Differential Gene Expression in Cultured Osteoblasts and Bone Marrow Stromal Cells From Patients With Paget's Disease of Bone


  • The authors state that they have no conflicts of interest.


Paget's disease is a focal condition of bone. To study changes in cells within pagetic lesions, we cultured osteoblasts and stromal cells from 22 patients and compared gene expression in these cells to cells from healthy bone. We identified several differentially regulated genes, and we suggest that these changes could lead to the formation of the lesions.

Introduction: Paget's disease is a focal condition of bone of unknown cause. Although it is regarded as primarily an osteoclast disorder, the tight coupling of the activity of osteoclasts and osteoblasts suggests that the osteoblast could play a key role in its pathogenesis. The aim of the study was to identify possible changes in pagetic osteoblasts and stromal cells that might contribute to the development of pagetic lesions.

Materials and Methods: Candidate genes were identified based on known bone cell regulators, supplemented with microarray analysis. Gene expression was determined by real-time PCR in primary cultures of osteoblasts and bone marrow stromal cells from pagetic patients and control subjects. Concentrations of secreted proteins were determined by ELISA.

Results: Dickkopf1 mRNA and protein levels were increased in both pagetic osteoblast and stromal cell cultures, and interleukin (IL)-1 and IL-6 were overexpressed in pagetic osteoblasts. These changes parallel recent findings in myeloma bone disease, which shares some clinical similarities with Paget's disease. Alkaline phosphatase was overexpressed, and bone sialoprotein and osteocalcin were underexpressed in pagetic osteoblasts, consistent with their circulating levels in pagetic patients. It is hypothesized that overexpression of Dickkopf1, IL-1, and IL-6 would result in stimulation of osteoclast proliferation and inhibition of osteoblast growth, leading to the development of the characteristic lytic bone lesions. By stimulating osteoblast differentiation, Dickkopf1 and IL-6 may also promote mineralization, leading to the conversion of lytic lesions to sclerotic.

Conclusions: These findings suggest that dysregulated gene expression in pagetic osteoblasts could cause the changes in bone cell number and function characteristic of Paget's disease.


Paget's disease is the second most common metabolic bone disease after osteoporosis. Its prevalence varies substantially between regions, but reaches 6–7% in the elderly population of Western Europe. The bone pain, skeletal deformity, pathological fractures, secondary arthritis, neurological complications, and deafness that may accompany this disease contribute to substantial morbidity in the elderly population. It is distinct from other metabolic bone diseases in that it is focal, the intervening regions of the skeleton being normal. The pagetic lesion develops initially as an area of intense osteoclast overactivity that appears on radiographs as a demarcated area of bone lysis. These lesions typically progress along bones at a rate of −1 cm/year, which is consistent with the histologically shown rate of advance of pagetic osteoclasts.(1) As the pagetic lesion advances, the area that was originally lytic becomes sclerotic, with marked osteoblast overactivity and greatly increased deposition of bone matrix and mineral. The fact that the lytic phase precedes the development of sclerosis has led to the hypothesis that the pagetic osteoclast is the site of the primary cellular abnormality in this condition. However, it is now clear that osteoblasts are intimately involved in the regulation of osteoclast function, through the elaboration of a variety of factors, including RANKL and osteoprotegerin (OPG). This raises the possibility that the fundamental abnormality in Paget's disease might reside in the osteoblast.

Paget's disease sometimes runs in families, and −10% of pagetic subjects are reported to have an affected relative.(2) This observation has led to much work seeking genetic associations of the condition. It is now apparent that mutations of the gene for sequestosome 1 are associated with Paget's disease in some families.(3) Other research has focused on a possible environmental cause of Paget's disease, and a slow viral infection has been suggested.(4) Evidence that the prevalence of Paget's disease has changed in recent decades would be consistent with altered exposure to an environmental agent.(5) However, both the genetic and environmental hypotheses fail to account for the focal nature of the condition. Understanding of this depends on identifying what distinguishes the cells in pagetic bone from those in normal bone, whether the normal bone is in the same individual or in an unaffected subject. This study seeks to address this question by comparing gene expression between pagetic and nonpagetic tissue, the latter sourced from both pagetic and normal individuals. A number of genes known to be involved in bone cell regulation have been studied, supplemented with microarray analysis to identify other candidate genes.


