The Paget's disease of bone risk gene PML is a negative regulator of osteoclast differentiation and bone resorption

ABSTRACT Paget's disease of bone (PDB) is characterized by focal increases in bone remodelling. Genome-wide association studies identified a susceptibility locus for PDB tagged by rs5742915, which is located within the PML gene. Here, we have assessed the candidacy of PML as the predisposing gene for PDB at this locus. We found that the PDB-risk allele of rs5742915 was associated with lower PML expression and that PML expression in blood cells from individuals with PDB was lower than in controls. The differentiation, survival and resorptive activity of osteoclasts prepared from Pml−/− mice was increased compared with wild type. Furthermore, the inhibitory effect of IFN-γ on osteoclast formation from Pml−/− was significantly blunted compared with wild type. Bone nodule formation was also increased in osteoblasts from Pml−/− mice when compared with wild type. Although microCT analysis of trabecular bone showed no differences between Pml−/− mice and wild type, bone histomorphometry showed that Pml−/− mice had high bone turnover with increased indices of bone resorption and increased mineral apposition rate. These data indicate that reduced expression of PML predisposes an individual to PDB and identify PML as a novel regulator of bone metabolism. This article has an associated First Person interview with the first author of the paper.


Introduction
Paget's disease of bone (PDB) is a skeletal disorder characterized by focal increases in disorganised bone remodelling with markedly increased osteoclast and osteoblast activity. Commonly affected sites include the pelvis, femur, lumbar spine, skull and tibia (Gennari et al., 2019;Ralston et al., 2019). Many patients are asymptomatic, but others suffer from various complications including bone pain, bone deformity, deafness, and secondary osteoarthritis (Tan and Ralston, 2014;van Staa et al., 2002).

Disease Models & Mechanisms • DMM • Accepted manuscript
Genetic factors are important in PDB and several predisposing genetic variants have now been identified by a combination of linkage studies in families and genome wide association studies (Albagha et al., 2010;Albagha et al., 2011;Laurin et al., 2002;Ralston and Albagha, 2014;Scotto di Carlo et al., 2020;Vallet et al., 2015). Followup functional studies are essential to identify the gene(s) responsible for association with PDB at these loci and define the mechanisms by which these genes regulate bone metabolism. For example, functional studies of the chromosome 10p13 susceptibility locus identified optineurin (OPTN) as the gene driving the association with PDB and elucidated the mechanism by which this gene regulates bone metabolism (Obaid et al., 2015;Wong et al., 2020). However, for many of these susceptibility loci, the genes responsible for driving the association with PDB are unknown. One of the predisposing loci identified for PDB by GWAS is located on chromosome 15q24. There are several genes at this locus (LOXL1, PML, STOML1, GOLGA6A,ISLR,ISLR2), but the strongest association is with a single nucleotide polymorphism (SNP) rs5742915 located within the coding region of the Promyelocytic Leukaemia gene (PML) which causes a phenylalanine to leucine amino acid substitution at codon 645 (p.Phe645Leu) (Albagha et al., 2011). The PML gene was so named as it was identified as a tumour suppressor gene that was disrupted in acute promyelocytic leukaemia where it is fused to retinoic acid receptor alpha (RARA) gene as a result of the chromosomal translocation t(15;17) (Nisole et al., 2013;Salomoni and Pandolfi, 2002).
Previous studies have shown that PML is involved in various biological processes including cell growth, senescence, apoptosis, protein degradation and antiviral response (Guan and Kao, 2015;Salomoni and Pandolfi, 2002). Until its discovery as a predisposing locus for PDB, PML had not been considered to play a role in bone metabolism but could be involved through its known effects on diverse bone-related signalling pathways such as NF-B, TGF-β, IFN-γ, p38 and Wnt (El Bougrini et al., 2011;Lin et al., 2004;Shin et al., 2004;Shtutman et al., 2002;Wu et al., 2002). In this study, we investigated the role of PML in bone cell function to gain an insight into the mechanisms by which PML affects bone metabolism and predispose to PDB.

