 . Recently, there has been increased focus on identifying putative mechanosensitive genes that could couple mechanical forces to downstream morphological responses 14,25,26 . Zebrafish, with their many transgenic lines marking various cell types of the musculoskeletal system 27 raise the prospect of using imaging to help unravel the cellular dynamics that underpin skeletal morphogenesis. We wanted to compare the effects of rigid paralysis and hyperactivity with flaccid paralysis and observe their impact on jaw joint morphology.

Zebrafish husbandry
Zebrafish were housed as previously described 28 . Animal experiments were ethically approved by the local ethics committee and by the Home Office.

Pharmacological treatment
Fish were treated from 3 days post fertilisation (dpf) to 5 dpf, with drugs replaced twice daily diluted in Danieau buffer in petri dishes 28 . Flaccid paralysis anaesthetic MS222 (Tricaine methanesulfonate), (Sigma) was used at 0.1 mg/ml. Decamethonium bromide (DMB) induces rigid paralysis and has been used to induce paralysis in chicks in ovo leading to alterations to joint patterning 12e14 . DMB was used at 8 mg/ml diluted into Danieau buffer. 4-amino-pyridine (4-AP), a potassium channel antagonist can induce hyperactivity in chick foetuses 29 . 4AP was tested at concentrations from 0.05 mM to 1.2 mM and 0.5 mM was selected for further analysis.
Tracking of fish swim motility Swim motility was measured by tracking individual control or 4AP-treated fish from movies. Tracking was performed using a manual tracking ImageJ plugin 30 , which when calibrated for pixel size and time interval between frames allows quantification of distance travelled and velocity (Sup. Vid. 1). Measurements were made on 10 fish per 4-AP dose per time period of drug application.

Recording frequency of jaw movement
Zebrafish were anaesthetised with MS222 and mounted laterally onto coverslips in 1% agarose. The agarose surrounding the head was removed and Danieau buffer flushed over the coverslip until jaw movements resumed. The number of mouth movements per minute was recorded from four fish per timepoint with 3 measurements taken per fish and mean values used (Sup. Vid. 2). 2tailed students t-tests were used to compare control with 4APtreated larvae and siblings with vhl mutants.

Measurement of jaw displacement
High-speed movies were made of jaw movements in wild type and vhl mutants; frames corresponding to maximum jaw displacements were selected, imported into ImageJ 31 and measurements taken on the distance between the tip of the Meckel's cartilage (MC) in the lower jaw and upper jaw in mm (Sup. Fig. 1).
Two-tailed Student t test was used to compare average displacements from vhl mutants to controls.

Wholemount immunohistochemistry
Immunohistochemistry was performed as previously described 24 . Larvae were fixed in 4% PFA and stored in 100% MeOH, rehydrated into PBS with 0.1% Tween (PBSTw), permeabilised using 15 mg/ml proteinase K, washed and blocked in PBSþ 5% horse serum for at least 2 h. The larvae were incubated with anti-myosin A4.1025 mouse IgG [1:200 dilution; Developmental Studies Hybridoma Bank (DSHB)] or rabbit anti-col2 IgG (1:500 dilution, Abcam) in blocking solution overnight at 4 C and washed a minimum of four times in 1ÂPBSTw. Larvae were incubated with secondary antibodies (Dylite 488 goat anti-mouse IgG and Dylite 550 goat antirabbit IgG, Molecular Probes, 1:500 dilution) then washed extensively in 1ÂPBSTw prior to visualisation. Controls were exposed to only secondary antibodies.

Analysis of shape variation
Changes to MC shape caused by flaccid paralysis (MS222, Myod), rigid paralysis and hyperactivity were quantified using twodimensional (2D) geometric morphometrics. 2D MC outlines [Sup. Fig. 2(A)] of 5dpf controls (n ¼ 15), vhl mutants (n ¼ 4), Myod mutants (n ¼ 8) were prepared (Adobe Illustrator) and compared with outlines from larvae treated from 3-5dpf with MS222 (n ¼ 18) and DMB (n ¼ 4). The outlines were converted into 200 XY coordinates with a common origin located at the anterior tip using TpsDig 2.25 39 [Sup. Fig. 2(B)]. Coordinates were converted to sine and cosine components using Hangle Fourier transformations 40 and superimposed using Procrustes superimposition. To assess shape variation qualitatively, the data were subjected to a between groups principal components analysis (PCA) and one-way nonparametric MANOVA. The analyses were performed using Paleontological statistics software (Past 2.17) 41 .

