Prostate Osteoblast-Like Cells: A Reliable Prognostic Marker of Bone Metastasis in Prostate Cancer Patients

The main aim of this study was to investigate the putative association among the presence of prostate cancer cells, defined as prostate osteoblast-like cells (POLCs), and showing the expression of typical morphological and molecular characteristics of osteoblasts, the development of bone metastasis within 5 years of diagnosis, and the uptake of 18F-choline evaluated by PET/CT analysis. To this end, prostate biopsies (n = 110) were collected comprising 44 benign lesions and 66 malignant lesions. Malignant lesions were further subdivided into two groups: biopsies from patients that had clinical evidence of bone metastasis (BM+, n = 23) and biopsies from patients that did not have clinical evidence of bone metastasis within 5 years (BM−, n = 43). Paraffin serial sections were obtained from each specimen to perform histological classifications and immunohistochemical (IHC) analysis. Small fragments of tissue were used to perform ultrastructural and microanalytical investigations. IHC demonstrated the expression of markers of epithelial-to-mesenchymal transition (VIM), bone mineralization, and osteoblastic differentiation (BMP-2, PTX-3, RUNX2, RANKL, and VDR) in prostate lesions characterized by the presence of calcium-phosphate microcalcifications and high metastatic potential. Ultrastructural studies revealed the presence of prostate cancer cells with osteoblast phenotype close to microcalcifications. Noteworthy, PET/CT analysis showed higher uptake of 18F-choline in BM+ lesions with high positivity (≥300/500 cells) for RUNX2 and/or RANKL immunostaining. Although these data require further investigations about the molecular mechanisms of POLCs generation and role in bone metastasis, our study can open new and interesting prospective in the management of prostate cancer patients. The presence of POLCs along with prostate microcalcifications may become negative prognostic markers of the occurrence of bone metastases.


Introduction
Metastasis to bone is a common feature in advanced prostate cancer (PCa) patients. PCa is one of the most frequent cancer in men and represents a great public health problem, with a total of 265,000 new diagnosis every year in both Europe and United States of America [1]. Frequently, prostate cancer patients show bone osteoblastic metastatic lesions at diagnosis [1,2]. e evidence that prostate cancer cells in patients enter the circulation in large numbers but still preferentially colonise to the bone has a number of implications. Prostate cancer cells have the ability to adhere at the main proteins of the extracellular matrix or at the bone marrow [2]. Also, the colonisation bone by prostate cancer cells suggests that metastatic cells have morphological and/ or molecular characteristics that make them capable to survive in the bone [2][3][4]. e type of bone metastases formed in prostate cancer is a reflection of the local interaction between tumour cells and the bone remodeling system-a complex mechanism which remains to be fully characterized. Bone metastases in prostate cancer are most often osteoblastic (involving the deposition of newly formed bone), but can also be osteolytic (characterized by destruction of normal bone) or mixed. e development of either osteolytic or osteoblastic lesions results from functional interactions between tumour cells and osteoclasts or osteoblasts, respectively [5]. However, the mechanisms responsible for the formation of prostate cancer metastasis to bone are complex and certainly involve both osteoclasts and osteoblasts activity [6]. In this context, the binary classification between osteoblastic and osteolytic lesions represents two extremes of a continuum which involves dysregulation of the normal bone remodeling process and which is yet to be fully understood. A detailed characterization of the osteoblastic-osteolytic spectrum and of premetastatic tumour cells could therefore pave the way for both the identification of early markers for bone metastasis and of novel drug targets to improve quality of life of patients with advanced prostate cancer.
As concerns the origin of metastatic cells, different hypotheses have been formulated. For a long time, the main theories of the formation of bone metastases contemplated the occurrence of specific genetics change in primary cancer cells that thus acquired the ability to spread to and thrive in distant organs [7,8]. In this context, the epithelial-tomesenchymal transition (EMT) could represent the key biological process adopted by epithelial cancer cells to promote tissue dissemination [9]. On note, in our recent study, we demonstrated a putative association between the occurrence of EMTand the development of breast cancer cells showing an osteoblast-like cells phenotype in lesions with microcalcifications [10,11]. In addition, we observed that the presence of breast osteoblast-like cells (BOLCs) in breast infiltrating cancer was associated with the formation of bone metastatic lesions within 5 years from diagnosis [12,13].
e main aim of this study was to investigate the putative association among the presence of prostate cancer cells, defined as prostate osteoblast-like cells (POLCs), and showing the expression of typical morphological and molecular characteristics of osteoblasts, the development of bone metastasis within 5years of diagnosis, and the uptake of 18 F-choline evaluated by PET/CT analysis.

