Osteosclerosis in advanced chronic idiopathic myelofibrosis is associated with endothelial overexpression of osteoprotegerin


Oliver Bock, MD, Institute of Pathology, Medizinische Hochschule Hannover, Carl-Neuberg-Strasse 1, 30625 Hannover, Germany. E-mail: bock.oliver@mh-hannover.de


Advanced chronic idiopathic myelofibrosis (IMF) with osteosclerosis and increase and thickening of bone trabeculae is typically contrasted by the absence or sparse presence of osteoclasts. Because osteoclast formation can be inhibited by osteoprotegerin (OPG) we investigated OPG expression in IMF with severe fibrosis and osteosclerosis, which expressed significantly higher (up to 71-fold) OPG mRNA levels when compared with prefibrotic cellular IMF and control cases. The receptor activator of nuclear factor kappaB ligand (RANKL), a positive regulator of osteoclast differentiation and putative antagonist of OPG was overexpressed by up to 34-fold exclusively in advanced IMF. Case-specific calculation of the RANKL/OPG ratio in advanced IMF showed a wide range without significant differences when compared with the prefibrotic IMF and non-neoplastic haematopoiesis. Immunohistochemical detection of OPG protein revealed strong labelling of endothelial cells within proliferating vessels in fibrotic IMF and heterogeneously labelled megakaryocytes, and fibroblasts. Osteosclerosis and impaired osteoclast function in IMF appears to be associated with upregulated endothelial OPG expression but concomitant reduction of the antagonist RANKL could not be demonstrated. We conclude that osteosclerosis in IMF is associated with increased endothelial OPG expression without concomitant RANKL downregulation.

Idiopathic myelofibrosis (IMF) belongs to the chronic myeloproliferative disorders (CMPD). Like the other CMPDs, it is caused by the clonal proliferation of a haematopoietic stem cell retaining the capacity to differentiate into all lineages (Tefferi, 2000). A feature that distinguishes IMF is the progression to bone marrow fibrosis and osteosclerosis whereby the proliferating fibroblasts are reactive and non-clonal in nature (Tefferi, 2000). In advanced stages with osteosclerosis there is an increase and thickening of bone trabeculae that characteristically lack rimming osteoblasts and osteoclasts (Ward & Block, 1971).

In normal bone homeostasis, osteoprotegerin (OPG) and the receptor activator of nuclear factor kappaB ligand (RANKL, also called TRANCE [tumour necrosis factor (TNF)-related, activation-induced cytokine)], participate in a cytokine axis that tightly controls the differentiation of osteoclasts from monocyte precursors. OPG, a member of the TNF-receptor family, acts as soluble decoy receptor for RANKL thereby limiting binding of RANKL to its functional receptor RANK (Hofbauer & Schoppet, 2004). RANKL is highly expressed in areas of trabecular bone remodelling and provides an important signal required for full osteoclast development, activation and survival (Hofbauer & Schoppet, 2004). Previous studies showed that, in patients with multiple myeloma and other diseases associated with bone destruction, an increase in RANKL expression along with a decrease in OPG expression triggered osteolysis by favouring osteoclast differentiation (Pearse et al, 2001; Grimaud et al, 2003). Accordingly, therapeutic studies using RANKL antagonists have been conducted in order to prove clinical benefit for patients with osteolysis and bone destruction (Sordillo & Pearse, 2003).

Recently, Chagraoui et al (2003a) demonstrated, in a murine model of IMF, that OPG induced inhibition of osteoclastogenesis might be responsible for the development of osteosclerosis by engrafting thrombopoietin-overexpressing haematopoietic cells into OPG-deficient recipients. In this animal model, stromal cells, but not haematopoietic cells, were identified as potential producers of OPG. Enhanced plasma levels of OPG were found in patients with manifest IMF (Wang et al, 2004). Expression of RANKL as well as the cellular source of enhanced OPG expression in IMF has not been studied to date. The aim of the present study was to (i) determine the expression level of RANKL and OPG mRNA in bone marrow cells derived from non-neoplastic haematopoiesis in comparison with IMF in the prefibrotic and advanced stage, (ii) demonstrate RANKL/OPG mRNA ratios in IMF and non-neoplastic haematopoiesis, and (iii) identify the cellular source of both RANKL and OPG protein in the bone marrow cells.