Bone samples

Bone samples were collected from consenting subjects (with or without Paget's disease) undergoing hip and knee replacements for arthritis. The study had the approval of the local institutional review board. Subjects taking steroids or suffering from other bone disorders were excluded, as were people with cancer, rheumatoid arthritis, kidney, liver, or thyroid disease. Samples collected from affected areas of the skeleton (judged by X-ray or scintigraphy) are described as “pagetic.” Nonpagetic samples were collected from subjects without Paget's disease and from unaffected bone of patients with the condition. Patient details are summarized in Tables 1 and 2.

Table Table 1.. Clinical Characteristics and Sample Details of Paget's Patients
original image
Table Table 2.. Sample Details of Nonpagetic Control Subjects
original image

Osteoblast cultures

Bone removed during orthopedic surgery was cut into small fragments that were washed several times with PBS and then with DMEM (Invitrogen) and were incubated in 1% collagenase (Sigma)/DMEM in a 37°C shaking water bath for 30 minutes. Bone fragments were incubated in T75 flasks at 37°C with 5% CO2 until outgrowth of cells was noted, usually 3–4 days later. At this stage, the bone fragments were transferred to new flasks and maintained in DMEM/10% FBS (Invitrogen) and 5 μg/ml l-ascorbic acid 2-phosphate (Sigma) at 37°C, 5% CO2 with media replacement every 3–4 days, until near confluence. Alkaline phosphatase (ALP) staining confirmed the phenotype of cultured cells as being of the osteoblast lineage and was comparable in pagetic and nonpagetic cells (Fig. 1). This was further supported by RT-PCR showing the presence of the osteoblast markers ALP, type 1 collagen α 1, and osteocalcin in RNA extracted from these cells. The cells had a uniform appearance under light microscopy, which was similar for pagetic and nonpagetic cells. Pagetic and nonpagetic osteoblasts took similar times to reach confluence in culture (−15 days). Seventy-two hours before the end of the culture period, all bone pieces were removed from the flasks, and media were renewed. Finally, conditioned media from each flask were collected and stored at −80°C, and the cells were harvested for RNA extraction.

Figure Figure 1.

Alkaline phosphatase staining of cultures of osteoblasts derived from pagetic (A) and nonpagetic (B) bone. The great majority of cells show some degree of staining and have typical osteoblast morphology. The variability of staining intensity reflects the differing degrees of differentiation of individual cells, as expected in short-term cultures. Appearances are not different between pagetic and nonpagetic cells. Horizontal bar, 200 μm.

Bone marrow stromal cell cultures

All the nonpagetic and eight of the pagetic bone marrow stromal cell (BMSC) samples were collected from subjects undergoing arthroplasty. Six additional pagetic BMSC samples were from iliac crest bone marrow aspirates. Bone marrow was collected in 1000 units/ml heparin in DMEM and aspirated to remove any clumps. After allowing the sample to settle, any large blood clots, fat tissue, or obvious bone fragments were removed. Bone marrow cells were separated on a Ficoll gradient, and the interface layer, enriched for mononuclear cells, was collected. Cells were seeded into T75 flasks at a density of 1.4 × 107 cells per flask in DMEM/10% FBS + 5μg/ml l-ascorbic acid 2-phosphate and maintained at 37°C in 5% CO2 until near confluence. Media were replaced every 3–4 days. As with the osteoblast cultures, ALP staining was present, and the osteoblast markers ALP, type 1 collagen α 1, and osteocalcin were expressed. Pagetic and nonpagetic cells were of similar appearance, but the pagetic cells tended to take longer to reach confluence (−16 days compared with 13 days for the nonpagetic cells). Seventy-two hours before the end of the culture period, all media were replaced and were subsequently collected and stored at −80°C. After trypsinization and removal of nonadherent cells for RNA extraction, some strongly adherent cells remained. These cells were both mono- and multinucleated and showed positive staining for TRACP, which is one of the markers of osteoclast cells.

Gene expression profiling

Total RNA was extracted from cells using the RNeasy Mini-prep Kit (Qiagen) and treated with RNase-Free DNase Set (Qiagen). RNA quality was determined by an Agilent bioanalyzer (Agilent Technologies). Twelve osteoblast RNA samples, five pagetic and seven nonpagetic, with RNA integrity number (RIN) values of 9.2–10 were chosen for microarray hybridization. RNA was reverse transcribed to cDNA using SuperScript II (Invitrogen), and cDNA second strand was synthesized with E. coli DNA polymerase I. Biotinylated cRNA was produced by in vitro transcription and fragmented to produce small RNA probes that were hybridized to the Human Genome U133A 2.0 GeneChip (Affymetrix), analyzing >22,000 probes that represent >14,500 genes. After hybridization, the gene chips were automatically washed and stained with streptavidin-phycoerythrin in a fluidics system and scanned with a Hewlett-Packard GeneArray Scanner (Hewlett-Packard).