Finite element models
The meshes for the FE models have been previously described 24 . Loads relating to the Protractor hyoideus (PH), Adductor Mandibulae (AM) and Intermandibularis were applied simultaneously to model rigid paralysis. The FE results are displayed as colourcoded strain contour plots to compare spatial distribution of tension and compression and their magnitude.
Cell orientation, area and shape Z-projections of 2e3 slices were created from images of MC chondrocytes labelled with Type 2 collagen. Individual cells were identified and selected by adjusting the threshold of the image, using Image J 31 . Measurements of cell area, length of the major and minor cell axis and angle of the longest axis of each cell were taken for cells at the MC. Major/minor axis was calculated to determine the cell circularity, with 1.00 indicating a perfect circle. The angle of the longest axis of the cell was adjusted in relation to the jaw midline. Graphs were produced and statistics performed using SPSS software (Version 23). A minimum of 9 cells located between the insertion points of the intermandibularis muscle were taken per fish with 3 fish used for each experimental condition.

Gap analysis
The interval between the MC and PQ cartilage elements of the jaw joint on their medial and lateral sides (typically the smallest and largest gaps between cartilage across the joint, respectively) were measured from confocal images using Leica LAS AF Lite software. Negative values were recorded for overlapped cartilage elements.
KruskaleWallis tests (used to make multi-comparisons between non-normal data) were performed to compare joint gap size.

Microscopy
Live zebrafish were mounted ventrally on coverslips in 1% low melting point agarose containing MS222. A confocal stack was produced by taking images at 1.6 mm intervals through the ventral jaw of 5dpf zebrafish carrying the Tg(Col2a1aBAC:mcherry) transgenic reporter or larvae immunostained for type II collagen using an SP8 or SP5 Leica confocal microscope using LAS capture software.