Materials and methods

Bone marrow study group

Formalin-fixed, paraffin-embedded bone marrow trephines with histopathologically proven IMF were retrieved from the bone marrow registry of the Institute of Pathology, Medizinische Hochschule Hannover. The study group (n = 102) comprised non-fibrotic, cellular IMF (n = 33), IMF with manifest fibrosis (n = 31), and 38 control cases showing no evidence of neoplastic disease. According to the World Health Organisation classification, IMF cases were thoroughly selected, re-evaluated and subdivided into two groups depending on the degree of myelofibrosis as essentially described (Buhr et al, 2003). In particular, subtyping of IMF cases depends on the type and amount of fibres (reticulin vs. collagen), the fibre spread, the appearance of sinusoidal sclerosis and intrasinusoidal haematopoiesis, and the existence of endophytic bone formation. Accordingly, cellular IMF cases displayed no (MF0) or only a minor deposit of reticulin fibres (MF1) whereas manifest IMF cases showed severe myelofibrosis with deposit of collagen, sclerotic sinusoids with sinusoidal haematopoiesis (MF3), and a remarkable osteosclerosis with bone apposition (MF4). Subsequently, two groups were assembled comprising IMF in the cellular stage (MF0, MF1) and IMF with severe myelofibrosis and osteosclerosis (MF3 and MF4). The summary of patients’ clinical characteristics are given in Table I.

Table I.  Description of the study group.
 No. of casesAge (years)Sex (female/male)Leucocyte count (109/l)Platelet count (109/l)Erythrocyte count (1012/l)Haemoglobin (g/dl)
  1. IMF, idiopathic myelofibrosis.

  2. Patients’ clinical data are briefly summarised with median values followed by ranges in parentheses.

Cellular IMF3368 (35–89)19/1410·8 (6·9–38·1)772 (183–1919)4·7 (2·7–6·3)12·2 (7·8–16·6)
IMF with manifest fibrosis/sclerosis3168 (49–86)13/189·5 (1·9–43·9)345 (48–935)3·8 (2·0–6·7)10·9 (5·1–15·8)

Real-time reverse transcription polymerase chain reaction and data evaluation

Total RNA was extracted from bone marrow sections as previously described (Bock et al, 2004), and 1 μg RNA, pretreated with RNase free (RNase) DNase (1 U/μg RNA, RQ1; Promega, Madison, WI, USA), was transcribed into the complementary DNA using 500 ng random hexamers (Amersham Pharmacia, Picattaway, CA, USA) and 200 U of SuperScript II RNase reverse transcriptase (Invitrogen, Karlsruhe, Germany) in a volume of 20 μl following the manufacturer's protocol. Negative controls were performed using water instead of reverse transcriptase.

Real-time reverse transcription polymerase chain reaction (RT-PCR) was performed on an ABI PRISM 7700 Sequence Detector (Applied Biosystems, Foster City, CA, USA). The sequences of PCR primers and TaqManTM probes (Roche Molecular Systems, Pleasanton, CA, USA) that amplify OPG, RANKL, the housekeeping genes β-glucuronidase (β-GUS) and heat-shock protein, HSP-70·1 are listed in Table II. The probes were purchased labelled with 6-carboxy-fluorescein (FAM) as the reporter dye and 6-carboxy-tetramethyl-rhodamine (TAMRA) as the fluorescent quencher. The real-time RT-PCR amplification was performed in a final reaction volume of 25 μl containing both primers (250 nmol/l), probe (150 nmol/l), 0·5 units of Platinum Taq Polymerase (Invitrogen, Karlsruhe, Germany), 200 μmol/l each of dATP, dCTP, dTTP and dGTP in 1x Platinum Taq reaction buffer and 4 μl cDNA. The reaction mixture was preheated at 95°C for 5 min, followed by 45 cycles at 95°C for 15 s and 60°C for 1 min.

