In vitro and in vivo approaches to study osteocyte biology

Abstract

Osteocytes, the most abundant cell population of the bone lineage, have been a major focus in the bone research field in recent years. This population of cells that resides within mineralized matrix is now thought to be the mechanosensory cell in bone and plays major roles in regulation of bone formation and resorption. Studies of osteocytes had been impaired by their location, resulting in numerous attempts to isolate primary osteocytes and to generate cell lines representative of the osteocytic phenotype. Progress has been achieved in recent years by utilizing in vivo genetic technology and generation of osteocyte directed transgenic and gene deficiency mouse models.

We will provide an overview of the current in vitro and in vivo models utilized to study osteocyte biology. We discuss generation of osteocyte-like cell lines and isolation of primary osteocytes and summarize studies that have utilized these cellular models to understand the functional role of osteocytes. Approaches that attempt to selectively identify and isolate osteocytes using fluorescent protein reporters driven by regulatory elements of genes that are highly expressed in osteocytes will be discussed.

In addition, recent in vivo studies utilizing overexpression or conditional deletion of various genes using dentin matrix protein (Dmp1) directed Cre recombinase are outlined. In conclusion, evaluation of the benefits and deficiencies of currently used cell lines/genetic models in understanding osteocyte biology underlines the current progress in this field. The future efforts will be directed towards developing novel in vitro and in vivo models that would additionally facilitate understanding the multiple roles of osteocytes.

Osteocytes represent more than 95% of the cellular component of mature adult bone [1–3]. They are stellate shaped cells enclosed within the skeletal lacuno-canalicular network. Morphologically they are characterized by numerous elongated cell processes and dendrites extending into channels in the matrix called canaliculi (~250–300 nm in diameter) [4]. The shape of embedded osteocytes is dependent on their location with cells present in trabecular bone exhibiting rounded morphology whereas osteocytes from cortical bone are elongated [5]. These cells display polarity in terms of the distribution of their cell processes with the majority coming from the cell membrane facing the bone surface [6]. Osteocytes are connected, via gap junctions, to neighboring osteocytes, to cells on the bone surface such as lining cells, osteoclasts and osteoblasts [7] and to pericytes of capillaries, which supply nutrients and oxygen to osteocytes and other bone cells [8]. Furthermore, Kamioka et al., suggest the existence of a direct signaling system between the osteocytes and the bone marrow compartment without involvement of the osteoblast/lining cells [9]. Dynamic imaging studies showed that the dendritic connections between cells in the bone matrix, and into marrow spaces, appear to be able to extend and retract [10, 11]. Osteocytes also appear to show undulating motion of their cell bodies within their lacunae, suggesting that they are not inactive cells, as previously thought.

During the transition from osteoblast to osteocyte, the cell undergoes a dramatic transformation that requires extensive restructuring of the cytoskeletal and intracellular machinery. Cells at the early stages of osteocyte differentiation are larger than mature osteocytes and have numerous ribosomes, a well-developed endoplasmic reticulum and a wide Golgi complex [3]. Once the osteoid mineralizes, they decrease protein synthesis and secretion [12]. Moreover, osteocytes lose the apical and basolateral plasma membrane polarization normally seen in osteoblasts [13]. The study of different markers shows that osteocyte differentiation is accompanied by the progressive reduction of numerous bone markers (bone sialoprotein, collagen type I, alkaline phosphatase (ALP), osteocalcin, Runx2), the maintenance of some others (osteopontin and E11/gp38 antigen), and the appearance of new markers (dentin matrix protein 1 (DMP1), CD44, matrix extracellular phophoglycoprotein (MEPE), phosphate-regulating gene with homologies to endopeptidases on the X chromosome (PHEX), sclerostin (SOST) and fibroblast growth factor 23 (FGF23)) [3, 12, 14]. Osteocytes have proved difficult to isolate and study due to their localization within the mineralized matrix, and the fact that they are terminally differentiated post-mitotic cells. This review will discuss in vitro and in vivo techniques and tools that have enabled great advances in our understanding of osteocyte biology in recent years.

Primary culture of osteocytic cells

Over the last 20 years a number of methods have been developed to study osteocytes in vitro and this has greatly expanded our knowledge on the autocrine and paracrine functions of these inaccessible cells. Van der Plas and Nijweide [15] were the first to report a method to isolate and culture osteocytes from embryonic chicken calvaria. After releasing cells by serial collagenase/EDTA digestion, they purified osteocytes by immunomagnetic separation with a monoclonal antibody MAb OB 7.3 later shown to target Phex [16]. Cells isolated with this technique are post-mitotic and show stellate morphology with extensive processes. This method generates a relatively pure culture of osteoid osteocytes (as the calvaria is hypomineralized at the age of extraction) that produce low levels of ALP and high osteocalcin [17, 18]. However, any contaminating fibroblastic cells will overgrow the mitotically inactive osteocytes making these culture suitable only for relatively short term experiments, and cell yield using this method is low [17].