Induction of paralysis
Flaccid paralysis, in the case of the myod mutant and following treatment with MS222 leads to the lower jaw 'hanging open' in the majority of cases, such that the jaw joint is subluxed, [ Fig. 1(A)]. As expected, larvae subjected to rigid paralysis by DMB maintained a tightly closed jaw throughout (data not shown).
We also wanted to test the impact of increased motility on the joint using 4-amino-pyridine (4-AP). While treatment with 4-AP at doses of 0.5 mM reliably induced hyperactivity measured both by swim motility and frequency of jaw movement for up to 6 h [Sup. Fig. 3(A)e(C)], treatment for longer than 6 h at 0.5 mM led to decreased swim motility, twinned with abnormal appearance of the fish even at lower doses [Sup. Fig. 3(D) and (E)]. We, therefore, concluded that long-term 4-AP treatment led to physiological changes that were not due to a direct effect on the skeleton.
We, therefore, made use of a zebrafish vhl mutant line previously reported to display increased frequency of jaw movements.
We quantified the number of jaw movements in vhl mutants and compared these to controls. Vhl mutants move their jaws significantly more than control fish by 3 dpf, with the increased frequency of movement sustained at 4 and 5dpf [ Fig. 1(B)]. We also tested the range of motion in vhl mutants (maximum displacement 48 mm n ¼ 3) and found that although the average value was higher than controls, it lay within the normal range of movements measured in control fish and was not significantly different (Average maximum displacement 38.5 mm from a range of 30.5 mm-57 mm n ¼ 5, To test whether the different conditions led to an alteration in muscle fibre development or configuration, we stained 5dpf larvae for skeletal myosin [ Fig. 1(C)]. At 5dpf the muscles that attach the lower jaw are the PH, the AM, the intermandibularis anterior (IA), the hyoideus superior (HS) and inferior (HI), with the sternohyoideus (SH) located more posteriorly. Functionally, the PH depresses the mandible leading to mouth opening while the AM closes the mouth 42,43 . The basic muscular configuration was similar in all mutants/treatments with the exception of the myod mutant, which as previously described lacks all lower jaw musculature except the sternohyoideus [ Fig. 1 34 . We, therefore, consider the myod mutant to be an extreme form of flaccid paralysis. Visible differences can be seen in the muscle fibres themselves, the fibres in the MS222-treated larvae appear 'baggy', whereas, by contrast, fibres in the DMB-treated larvae appear tauter, and muscles in vhl mutants appear enlarged [ Fig. 1 To test the impact of changes to movement to the musculoskeletal structure of the lower jaw, we visualised the cartilage structures of the lower jaw by immunostaining for type II collagen [ Fig. 1(D)]. We saw, as previously reported that flaccid paralysis (MS222) or lack of muscle (myod mutants) showed altered jaw morphology such that the MC overlapped the palatoquadrate (PQ) on the medial side [Figs. 1(D) and 2(A) and 24 ]. Interestingly both the rigidly paralysed (DMB-treated) larvae and the hyperactive vhl mutants also showed alterations to joint morphology.
Higher magnification images of the joint region [ Fig. 2(A) (top panels: max projection) and 2A (bottom panels: single z-plane)] revealed that while control larvae have a complementary shape between the MC and the palatoquadrate with an evenly sized interzone between the two elements, all other conditions showed abnormalities. The muscle-less myod mutants, the flaccidly paralysed MS222 treated and the rigidly paralysed DMB-treated larvae all showed a failure to align the cartilage elements such that there was an overlap on the medial surface of the joint and an enlarged gap at the lateral edge [ Fig. 2 Outlines of the region allow this overlap to be visualised, these outlines demonstrate that the MC shape is more plastic than the PQ [ Fig. 2(B)]. To make a fuller assessment of the shape of the whole MC we converted each outline [ Fig. 3(A)] into 200 landmarks. These were transformed using the Hangle Fourier transformation and the Procrustes superimposition analysed using a non-parametric manova (npManova) ( Table I) and between groups Principle Component Analysis (PCA) (PAST 2.17) (Fig. 3). Results of the npManova (Table I) revealed the MC of drug treated groups and vhl and myod mutants were significantly different from controls. PCA was used to describe the variation in MC shape. Each principal component (PC) captures a trend in the variation of MC shape, for example, PC1 describes 70.6% of the variation between the groups [ Fig. 3(E)] and separates the groups by the increasing width of the MC at the midline (double arrow) from the narrow shape in controls to the wider myod and vhl mutants on the positive side. PC2 captures 13.9% of the variation and describes the differences in the angle of the arch between controls and the more tightly angled vhl mutants and MS222 treated larvae [ Fig. 3(E) double arrow]. The overlap in the medial side of the MC is captured by PC3 and accounts for 10.2% of the variation; while PC4 accounts for the final 5.14% of the variation. We generated 'morphospace' plots of the shapes represented by these components to explore the similarities and differences between the changing MC shape [ Fig. 3 Table I).
Morphospace analysis revealed differences to the morphology of the anterior tip of the MC. We took high magnification images of this region (Fig. 4), which showed differences between the behaviour of cells in this region; such that cells in flaccid paralysis and muscle-less mutants appeared small and rounded, whereas cells in DMB-treated fish appeared similar to controls with an enlarged, elongated shape; suggestive of increased maturity [ Fig. 4(A)]. To test this we measured cell 'circularity', taking the ratio between the cell's major and minor axis [ Fig. 4(B) and (C)]. This showed cells from myod mutants and anaesthetised fish were significantly more rounded than controls; whereas DMB-treated fish did not differ from controls [ Fig. 4(C)].   Another measure of maturity is increased cell volume. We tested the cell area [ Fig. 4(D)]; myod and flaccidly paralysed fish had significantly smaller cells than controls, whereas, cell size in vhl mutants and the DMB-treated fish did not significantly differ from controls [ Fig. 4(D)].
Finally, another sign of maturity is cell intercalation as they stack into the mature cartilage shape. We, therefore, measured the angle of cells at the MC tip and found while anaesthetised and myod mutants had significant differences to cell orientation to controls, rigidly paralysed and hypermobile larvae did not [ Fig. 4(E)].
To explore the alterations to mechanical environment of the cells in the MC we generated Finite Element (FE) models for jaw opening and closure and to replicate rigid paralysis [ Fig. 5 (A)e(C)]. These models show that muscular strain is concentrated around the tip of the MC during jaw opening and in rigid paralysis [ Fig. 5 (A)e(C)]. High strains are located at the joint regions during jaw closure and during rigid paralysis. The most parsimonious explanation for the alterations to cell size and morphology at the MC tip in flaccidly paralysed larvae and the difference between flaccid and rigid paralysis is that the 'small, round' cells at the MC tip are less differentiated than the larger and more polarised cells that have intercalated in the wild type and DMB-treated fish. To test this we used the transgenic reporter, sox10:eGFP. Sox10 marks all migratory neural crest precursors and therefore, marks cells of the interzone and precursor cells, which form part of the element prior to expression of type II collagen 36 . We crossed sox10:GFP with the col2a1:mCherry line, in this cross cells which are less differentiated are green or pale yellow whereas more mature cells will be orange or red. We imaged the ventral jaw of control and MS222 immobi-    the MC we generated FE models in which the IM muscle alone was applied to test whether biomechanical strain patterns from this muscle could explain the differences to cell behaviour between flaccidly and rigidly paralysed larvae [ Fig. 4 . Therefore, the differences in cell morphology and therefore element shape at the anterior of the MC between rigid and flaccid paralysis can likely be explained by the requirement for tension from contraction of the IM.

