Archives of Biochemistry and Biophysics 561 (2014) 22–28

Contents lists available at ScienceDirect

Archives of Biochemistry and Biophysics

journal homepage: www.elsevier .com/ locate /yabbi

Review

Regulation of cortical and trabecular bone mass by communicationbetween osteoblasts, osteocytes and osteoclasts

http://dx.doi.org/10.1016/j.abb.2014.05.0150003-9861/� 2014 Elsevier Inc. All rights reserved.

⇑ Corresponding author. Address: 9 Princes St, Fitzroy, Victoria 3122, Australia.Fax: +61 3 9416 2676.

E-mail address: nsims@svi.edu.au (N.A. Sims).

1 Abbreviations used: BMU, basic multicellular unit; RANKL, Receptor ActNFjB Ligand; OPG, osteoprotegerin; PTH, parathyroid hormone; BMD, bonedensity; CTHRC1, collagen triple helix repeat containing 1; CT-1, CardiotrVDR, vitamin D receptor; PTHR1, PTH receptor; LIF, leukemia inhibitory factoncostatin M; CNTF, ciliary neurotrophic factor.

Natalie A. Sims ⇑, Christina VrahnasSt. Vincent’s Institute of Medical Research, Fitzroy, Victoria, AustraliaDepartment of Medicine at St. Vincent’s Hospital, The University of Melbourne, Fitzroy, Victoria, Australia

a r t i c l e i n f o a b s t r a c t

Article history:Received 20 March 2014and in revised form 15 May 2014Available online 26 May 2014

Keywords:OsteoblastOsteoclastOsteocyte

The size and strength of bone is determined by two fundamental processes. One process, bone remodel-ling, renews the skeleton throughout life. In this process existing bone is resorbed by osteoclasts andreplaced, in the same location, by osteoblasts. The other process is bone modelling, where bone formationand resorption occur at different sites so that the shape of bone is changed. Recent data suggests thatboth remodelling and modelling are controlled by signals between the cells that carry out these two pro-cesses. Osteoclasts both resorb bone, and provide inhibitory and stimulatory signals, including cardiotro-phin-1 and sphingosine-1-kinase, to the osteoblast lineage thereby regulating their differentiation andactivity on both trabecular and cortical surfaces. In addition, the osteoblast lineage, including osteoblastprogenitors, matrix-producing osteoblasts, bone lining cells, and matrix-embedded osteocytes, produceboth inhibitory and stimulatory factors that stimulate osteoclast differentiation. We will discuss the rolesof osteoblast- and osteocyte-derived RANKL, and paracrine, autocrine and endocrine factors, such as eph-rinB2, the IL-6/gp130 family of cytokines, parathyroid hormone, and its related peptide, PTHrP. These fac-tors not only stimulate RANKL production, but also stimulate osteoblast differentiation and activity. Thisreview will focus on recent data, generated from pharmacological and genetic studies of mouse modelsand what these data reveal about these pathways at different stages of osteoblast differentiation andtheir impact on both bone remodelling and modelling in trabecular and cortical bone.

� 2014 Elsevier Inc. All rights reserved.

Introduction

The skeleton is constantly remodelled by repeated cycles of cel-lular activity occurring asynchronously throughout the skeleton inwhich tiny packets of bone are resorbed and then replaced. Thisprocess always occurs in the same sequence: bone resorption byosteoclasts followed by bone matrix production by osteoblasts.This is the fundamental process by which the skeleton changesin response to hormonal and mechanically-induced stresses.

In addition to this process, bone also adapts by the process ofmodelling; here bone formation and resorption do not occur insequence at the same site. Modelling occurs during growth, andin response to mechanical loading; it can also be induced by phar-macological agents that promote bone formation without arequirement for prior resorption [1]. Modelling is also responsiblefor cortical expansion, where osteoblasts on the periosteal surface

continue to form bone at the diaphysis of the long bones (Fig. 1).The mechanisms that determine why some bone surfaces remodelwhile others model are not known, but understanding the relation-ships between the cells involved in modelling and remodellingholds great potential for developing therapeutics that can restorebone strength in osteoporosis.

Originally, the basic multicellular unit (BMU)1 responsible forremodelling was thought to consist of two classes of specialized cellson the bone surface, osteoclasts and osteoblasts, which contribute toremodelling by bone resorption and formation, respectively.Although osteoclasts are derived from the hemopoietic lineage,and osteoblasts from the mesenchymal lineage, these cell types actin close apposition and regulate the function of the other lineageby appropriate production of both inhibitory and stimulatory factors[2,3]. Over the past 50 years, this concept has been refined, and a

ivator ofmineral

ophin-1;or; OSM,

Epiphysis

Metaphysis

Diaphysis

Metaphysis

Epiphysis

Trabecular (cancellous) bone

Epiphyseal (growth) plate

Marrow cavity

Cor�cal (compact) bone Periosteal Surface

(periosteum)

Endocor�cal surface 1

3 4

Osteoid Bone

Endocor�cal/ trabecular (remodelling)

Periosteum (modelling)