Table II.  Sequences of primers and probes used in this study.
Primer and probesSequence (5′–3′)Reference/product
  1. OPG, osteoprotegerin; RANKL, receptor activator of nuclear factor kappaB ligand; β-GUS, β-glucuronidase; FAM, 6-carboxy-fluorescein; TAMRA, 6-carboxy-tetramethyl-rhodamine.

ForwardGCAGCGGCACATTGGACGenBank NM002546, 69 bp
ForwardCCGGTGGTGCAGTCGGGenBank NM005345, 62 bp

Amplification for OPG, RANKL, β-GUS, and HSP-70·1 could be demonstrated to be linear over a broad concentration range allowing relative quantification of OPG and RANKL in three-independent runs using the ΔΔCT – method as described (Livak & Schmittgen, 2001; Bock et al, 2004). RANKL/OPG ratios were calculated by using the respective case-specific relative RANKL and OPG mRNA expression level. To statistically analyse the gene expression of OPG and RANKL in cellular IMF, advanced IMF with osteosclerosis, and non-neoplastic haematopoiesis, the non-parametric Kruskal–Wallis test was applied.


Cellular protein expression was determined using tissue microarrays (TMA) as described (von Wasielewski et al, 2002; Bock et al, 2004). After histomorphological evaluation of bone marrow sections a representative area of haematopoiesis was marked on the paraffin block. Using a biopsy needle, tissue cores of c. 1·4-mm diameter were punched out of the marked area. Representative tissue cores of haematopoiesis from up to 60 bone marrow trephines were melted together in a checkerboard pattern into one tissue array block. Serial sections derived from tissue arrays were mounted on customary glass slides. For immunohistochemistry, tissue array slides were pretreated with protease XXIV (21·4 mg/50 ml aqua dest; Sigma-Aldrich, München, Germany) for 10 min at 37°C, followed by a tyramine amplification method according to standard protocols (von Wasielewski et al, 1997). The antibodies used were a monoclonal anti-human OPG/TNFRSF11B antibody [mouse-anti human monoclonal antibody (mAb), clone 69146·11, MAB805; R&D systems, Minneapolis, MN, USA], and a monoclonal anti-human TRANCE/TNFSF11 antibody (mouse-anti human mAb, clone 70525, MAB626; R&D systems). Both antibodies were applied at a dilution of 1:100. Specificity of the OPG antibody was tested in two bone marrow trephines from patients with Paget's disease, the RANKL antibody was tested in a lymph node showing reactive hyperplasia, according to the distributor's instructions. To ensure representation of the cellular origin of OPG as detected by TMA immunohistochemistry, corresponding whole bone marrow sections (n = 15) were stained accordingly.


Advanced IMF with fibrosis and osteosclerosis expressed significantly increased OPG mRNA levels when compared with the prefibrotic phase and non-neoplastic haematopoiesis

Bone marrow cells from patients with advanced IMF (n = 31) that progressed to fibrosis and osteosclerosis expressed significantly higher OPG mRNA levels (median 4·0, range 0·42–71·4, P < 0·001) when compared with IMF in the cellular, prefibrotic phase (median 1·1, range 0·13–5·1, n = 33). No differences could be found between IMF in the cellular phase and non-neoplastic control cases (median 0·9, range 0·14–6·8, n = 38) (Fig. 1).

Figure 1.

