Hypomorphic conditional deletion of E11/Podoplanin reveals a role in osteocyte dendrite elongation

The transmembrane glycoprotein E11/Podoplanin (Pdpn) has been implicated in the initial stages of osteocyte differentiation. However, its precise function and regulatory mechanisms are still unknown. Due to the known embryonic lethality induced by global Pdpn deletion, we have herein explored the effect of bone‐specific Pdpn knockdown on osteocyte form and function in the post‐natal mouse. Extensive skeletal phenotyping of male and female 6‐week‐old Oc‐cre;Pdpn flox/flox (cKO) mice and their Pdpn flox/flox controls (fl/fl) has revealed that Pdpn deletion significantly compromises tibial cortical bone microarchitecture in both sexes, albeit to different extents (p < 0.05). Consistent with this, we observed an increase in stiffness in female cKO mice in comparison to fl/fl mice (p < 0.01). Moreover, analysis of the osteocyte phenotype by phalloidin staining revealed a significant decrease in the dendrite volume (p < 0.001) and length (p < 0.001) in cKO mice in which deletion of Pdpn also modifies the bone anabolic loading response (p < 0.05) in comparison to age‐matched fl/fl mice. Together, these data confirm a regulatory role for Pdpn in osteocyte dendrite formation and as such, in the control of osteocyte function. As the osteocyte dendritic network is known to play vital roles in regulating bone modeling/remodeling, this highlights an essential role for Pdpn in bone homeostasis.

Although it is well recognized that osteocytes are derived from osteoblasts, the mechanisms which govern this transition (osteocytogenesis) are yet to be fully elucidated. Several different genes have been implicated in influencing osteocytogenesis, one of which encodes for E11/Podoplanin (Pdpn) (Zhang et al., 2006). Pdpn is a mucin-like, transmembrane glycoprotein, which undergoes Oglycosylation leading to the production of different glycoforms. Pdpn is up-regulated by hypoxia in the lung , IL-3 and PROX-1 in the lymphatic system (Groger et al., 2004;Hong et al., 2002), and by TGF-β in fibrosarcoma cells (Suzuki et al., 2005).
Few studies have investigated the precise function of Pdpn in osteocytes. We have previously shown that Pdpn is expressed by early embedding osteocytes, thus identifying it as a factor which likely contributes to the vital, early stages of osteocyte differentiation (Staines, Prideaux, et al., 2016). It is known that Pdpn expression in osteocytes is up-regulated in response to mechanical strain in vivo (Zhang et al., 2006) and that increased Pdpn expression, through overexpression in ROS 17/2.6 cells and through stabilization by proteasome inhibitors in MLO-A5 cells, leads to the formation of long dendritic processes (Sprague, Wetterwald, Heinzman, & Atkinson, 1996;Staines, Prideaux, et al., 2016). The formation of these cytoplasmic processes is abrogated in cells pre-treated with siRNA targeted against Pdpn (Zhang et al., 2006). Although these data support an important role for Pdpn in dendritic process formation, a key feature of the differentiating osteocyte, the underpinning mechanisms remain to be fully defined.
While in vitro studies are informative, to fully disclose the biological function of Pdpn during in vivo osteocytogenesis and bone modeling/remodeling, it is essential to study an in vivo model of Pdpn deletion to enable a thorough examination of its functional role. Global deletion of Pdpn is perinatally lethal due to the essential role of Pdpn in lung and epithelial cell function (Zhang et al., 2006). We have therefore generated a Pdpn conditional knockout mouse to examine how deletion of Pdpn influences osteocytogenesis and the skeletal phenotype of these mice. We have used the well characterized osteocalcin (OC)-cre promotor mouse (Zhang et al., 2002) as the expression of osteocalcin by late osteoblasts ensures that we eliminate Pdpn expression from late osteoblasts as they transition to the osteocyte phenotype. Using an osteocyte-specific cre mouse would have invalidated our experimental approach as it would have precluded our ability to study the role of Pdpn in osteocyte formation; osteoblasts would already have undergone differentiation into osteocytes. It is also pertinent to add that an osteocyte-specific crepromotor mouse does not currently exist. Here, we show that a significant reduction in the expression of Pdpn in mice affects tibial microarchitecture, compromises osteocyte dendrite elongation, and thereby implicating Pdpn as a regulator of both osteocyte form and function.