PML is expressed in osteoclasts and osteoblasts
We found that PML protein was expressed in the mouse monocyte-macrophage cell line RAW 264.7 as well as in primary mouse bone marrow derived macrophages (BMDMs). Expression of PML was detected at all stages during osteoclast differentiation following stimulation with RANKL ( Figure 1A, B). We also found that PML was expressed in mouse calvarial osteoblasts and during their differentiation to the stage of bone nodule formation ( Figure 1C).
To determine whether PML is expressed in human osteoclasts we conducted immunostaining for PML protein in human osteoclasts from giant cell tumour of bone as well as bone sections from patients with PDB as well as bone sections from patients unaffected by PDB. This showed that PML protein was expressed in the nuclei of osteoclasts in all sections examined ( Figure 1D) as well as in osteoblasts which were visible in the PDB sample.
We also detected PML mRNA in peripheral blood mononuclear cells (PBMC) and found that levels of expression were significantly lower in PBMC from PDB patients (n=18) compared with unaffected controls (n=7) (P = 0.01; Figure 1E). Two out of the 18 patients with PDB were positive for P392L mutation in SQSTM1 but levels of PML mRNA expression in these subjects did not differ from the rest of the PDB cohort (data not shown). The number of PDB patients was too small to perform an expression quantitative trait locus (eQTL) analysis for PML in PBMC but four samples of the PDB group were T/C heterozygotes at the rs5742915 SNP which allowed us to investigate allele specific gene expression. This showed that the mean  SD expression from the C allele was 19.0  3.8 % lower than the expression from the T allele (P=0.0002); consistent with the hypothesis that allelic variants at rs5742915 are associated with reduced PML expression.
In order to confirm whether allelic variation at the rs5742915 SNP on 15q24 was an eQTL for PML, we scrutinised the GTEx portal (GTExPortal, 2020) and found that carriage of the C-allele at this SNP which is associated with a 1.34-fold increased risk of PDB (Albagha et al., 2011) is also associated with reduced PML mRNA expression levels in skin cells ( Figure 1F). This indicates that reduced expression of PML increases the risk of PDB.

Effect of PML Over-expression on osteoclast differentiation in RAW264.7 cells.
Given that PML expression was reduced in PDB patients, we studied the effect of altered PML expression on osteoclast differentiation in RAW 264.7 cells, a mouse monocyte-macrophage-like cell line that differentiates into osteoclast-like cells upon RANKL treatment. Over-expression of PML resulted in a significant reduction in the number and size of osteoclasts formed compared to empty vector (Figure 2A-D).

Effects of targeted inactivation of PML on bone metabolism
To investigate the effects of PML on skeletal phenotype in vivo we compared the characteristics of mice with targeted inactivation of PML (Pml -/-) (Wang et al., 1998) and wild type (WT) littermates. Although the susceptibility alleles on the chromosome 15q24 locus predispose to PDB similarly in both men and women (Albagha et al., 2011), we decided to focus our analysis on male Pml -/mice since PDB is about 40% more common in men than in women (van Staa et al., 2002).
The Pml -/and WT mice were phenotypically normal. There were no differences between genotypes in body weight, body habitus, dentition, gait or survival. In contrast to a previous study (Lunardi et al., 2011), we found no evidence to suggest that Pml -/mice had increased susceptibility to infection.

Osteoclast function in Pml -/mice
Osteoclasts generated in vitro from bone marrow derived macrophages from Pml -/mice were significantly greater in number and larger in size as compared with those from WT littermates ( Figure 3A-C). Survival of osteoclasts from Pml -/mice was also significantly prolonged following RANKL withdrawal as compared with WT ( Figure   3D). Furthermore, osteoclasts generated from Pml -/mice showed higher resorption activity compared to those from WT ( Figure 3E-G). Taken together, these data indicate that absence of Pml results in a significant increase in osteoclast formation, activity, and survival.