Other investigators have used different fractions of cells digested from newborn rat calvaria, or outgrowth cells from predigested tissue to isolate osteocytes based on their morphology when cultured on matrigel or type I collagen-coated surfaces [19, 20]. A similar methodology has also been employed using human alveolar bone and neonatal rodent long bones [13, 21]. The osteocyte-enriched fractions show lower ALP activity and higher osteocalcin expression than osteoblast-enriched fractions, however the osteocyte-like cultures show proliferation after a week in culture suggesting the presence of contaminating proliferative cells [19]. More recently a method has been reported for culturing long bone osteocytes from adult mice which should enable in vitro investigation of age-related changes in osteocyte function as well as characterization of isolated osteocytes from various genetically modified mice [22].

Osteocyte-like cell lines

Development of osteocyte-like cell lines has greatly aided investigations of osteocyte biology. A summary of the sources and phenotypic characteristics of cell lines that have been reported is shown in Table 1. The most widely used cell lines, MLO-Y4 and MLO-A5, were developed by Lynda Bonewald’s group [23, 24]. These are immortalized cells and undergo proliferation which is not a characteristic of osteocytes, however they show stellate morphology and expression of selected osteocyte marker genes. MLO-Y4 cells appear to have an osteocytic phenotype, with low ALP, high osteocalcin, E11 and connexin 43 expression, however, they do not mineralize or produce Sost [25]. These cells have proved useful for studies of modulation of osteocyte apoptosis and cell death [26–33], effects of fluid flow and other mechanical stimuli [30, 34–36], and the function and regulation of gap junctions and hemichannels [37–39]. MLO-Y4 cells also support osteoclastogenesis in the absence of osteotropic factors [40–42] and modulate the activity of osteoblasts [43], mesenchymal progenitor cells [44], and endothelial cells [45]. MLO-A5 cells are considered to be late osteoblastic/preosteocytic cells, and form mineralized sheets very rapidly in culture, even in the absence of added ascorbate and β-glycerophosphate which are usually added to induce mineralization [23]. They have therefore been useful in studies of osteoblast to osteocyte transition, mineralization, and the effects of mechanical stimuli on these processes [11, 46–48].

Table 1.

Osteocyte cell lines

Cell line	Source	Method of immortalization	Proliferation	Mineralization	Marker gene expressiona							Ref.
					ALP	OC	E11	Cx43	Dmp1	Sost	FGF2 3
MLO-Y4	14 d.o mouse long bone	OC-SV40 TAgb	Yes	No	+/−	++	+++	+++	++	−	+	[24]
MLO-A5	14 d.o mouse long bone	OC-SV40 TAgb	Yes	Yes	+++	+++	++	++	++	+	+	[23]
HOB-01-C1	Human trabecular bone	Adenoviral SV40 TsA 209c	Nod	Yes	+/−	++	ND	ND	ND	ND	ND	[49]
OC1/14/59	E18.5 mouse calvaria, PTH1R−/−	Immortomousee	Nod	Yes	+/−	+++	ND	+++	ND	ND	ND	[50]
IDG-SW3	3 m.o mouse long bone, Dmp-GFP+	Immortomousee	Nod	Yes	++	ND	+++	ND	+++	+++f	+f	[51]

^aND: not determined; − not expressed; +/− very low expression; + low expression; ++ intermediate expression; +++ high expression. Note that only some of these genes were directly compared to each other.

^bCells derived from a mouse expressing the SV40 T antigen under the control of an osteocalcin promoter [24]

^cThermolabile SV40 T antigen contruct, active at 34°C, inactive at 39–40°C allowing conditional immortalization

^dAt non-permissive temperature (37–40°C)

^eMice have a IFNγ-inducible promoter driving thermolabile SV40 T antigen expression allowing conditional immortalization

^fExpression low/absent initially but present after 1–2 weeks in differentiation conditions