Microarray data analysis

After scanning, the Affymetrix GeneChip Operating Software (GCOS) was used to produce fluorescence intensity values for each of the perfect-match and mismatch probes on each array. This information was exported to CEL data files, which were in turn imported into the R statistical computing environment(6) for analysis. The Affymetrix package(7) from the Bioconductor suite of analysis tools(8) was used to apply the Robust Multichip Analysis (RMA) algorithm(9) to the data from each array. The Limma package(10) was used to perform a linear model–based analysis of the RMA normalized data, with the goal of detecting probe sets undergoing significant changes in expression level between pagetic and nonpagetic RNA. For each comparison, the false discovery rate controlling method of Benjamini and Hochberg(11) was used to produce adjusted p values for each probe set, based on a significance level of α = 0.05.

Real-time PCR

cDNA was synthesized using Superscript III (Invitrogen), and used for multiplex real-time PCR in ABI PRISM 7700 or 7900 Sequence Detection Systems (Applied Biosystems). Primers and probe sets were purchased as TaqMan Gene Expression Assays (Applied Biosystems) containing a forward and reverse unlabeled PCR primer pair and a fluorescent reporter dye-labeled TaqMan MGB probe. All probes used to detect candidate genes were labeled with FAM, whereas the 18S rRNA endogenous control probe was VIC labeled. A seven-point validation curve was performed for each probe set and used to determine the method of analysis in all further experiments with this probe set. If the slope was >0.3, the standard curve method was used; otherwise, the ΔΔCt method was used.(12)

Expression levels were determined by real-time PCR in 14 pagetic osteoblast samples and 28 controls. Eight of these samples were paired'taken from a pagetic lesion and from healthy bone of the same patient. We also studied the expression of these candidate genes in RNA extracted from BMSCs taken from 14 pagetic and 21 nonpagetic samples. For both osteoblasts and BMSCs, samples were grouped into pagetic and nonpagetic, and the average of the nonpagetic values was used as a denominator to determine relative expression. There was no dependence of the level of expression of any of the differentially expressed genes on the age or sex of the subjects. Each experiment was repeated at least twice with similar results, and the figures show results of one representative experiment. Statistical significance was determined by two-tailed Student's t-test. For the analysis of the subgroup of paired samples, expression levels were normalized using the sample from the uninvolved site, and significance was determined by a paired Student's t-test.


A modification of the method described by Tian et al.(13) was used to determine Dickkopf1 (Dkk1) protein concentrations. Microtiter plates (Greiner) were coated with 100 μl of anti-Dkk1 polyclonal antibody (R&D Systems) at a concentration of 1 μg/ml in PBS, pH 7.2. After an overnight incubation at 4°C, the reaction was blocked with 4% BSA. Conditioned media and serum samples were diluted up to 1:8 in dilution buffer (1× PBS, 0.1% Tween-20, and 1% BSA). Recombinant human Dkk1 (R&D Systems) diluted in dilution buffer to concentrations from 0.3125 to 20 ng/ml was used as a standard curve. A total of 100 μl of standard or sample was loaded per well and incubated overnight at 4°C, washed, and incubated with biotinylated goat anti-human Dkk1 IgG (R&D Systems) diluted to a concentration of 0.2 μg/ml in dilution buffer. This was followed by the addition of 100 μl of a 1:200 dilution of streptavidin-horseradish peroxidase (R&D Systems) and Substrate Reagent Pack (R&D Systems). The reaction was stopped with 1 M sulfuric acid, and absorbance was measured at 450nm. The concentration of secreted OPG protein was determined with The Human Osteoprotegerin/TNFRSF11B DuoSet (R&D Systems), used with the same Substrate Reagent pack as above.

Experiments were carried out in duplicates and repeated three times. The figures present the average values of pooled results from the three experiments. Statistical significance was determined by two-tailed Student's t-test.