Discussion
We have previously shown that lack of jaw muscle activity in flaccidly immobilised larvae leads to altered joint shape 24 . In this paper we explored the differences and similarities in skeletal development under continuous or absent muscle load compared with control and hyperactive zebrafish.
Flaccid and rigid paralysis of zebrafish caused similar changes to the morphology and function of the jaw joint. However, in fish as in chicks, there were overall differences between flaccid and rigidly paralysed skeletal morphologies. In chicks flaccid paralysis led to a wider range of phenotypes including abnormal cartilage morphology, which we also see 29 . In fish, these morphological abnormalities were most pronounced at the tip of the MC. In the presence of dynamic muscle force, chondrocytes at the tip of the MC in control zebrafish have an ordered structure and the jaw joints develop correctly. In the absence of load (MS222 treated and myod mutants) the chondrocytes remain immature as indicated by the large number of sox10 positive cells and fail to orient correctly at the tip of the MC. The jaw joints also have a characteristic overlapping morphology. Under rigid paralysis (DMB-treated) where muscle are continuously contracting the orientation of the chondrocytes at the tip of the MC were not significantly different from control and appear mature. However, the general shape of the MC and the joints were altered. This suggests that dynamic movement, rather than muscle forces are required for normal joint morphogenesis, whilst successful chondrocyte maturation at the tip of the MC requires only the presence of muscle force. Our examination of the mechanics through FE modelling strongly suggests that tension exerted by contraction of the intermandibularis muscle provides a mechanical stimulus for the chondrocytes in this region to mature. This explains both the changes to cell morphology and the increased expression of sox10 relative to col2a1, a marker of mature chondrocytes, in immobilised zebrafish.
We also show the effects of hyperactivity on skeletal development. Interestingly, while hyperactivity led to altered skeletal behaviour; generally the effects were less dramatic than those elicited by paralysis. This suggests that movement and force are required to shape skeletal development and that normal levels of activity are optimal 18,29 . We show in vhl mutants that although the frequency of movement is approximately twice that of controls their range of jaw movement was normal. Perhaps if the range of motion had differed this might have a more severe impact on joint morphology, in monkeys surgical hyperextension of the jaw led to significant alterations to cartilage morphology 44,45 . One way in which our hypermobility results differ from those in chicks is that we show the size of the interzone is diminished in hypermobile vhl mutants, whereas in chicks hypermobility increased joint cavity size 18,29 . These differences in joint structure may be explained by the timing of hypermobility as our experiments are all undertaken at precavitation stages, or alternatively the effect of hypermobility on joint structure may differ between species.
Taken together the data suggest that dynamic mechanical forces control cartilage element morphology and joint shape by controlling a number of cellular behaviours, including proliferation, orientation, migration and differentiation and that these differ dependent on their local mechanical forces. This fits with in vitro literature that shows different mechanical forces can differentially regulate markers of chondrocyte maturation, with tensile strain pushing cells towards maturation, while hydrostatic pressure slowed chondrocyte maturation 46 , subsequent work in stem cells has demonstrated that hydrostatic pressure stabilises the chondrocyte phenotype preventing entry to hypertrophy 47 . In this context it suggests that tension at the Meckel's tip is required for chondrocyte maturation in this region, while different forces resulting from dynamic movement prevent premature maturation of interzone cells, maintaining them in a state where they retain the ability to migrate or reorient, allowing optimal joint morphology to be sculpted by movement so that joint surfaces are complementary. We have previously shown that joint cells in flaccidly paralysed larvae are incorrectly oriented 24 , the prediction from these data would be that rigidly paralysed larvae would also lead to abnormal cell orientation at the joint due to the lack of dynamic movement during paralysis.
Recently there have been significant efforts to identify mechanosensitive genes from the skeletal system; including the correlation of gene expression patterns to mechanical stimuli in chick limbs 14,25 , and transcriptomics to identify all genes that show significant changes to expression in the humerus of control and muscle-less mice 26,48 . Human GWAS have also identified a number of genetic associations with joint geometry 49,50 . However, it remains difficult to test the functional effects of gene manipulation in a model system. Zebrafish have the advantage of being genetically tractable, therefore, a better understanding of the dynamic process of joint morphogenesis should in future allow the functional testing of putative mechanosensitive genes to test their ability to elicit certain cell behaviours.