Bodine et al reported the development of a human osteocyte-like cell line, HOB-01-C1 that is conditionally transformed depending upon the temperature of culture [49]. These cells mineralize rapidly, and are responsive to parathyroid hormone (PTH), TNFα and IL1β, however they have not been used in further studies. Divieti et al. also produced three lines of conditionally transformed osteocyte-like cells from PTH/PTH related protein (PTHrP) receptor (PTH1R) knockout mice that showed high binding affinity to the C-terminal end of PTH [50]. These cells have not been further utilized in the study of osteocyte biology, but the absence of PTH1R makes them suitable to study PTH signaling through other receptors. The newest osteocyte-like cell line reported, IDG-SW3, was isolated from mice carrying the Dmp1-GFP reporter transgene and a conditionally immortalizing, temperature sensitive SV40 Tag (Immortomouse) [51]. Cells derived from these animals can be expanded at 33°C in the presence of IFNγ (Tag permissive temperature), but they stop proliferating and start to differentiate in non-permissive conditions (37–39°C). IDG-SW3 appear to progress from a late osteoblastic phenotype, to a mature osteocyte-like phenotype with production of mineralized matrix. Induction of various osteocyte marker genes including Dmp1, MEPE, Sost and Fgf23 is observed over 3–4 weeks in culture at 37°C [51]. These cells should prove useful for further study of osteocytes at various differentiation stages. However, it should be noted that these cultures still contain a mixture of osteoblasts and osteocytes at different stages of differentiation and for studies requiring enriched osteocytes, the Dmp1-GFP cells will need to be isolated by flow cytometry (FACS).

Use of osteocyte-like cell cultures to understand mechanobiology in bone

In healthy individuals bone accrues in response to weight bearing activity, and is lost in the absence of mechanical stimuli. The osteocyte network is thought to be the main mechanosensor orchestrating this process, most likely by sensing changes in fluid flow that occur in the canalicular network in response to loading [52, 53]. It is not clear how this occurs, but it may involve the dendritic processes [54, 55], the cell body and/or the primary cilium [56, 57]. Modeling in vivo mechanical conditions with in vitro models is not trivial, but cell culture experiments have demonstrated molecules and pathways that are likely to be important in the response to mechanical signals. Primary chicken osteocytes and MLO-Y4 cells respond rapidly to fluid flow shear stress (FFSS), releasing nitric oxide (NO), prostaglandin E₂ (PGE₂), Ca²⁺ and ATP [36, 38, 58–61]. These responses tend to occur to a greater degree, or at lower magnitudes of stress in osteocytes than in comparable osteoblast cultures, suggesting that osteocytes are particularly sensitive to this type of stimulus. For example, primary chicken osteocytes produced greater amounts NO and PGE₂ than osteoblasts under the same conditions [58–60, 62], and nuclear translocation of β-catenin occurred at lower levels of stress in MLO-Y4 than in 2T3 cells [63]. In addition, Bacabac et al. found that partially adherent rounded MLO-Y4 cells showed much greater NO response to mechanical force than flat adherent cells suggesting that studies of adherent cells may underestimate the differences in responsivity to mechanical signals between osteoblasts and osteocytes [64]. Rapid release of signalling molecules such as PGE₂ can occur via connexin 43 hemichannels [35] whereas stretch-activated cation channels and integrins have been shown to be important for Ca²⁺ fluxes [65, 66]. Conditioned medium from osteocytes exposed to FFSS inhibits osteoclastogenesis and resorption compared to medium from static cultures [40, 67]. Conditioned medium from primary osteocytes exposed to FFSS also inhibited osteoblast proliferation and stimulated ALP production, while medium from static cultures had no effect [68]. FFSS has also been shown to protect osteocytes but not osteoblasts from undergoing apoptosis [30, 32, 65, 69]. In vitro study of osteocytic cells is now an active field in numerous labs, and has greatly contributed to our knowledge of the extensive activities these cells are likely to be performing in vivo.

Transgenic approaches to identify and characterize osteocytes

Osteoprogenitor cell differentiation is a process that is much easier to document in vitro than in an in vivo context [70]. Becoming an osteocyte represents terminal differentiation of the osteoprogenitor lineage. Until recently, osteocyte isolation was performed using only few osteocyte-specific antibodies such as monoclonal antibodies MAb OB7.3 [71], MAb SB5 [72] and MAb OB37.11 [73], specific for avian osteocytes. Breakthroughs in osteocyte isolation were made thanks to the identification of new osteocyte selective markers such as Dmp1, Mepe, Phex, E11/gp38 and Sost. Dmp1 belongs to the SIBLING (small integrin-binding ligand N-linked glycoprotein) family and is thought to be involved in the regulation of matrix mineralization and phosphate homeostasis [74]. Sclerostin is a secreted glycoprotein that appears to be highly expressed in mature but not early osteocytes [14]. It functions as a Wnt antagonist by binding Lrp4/5/6, acting as a regulator of bone mass [75–77].

In the last decade, a number of visual transgenes have been developed that have been very useful for dissecting different populations of osteoblast lineage cells including osteocytes. Early studies utilized collagen type I promoters (Col3.6kb and Col2.3kb) to drive expression of green fluorescent proteins (GFP) in preosteoblasts and ostreoblasts/osteocytes respectively [78]. This approach was followed by generation of other bone directed promoter-GFP reporters like osteocalcin (Oc-GFP) that labeled a population of mature osteoblasts [79, 80]. Despite the usefulness of these transgenic mice in bone biology, they did not show enough specificity to target partially embedded (preosteocytes) or matrix embedded cells (osteocytes).