| Oc-cre mediated bone-specific deletion of Pdpn
The genotypes from the breeding strategy were born at the expected Mendelian frequency, and all cKO mice exhibited survival indistinguishable from that of fl/fl control mice. To confirm that the Pdpn floxed allele was selectively deleted in bone in cKO mice, we performed PCR analysis using primers designed to specifically detect Pdpn alleles before (Tm1c) and after (Tm1d) cre-recombination. As 2.2 | Bone-selective ablation of Pdpn has no effect on the gross skeletal phenotype Analysis of total body weight (g) of male or female 6-week-old mice showed no significant differences between genotypes (male fl/fl: 20.4 ± 0.5; male cKO: 19.1 ± 0.4; female fl/fl: 16.3 ± 0.6; female cKO: 16.2 ± 0.5; n>4/genotype/sex). Similarly, no differences were observed in the tibia lengths (mm) in cKO and fl/fl mice (male fl/fl: 16.2 ± 0.1; male cKO: 16.1 ± 0.2; female fl/fl: 15.6 ± 0.2; female cKO: 15.5 ± 0.1; n > 4/genotype/sex). Whole mount skeletal staining of mice also revealed no obvious gross differences in Alcian blue or Alizarin Red staining between cKO and fl/fl mice (Figure 1g,h). Gait parameters of freely moving mice using the CatWalk gait analysis system were also unchanged (Supplementary Figure S1). Together these data suggest that the bone-selective deletion of Pdpn has no effect on the gross skeletal phenotype of mice.

| Pdpn deletion significantly alters tibial cortical bone microarchitecture
Assessment of cortical bone mass by μCT (CSA/mean cortical thickness) revealed that Pdpn deletion produces statistically significant alterations in tibial cortical mass and shape, to differing extents, STAINES ET AL.
FIGURE 1 (a) Schematic of the Pdpn floxed allele before and after deletion of the loxP cassette containing exon 3 via osteocalcin cre (Oc-cre) mediated recombination. (b) PCR analysis of genomic DNA from the long bones of fl/fl, and cKO mice with primers for the Tm1c allele (before cre recombination), and the Tm1d allele (∼440 bp, after cre recombination). (c) Immunohistochemical labeling of Pdpn in the lung, kidney, spleen, heart, liver, muscle, and growth plate, of 6-week-old mice. Images are representative of n = 4/sex/genotype. Scale bar = 20 μm. (d) Immunohistochemical labeling of Pdpn in the trabecular bone and cortical bone of 6-week-old mice. Arrows are pointing at embedded osteocytes within the trabecular and cortical bone and their dendritic processes projecting from the cell bodies. Images are representative of n = 4/sex/genotype. Scale bar = 20μm. (e) Quantification of osteocytes positive for Pdpn immunolabeling relative to negatively labeled osteocytes (n = 3/genotype), p < 0.001***. (f) Western blotting for Pdpn (∼37 kDa) in cortical bone protein lysates from 6-week-old mice. β-actin was used as a loading control. Whole mount Alcian Blue and Alizarin Red stained skeletal preparations of 6-weekold male (g) fl/fl, and (h) cKO mice (scale bar = 10 mm) including hindlimb and calvaria preparations (scale bar = 5 mm) in both male and female cKO compared to fl/fl control mice.
Specifically, CSA is unaffected in female cKO mice but is significantly lower in male cKO compared with male fl/fl mice at ∼40-65% of tibial length (p < 0.05; Figure 2a). In addition, mean cross-sectional thickness was significantly lower at several regions in male cKO compared to male fl/fl mice (p < 0.05); however, Pdpn deletion resulted in an increase in mean cross-sectional thickness at several regions along the tibial length in female cKO compared with fl/fl control mice (p < 0.05; Figure 2b). Together these data indicate that the conditional deletion of Pdpn produces a deficit in the cortical tibial microarchitecture in both male and female mice, with a greater effect observed in the male mice. Despite this, no significant differences were observed in the trabecular bone volume/tissue volume (BV/TV), trabecular number (Tb.N), trabecular thickness (Tb. th), trabecular separation (Tb.Sp), or trabecular pattern factor (Tb.Pf) in male or female cKO mice in comparison to age-matched fl/fl control mice (Table 1).