Bone marrow cells derived from idiopathic myelofibrosis (IMF) with evident fibrosis and osteosclerosis significantly overexpressed osteoprotegerin (OPG) mRNA by up to 71-fold (median 4·0) when compared with IMF in the cellular prefibrotic phase of the disease and non-neoplastic controls. Cellular IMF (median 1·1) and control group (median 0·9) exhibited nearly identical OPG gene expression levels. Note that the two highest OPG expressing cases (64·8- and 71·4-fold) were not included in this figure in order to better illustrate the expression level of the total study group.

Advanced IMF expressed significantly higher RANKL mRNA levels when compared with the prefibrotic phase of IMF and non-neoplastic haematopoiesis

Bone marrow cells in advanced IMF (n = 27) expressed significantly higher RANKL mRNA levels (median 5·0, range 1·17–34·6, P < 0·001) in comparison with the prefibrotic phase of the disease (median 0·9, range 0·2–4·7, n = 14), and non-neoplastic haematopoiesis (median 0·9, range 0·34–3·84, n = 20). RANKL mRNA expression by bone marrow cells in prefibrotic IMF did not significantly differ from non-neoplastic haematopoiesis (Fig. 2).

Figure 2.

Advanced idiopathic myelofibrosis (IMF) also showed significantly higher receptor activator of nuclear factor kappaB ligand (RANKL) mRNA expression by up to 34-fold (median 5·0) when compared with cellular prefibrotic IMF (median 0·9) and the control group (median 0·9). Note that the highest RANKL expressing outlier (34·6-fold) was not illustrated in this figure in order to better demonstrate RANKL mRNA level in the total study group.

Increased ratio for RANKL/OPG in prefibrotic and advanced IMF when compared with non-neoplastic haematopoiesis

The case-specific ratio for RANKL/OPG was increased in prefibrotic IMF (n = 14, median = 1·6, range 0·27–5·5) and advanced IMF (n = 20, median 1·7, range 0·09–21·9) when compared with non-neoplastic haematopoiesis (n = 18, median 0·9, range 0·26–3·17) but this did not reach statistical significance (Fig. 3).

Figure 3.

The ratio of receptor activator of nuclear factor kappaB ligand (RANKL) vs. osteoprotegerin (OPG) could be demonstrated to be higher in prefibrotic (median 1·6) and advanced idiopathic myelofibrosis (median 1·7) when compared with the control group (median 0·9) but this was not statistically significant. Note that RANKL/OPG ratios are illustrated in a log manner.

Case-specific RANKL and OPG mRNA levels did not reveal a definite shift towards an aberrant RANKL/OPG ratio in advanced IMF

The 20 cases available for case-specific analysis of RANKL and OPG mRNA expression in advanced IMF showed RANKL/OPG ratios in the range of 0·09–21·9 (median 1·7). In order to allow a comprehensive interpretation of the underlying data, respective RANKL mRNA level, OPG mRNA level, and RANKL/OPG ratio were displayed case-by-case in a point plot (Fig. 4). Corresponding levels were connected by solid lines. Even though nine of 20 cases showed a ratio <1·0 the remaining 11 cases clearly showed increased ratios representing higher RANKL than OPG expression. The absolute values underlying the RANKL/OPG ratios in advanced IMF are listed in Table III.

Figure 4.

Case-specific comparison of receptor activator of nuclear factor kappaB ligand and osteoprotegerin ratios in advanced idiopathic myelofibrosis (n = 20) showed variable patterns with ratios ranging from 0·09 up to 21·9 (median 1·7). Corresponding case-specific levels were connected by solid lines. Absolute values in Table III comprehensively completed the data in this figure.

Table III.  RANKL/OPG ratio based on analysis of case-specific relative RANKL and OPG mRNA expression in advanced idiopathic myelofibrosis.
Case no.RANKL factorOPG factorRANKL/OPG ratio
  1. OPG, osteoprotegerin; RANKL, receptor activator of nuclear factor kappaB ligand.