Collection of Prostate Samples.
In this study, we enrolled 110 patients undergoing prostate biopsies. From this selection, we collected prostate biopsies from each patient and, when available, data of PET/CT analysis. e study was approved by Institutional Ethical Committee of the "Policlinico Tor Vergata." Experimental procedures here reported were performed in agreement with the e Code of Ethics of the World Medical Association (Declaration of Helsinki).
All patients have signed the informed consent prior to surgical procedures. From each sample, paraffin serial sections were used for both histological and immunohistochemical investigation. Also, 1 mm 3 of tissue were studied by transmission electron microscopy and microanalytical analysis. Exclusion criteria were history of previously or concomitant other neoplastic diseases, autoimmune diseases, viral chronic infections (HBV, HCV, and HIV), and any antitumoral treatment received before biopsy.

Immunohistochemistry.
To study the immunophenotypical profile of prostate metastatic cells, we performed immunohistochemical reactions to investigate the expression of the following biomarkers: vimentin (EMT), BMP-2, PTX-3, RUNX2, RANKL, and VDR (mineralization process). For antigen retrieval, 3 μm thick paraffin sections were treated with citrate pH 6.0 or EDTA citrate pH 7.8 buffers (95°C for 30 min). en, primary antibodies listed in Table 1 were incubated for 1 hour at room temperature. HRP-DAB Detection Kit (UCS Diagnostic, Rome, Italy) was used to reveal the reaction of primary antibodies with their specific target. Immunohistochemical signal was assessed independently by two investigators by counting the number of positive cancer cells (out of a total of 500 in randomly selected regions).

18F-Choline PET/CT Analysis.
Among patients enrolled in the study, 11 were subjected to 18 F-methylcholine ( 18 Fcholine) PET/CT analysis. Results of 18F-choline PET/CT were collected to verify a possible correlation between 18 Fcholine uptake in prostate tumours and the presence of POLCs. 18 F-choline PET/CT analysis was performed as previously described [19,20]. From each patient, standardized uptake value (SUV) max and SUV average were recorded.

Histological Classification.
Prostate biopsies were classified in 44 benign lesions (BL) and 66 malignant lesions according to EAU-ESTRO-SIOG Guidelines 2017 [23]. We subdivided the malignant lesions in those (BM+, n � 23) taken from patients with clinical evidence of bone metastasis and those (BM−, n � 43) from patients without clinical evidence of bone metastasis after 5 years from diagnosis. From radiological point of view, all metastatic sites showed typical characteristics of osteoblastic lesions. Calcifications were present in 38 out of 110 prostate biopsies. In particular, we observed psammoma bodies in 32% of BL, 87% of BM+, and 79% of BM−. Patient baseline characteristics are reported in Table 2.

EMT Characterization.
Immunohistochemical analysis of vimentin expression was performed in order to evaluate the number of prostate cells that acquire mesenchymal phenotype (Figure 1(a)-1(c)). As shown in a recent study [24] (Figure 2(g)). In particular, the signal in BM+ appeared very intense both in nucleus and in cytoplasm ( Figure 2(h)), while it was less intense and mainly nuclear in BM− (Figure 2(i)).

Expression of POLC Biomarkers in Gene Expression
Datasets. We examined expression of the EMT and bone markers studied by IHC in public datasets comprising gene expression profiling data from patients with primary tumours or metastatic castration-resistant prostate cancer (CRPC). Individual gene comparisons did not show a univocal behaviour (Figure 3(a)). Only VDR expression was  significantly upregulated in CRPC compared to primary tumours. Interestingly, however, the small set of genes was able to discriminate most of the primary tumours from metastatic CRPCs in unsupervised clustering (Figure 3(b)). Furthermore, analysis of a second dataset with annotated metastatic sites showed that the gene set expression was remarkably higher in tumour specimens taken from bone metastases compared to primary tumours and other metastatic sites (Figure 3(c)). ( Figure 4(d)). Finally, significant differences in VDR expression were observed (Micro+ 192.00 ± 18.02 vs Micro− 109.30 ± 19.14; p � 0.001) (Figure 4(e)).

Ultrastructural Characterization of Prostate Cancer Cells.
TEM analysis allowed us to characterize ultrastructure of prostate cells in malignant lesions. Specifically, we observed both cuboidal and large spindle-shaped cells with abundant clear cytoplasm in BM+ ( Figure 5(a)). Moreover, in these lesions, we identified several calcifications and prostate cancer cells with morphological appearance of osteoblasts containing cytoplasmic electrondense granules made of HA ( Figure 5(b)). In addition, EDX microanalysis demonstrated that all calcifications here detected were made of calciumphosphate (hydroxyapatite) (Figure 5(b)).