Local regulators of bone remodeling

Expression of RANKL mRNA in pagetic osteoblasts and BMSCs was about one fifth of that found in nonpagetic cells (Fig. 2A). OPG mRNA levels were significantly higher in pagetic BMSCs, but there was no significant difference between pagetic and nonpagetic osteoblasts. In both osteoblasts and BMSCs, the RANKL/OPG ratio was lower in the pagetic samples (Fig. 2A). This result was unexpected, because a low RANKL/OPG ratio is usually associated with a decrease in bone turnover. Concentrations of secreted OPG protein corresponded to the levels of OPG mRNA, being higher in the media of pagetic BMSC cultures but not in pagetic osteoblasts (Fig. 2B). OPG protein levels were closely correlated with expression of mRNA for this protein (Fig. 2C).

Figure Figure 2.

Expression levels of RANKL, OPG, IL-1β, and IL-6 in pagetic and nonpagetic osteoblasts and BMSCs. RNA was extracted from osteoblasts cultured from bone samples of 28 nonpagetic and 14 pagetic subjects and BMSCs cultured from 21 nonpagetic and 14 pagetic bone marrow samples. The relative expression of the genes was determined by real-time PCR. (A) mRNA levels of OPG and RANKL and the ratio between RANKL/OPG expression levels. The mean values of the nonpagetic group were normalized to 1, and the mean values of the pagetic group are expressed relative the nonpagetic group. (B) Concentration of secreted OPG protein in conditioned media samples as determined by ELISA. The conditioned media samples were collected during the last 72 h of incubation from the cell cultures described above. (C) Correlation between levels of OPG mRNA in cells and secreted OPG protein in conditioned media. Levels of OPG mRNA are expressed relative to the expression of ribosomal RNA. The correlation was calculated using Deming (model II) linear regression (r = 0.89, p < 0.0001). (D) Relative mRNA levels of IL-1β and IL-6 were determined by real-time PCR, and the mean values of the nonpagetic group were normalized to 1. Data in A, B, and D are presented as mean ± SE. ap < 0.05 vs. control.

The pro-inflammatory cytokines interleukin-1β (IL-1β) and IL-6 are produced in the bone microenvironment and affect the rate of bone remodeling, promoting osteoclast differentiation and activity. mRNA levels of IL-6 were >2-fold higher in the pagetic osteoblasts and BMSCs (Fig. 2D). IL-1β expression level was significantly higher in the pagetic osteoblasts, but in BMSCs, there was no difference between the pagetic and nonpagetic groups (Fig. 2D).

Other genes studied in osteoblasts were CSF1, IL-11, COX2, SHIP, VEGF, SQSTM1, and BCL-2. Their expressions in pagetic and nonpagetic osteoblasts were similar. In BMSCs, these genes were studied, as well as TNF and MIP-1α. Again, there were no significant differences between pagetic and nonpagetic cultures.

Gene expression profiles of pagetic osteoblasts

We used microarray analysis to compare gene expression between pagetic and nonpagetic primary osteoblasts, as a way of generating further candidate genes that could be studied by real-time PCR. Twelve samples were analyzed: three paired (taken from involved and uninvolved bones of the same patient), plus two additional pagetic and four nonpagetic samples. Lists of genes that were differentially expressed in the pagetic osteoblasts are presented in Tables 3 and 4. Table 3 shows those that were significantly different between groups, and Table 4 shows those that had the greatest fold change in expression. Of the latter, not all were statistically significant, because of the small number of samples subjected to this analysis and the need to adjust for multiple statistical comparisons. Of the genes listed in Tables 3 and 4, some play important roles in osteoblast biology and were studied further.

Table Table 3.. Genes Showing Statistically Significant Changes in Expression in Microarray Analysis of Pagetic and Nonpagetic Osteoblast RNA Samples
original image
Table Table 4.. Top 10 Upregulated and Downregulated Genes in Microarray Analysis of Pagetic and Nonpagetic Osteoblast RNA Samples Ranked According to Fold Change
original image

Dickkopf 1

Dickkopf 1 (Dkk1) is a soluble inhibitor of the Wnt signaling pathway that plays a critical role in osteoblast differentiation. Microarray analysis indicated a 4-fold upregulation of Dkk1 in pagetic osteoblast RNA. This was confirmed by real-time PCR, which showed a 3-fold increase in Dkk1 mRNA levels in pagetic osteoblasts and a >5-fold increase in pagetic BMSCs (Fig. 3A).

Figure Figure 3.