Dmp1 gene expression is characteristic of odontoblasts, but in bone, expression is restricted to preosteocytes/osteocytes as assessed by in situ hybridization in chicken, rat and murine models [81, 82]. To study the activation of Dmp1 in the osteoblast lineage, different lengths of Dmp1 promoter have been utilized to drive transgene expression (GFP or lacZ) [83]. Their temporal activation was evaluated in an in vitro model using BMP-induced differentiation in 2T3 osteoblast-like cells. Constructs containing larger Dmp1 promoter fragments −9624bp (10kb) or −7892bp (8kb) showed a dramatic increase in expression following the onset of mineralization. The majority of 2T3 cells harboring the −2433 (2.5kb) promoter showed little or no GFP expression. In addition, FFSS activated the endogenous Dmp1 gene and the 8kb Dmp1-GFP in mineralized 2T3 cells, suggesting that Dmp1 is responsive to mechanical loading. Further in vivo studies utilizing the 8kb Dmp1 promoter show activation of the reporter gene following mechanical loading of ulnae [83].

Subsequently, to utilize the advantage of the strong expression with the onset of mineralization of 8kbDmp1 promoter in 2T3 cells, a transgenic model was generated that expressed a topaz variant of eGFP protein (excitation 514nm, emission 527nm) under the control of the 8kbDmp1 promoter [82]. The expression of GFP was restricted to populations of preosteocytes and osteocytes within the bone (Figure 1A–B). Interestingly, expression was also detected in cells with dendritic processes within the brain (Figure 1C), and in peritubular areas of the kidney (Figure 1D). Dmp1-GFP expression is also activated in primary bone marrow stromal cell cultures under osteogenic conditions in cells residing within mineralized nodules. They exhibit an osteocyte-like phenotype characterized by an extensive network of dendritic processes (Figure 2A) [82].

Figure 1. Histological evaluation of 8kbDmp1-GFP expression.

6–8 week old 8kbDmp1-GFP transgenic mice were sacrificed and cryosections of different tissues prepared. GFP expression was observed in osteocytes within cortical and trabecular bone (A–B, arrowheads). Osteoblasts on bone surface were GFP negative (A–B, arrows). We have detected GFP expression in brain in cells that have dendritic processes (C), and in the cells located in the peritubular areas of the kidney (D). (BM, bone marrow; CB, cortical bone) Tissue samples were dissected, fixed in 10% formalin, bones decalcified for 1 week in 15%EDTA, placed in 30% sucrose overnight and embedded in cryomedium. Sectioning has been completed using a cryostat and a tape transfer system.

Figure 2. Osteocytic expression of Dmp1-GFP in vitro and in vivo.

Mesenchymal progenitor cells in primary bone marrow stromal cultures differentiate into mineralized nodules. Osteocyte-like cells expressing Dmp1-GFP can be detected within the mineralized colonies. Cells exhibit dendritic features reassembling osteocytes (A). Histological analysis of cortical bone derived from Col2.3cyan/Dmp1-GFPtopaz dual transgenic mice (B). The extensions from osteocytes (green-arrowhead) connect to osteoblasts on bone surface (blue-arrows). Images 2A and 2B were a part of published studies [82, 89].

New transgenic reporter mice for studying osteocyte biology

New recombineering approaches have been used to make transgenic mice for studying the complete cis-regulatory regions of genes expressed in osteocytes. Patients with Van Buchem disease lack mutations in SOST coding regions, but they have a homozygous 52kb noncoding deletion downstream of SOST gene and upstream of the MEOX1 gene. The first study to utilize the human SOST BAC sequences confirmed that this deletion affected a SOST-specific regulatory element resulting in a phenotype with characteristics of Van Buchem disease [84].

Recently, we have generated transgenic mice for studying the complete cis-regulatory regions of the Dmp1 and Sost genes using different variants of fluorescent proteins (Fig 3A). Cortical long bone histology from a mouse containing 160Kb of the Sost gene with an EGFP reporter [85], and a Dmp1 BAC clone containing 165Kb of the Dmp1 gene with a DsRed reporter, is shown in Fig 3B–C. Notably, only a fraction of the osteocytes are positive for the Sost-GFP but many more osteocytes are positive for the Dmp1-DsRed. These results suggest that for any bone or condition, there is heterogeneity in expression of osteocyte-selective or specific genes within the osteocyte population.

Figure 3. SostBAC-EGFP and Dmp1BAC-DsRed transgenic reporter mice.

BAC constructs used to generate reporter mice (A). Confocal image of cryosectioned cortical bone of the femoral diaphysis from one-month old dual transgenic animals (B). Dmp1 red cells are overlayed with the Sost green cells. Nuclei are stained with DAPI. The same image showing Sost but not Dmp1 reporter (C).