Disease Models & Mechanisms • DMM • Accepted manuscript
In order to gain insights into the molecular mechanisms of osteoclast activation in Pml -/mice, we compared the expression of key osteoclast-related genes Nfatc1, Dcstamp and Ctsk in Pml -/and WT during RANKL-induced osteoclast differentiation.
These genes were chosen as being representative of a range of genes that are activated during osteoclast differentiation (Asagiri and Takayanagi, 2007). This revealed higher levels of Nfatc1 expression in osteoclast precursors from Pml -/mice compared to WT ( Figure 3H). We also observed significant increase in Dcstamp expression during osteoclast differentiation in both WT and Pml -/but the expression of this gene was significantly higher in Pml -/compared to WT during later stages of osteoclast differentiation. Additionally, the expression of the osteoclast marker gene Cathepsin K (Ctsk) was significantly higher in Pml -/compared to WT in osteoclast precursors as well as during their differentiation into osteoclasts ( Figure 3H).

PML regulates the inhibitory effect of IFN-γ on osteoclast differentiation.
Interferon gamma (IFN-γ) is a critical regulator of osteoclast differentiation (Takayanagi et al., 2000) and previous studies have shown that PML positively regulates IFN-γ signalling (El Bougrini et al., 2011). In view of this, we investigated the effect of IFN-γ on osteoclasts differentiated from bone marrow derived macrophages in Pml -/mice. These studies showed that while IFN-γ inhibited osteoclast differentiation in WT (P = 7.1 x 10 -5 ) and Pml -/mice (P =0.043). The inhibitory effect of IFN-γ was significantly blunted in Pml -/mice compared with WT, particularly with regard to large osteoclasts (≥10 nuclei) ( Figure 4A -B). Treatment of cultures with IFN-γ resulted in a significant decrease in the number of osteoclasts from WT (56.8%) compared to Pml -/-(17.1%; P= 4.2 x 10 -5 ) ( Figure 4C). Similarly, the reduction in the number of large osteoclasts upon treatment with IFN-γ was significantly higher in cultures from WT (82.1%) compared to those from Pml -/-(12.2%; P=1.4 x 10 -5 , Figure 4D). These observations indicate that the inhibitory effect of IFN-γ on osteoclast generation is partly dependent on PML.

Osteoblast function in Pml -/mice
We investigated the role of PML in osteoblast function by conducting mineralising bone nodule assays in calvarial osteoblasts derived from Pml -/and WT mice after 18 days culture in osteogenic medium. This showed that bone nodule formation was significantly greater in Pml -/mice compared with WT ( Figure 5A-B). Additionally, the expression of the osteoblast marker gene Alkaline phosphatase (Alpl) was significantly higher in proliferating osteoblasts (day 1 of culture) from Pml -/mice compared with WT (P=0.015; Figure 5C). Also, expression of Col1a1 was higher in osteoblasts from Pml -/compared to WT but this was of borderline significance (P=0.05; Figure 5D)

Bone turnover in Pml -/mice
To investigate the effect of PML inactivation on bone turnover in vivo, we performed bone histomorphometry in Pml -/and WT mice. Static bone histomorphometry parameters showed increased bone resorption parameters in male Pml -/mice as shown in Table 1  increase in the mineral apposition rate (MAR) in Pml -/compared to WT mice (Table   1 and Supplementary Figure 1). There was also a trend for higher bone formation rate per bone surface (BFR/BS) in Pml -/mice compared to WT but this did not reach statistical significance (P = 0.09; Table 1).

Bone volume and structure in Pml -/mice
We analysed bone volume and structure of Pml -/and WT mice using microcomputed tomography (µCT). We found no significant differences in trabecular or cortical bone parameters in the hind limbs of male Pml -/compared to WT at 4 months of age (Table 2 and 3.). We also searched for evidence of Pagetic-like lytic lesions in male Pml -/and WT mice at 4 months but none were detected (data not shown). We went onto study aged Pml -/and WT at 14 months of age. We found no difference in Disease Models & Mechanisms • DMM • Accepted manuscript BV/TV, trabecular separation, or number. However, we observed a significant 16% reduction in trabecular tissue volume (TV) in Pml -/mice compared to WT (Table 2).
Likewise, TV was also significantly lower in cortical bone of 14-month-old Pml -/mice compared to WT, although cortical thickness was not affected (Table 3). However, both the periosteal and endosteal perimeters were decreased by 7% and 10% respectively in Pml -/mice compared to WT, which, together with the decreased tissue volume, indicates a reduction in bone size (Table 3). This resulted in a significant 20% reduction in moment of inertia along all axes in Pml -/compared to WT. However, the µCT scans of the hind limbs of the 14-month-old mice did not reveal any evidence of Pagetic-like bone lesions. Supplementary Table S2 provides a breakdown of number of mice analysed in each age group.