Upregulation of Dkk1 expression in pagetic osteoblasts and BMSCs. (A) The relative expression of Dkk1 was determined by real-time PCR, and values are expressed relative to the mean level of expression in the nonpagetic groups, which have been normalized to 1. The number of samples was the same as in Fig. 2. (B) The concentration of secreted Dkk1 protein in conditioned media samples as determined by ELISA. The conditioned media samples were collected during the last 72 h of incubation from the cell cultures described above. Data in A and B are presented as mean ± SE. ap < 0.05 and bp < 0.01 vs. control. (C) Correlation between levels of Dkk1 mRNA in cells and secreted Dkk1 protein in conditioned media. Levels of Dkk1 mRNA are expressed relative to the expression of ribosomal RNA. The correlation was calculated using Deming (model II) linear regression (r = 0.88, p < 0.0001).

There was a 2- to 3-fold increase in the levels of secreted Dkk1 protein in the pagetic media samples (Fig. 3B). The levels of Dkk1 mRNA and secreted Dkk1 protein were closely related (Fig. 3C).

To study whether the higher level of expression of Dkk1 is associated with changes in expression of other genes of the Wnt signaling pathway, we constructed a combined list of genes from Wnt pathway (hsa04310) in the Kyoto Encyclopedia of Genes and Genomes (KEGG) and Wnt signaling target genes listed on the Wnt home page (http://www.stanford.edu/-rnusse/wntwindow.html). Dkk1 itself was removed from the lists. A global assessment of the microarray data showed that the expression values for the genes in this list exhibited greater differences between the Paget's and control samples than would be expected by chance, as assessed by two methods for gene set analysis: Significance Analysis of Function and Expression(14) (p < 0.01) and globaltest(15) (p < 0.01). These results suggest that the differences in levels of expression observed in genes relating to Wnt signaling are in some way related to the mechanisms underlying Paget's disease.

Extracellular osteoblast proteins

Two bone matrix proteins seemed to be downregulated in the pagetic osteoblasts: bone sialoprotein and osteocalcin. These proteins are expressed in osteoblasts at late stages of differentiation when matrix mineralization is taking place. Real-time PCR showed a 10-fold lower expression of bone sialoprotein in pagetic osteoblasts and a >2-fold decrease in pagetic BMSCs (Fig. 4A). Osteocalcin showed a 5-fold lower expression in pagetic osteoblasts (Fig. 4B), with no significant change in pagetic BMSCs.

Figure Figure 4.

Differential expression of mRNA of extracellular osteoblast proteins in pagetic osteoblasts and BMSCs. The relative expression levels of (A) bone sialoprotein, (B) osteocalcin, and (C) alkaline phosphatase (ALP-L) were determined by real-time PCR. The mean values of expression in the nonpagetic groups were normalized to 1, and the mean values of the pagetic groups are expressed relative to the nonpagetic groups. The number of samples was the same as in Fig. 2. Error bars represent SE. ap < 0.05 and bp < 0.01 vs. control.

Although not identified by the microarray analysis as a differentially regulated gene, we also compared the expression levels of bone-specific ALP (ALP-L), because this is an important marker of osteoblast differentiation and function. The relative expression of this gene was −2.5 times higher in pagetic osteoblasts, whereas no significant difference was observed between pagetic and nonpagetic BMSCs (Fig. 4C).

Other differentially regulated genes in pagetic osteoblasts

Keratin 18 (KRT18), a type I cytoplasmic intermediate filament protein, was identified by the microarray analysis as having a significantly higher expression in pagetic osteoblasts (fold change: 6.76; p = 0.04). The upregulation of KRT18 in pagetic osteoblasts was confirmed by real-time PCR (Fig. 5A). Interferon α-inducible protein 27 (IFI27), one of many genes induced by IFN-α and IFN-β, showed a 6.86-fold upregulation in the microarray analysis. Real-time PCR also showed a significant upregulation of IFI27 in the pagetic osteoblast samples (Fig. 5B).One of the genes that was found to be significantly downregulated in the pagetic mRNA samples was the transcription factor MAF-B. Real-time PCR analysis indicated an expression level that was 5-fold lower in the pagetic osteoblasts as compared with the nonpagetic controls (Fig. 5C). No significant changes in the levels of expression of these three genes were found in the BMSCs.

Figure Figure 5.