18F-Choline PET/CT Analysis.
We collected PET/CT data of 11 patients: 5 BM+ and 6 BM− ( Figure 5(c)). Despite the low number of patients, we found significant differences between both SUV max and SUV average between BM+ and BM− (

Discussion
Prostate metastasis to the bone more often results in osteoblastic lesions, though it is known that prostate bone metastases can display both blastic and lytic characteristics during the early phases of their formation [25]. In addition, there is evidence that during the early phases of osteoblastic metastases formation, it is possible to observe osteolytic lesions, suggesting an overall increase of bone remodeling at these sites. e pathophysiology of bone metastases is frequently explained by the theory of the vicious cycle proposed for the first time by Mundy and Guise [26]. According to this theory, cancer cells resident in bone cause bone destruction because they are capable to stimulate osteoclast activity. In return, cancer cells receive positive feedbacks from humoral factors released by the bone microenvironment during bone destruction and remodeling [27]. Indeed, it is widely accepted that the bone microenvironment is crucial to the success of cancer cells in bone.
In a recent study, we described for the first time the characteristics of prostate cells involved in the production of prostate calcifications demonstrating their similarity with osteoblasts [24]. In addition, our research group described the presence of osteoblast-like cells in breast cancer (BOLCs) showing a correlation between the appearance of BOLCs in primary lesions and development of bone metastases. Based on these studies, the main aim of this study was to investigate the possible correlation between the presence of prostate cancer cells showing expression of typical morphological and molecular markers of osteoblasts and the development of bone metastasis in prostate cancer patients within 5 years from diagnosis of primary lesion. To this end, we collected 110 prostate biopsies (44 benign and 66 malignant lesions). Malignant lesions were subdivided in biopsies from patients with clinical evidence of bone metastasis (BM+, n � 23) and those from patients without clinical evidence of bone metastasis (BM−, n � 43).
As already reported by Scimeca et al., we found a significant correlation between vimentin expression, one of the most important markers of mesenchymal cells [14], and the presence of prostate osteoblast-like cells (POLCs). Specifically, our data showed a significant increase of positive cells   WA25  WA46  WA47  WA19  WA11  WA30  WA18  WA35  T31  T37  WA32  WA42  WA7  WA53  WA28  WA10  WA14  T65  T45  WA54  WA40  WA6  WA24  WA13  WA3  WA37  WA20  WA4  WA55  WA31  WA22  T54  WA23  T59  T51  T5  WA39  T47  T64  WA5  WA16  WA29  T1  T49  T42  T20  T62  T32  T63  T41  T44  T85  T43  T75  T68  T82  T52  T83  T70  T73  T12  T11  T7  T60  T56  T67  T57  T27  T55  T17  T19  T39  T46  T26  WA26  T69  WA33  T24  T21  T29  T25  T40  T50  T8  T61  T53  T10  T48  T3  T66  T58  T6   in prostate cancer of BM+ group as compared with BM−. In addition, we proved that primary prostate cancer lesions of BM+ patients were characterized by the expression of osteogenic molecules able to induce osteoblast differentiation and to increase osteoblast function such as mineralization. Among them, BMP-2 is a potent inducer of bone formation through the stimulation of osteoblast differentiation. BMP-2 exerts this effect via two types of serine/threonine kinase receptors: BMP-2 binds the type II receptor, which subsequently activates the type I receptor by a direct association [28]. Our results showed an increase of BMP-2 expression in prostate malignant lesions. Conversely, the absence of significant differences of BMP-2 expression between BM+ and BM− suggests that it could be involved in the early phases of cancer transformation rather than during metastatic process. In support of this, several studies demonstrated the ability of BMP-2 to induce malignant transformation of epithelial tissues [29][30][31]. However, we also demonstrated the association between BMP-2 expression and the presence of prostate calcifications, regardless of the lesion type. us, BMP-2, in association to EMT phenomenon, can participate to induce mesenchymal-like cells to acquire osteoblast phenotype. As concerns PTX-3, PTX-3 is a multifunctional glycoprotein produced by a variety of cells [32,33]; our results displayed a significant correlation between the presence of PTX-3-positive prostate cells and bone metastasis formation. Also, it is important to emphasize that BM− group showed the same number of PTX-3-positive cells of BL, suggesting that the presence of PTX-3-positive cells could represent a reliable predictive element for the development of bone metastasis from prostate cancer. ese data are in line with recent studies that demonstrated the involvement of PTX-3 in osteoblast proliferation, differentiation and function [34][35][36], and bone metastasis from breast cancer formation.
To further characterize the phenotype of POLCs, we investigate the expression of the main markers of osteoblasts, RUNX2, RANKL, and VDR. RUNX2 is the first transcription factor required for the determination of the osteoblast lineage [37]. In particular, RUNX2 is detected first in preosteoblasts and its expression is upregulated during the early phases of osteoblast differentiation. In line with this, our results displayed an increase of RUNX2-positive prostate cells in malignant lesions respect to BL, but no difference was observed between BM+ and BM−. erefore, the acquisition of RUNX2 expression by prostate cells seems to be linked to cancer transformation rather than to metastatic process.
In agreement with the physiological role of RUNX2 in osteoblast function [38], we did not observe an increase of RUNX2-positive cells in Micro+ with respect to Micro− lesions. Indeed, mature osteoblasts lose the expression of RUNX2 during the mineralization phase of bone formation. Conversely, analysis of RANKL and VDR showed a putative correlation among the presence of RANKL and/or VDR positive prostate cancer cells, bone metastasis formation, and microcalcifications. As regards the formation of bone metastasis, the presence of RANKL-positive prostate cancer cells can trigger osteoclast activity by binding to the osteoclast receptor RANK [39]. Indeed, RANKL is a type II membrane protein expressed by osteoblasts that is able to induce osteoclasts proliferation and function. In addition, at the primary lesion site, RANKL expression can reflect the presence of cells   analysis of this set of genes in patients with primary and metastatic prostate cancer further showed that deregulated expression of these markers of EMT, bone mineralization, and osteoblastic differentiation occurred preferentially in the setting of metastatic disease and particularly at metastases in bone, further supporting their relevance as adverse prognostic markers.
It is important to highlight that in this study, POLCs were also characterized from the ultrastructural point of view. In particular, we observed the presence of cytoplasmic vesicles containing HA granules in prostate cancer cells showing osteoblast phenotype (POLCs) [24,[40][41][42].
Of note, although preliminary, our data showed a significant correlation between the uptake of 18F-choline PET/ CT and the presence of POLCs in prostate cancer tissues. If confirmed in a larger patient cohort, this evidence could provide the scientific rationale for the development of algorithms able to predict the metastatic potential of primary prostate cancer lesions by 18F-choline PET/CT analysis [43].
is study proposes a new cell type generated by a process of cell transdifferentiation and related to formation of bone metastasis: the POLCs. Although our data require further investigations about the molecular mechanisms of both POLCs generation and metastasization to the bone, this study opens new and interesting prospective for the management of prostate cancer patients. e presence of POLCs could become prognostic markers for occurrence of bone metastatic disease.