Endothelial cells in advanced IMF with fibrosis and osteosclerosis showed prominent labelling for OPG protein

As shown in Fig. 5A–C the strongest labelling was observed in endothelial cells of extended and proliferating vessels. Besides endothelial cells, fibroblasts in areas with manifest fibrosis were also occasionally labelled Co-OPG. Megakaryocytes heterogeneously exhibited a rather dot-like and faint labelling (Fig. 5A, insert) or no labelling at all (Fig. 5B). In reactive states and normal bone marrow, endothelial cells also displayed OPG labelling but to a lesser degree corresponding to the lower level of OPG mRNA.

Figure 5.

Osteoprotegerin (OPG) expression could be predominantly demonstrated in endothelial cells in idiopathic myelofibrosis (IMF) (A–C). In the abnormal extended sinus lined by intensely OPG-expressing endothelial cells, haematopoietic precursor cells can be perceived with some positively stained megakaryocytes in the vicinity (arrows), (A, 200×). Megakaryocytic OPG expression (A, insert, 630×) could not be observed in all cases as shown in (B) (200×, arrows indicate negative megakaryocytes); here, almost exclusively endothelial cells display labelling. Areas with fully developed fibrosis and osteosclerosis exhibited predominant endothelial OPG expression, (C, 200×). Rarely, OPG-labelled osteoblasts (solid arrows) and osteoclasts (dotted arrow) lining the trabecular bone (asterisk, dislocated because of preparation) could be demonstrated in advanced IMF (C, insert, 630×).

Immunohistochemistry for RANKL failed to detect relevant cellular sources in the bone marrow in IMF and control haematopoiesis other than plasma cells, which were also positively labelled in the, likewise and simultaneously stained, reactive lymph nodes (not shown).


Our study revealed increased OPG mRNA and protein expression in advanced IMF, which is in close agreement with a recent study demonstrating elevated OPG plasma levels in patients with IMF in comparison with normal volunteers (Wang et al, 2004). Initially identified as a novel member of the TNF receptor superfamily (Simonet et al, 1997), OPG represents a key molecule in the regulation of bone formation and turnover through inhibition of osteoclast differentiation. Recent data showed that OPG, upregulated in vivo in mice overexpressing TPO and TGFβ-1, led to severe osteosclerosis (Chagraoui et al, 2003a). In contrast, mice deficient for OPG developed extensive osteoporosis (Bucay et al, 1998).

Studies in a murine model of IMF combined with induced OPG deficiency suggested that not haematopoietic but stromal cells provide the relevant source for OPG production although megakaryocytes in mice have the potential to release OPG (Chagraoui et al, 2003b). The cellular origin of OPG in patients with IMF has not been elucidated. Therefore immunohistochemistry was performed to precisely delineate OPG overexpressing cells in the bone marrow. As shown in Fig. 5 the strongest labelling was observed in endothelial cells of extended and proliferating vessels. Besides endothelial cells, fibroblasts in areas with manifest fibrosis were also occasionally labelled for OPG. Megakaryocytes heterogeneously exhibited a rather dot-like and faint labelling (Fig. 5A and insert). Osteoblasts and osteoclasts in advanced IMF were infrequently demonstrated. Positive OPG labelling was detectable in both cell types (Fig. 2C, insert). With regard to endothelial cell labelling, reactive states and normal bone marrow endothelial cells also displayed OPG, but to a lesser degree, corresponding to the lower OPG mRNA level.

Endothelial cells have been identified as potential producers of OPG (Collin-Osdoby et al, 2001). Besides a role in bone homeostasis, in that endothelial cells may be involved (Simonet et al, 1997), OPG has also been suggested to be important for endothelial proliferation and survival (Malyankar et al, 2000). Neoangiogenesis is a prominent feature in progressed IMF (Mesa et al, 2003). Exaggerated production of OPG therefore might have a dual pathogenetic function in progressed IMF. Besides impairment of osteoclast formation, thereby inducing osteosclerosis, OPG might contribute to endothelial growth and neoangiogenesis.