| Hypomorphic deletion of Pdpn results in gender-dependent effects on tibial structural parameters
To provide an estimate of tibial resistance to bending forces, we calculated second moment of area around minor (I min ) and major axes (I max ) (Figure 3a,b). This showed a reduction in I min along the tibia length in female cKO compared to female fl/fl which was most pronounced in midshaft and distal tibia (p < 0.05, Figure 3a). Statically significant reduction in I min was also apparent in the tibia of male cKO compared with their male fl/fl controls, less so at the midshaft but more proximally (p < 0.05, Figure 3a). Overall, the trend in cKO mice was for smooth lowering of I min proximodistally, which contrasts markedly from fluctuations in I min along the tibia length of age-matched fl/fl control mice. For I max , the trend was similar for female cKO and their fl/fl controls (p < 0.05, Figure 3b); however, the effect in male cKO seemed to be more proximal and midshaft (p < 0.05, Figure 3b). Tibial ellipticity FIGURE 2 Whole bone analyses of cortical bone between 10% and 90% of total tibial length, excluding proximal and distal metaphyseal bone, of female fl/fl (brown), female cKO (orange), male fl/fl (black), and male cKO (gray) tibia at 6 weeks of age showing (a) cross-sectional area (CSA; mm 2 ) and (b) mean cross-sectional thickness (mm). Graphs represent mean ± SEM, n = 4/group. p < 0.05 was considered to be significant and p ≤ 0.01-0.05 was noted as green, p ≤ 0.001-0.01 as yellow, and p ≤ 0.000-0.001 as red. Not significant is noted as blue STAINES ET AL.   Whole bone analyses of cortical bone between 10% and 90% of total tibial length, excluding proximal and distal metaphyseal bone, of female fl/fl (brown), female cKO (orange), male fl/fl (black), and male cKO (gray) tibia at 6 weeks of age showing (a) I min (mm 4 ), (b) I max (mm 4 ), (c) ellipticity, and (d) resistance to torsion (J; mm 4 ) Graphs represent mean ± SEM, n = 4/group. p < 0.05 was considered to be significant and p ≤ 0.01-0.05 was noted as green, p ≤ 0.001-0.01 as yellow, and p ≤ 0.000-0.001 as red. Not significant is noted as blue   Table 3) or lacunar volume (Lc.V/Ct.TV; Table 3). Interestingly, no significant effect on vascular porosity; canal number (N.Ca; Table 3), density (N.Ca/Ct.TV; Table 3) and volume (normalized by cortical tissue volume; Ca.V/Ct.TV; Table 3) was also evident in cKO bones.
Due to the reported function of the osteocyte in regulating bone remodeling and phosphate homeostasis, we next sought to examine whether our cKO mice exhibit differential expression of the osteocyte factor sclerostin, as well as serum phosphate. We observed no In the light of the relatively normal osteocyte organization in cKO mice as disclosed by nano-CT (Table 3)