Differential expression of mRNA for keratin 18, IFI 27, and MAFB in pagetic osteoblasts. The mean values of mRNA expression in the nonpagetic groups were normalized to 1, and the mean values of the pagetic groups are expressed relative the nonpagetic groups. Error bars represent SE. bp < 0.01 vs. control.

Differential gene expression in paired samples

Comparisons of cells from the pagetic patients with those from the group of nonpagetic controls could be influenced by the exposure of the patients to bisphosphonate therapy. Therefore, we analyzed gene expression in the eight paired osteoblast samples, comparing involved and uninvolved bone from the same patients. This permitted comparison of pagetic and nonpagetic cells with the same history of bisphosphonate exposure. As shown in Fig. 6, the changes within the paired samples were all consistent with those found in the comparisons of the groups of pagetic and nonpagetic samples, although the differences were less significant because of the smaller sample size. This suggests that the changes found in this study are more likely to reflect the pathology in the pagetic lesion rather than the effects of treatment.

Figure Figure 6.

Gene expression (by real-time PCR) in the subset of eight paired osteoblast samples. The broken horizontal lines represent the level of expression of each of the genes in the uninvolved site, which has been normalized to 1. The bars represent the relative levels of expression in the involved sites. (A) Relative expression levels of genes that were found to be upregulated in the full set of samples. (B) Relative expression levels of genes that were found to be downregulated in the full set of samples. ap < 0.05 and bp < 0.01 vs. control.


The elucidation of pathways by which osteoblasts regulate osteoclast activity provided the original stimulus to this work. The most important of these pathways is that involving osteoblast production of RANKL and its blocking receptor, OPG. Osteoblasts could produce the bone lysis seen in Paget's disease either by overexpressing RANKL or underexpressing OPG. Our hypothesis that changes in osteoblast expression of either of these might underlie the development of pagetic lesions is not supported by the present evidence. In fact, the trends for both are quite the opposite, and the RANKL/OPG ratio is, if anything, reduced in isolated pagetic osteoblasts. This finding is apparently at variance with the results of Menaa et al.,(16) who reported increased RANKL mRNA expression in pagetic mixed marrow cultures. However, this may have arisen from increased IL-6 activity, most of which would be of osteoclast origin, rather than reflecting the intrinsic activity of the osteoblast, which our studies address. Our findings suggest that other factors are driving the development of the pagetic lesion and that the changes in the RANKL/OPG system are reactive rather than causative.

A second well-characterized mechanism by which osteoblasts influence osteoclasts is through the production of the cytokines IL-1 and IL-6. Each of these is well established as a direct upregulator of osteoclastogenesis, and IL-1 can also increase IL-6 release from stromal or osteoblast cells.(17) Thus, the overexpression of these two cytokines in pagetic osteoblasts provides a synergistic mechanism for the local stimulation of osteoclastogenesis. These findings are consistent with the earlier reports of Roodman et al.,(18) that IL-6 concentrations were increased in the bone marrow plasma from patients with Paget's disease and that the osteoclastogenic effect of conditioned media from pagetic bone marrow cultures was blocked by a neutralizing antibody directed against IL-6.(16) Hoyland et al.(19) also showed, using in situ hybridization, that osteoblasts in Paget's disease express higher levels of IL-6 and IL-6R mRNA than do osteoblasts from osteoarthritis patients. More recently, Neale et al.(20) found that the addition of serum from untreated Paget's patients dose-dependently increased osteoclast formation and bone resorption in normal monocyte cultures. Elevated IL-6 levels were found in the supernatant, and the addition of a specific antibody to human IL-6 blocked the increase in osteoclast formation and resorption. In normal subjects, the osteoclast is the most abundant source of IL-6 in the body, so the earlier finding of high concentrations in mixed bone marrow cultures from Paget's patients could be explained simply by the increased numbers of osteoclasts. The demonstration in this study, that IL-6 production is increased in isolated osteoblasts and could therefore be a primary abnormality in Paget's disease, is an important addition to our knowledge in this area. In addition, these studies use quantitative methods to assess gene expression and are based on considerably larger numbers of patient samples than the earlier work.

It should be noted that IL-6 also has effects on osteoblasts. In the context of Paget's disease, these could be both autocrine (involving IL-6 of osteoblast origin) and paracrine (involving IL-6 of osteoclast origin). There is now a large body of literature indicating that IL-6 modulates osteoblast proliferation, differentiation, and apoptosis. The directions of these effects are dependent on the particular model used, and whether the IL-6 soluble receptor is present, but in the context of high bone turnover, IL-6 clearly supports osteoblast generation.(21) Thus, IL-6 excess may be a key contributor to the pagetic overactivity of both osteoblasts and osteoclasts, similar to the pivotal role that it plays in the development of osteolytic lesions in multiple myeloma, prostate, and breast cancer (see below).