Overexpression of OPG was not accompanied by a decreased expression of its putative antagonist RANKL. Indeed, RANKL mRNA was also overexpressed in advanced IMF. Case-specific RANKL/OPG ratios in advanced IMF revealed clearly increased and decreased, as well as balanced, subsets. Also, prefibrotic IMF and control haematopoiesis showed an almost similar pattern of ratios even though outliers could not be demonstrated to the same extent. Studies on osteolytic bone lesions suggested the principle of a shift towards an increase of the RANKL/OPG ratio (Grimaud et al, 2003). Considering this principle, the ratio should decrease inversely in a disease that displays bone apposition, such as IMF. Even though nearly half of the cases showed a RANKL/OPG ratio below 1, therefore favouring OPG, a straightforward decrease was not demonstrable. With regard to comparison of RANKL/OPG ratios in the total study group, both IMF in the prefibrotic phase as well as advanced stages in the median exhibited higher ratios when compared with non-neoplastic haematopoiesis.

As demonstrated for multiple myeloma the imbalance of increased RANKL and decreased OPG consecutively led to osteolysis and progression of the disease (Pearse et al, 2001). Accordingly, an opposite ratio that favours OPG might explain decreased bone turnover and bone apposition in advanced IMF. As shown, the RANKL/OPG ratios were rather higher for prefibrotic and advanced IMF when compared with the control cases even though this did not reach statistical significance. Indeed, variable intraindividual patterns for both factors under investigation led to a remarkable range of ratios. In IMF, deregulated bone remodelling therefore could not solely be explained by opposite expression patterns of RANKL and OPG. Both factors are generally synthesised and secreted by identical cell types, including those derived from stroma, bone matrix, and diverse haematopoietic progenitors. Recently, the important role of megakaryocytes in bone remodelling has been elegantly demonstrated in a co-culture system of osteoblasts and megakaryocytes (Bord et al, 2005). In this model, megakaryocytes increased type-1 collagen and OPG expression by osteoblasts along with a remarkable decrease of RANKL (Bord et al, 2005). The same group had previously demonstrated that megakaryocyte expression of OPG and RANKL is inversely modulated by oestrogens (Bord et al, 2004). It should be noted that in this study, the megakaryocytes under investigation were non-neoplastic. The complexity of mechanisms involved in bone remodelling has been recently demonstrated through the essential role of the dendritic cell-specific transmembrane protein (DC-STAMP) in osteoclastogenesis (Kukita et al, 2004). Indeed, RANKL induced DC-STAMP expression in osteoclast precursors and differentiation whereas small interfering RNAs and specific antibodies markedly suppressed formation of osteoclasts. In addition, the crucial role of TNF receptor-associated factors (TRAF), such as TRAF2 and TRAF6, in RANKL-induced osteoclast differentiation could be demonstrated (Kanazawa & Kudo, 2005; Lee et al, 2005). Accordingly, RANKL per se no longer appears to be the essential factor for osteoclastogenesis. IMF represents a neoplastic stem cell disease with atypia occurring in many cell lines. Hence, disturbance of the signal transduction in the monocyte lineage affecting the tight control of cellular differentiation cannot be excluded with certainty.

We conclude that OPG appears to be involved in reduced osteoclastogenesis in advanced IMF through overexpression by endothelial cells. In addition, an important role for OPG in endothelial cell survival and neoangiogenesis in advanced IMF must be considered. In contrast to other diseases showing lytic bone lesions because of RANKL upregulation, such as multiple myeloma, the RANKL/OPG ratio was not demonstrated to be notably shifted in IMF that showed decreased bone turnover. Future studies of other factors involved in bone remodelling downstream of RANKL, such as DC-STAMP, will shed light on the complex regulation of bone apposition and turnover in IMF.


The authors are indebted to Ms Henriette Bruchhardt and Ms Christina Koop for their excellent technical assistance.

This work was supported by grants from Deutsche Krebshilfe, Dr Mildred Scheel Stiftung 10-2191 to O.B. and H.K.