These double reporter mice have been used to develop a calvarial osteocyte culture and a long bone cortical osteocyte culture system for ex vivo studies of osteocytes [85]. There is evidence that mineralized matrix around the osteocyte is required to maintain osteocyte characteristics. Many of the osteocyte cell models, such as MLO-Y4 make no mineralized matrix and express few osteocyte signature genes at any appreciable level. When bone chips are cultured, Dmp1 labeled cells outgrow and lose Dmp1 expression, showing evidence for dedifferentiation and loss of osteocytic phenotype in vitro [86]. Therefore the hypothesis was that to maintain the most “natural” osteocyte state, leaving them in the bone matrix and removing other cellular elements was a reasonable approach to study osteocytes in mature cortical bone. To achieve this a “bone tube” system was employed. Briefly, following dissection and cleaning of the attached muscles, cortical bones from femur, tibia, and humerus were isolated, the growth plate region was removed, and the bone marrow and periosteal and endosteal osteoblasts were eliminated by five serial collagenase-trypsin digests. The SostBAC-GFP and Dmp1BAC-DsRed reporters could be monitored and demonstrated that the osteocytes remained alive and functional for up to 10 days in these ex vivo cultures [85].

The “bone tube” osteocyte system can also be used to image osteocytes under different conditions. An example of this approach is shown in Figure 4. Tibial bone tubes from Sost-GFP mice were loaded to 3N and underwent imaging in culture. The load was removed and samples reimaged. We have used a DISMAP approach and evaluated GFP labeled osteocytes, before and after loading [87, 88]. The strain fields could be calculated using the position of the osteocytes as a coordinate map. It was observed that the local strains throughout the bone are very discontinuous, with strains as high as 10,000 μstrain around the osteocyte lacunar region. We anticipate that in future bones from these transgenic mice will be used with high resolution confocal microscopy to map deformation changes at the osteocyte-lacunar-canalicular level in physiological or genetic mouse models [85].

Figure 4. Mechanical loading of tibia from SostBAC-EGFP reporter mice.

Reporter mice at 2 months of age were used to generated ex vivo osteocyte-enriched tibial “bone tubes” with Sost-EGFP expressing osteocytes. The tibiae were loaded to 3N while cultured in 10% FCS/αMEM. Bone tubes were imaged and the load removed and reimaged. Using DISMAPP, a computational tool to map strain fields, the osteocytes were used as landmarks and the local deformation of the bone was calculated. Note the high strains around the osteocytes exceeding 10,000 μstrain, and the discontinuous strains throughout the bone at this local level.

The osteoblast versus the osteocyte

The mechanism of transition of osteoblasts into osteocytes remains poorly defined. To better understand this process and define the gene expression changes between osteoblasts and osteocytes, dual transgenic mice expressing Dmp1-GFP (topaz) and Col2.3CFP (cyan, excitation 433nm, emission 475nm) were utilized. Osteocytes were identified as Dmp1-GFP⁺ cells, while osteoblasts are represented by the Col2.3CFP⁺/Dmp1-GFP⁻ population [89]. This combinatorial approach allows for more detailed histological analysis (Figure 2B) and allows for additional characterization and cell isolation by FACS. FACS-separated osteocytic and osteoblastic populations from calvarial digests were used for microarray analysis. The genes most characteristic of osteoblasts included keratocan (Kera) and tenomodulin (Tnmd), while reelin, Dmp1 and Phex were strongly expressed in osteocytes [89]. In addition, this study points to the high expression levels of selected muscle-related genes, neuronal regulators such as Npy, and ion channels in the osteocytic population. The expression of these genes might be important in the maintenance of the osteocyte phenotype (dendritic extensions) and intercellular communication.

Models to evaluate effects of osteocyte-targeted gene deletion or overexpression

In the past decade the use of the Cre/loxP technology became a popular method to evaluate gene deletion or to target genes to tissues or particular stages within specific cell lineages. The efficacy and specificity of the recombination depends on the promoter utilized as well as the particular gene locus that is targeted. To direct recombination to the osteprogenitor lineage, multiple promoter-Cre transgenic mice have been generated. Cre has been utilized to target earlier stages of differentiation within the bone lineage (Osterix-Cre, Col3.6kb-Cre, Col2.3kb-Cre, Oc-Cre) [90–92]. Most of these transgenes are also expressed in the osteocytes, and as osteoblast lineage cells eventually mature into osteocytes the gene of interest will also be deleted in osteocytes. Due to these complexities, the interpretation of these experiments requires extensive work, and it can be very difficult to clarify which stage(s) of the lineage are primarily responsible for the effects of the gene deficiency. A number of Cre recombinase transgenes targeting osteocytes have been reported and their specificity will be discussed here.