Perhaps the most interesting finding to emerge from this study, is the increased expression of the Wnt signaling antagonist, Dkk1, in both osteoblasts and in BMSC cultures. The canonical Wnt pathway plays a key role in regulating osteoblast proliferation and differentiation.(22) This pathway is activated in osteoblasts by the binding of Wnt ligands to the seven-transmembrane receptor Frizzled and to the low-density lipoprotein receptor–related proteins 5 (LRP5) co-receptor. Dkk1 inhibits Wnt signaling by binding and inactivating LRP5.(23) Dkk1 overproduction has recently been implicated in the development of focal bone lesions in multiple myeloma,(13) which are radiologically similar to lytic pagetic lesions. Tian et al.(13) have suggested that the release of Dkk1 from malignant plasma cells in multiple myeloma results in an inhibition of osteoblast proliferation, accentuating the imbalance between bone formation and bone resorption and facilitating local bone loss. Dkk1 overexpression has also been observed in prostatic(24) and breast cancer(25) cell lines that produce lytic bone lesions when inoculated into animals. In the context of Paget's disease, overproduction of Dkk1 in the osteoblast itself could have a similar effect on bone formation. Gunn et al.(26) have extended this work by showing that Dkk1 from myeloma cells increases IL-6 expression from undifferentiated mesenchymal stem cells, which promotes further myeloma cell proliferation. In the context of Paget's disease, the stimulation of IL-6 production as a result of overexpression of Dkk1 will contribute to the development of the lytic pagetic lesion through further accelerating local bone turnover, as discussed above.

Subsequent studies have indicated that the role Dkk1 (and Dkk2, which was not expressed in pagetic or nonpagetic osteoblasts in these studies) plays in osteoblast development is more complex. Using mouse bone marrow and the KS483 mesenchymal cell line, van der Horst et al.(27) showed that inhibition of Wnt signaling is necessary for late-stage osteoblast differentiation and mineralization and that this inhibition is achieved by the upregulation of the expression of Dkk1 and Dkk2. Similar results were found in studies of the role of Dkk2 in terminal osteoblast differentiation in mice.(28) Thus, Dkk overexpression could both promote the development of the lytic pagetic lesion (similar to its role in myeloma) but ultimately also contribute to the sclerotic stage of Paget's disease, to which most lytic lesions progress. In vivo, this transition takes place over a period of months to years, and often there is co-existence of both, with sclerosis in the older part of a lesion, whereas the advancing edge shows a lytic process. Possibly, Dkk1 inhibits the proliferation of osteoblasts in the lytic pagetic lesion and facilitates local mineralization as osteoblast differentiation takes place; this transition is simply a time-dependent process within a given cohort of pagetic cells.

The changes in expression of the genes for ALP and osteocalcin are consistent with clinical experience in Paget's disease, in which serum ALP activity is a sensitive indicator of disease activity, whereas serum osteocalcin is frequently in the normal range despite the other evidence of osteoblast overactivity.(29) Osteocalcin and bone sialoprotein are produced in well-differentiated osteoblasts. Their downregulation in the pagetic osteoblast cultures implies that these cells are of a relatively immature phenotype. The consonance between the gene expression in vitro and the circulating levels of these markers in vivo provides reassurance that the phenotype of the pagetic osteoblast has been maintained in the long-term cultures used in these studies. Changed expression of a number of other genes was also found in this study. At this time, these do not have recognized functions in osteoblast regulation. Their altered expression could reflect differences in differentiation or proliferation between pagetic and nonpagetic osteoblasts or they could contribute to the pagetic phenotype by mechanisms that remain to be elucidated.

This study does not address the possible genetic contribution to the development of Paget's disease. In a minority of patients with this condition, there is a family history, and several genetic loci have been associated with the development of Paget's disease, including mutations in the gene for sequestesome-1. Expression of this gene was moderately, but nonsignificantly, upregulated in this study (data not shown). However, a germline mutation can only provide a partial explanation for a focal condition such as Paget's disease, because the pagetic phenotype is expressed only in affected areas of the skeleton, whereas the mutation will be present in all or none of the skeletal cells. An inherited mutation could provide a predisposition to an infective agent or to the acquisition of some other form of abnormality that results in the fully expressed cellular phenotype.