Conclusion
e clinical course of metastatic bone disease in prostate cancers is often long, with patients experiencing sequential skeletal complications over a period of several years. ese include bone pain, fractures, hypercalcemia, and spinal cord compression, all of which may profoundly impair patient's quality of life. In addition, once prostate tumour cells are engrafted in the skeleton, curative therapy is no longer possible and palliative treatment becomes the only option. us, the identification of early markers of bone metastasis and especially the characterization of the cells involved in the metastatic process can lay the foundation for the identification of new tools for monitoring, prevention, or cure of bone metastatic diseases and providing support to the physicians in the management of prostate patients. In this context, positron emission tomography (PET)/computed tomography (CT) has emerged as a significant and promising staging modality for primary, recurrent, and metastatic prostate cancer. Much more important, the identification of highly sensitive and specific radiotracers can implement the therapeutic/diagnostic perspectives for prostate cancer patients "opening the way" for the development of new theranostic approaches. PSMA PET/ CT ligands labelled with 18 F and 68 Ga have certainly revolutionized the management of metastatic prostate cancer selecting patients who may benefit from targeted systemic radionuclide therapy. In a nuclear oncology theranostic design, 68 Ga-PSMA already constitutes the diagnostic positronemitting of beta − emitter Lutetium-177 PSMA ( 177 Lu-PSMA) [44] and alpha-emitter Actinium-225 PSMA ( 225 Ac-PSMA) [45]. Finally, the results reported here about the phenotypic characterization of POLCs could provide a scientific rationale for the development of theranostic anti-POLC radiomolecules for the cure and prevention of prostate cancer bone metastasis.

Data Availability
e data used to support the findings of this study are included within the article. Gene expression data from two studies in prostate cancer patients (reference [21] and [22]) were retrieved from the cBioPortal platform. Expression of the selected genes was compared between primary tumours and metastatic CRPC and, for the second dataset, among primary and different metastatic sites.

Conflicts of Interest
e authors declare that there are no potential conflicts of interest relating to the manuscript.