| DISCUSSION
We, and others, have previously shown that Pdpn promotes osteocytogenesis and dendrite formation in vitro (Staines, Prideaux, et al., 2016;Zhang et al., 2006). Our studies, described herein, have for the first time successfully generated a bone-specific conditional Pdpn hypomorphic mouse and we have confirmed a role for Pdpn in the attainment of fully elongated osteocyte dendrites.
Previous attempts to decipher the in vivo role of Pdpn in osteocytogenesis have involved global deletion of Pdpn in mice. In the lung, Pdpn is known as T1alpha or RT140 and it is expressed on the apical surface of the lung epithelial cells (Dobbs, Williams, & Gonzalez, 1988;Rishi et al., 1995). In mice, global deletion of this gene results in death at birth due to respiratory failure. This is associated with the as such, the lungs of these mice are unable to inflate as normal . T1alpha deletion in mice also produces lymphatic defects with pronounced lymphedema resulting in swelling of the limbs at birth (Schacht et al., 2003). A previous comprehensive study has attempted to analyze the effect of Pdpn deletion on the skeleton by generating a global Pdpn knock-out mouse by targeting exon 1 (Zhang et al., 2006).
While these mice did not express Pdpn in bone, the animals died soon they were able to make some important observations. The general appearance of the embryonic Pdpn null and wild-type mice was similar.
More specifically, they found no significant differences in the length or diameter of the femur, or in the femur cortical thickness. They did report an increase in the body weight of the Pdpn null mice, however this was attributed to lymphedema resulting from limb swelling (Zhang et al., 2006).
By using the cre-LoxP system targeted to exon 3 of the Pdpn gene, we have generated bone-specific conditional knockdown mice which exhibited survival indistinguishable from that of fl/fl control mice. This has for the first time allowed us to study the role of Pdpn in bone development in the post-natal mouse. The normal survival rates of cKO mice contrasts with perinatal lethality of global Pdpn -/mice (Zhang et al., 2006). Here, we used the well characterized osteocalcindriven cre promotor to drive Pdpn deletion. Oc-cre expression has previously been reported to be exclusive to late osteoblasts, with onset of expression just before birth and continuing throughout the mature osteoblast lineage (Zhang et al., 2002). This has therefore allowed us to examine the structure and function of the postnatal skeleton in the presence of reduced Pdpn expression in bone, where we observed significant differences in tibial cortical bone microarchitecture and in the volume and length of osteocyte dendrites likely leading to downstream effects on the tibia's anabolic response to loading.
Osteocytes play a vital role in regulating bone remodeling and their vast dendritic network is critical to cell-cell communication, maintaining cell viability and allowing the transfer of nutrients and waste products (Dallas et al., 2013). Accumulating evidence has suggested that Pdpn may play a critical role in dendrite formation; in MLO-Y4 cells, the deletion of Pdpn with siRNAs abrogated dendrite formation (Zhang et al., 2006). Conversely, the ectopic overexpression of Pdpn in keratinocytes induces plasma membrane extensions (Scholl, Gamallo, Vilaro, & Quintanilla, 1999), and in endothelial cells, the reorganization of the actin cytoskeleton and the formation of long tube like structures (Schacht et al., 2003). Similarly, we have previously shown that stabilization of Pdpn protein, through inhibition of endogenous proteasome activity, promotes dendrite formation through the RhoA/ROCK/ERM pathway (Staines, Prideaux, et al., 2016). RhoA is a small GTPase and a master regulator of various cellular processes such as cytokinetics, cytoskeletal regulation, and cell migration (Takai, Sasaki, & Matozaki, 2001). Pdpn is able to co-localise with the ezrin, radixin and moesin (   loading. Moreover, as the bone mechanical properties, including stiffness, depend upon the geometry of the bone shaft, it is likely that the reduction of the osteocyte dendritic network affects the mechanical properties as a secondary effect through these changes in the microarchitecture (Saffar, Jamilpour, & Rajaai, 2009).
Here, we used whole bone μCT analysis to fully define these changes in microarchitecture and determine the estimated strength and rigidity of the bone through the study of the bone's cross sectional geometry. CSA is directly related to a bone's strength against compressive forces applied equally throughout the bone; factors such as bone shape and the effects of muscle contraction, however, result in long bones experiencing bending and torsional forces . We measured indices of rigidity, including maximum and minimum resistance against bending forces in the cross section, second moment of area around minor axis (I min ), and second moment of area around major axis (I max ) in male and female cKO tibia and their age-matched control mouse tibia. Our data, using this approach, revealed sex-dependent differences in cortical bone microarchitecture and in our estimation of tibial resistance to bending forces in cKO mice in comparison to control mice. These differences suggest that bone responses to dysfunctional osteocyte form and function may be sex-dependent. Similarly, the differences in the regional responses in both male and female bones suggest that Pdpn may play differential roles in bone development and function.
Interestingly, Bonewald and colleagues found osteocyte Pdpn expression was increased in an ulna loading model, but that this increase was not consistent along the length of the diaphysis (Zhang et al., 2006). This regional requirement for Pdpn function during and can bind to hydroxyapatite (Staines, Macrae, & Farquharson, 2012). DMP1 is highly expressed by osteocytes and is restricted to the dendritic processes. Deletion of DMP1 in mice causes remarkable defects in both tooth and bone (Ye et al., 2004(Ye et al., , 2005. DMP1 knockout mice also display an abnormal lacuna-canalicular system, with a reduction in the number of canaliculi and a twofold expanded osteocyte lacunae with rough, not smooth, lacunar walls (Lu et al., 2011). The authors of this elegant study also showed that this phenotype can be rescued by the re-expression of the 57-kDa C-terminal fragment of DMP1 (Lu et al., 2011). DMP1 has been demonstrated to bind to the hyaluronan receptor CD44, a membrane bound protein thought to interact with the ERM family of proteins that are involved in actin cytoskeleton rearrangement and as such, the formation of the osteocyte dendritic processes (Jain, Karadag, Fohr, Fisher, & Fedarko, 2002).
As we did not achieve complete cre-recombination and consequently not a complete knock-out of Pdpn expression (∼70% reduction), it is also possible that a low level of Pdpn expression is sufficient to drive osteocytogenesis. A more complete knock-out of Pdpn expression may be required to reduce dendrite formation to a level where bone function is pathologically compromised. Our decision to use the osteocalcin-cre promotor mouse to drive deletion of Pdpn selectively in bone was based on the onset of OCN expression. Despite this, there have been previous reports of incomplete recombination when using these mice (Xiao et al., 2010). As such it might be prudent to attempt another cre driver mouse. Such examples would include the col 2.3 kb (Col 2.3-Cre) and 3.6 kb (Col 3.6-Cre) fragments of the rat Col1a1 promoter or the osterix-cre mouse (Liu et al., 2004;Rodda & McMahon, 2006).