There are a number of cautions that need to be borne in mind when interpreting the results of this study. Microarray analysis has the capacity to assess expression of a very large number of genes. This brings with it the likelihood that differences will be found by chance alone, particularly so when the microarrays have been carried out on a relatively small number of samples. For this reason, we used the microarray analysis as a way of looking for potential candidate genes and assessed the expression of these genes in a larger number of samples, using real-time PCR. In some cases, we have also been able to measure protein levels in the culture media, providing independent verification of the results. Many of the genes discussed here in which differential expression has been identified are known to play key roles in osteoblast biology, and the direction in which their expression is changed is consistent with the bone phenotype in Paget's disease, lending biological credibility to the findings. In any condition for which patients receive pharmaceutical treatment, differences in gene expression could result from either the condition itself or its treatment. For this reason, we collected affected and unaffected bone from the same subjects. This permitted us to assess gene expression in pagetic and nonpagetic osteoblasts, both of which had been exposed to similar doses of bisphosphonates over the same time-course. Paired samples were only available in the minority of cases, so the results are not as statistically significant as those for the total dataset. However, the direction of the changes is consistent with those in the main dataset, which suggests that the principal findings reflect the nature of Paget's disease rather than the effects of its treatment. Finally, both pagetic and nonpagetic bone samples were obtained from sites of joint replacement, so both groups of samples may reflect local arthritic change, but this should not contribute bias to the comparisons made.

Figure 7 provides a diagrammatic representation of the key findings to emerge from this study and how they might work together to produce the overactivity of osteoblasts and osteoclasts characteristic of Paget's disease. The central role of IL-6 is consistent with the earlier work of Roodman et al.,(18) and the highly significant changes in Dkk1 expression mirror the recent findings in multiple myeloma and other osteolytic malignancies. There may well be other major players that will need to be added to this schema in the future, and these studies do not permit determination of which changes are primary and which are secondary. Perhaps the greatest significance of this schema is that it gives the pagetic osteoblast a central role in the genesis of this condition. The fact that there are substantial differences in gene expression of pagetic osteoblasts, that persist after several weeks in culture, is highly suggestive of an intrinsic abnormality in this cell. The fact that some of these alterations reflect the in vivo findings (e.g., altered expressions of ALP and osteocalcin genes) indicates that these do not simply reflect artifacts from long-term culture. These findings do not prove that the osteoblast is the site of the primary abnormality in this condition, but they certainly require that this possibility be seriously considered. Hitherto, many workers have regarded the osteoblast as a bystander cell, responding to the mayhem created by rampantly overactive osteoclasts. These data raise the possibility that a pagetic lesion could arise from dysregulated gene expression, in particular of Dkk1 and IL-6, in a group of osteoblasts and that this results in the abnormalities in osteoclast number, morphology, and function that are characteristic of this condition. Further studies will be necessary to test this hypothesis.

Figure Figure 7.

Schema showing possible effects and interactions of the changes in osteoblast gene expression shown in this study. Solid lines denote production of a factor by that cell, and broken lines indicate a regulatory influence of a factor. The overproduction of IL-1 and IL-6 by the pagetic osteoblast will result in osteoclast proliferation, leading to further increases in IL-6 levels in the bone marrow microenvironment. Increased production of Dkk1 by the pagetic osteoblast will further increase IL-6 levels and reduce osteoblast proliferation. This combination of effects could account for the development of lytic lesions in early phase Paget's disease. Over time, both the excess of Dkk1 and of IL-6 will result in increased differentiation of osteoblasts, thus promoting mineralization. This could contribute to the development of sclerosis in longer-standing pagetic lesions.


The authors thank the patients who donated bone tissue. In addition, the authors thank the surgeons who collected the tissue (including Stuart McCowan, Anthony Hardy, Mark Clatworthy, Richard Lander, Garnet Tregonning, and Preston Moorcroft), Liam Williams from the School of Biological Sciences for assistance with microarrays, and the staff at Middlemore Hospital, Manukau Super Clinic, and Queen Elizabeth Hospital, Rotorua. The study was funded by the Health Research Council of New Zealand, Paget's Disease Charitable Trust Foundation of New Zealand, and The Paget Foundation, USA.