Evaluation of the 10kbDmp1-Cre using Ai9 reporter mice

Based on the expression of Dmp1 in osteocytes, Lu et al. utilized a 10kb fragment of the Dmp1 promoter (−9624 ~ +4439) to direct expression of Cre recombinase [93]. Using a Rosa26R reporter model in which a lacZ expression is activated following excision of a stop cassette, recombination was detected in osteocytes and odontoblasts [94]. The authors did not observe any Cre activity during embryonal development, and little or no expression in other tissues like kidney and brain. This result could be due to low Cre efficiency or be specific to the floxed gene locus (Rosa26R). In addition, detection of recombination is highly dependent on the variability of the β-galactosidase detection methodology. The incomplete penetration of fixative, shorter staining to avoid background and variable length of the development of the lacZ reaction, can significantly affect the extent of the staining and intensity.

To evaluate the Cre activity we have utilized a different reporter model in which a recombination event causes activation of tdTomato fluorescent protein (Jackson labs, #007905, termed Ai9). The Rosa-CAG-LSL-tdTomato-WPRE targeting vector contains a CMV-IE enhancer/chicken beta-actin/rabbit beta-globin promoter (CAG), a loxP-flanked STOP cassette, tdTomato sequence, and a woodchuck hepatitis virus post-transcriptional regulatory element (WPRE; to enhance the mRNA transcript stability), followed by a polyA signal. This entire construct was inserted between exons 1 and 2 of the Gt(ROSA)26Sor locus [95]. This design provides additional sensitivity for recombination detection. In addition, it allows straightforward detection of the tdTomato using frozen sections and epifluorescent imaging, without the requirement for additional staining.

Utilizing 10kbDmp1-Cre, recombination was detected within brain and kidney (data not shown), muscle tissue, including mature myofibers, and in the majority of the osteoblasts on both trabecular and endocortical surfaces (Figure 5C–D, arrows indicate osteoblasts). tdTomato expression was also observed within a population of bone marrow cells (Figure 5C–D, indicated by arrowheads). Despite improved detection of the Dmp1-Cre directed recombination using tdTomato reporter mice, a number of potential difficulties can be experienced using fluorescent reporter genes. Detection of GFP in formalin-fixed, paraffin embedded tissue sections can be obtained using only certain GFP variants (topaz, emerald), and in high-expressing transgenic lines. Tissue autofluorescence can be an obstacle, which can be eliminated by careful selection of fluorescent filters [96, 97]. Use of cryosectioning as a gold standard for GFP detection significantly improves transgene detection.

Figure 5. Evaluation of Dmp1-Cre directed recombination.

Histological sections from Dmp1-Cre negative mice showing background fluorescence signal using dual filters (TRITC/FITC) (A–B). 10kbDmp1-Cre activity was evaluated using cryosections of adult femurs of dual transgenic mice. Dmp1-Cre-activated tdTomato expression was detected in the majority of osteocytes within cortical (CB) and trabecular bone (C–D). Osteoblasts on the bone surface are tdTomato+ (C–D, arrows). Cre activity was observed within the muscle and in cells within bone marrow (C–D, see arrowheads).

The expression of 8kbDmp1-Cre is similar to the 10kbDmp1-Cre activity, and is present in muscle, osteocytes, and in a proportion of cells on the bone surface and within bone marrow (BM) (E–F).

Evaluation of the 8kbDmp1-Cre using Ai9 reporter mice

Due to the potential issue of non-specific expression of 10kbDmp1-Cre, and based on the more osteocyte-restricted expression of 8kbDmp1-GFP, another construct had been recently developed that harbors the 8kbDmp1 promoter directing Cre [98]. Similar to the 10kbDmp1-Cre model, Cre activity in the 8kbDmp1-Cre was observed within osteocytes. However, a proportion of osteoblasts showed Cre activity on the endocortical and cancellous surfaces in the 8kbDmp1-Cre (Figure 5E–F).

Use of the inducible 10kbDmp1-CreERT2 to target osteocytes

Another approach used to avoid non-specific expression of 10kbDmp1-Cre potentially present during developmental stages, is to allow for control of the Cre activity using a tamoxifen-inducible Cre system (CreERT2) in which a mutated estrogen ligand-binding domain has been fused to the Cre recombinase [99]. In this model, Cre activates following induction with tamoxifen treatment. Histological evaluation showed the presence of spontaneous (non-tamoxifen induced) Cre activity in a small proportion of osteocytes (Figure 6C–D, see arrowheads). No spontaneous activation was detected in the osteoblasts on bone surface (Figure 6C–D, see arrows). To test whether the various dosing of tamoxifen can modulate Cre activity specifically to osteocytes and not osteoblasts, tamoxifen was used as a single dose of two concentrations, high, 75μg/g ip and low 7.5μg/g ip. Two days after induction with high dose tamoxifen, the majority of osteocytes and osteoblasts were tdTomato positive (data not shown), while the lower concentration showed more restricted Cre expression to osteocytes (Figure 6E–F). Similar findings were previously reported by Powell et al. [99] using Rosa26R reporter mice (Figure 6A–B). Interestingly, under both treatment regimens we did not observe any Cre activity within the muscle. The presence of transient expression of the Dmp1 promoter during developmental stages could be a potential explanation, as this activation is circumvented using the inducible system. As mentioned above, the differences in Cre-recombinase activity using different reporter mice highlight the importance of carefully assessing not only the expression but also the functional activity of any gene of interest that is targeted for ablation.