| Primary osteoblast isolation and culture
Primary calvarial osteoblasts were obtained from 3-day-old Pdpn cKO and fl/fl mice by sequential enzyme digestion of excised calvarial bones using a four-step process as has previously been described (Staines, Zhu, Farquharson, & Macrae, 2013) (1 mg/ml collagenase type II in Hanks' balanced salt solution [HBSS] for 10 min; 1 mg/ml collagenase type II in HBSS for 30 min; 4 mM EDTA for 10 min; 1 mg/ml collagenase type II in HBSS for 30 min). The first digest was discarded and the cells were resuspended in growth medium consisting of a-

MEM (Invitrogen, Paisley, UK) supplemented with 10% (v/v) FCS
(Invitrogen) and 1% gentamycin (Invitrogen). Osteoblasts were seeded at a density of 1 × 10 4 cells/cm 2 and cultured for up to 21 days with the addition of 2.5 mM β-glycerophosphate and 50 µg/ml ascorbic acid. At days 0, 7, 14, and 21, cells were either processed for RNA extraction or fixed in 4% paraformaldehyde and stained with 2% alizarin red (pH 4.2) for 5 min at room temperature. Alizarin red-stained cultures were extracted with 10% cetylpyridinium chloride for 10 min and optical density was measured at 570 nm.

| Gait analysis
Gait parameters of freely moving male mice were measured using the CatWalk gait analysis system (Noldus Information Technology, the Netherlands) as described previously (Hamers, Lankhorst, van Laar, Veldhuis, & Gispen, 2001;Masocha & Parvathy, 2009). Each mouse was placed individually in the CatWalk walkway and allowed to walk freely and traverse from one side to the other of the walkway glass plate. Mice were habituated every day for 2 weeks prior to the test run, in which the gait of all mice was recorded three times and analyzed using the CatWalk system. Analysis of the recording generated a wide range of parameters; those analyzed are detailed in Supplementary Table S1.