Discussion
The chromosome 15q24 locus was detected as a susceptibility locus for PDB by an extended genome wide association study (Albagha et al., 2011;Ralston and Albagha, 2014). Though there are multiple candidate genes at this locus, the strongest association was with rs5742915 which is located within the coding region of PML causing and amino acid change (p.Phe645Leu). Here, we investigated the role of PML in bone metabolism using Pml knock out mice and identified a new role for this gene in regulating bone cell function.
We found that PML was expressed in both osteoclasts and osteoblasts as well as RAW 264.7 cells. We also gained robust evidence to show that PML acts as a negative regulator of osteoclast differentiation and function. In RAW 264.7 cells, we found that overexpression of Pml supressed osteoclast differentiation. The negative regulatory role of PML on osteoclast was confirmed by studies of osteoclast function in mice with targeted inactivation of Pml. We found evidence of increased osteoclast differentiation, increased osteoclast size and multinuclearity, increased survival and There are other potential mechanisms by which PML could influence osteoclast and osteoblast function. One is through its effect as an inhibitor of the p38 MAPK pathway (Shin et al., 2004). This pathway plays an important role in bone metabolism by stimulating osteoclast formation, maturation, and bone resorption as well as by regulating osteoblast differentiation, extracellular matrix deposition and Disease Models & Mechanisms • DMM • Accepted manuscript bone mineralisation (Greenblatt et al., 2010;Thouverey and Caverzasio, 2015). It is therefore possible that a reduction in Pml expression could increase p38 MAPK signalling thereby contributing to the high bone turnover seen in the Pml -/mice and in humans with PDB. Another possibility would be through an autophagy mediated mechanism since PML has been shown to interact with both p62 and the autophagy effector protein LC3 (Li et al., 2020). Further work would be required however to determine whether cross-talk between PML and p62 plays a role in regulating bone cell function and whether reduced levels of PML may affect this process.
We also found that PML negatively regulates osteoblast function as reflected by the increase in bone nodule formation on osteoblast cultures from Pml -/mice compared with WT. This was accompanied by increases in key osteoblast markers such as alkaline phosphatase (Alpl) and Col1a1 in Pml -/osteoblasts compared to WT. These changes in early stages of osteoblast differentiation could lead to increased osteoblast formation, differentiation and activity thereby resulting in increased bone formation. There have been very few previous studies in the effects of PML on osteoblast function but in a study by Sun et al (Sun et al., 2013) it was reported that over expression of PML in human mesenchymal stem cells inhibited cell proliferation by causing apoptosis but also increased alkaline phosphatase activity. These findings differ somewhat from the present study where we found that Pml -/osteoblasts had an increased propensity to form bone nodules in vitro and that Pml -/mice had evidence of increased bone formation. We speculate that these differences may be accounted for by the differences in experimental design and the use of stem cells in the previous study and clavarial osteoblast cells in this study.
Skeletal phenotyping of young adult male mice (4-month-old) using µCT revealed no significant differences in trabecular or cortical bone volume or structure between Pml -/mice and WT littermates. Although the histomorphometric studies showed evidence of increased bone resorption and bone formation in Pml -/mice it seems that the tight coupling between these processes resulted in no overall change in bone mass or bone structure in the different genotype groups. Analysis of aged male mice (14-months-old) using µCT similarly revealed no significant differences in trabecular bone volume or structure between Pml -/mice and WT but there was a significant reduction in bone size in Pml -/mice. This was evident from lower trabecular and cortical tissue volume and the decreased periosteal and endosteal