Figure 6. Evaluation of Dmp1 directed inducible Cre system.

X-gal staining of tibia sections from Rosa26:10KbDmp1-CreERT2-positive mice (A-B). Animals were injected with a low dose of tamoxifen (7.5μg/g) at days 3, 5, 7, 14 and 21 days postnatal and sacrificed at 23 days (2 days after last tamoxifen injection). Images of osteocyte staining in cortical (A) and trabecular (B) bone are shown.

Cre activation of tdTomato reporter without tamoxifen treatment in cortical and trabecular areas (C–D). 10kbDmp1-Cre activity was induced in four-week old mice by a single dose of tamoxifen (7.5μg/g) and mice were sacrificed two days later (E–F). Cre activation was observed in osteocytes, and some osteoblasts. Arrows indicate osteoblasts, and arrowheads osteocytes.

Methods to evaluate the osteocyte specificity of a gene deletion

Use of osteocyte enrichment by serial digestion of bone

To evaluate the specificity of gene deletion in osteocytes, Kramer et al. utilized serial digestions of long bones [100]. The third digestion presented an osteoblastic phenotype, while the sixth digestion contains cells that expressed Dmp1 and Sost, defining it as an osteocyte population [100, 101]. This approach can be utilized to evaluate the effects of Cre recombination on the expression of a targeted gene within osteocyte and osteoblast populations.

Breeding Dmp1-GFP into the Dmp1-Cre targeted deletion model

To confirm that Dmp1-Cre deletion was targeted specifically to osteocytes, breeding with the Dmp1-GFP mouse was completed [98, 99]. This allows for a combination of histological examination of the Dmp1-GFP expression, and cell isolation by FACS to confirm that Dmp1-GFP negative cells do not express Cre transgene.

Immunohistochemical detection of recombination

Some of the studies utilized immunohistochemistry to detect osteocyte-specific deletion of the gene of interest [98]. As recombination effects of Cre are also locus dependent [102–104], confirmation of the deletion of the gene at protein level serves as ultimate proof of recombination efficiency. This approach represents an important step to test whether Dmp1-Cre targeted gene deletion is specific to osteocytes.

Strategies to target overexpression or gene deletion in osteocytes

Deleting genes in a lineage-specific manner using the Cre/loxP system has provided insights into numerous aspects in bone biology. Following reports using Oc-Cre and Col2.3kb-Cre models that act on osteoblasts and osteocytes, the Dmp1 promoter has become the main tool to target the expression or deletion of genes to the most mature stages of the osteoblast lineage (Table 2). Transgenic mice expressing a constitutively active PTH receptor (caPTH1R) under the control of the 8kbDmp1 (8kbDmp1-caPTH1R) promoter exhibit increased bone mass and bone remodeling. They show reduced expression of sclerostin resulting in increased Wnt signaling, increased osteoclast and osteoblast number, and decreased osteoblast apoptosis [105]. This phenotype bears similarities to mice with Col2.3kb-directed caPTH1R expression without showing accumulation of the fibrotic cells within bone marrow. The striking difference within these two models is a marked stimulation of periosteal bone formation in the long bones and calvaria of 8kbDmp1-caPTH1R mice [106]. In the Col2.3-caPTH1R, periosteal bone formation is reduced and calvarial thickness is decreased [107]. In addition, despite increased numbers of osteoblastic cells within bone marrow, activation of PTH signaling in osteocytes does not expand hematopoietic stem cells (HSCs) [108]. This is in contrast to the mice with constitutive activation of the PTH1R in immature and maturing osteoblasts (Col2.3-caPTH1R) which exhibited increased osteoblastic cell numbers with a corresponding expansion of HSCs. In line with this, osteoblast ablation directed by the Col2.3kb promoter revealed the importance of the earlier stages of the osteoblast lineage on hematopoietic lineage regulation [109].

Table 2.

Transgenic mice targeting osteocyte.