| In vivo loading
Twelve-week-old male mice were isoflurane-anesthetized and the right tibia loaded as described previously using a well-established model for comparing architectural load-induced changes in tibiae in control and mutant mice in which the contralateral left tibia is used as control (De Souza et al., 2005). Briefly, axial compressive loads were applied by a servo-hydraulic materials testing machine (Bose, Framingham, Massachusetts, UK) via custom-made cups which hold knee and ankle joints flexed and the tibia vertically. The loading pattern consisted of a trapezoidal wave, with peak 11N loads for the cKO and 12N for the WT mice for 0.05 s, rise and fall times 0.025 s each and baseline hold time of 9.9 s at 2 N (calibrated for peak strain level by finite element analysis) (Pereira, Javaheri, Pitsillides, & Shefelbine, 2015). Forty cycles were applied in each loading episode. changes. Soft tissues and legs were dehydrated and processed to paraffin wax using standard procedures. Sections (5 μm) were cut and used for histological and immunohistochemical analysis. To analyze osteoclast and osteoblast numbers, sections were reacted for tartrateresistant acid phosphatase activity (TRAP) and stained by H&E respectively, as described previously (Erlebacher & Derynck, 1996).
Osteoclasts were quantified per mm 2 of trabecular bone, and osteoblasts per mm of bone surface, in an identical region at an equivalent distance beneath the growth plate in all samples.

| Immunohistochemistry
For immunohistochemical analysis, sections were dewaxed in xylene and rehydrated.

| Whole bone cortical μCT analysis
Whole bone analysis was performed on datasets derived from whole CT scans using BoneJ (version 1.13.14) a plugin for ImageJ, as previously described . Following segmentation, alignment, and removal of fibula from the dataset, a minimum bone threshold was selected using a histogram-based method in ImageJ which utilizes all pixels in a stack to construct a histogram and was further confirmed using ImageJ "threshold function." The gray level threshold ranged between 22,000 and 22,100 and was applied to all datasets to separate higher density bone from soft tissues and air. This threshold was used in "Slice Geometry" function within BoneJ to calculate bone cross sectional area (CSA), second moment of area around the minor axis (I min ), second moment of area around the major axis (I max ), mean cortical thickness determined by local thickness in two dimensions (Ct.Th), ellipticity and resistance to torsion (J). The most proximal and distal (10%) portions of the tibial length were excluded from analysis, as these regions include trabecular bone.

| Nano-computed tomography analysis
The samples were placed in Orthodontic Wax (Kerr, CA) at 50 kV and 200 µA, 9,800-ms exposure time with a 0.25-mm aluminum filter (99.999% purity, Goodfellow, Huntington, UK), voxel size of 0.6 µm, 360°at a rotation step of 0.25°. Two-frame averaging was used to improve the signal-to-noise ratio. We analyzed 300 consecutive images from the tibia-fibula junction from each sample. Using the CtAn software, osteocyte lacunar indices were calculated by measuring the 3D parameters of each discreet object within the volume of interest after segmentation as described previously . Shape analysis of the lacunae was conducted utilizing "Analyze Particles" function in BoneJ.

| Mechanical testing
A Lloyd LRX5 materials testing machine (Lloyd Instruments, West Sussex, UK) fitted with a 100 N load cell was used to determine bone stiffness and point of failure of tibiae. The span was fixed at 10 mm, and the cross-head was lowered at 1 mm/min. Data were recorded after every 0.2-mm change in deflection. Each bone was tested to failure, with failure points being identified as the point of maximum load from the load-extension curve. The maximum stiffness was defined as the maximum gradient of the rising portion of this curve (Huesa et al., 2011

| Statistical analysis
Data are expressed as the mean ± SEM of at least three replicates per experiment. Results were analyzed blinded. For cortical bone, graphs were developed using the R programming language "R," version 3.1.3 (R Foundation for Statistical Computing, Vienna, Austria; http://www. r-project.org). Normality and homogeneity of variance of all the data were checked using the Shapiro-Wilk and the Bartlett's test in R 3.1.3, respectively. Two-sample Student's t-test was used to compare means between female cKO and fl/fl, and between male cKO and fl/fl.
Kruskal-Wallis test was employed if either the normality or the homogeneity of variance assumptions were violated (p ≥ 0.05).
p < 0.05 was considered to be significant and noted as *; p-values of <0.01 and <0.001 were noted as ** and ***, respectively.