Disease Models & Mechanisms • DMM • Accepted manuscript
perimeters in Pml -/mice, which is accompanied by reduction in moment of inertia (MMI) at all axes. MMI is a geometry-dependent parameter which predicts resistance to bending (or torsion in the case of the polar moment of inertia) of a structure.
Bones with a smaller cross-sectional area, but the same cortical thickness will have a lower MMI. During long bone growth, bone is shaped by resorption on the outside (periosteum) and formation on the inside (endosteum). Therefore, increased osteoclastic bone resorption could lead to excess resorption on the outside, compensated by increased endosteal bone formation, leading to a decrease in bone perimeter.
Despite the overall increase in bone turnover, we observed no evidence of focal bone lesions in these mice as occurs in human PDB. This indicates that deletion of PML is not sufficient to cause PDB-like bone lesions in mice, contrasting with mouse models of two other PDB-susceptibility genes; SQSTM1 P394L knock-in mice Although we have no reason to suspect that the effects of PML on bone metabolism differ in males and females, a limitation of the preclinical studies described here was that the skeletal phenotyping was restricted to male mice. We chose to study males because PDB is more common in men but acknowledge that further studies to investigate bone metabolism in female Pml -/mice would be of interest.
In contrast to a previous report (Lunardi et al., 2011) we did not observe any difference in survival between Pml -/mice and littermates, nor did we observe increased susceptibility to infections. The reasons for are not entirely clear but are likely to be due to differences in animal husbandry since the increased risk of infections noted by Lunardi was only observed when the mice were kept in non-

Disease Models & Mechanisms • DMM • Accepted manuscript
pathogen free conditions or were directly challenged with micro-organisms. In this study however, the mice were kept in specific-pathogen free conditions.
In summary, our findings are consistent with a model whereby genetic variations at the 15q24 locus predispose to PDB by reducing expression of PML which stimulates osteoclastic bone resorption and bone formation. Although the PML variant associated with increased risk of PDB (rs5742915_T>C) results in an amino acid change, it is not predicted to be pathogenic. However, further studies are warranted to determine how this variant regulates the expression of PML and to determine if it does so directly or through linkage disequilibrium with another nearby functional variant.

Mice
The Pml -/mice were obtained from the National Cancer Institute, USA and were generated as previously described (Wang et al., 1998). Briefly, Pml was disrupted in the mouse germ line and knockout generated by deleting part of exon 2 (94bp) which encodes the RING finger domain. Complete absence of Pml in these mice was verified by southern and northern blotting as well as immunofluorescence staining (Wang et al., 1998). The animals bred and maintained for the study were genotyped to confirm their status as per the protocols specified by the National Cancer Institute's mouse repository. The Pml -/and WT mice used in the experiments were littermates on a C57BL/6 background. The skeletal phenotyping experiments described were conducted on male Pml -/mice. The mice were housed in a standard animal facility (specific-pathogen free) with free access to food (pelleted RM1; SDS diets, UK) and water. All experiments on mice were performed according to institutional, national, and European animal regulations.

Microcomputed tomography
Mouse hindlimbs were imaged by micro computed tomography (µCT) using a Skyscan 1272 µCT scanner (Bruker, Belgium) as described previously (van 't Hof and Dall'Ara, 2019). Briefly, hind limbs of mice were dissected free of most soft tissue, fixed in 4% buffered formaldehyde for 24 h, stored in 70% ethanol and scanned at a resolution of 5 µm (60 kV, 150 µA, rotation step size 0.3°, 0.5 mm aluminium filter). Image reconstruction was performed using the Skyscan NRecon package. Skyscan Dataviewer software was used to orientate the image stacks and create subvolumes for subsequent image analysis. The reconstructed CT images were also subjected to 3D analyses using Skyscan Dataviewer software to screen for bone lesions and CT Vol software in addition was used to generate 3D model images of bones.
Trabecular bone was analysed in a stack of 200 slices starting 100 m from the distal femoral growth plate. Cortical parameters were measured in 100 slices at the midshaft of the femur. Trabecular and cortical bone parameters were measured in Skyscan CTAn software using a fixed threshold and automated separation of cortical and trabecular bone using a custom macro.