Gene	Symbol	Type	Effect	Bone phenotype	Reference
Parathyroid hormone receptor	ca PTHR1	8kbDmp1	Expression	⇑bone mass/decrease in SOST expression	[105]
β-catenin	Ctnnb1	10kbDmp1-Cre	Deletion	⇓bone mass/increase	[101]
low-density lipoprotein receptor-related protein 5	Lrp5	10kbDmp1-Cre	Deletion/Activation	⇑bone mass with deletion/⇓bone mass with activation of high bone mass mutation	[111]
Polycystin 1	Pkd1	10kbDmp1-Cre	Deletion	⇓bone mass/lack of anabolic response to loading	[116]
Vitamin D receptor	Vdr	10kbDmp1-Cre	Deletion	No change in bone mass on normal calcium diet	[117]
Receptor activator of nuclear factor κB ligand	Tnfrsf11	10kbDmp1-Cre	Deletion	⇑bone mass/regulation of osteoclast function	[114, 115]
Parathyroid hormone and parathyroid hormone related protein receptor	Pth/Pthrp receptor (PPR)	10kbDmp1-CreERT2	Deletion	⇓bone mass/increase in SOST expression	[99]
Myocyte enhancer factor 2	Mef2c	10kbDmp1-Cre	Deletion	⇑bone mass/decrese in SOST expression	[118]
Connexin 43	Cx43	8kbDmp1-Cre	Deletion	⇑osteocyte apoptosis/widening of long bones	[98]
Sclerostin	SOST	8kbDmp1	Expression	⇓bone mass/modulation of response to loading	[110]

A number of studies have utilized DMP1-Cre transgenic models to evaluate the effects of activation or deletion of Wnt-signaling related molecules in mature osteoblast/osteocytes. A useful model to study the role of SOST was developed by overexpressing human SOST in osteocytes and evaluating the response to mechanical loading [110]. Load-induced bone formation was reduced in Dmp1-SOST transgenic mice, due to lower Wnt signaling in osteoblast lineage cells. This suggests that downregulation of Wnt signaling modulator Sost in osteocytes is required for the response to mechanical loading. An additional insight into regulation of the Wnt pathway has been provided by studies in which Lrp5, a co-receptor in canonical Wnt signaling has been deleted using Dmp1-Cre. While activation of a high bone mass mutation in Lrp5 by Dmp1-Cre caused an increase in bone mass, Dmp1-Cre targeted deletion of Lrp5 resulted in bone loss. This work provided compelling evidence that Lrp5 signaling functions locally through mature osteoblast/osteocyte populations [111].

To evaluate the role of β-catenin-dependent canonical Wnt signaling, a study was completed using Dmp1-Cre directed deletion of β-catenin [101]. Similar to the conditional deletion of β-catenin in osteoblasts using Col1a1-Cre or Oc-Cre, Dmp1-directed deletion of β-catenin resulted in low bone mass due to increased osteoclast activity [112, 113]. The phenotype progressively developed in early adulthood, suggesting a lack of Dmp1-Cre expression prior to formation of mature bone and osteocytes. Increased osteoclastogenesis was due to decreased OPG expression by osteocytes, while changes in RANKL expression were minimal [101]. More recent work by the O’Brien and Takayanagi groups evaluated the role of the osteocyte in RANKL production and presented evidence that osteocyte-derived RANKL is a critical regulator of osteoclastogenesis [114, 115]. The deficiency of these studies is their heavy reliance on Dmp1-Cre to target RANKL specifically in osteocytes and not in osteoblasts. In addition, the ability to detect gene deletion is limited by a lack of good Rankl antibodies. In a recent webinar http://www.nature.com/bonekey/webinars/index.html?key=webinar15, Dr. O’Brien summarized his findings and outlined the importance of generating new Cre transgenes that are more osteocyte specific.

Summary.

We have described the In vitro and In vivo models that are currently available to study osteocyte biology. The identification of several osteocyte specific genes has allowed the generation of a plethora of new data on osteocyte functions. Researchers have shown that the Dmp1 promoter can be successfully utilized to overexpress or delete genes in the most mature cells of the osteoblast lineage. Currently, a DMP1-GFP model is used to isolate osteocyte cell populations, while few models are available for targeting gene deletion including 10kbDmp1-Cre, 8kbDmp1-Cre and a tamoxifen inducible 10kbDmp1-CreERT2. It is important to note that the use of restricted promoters rarely achieves absolute specificity and efficiency. Moreover, detection of Cre recombinase activity using reporter mice should be complemented with the demonstration of the deletion of the gene of interest in order to assure successful tissue/cell specific gene ablation. At present, Dmp1 directed expression represents a useful approach in combination with other Cre models where promoters are targeted to earlier stages of the osteoblast lineage. In conclusion, we have presented the pros and cons or benefits and deficiencies of currently available cell lines and genetic models used to investigate and understand osteocyte biology. We have outlined current progress in this field. Future efforts will be directed towards the generation and development of novel osteocyte-specific approaches that will enhance our understanding of the multiple roles and functions of osteocytes.