Bone Histomorphometry
Bone histomorphometry was performed at the distal femoral metaphysis essentially as described in van 't Hof RJ et al (van 't Hof et al., 2017). Briefly, mice received intraperitoneal calcein injections (2mg/ml, 150l) 5 days and 2 days before culling.
Hind limbs were fixed for 24 h in 4% buffered formalin and embedded in methyl methacrylate (MMA). Five m sections were cut using a tungsten steel knife on a Leica motorized rotary microtome, and stained for TRAcP to visualise osteoclasts and counterstained with Aniline Blue. For analysis of calcein double labelling, sections were counterstained with Calcein Blue. The only difference from the methods described in van 't Hof et al (van 't Hof et al., 2017), is that for the TRAcP stain, the slides were coverslipped using 80% glycerol rather than Apathy's serum.

Bone resorption assay
The bone resorption activity of osteoclasts was determined using Osteo Assay surface 24-well plates (Corning). BMDMs were plated in Osteo Assay plates (50,000 cells/well in 500ul/well supplemented αMEM) and differentiated into osteoclasts as described above in osteoclast cultures. Wells were then treated with 2% sodium hypochlorite solution for 5 minutes, washed with distilled water and air-dried. For modified Von-Kossa staining, plates were treated, away from light, with 5% (w/v) silver nitrate solution for 30 minutes and then rinsed for 5 minutes with distilled water. Wells were then incubated in 5% (w/v) sodium carbonate in formalin solution

Disease Models & Mechanisms • DMM • Accepted manuscript
for 5 minutes and then washed twice with distilled water followed by drying them at 50°C for one hour. Plates were then imaged with a Zeiss inverted microscope and resorption areas in each well were analysed using ImageJ software (Schneider et al., 2012). Each experiment was repeated using bone marrow cells from at least 3 different mice with 4-6 technical replicate wells per experiment.

Primary osteoblast cultures
Osteoblasts were isolated from the calvarial bones of 2-4 day-old mice by sequential

Disease Models & Mechanisms • DMM • Accepted manuscript
Immunoblotting Cells were lysed using radioimmunoprecipitation assay (RIPA) buffer, centrifuged and protein concentration was measured using the Pierce protein assay. Proteins were loaded on Mini Protean TGX Precast gel electrophoresis system and electroblotted onto Biorad Mini PVDF membranes using Transblot Turbo transfer system. Membranes were blocked with 5% (w/v) non-fat milk in Tris buffered saline (Thermoscientific Pierce) with Tween-20 (TBST: 50 mM Tris, 150 mM NaCl, 0.1% [v/v] Tween-20) and probed with relevant primary antibody. After washing with TBST, membranes were incubated with anti-rabbit horseradish peroxidase conjugated secondary antibody (1:5000, Cell Signaling) washed and visualized using Clarity Western ECL kit (BioRad) on a Licor Odyssey imager.

Quantitative Real-Time PCR (qRT-PCR)
Total RNA was isolated using GenElute Mammalian Total RNA Kit and RNA was quantified using the Nanodrop 1000 Spectrophotometer. Complementary DNA was generated by RT-PCR using the qScript cDNA SuperMix kit following the manufacturer's instructions. Primers and fluorescently-labelled probes were designed using the Primer 3 and the Roche Diagnostics website (Roche). Table S1 describes primer sequences and other details for target genes analyzed by qPCR.
Real-time PCR was performed on diluted cDNA using SensiFAST Probe No-ROX kit on a Chromo 4TM Detector/ Bio RAD CFX Connect system and analysed using the Opticon MonitorTM software version 3.1 or Bio RAD CFX Manager V1.0. Samples were normalized to 18s rRNA expression. 18s cDNA was amplified with the VIClabelled predesigned probe-primer combination from Applied Biosystems (4319413E) allowing two channel detection of one cDNA. rs5742915 allele-specific expression of PML was performed using fluorescently labelled TaqMan probes (Applied Biosystems cat number 4351379) by following the manufacturer's protocol.

Study subjects
We

Statistical analyses
Analysis was performed using SPSS (IBM, USA) and Prism ver 8.4 (GraphPad software, USA). Box and whiskers plots show the interquartile range (boxes), median (line inside boxes) and range (whiskers). Tow-tailed unpaired student t-test was used for comparisons between two groups. P value < 0.05 was considered to indicate statistical significance. PML mRNA expression data in PDB patients and controls were analysed using linear regression adjusting for age and gender. Experiments were performed as independent replicates and data presented as indicated in figure legends.