Coupling factors

Osteoid

Mature osteoblasts

Osteocytes

Lining cells Reversal Phase

5

IL-6 family cytokines

Osteoclast

2

Fig. 1. Structure of trabecular and cortical bone, the process of bone remodelling, and cellular signals from coupling factors and the IL-6 cytokine family that regulate boneremodelling and periosteal modelling. The proximal and distal ends of the growing murine femur (epiphyses) contain a high proportion of trabecular (cancellous) bone.Trabecular bone is also prevalent in the metaphyses, which is separated from the epiphysis by the epiphyseal (growth) plate; this is the site of longitudinal bone growth. Themidshaft of the femur, diaphysis, contains a high proportion of cortical (compact) bone that surrounds the inner marrow space and trabecular regions. Bone remodellingoccurs on trabecular surfaces, and on the endocortical surface (both surfaces together are termed the endosteum). Bone modelling occurs throughout life in murine bones onthe outer periosteal surface (periosteum). During bone remodelling on the endosteum, (1) osteoclasts attach to the bone surface, resorb bone and release coupling factors thatstimulate osteoblast differentiation on the endosteal surface. These coupling factors also signal to periosteal osteoblasts, perhaps through the osteocyte canalicular network.After the reversal phase, which remains poorly understood in murine bones, (2) pre-osteoblasts mature, attach to the bone surface and fill the cavity created by osteoclastswith bone matrix, termed osteoid. (3) Mature osteoblasts, when their task of producing osteoid is completed, become lining cells or (4) become embedded within the osteoidas it is mineralised. These osteoblasts become osteocytes and release factors that regulate mineralisation. IL-6 family cytokines are released by the osteoblast lineage and actto stimulate osteoblast differentiation and bone matrix production on endosteal surfaces, but limit osteoblast activity on the periosteum.

N.A. Sims, C. Vrahnas / Archives of Biochemistry and Biophysics 561 (2014) 22–28 23

number of regulatory factors have been identified, some of which wediscuss below [4–7]. The best understood example of this intercellu-lar regulation is the production of both the osteoclast stimulusReceptor Activator of NFjB Ligand (RANKL) and its decoy receptorinhibitor osteoprotegerin (OPG) by cells of the osteoblast lineage[8]. It is therefore, the same cell lineage that both forms bone matrixand regulates osteoclast differentiation in response to paracrineand endocrine stimuli, including parathyroid hormone (PTH),1,25-dihydroxyvitamin-D3 and cytokines [9–11]. The osteoblastlineage includes committed osteoblast precursors, matrix-producingosteoblasts, lining cells and matrix-embedded osteocytes; the majorcontributing cells to these two activities are unlikely to be at thesame stage of differentiation within the lineage, and this conceptis discussed below. Similarly, osteoclasts produce a range of‘‘coupling factors’’. This is achieved both by releasing factors fromthe bone matrix itself during the process of resorption, and byproduction of soluble, and possibly membrane bound, regulators ofbone formation (for recent reviews see [4,7,12]).

Although the initial concept of remodelling focussed on thecells on the bone surface, we now understand that there are manyother cellular contributors that regulate bone formation andresorption within the BMU. These include osteocytes, terminallydifferentiated osteoblast lineage cells that reside in an intercon-nected network that extends throughout the bone matrix, andmultiple cell types in the marrow space (e.g., haemopoietic precur-sors, macrophages, T-cells, natural killer cells and adipocytes) [13].Furthermore, different stages of osteoblast differentiation are nowunderstood to play distinct roles in regulating the activities ofosteoclasts [13], and each other [14]. This is particularly relevant

for the initiation of the bone remodelling cycle, where osteocytesand osteoprogenitors produce the RANKL required for osteoclasto-genesis [7].

The identification of a bone remodelling canopy that lifts fromthe bone surface when osteoclastic resorption initiates the remod-elling cycle to enclose the BMU in an isolated environment is a con-cept that has been explored at length in human specimens by theDelaissé laboratory [15,16]. This would provide a controlled localein which osteoblast lineage cells, osteoclasts, and potentially othercontributing marrow cells, may exchange factors and influenceprecursors provided by the associated vasculature. However,experimental interrogation of its contribution to the actions of spe-cific coupling factors using genetically altered mouse models islimited because this anatomical structure has not been observedin the mouse, the model that has been used most extensively fordefining the intercellular signalling pathways that modify boneremodelling.

Much work using genetically altered mice has focussed on theoverall influence of these pathways on the internal trabecular net-work (Fig. 1), including the quantity of trabecular bone and thelevel of trabecular remodelling. However, major questions remainabout the effects of the intercellular signalling pathways that reg-ulate the cortical bone (Fig. 1), and cortical bone matrix quality andstrength. Since it is now understood that intra-cortical remodellingand cortical bone loss are contributors to skeletal fragility [1], moreattention is beginning to be paid to differences in effect of signal-ling pathways in cortical vs. trabecular bone. This review will focuson some notable intercellular pathways that control bone massand bone strength in cortical and trabecular bone, as examples of

24 N.A. Sims, C. Vrahnas / Archives of Biochemistry and Biophysics 561 (2014) 22–28

a wider range of factors at play: osteoclast-derived coupling factors(cardiotrophin-1 and sphingosine-1-phosphate), osteoblast line-age-derived RANKL, IL-6 family cytokines, and ephrinB2.

Interpreting data on bone mass and remodelling in mousemodels

Studies of bone strength, mass and remodelling in humans relymainly on surrogate markers such as bone mineral density (BMD)and serum biochemical markers since biopsies are rarely obtained.The use of mouse models, particularly genetically altered mice,allows direct measurement of changes in bone mass, remodellingand strength when specific factors are removed from the body.Now that cell-specific deletion is possible, the contributions ofindividual cell types to bone mass can also be determined. It isnow commonly accepted that genetically altered mice should bebackcrossed onto an inbred strain before analysis, and that litter-mate controls should be used for all skeletal analysis, since differ-ent inbred strains have significantly different bone trabecularand cortical bone mass, shape, strength, and level of remodelling[17–19], but equal consideration needs to be given to the sexand age of the animals analysed.

Male and female mammals, including humans, differ signifi-cantly in their rates of bone remodelling, their cortical dimensions(determined by modelling) and their bone strength. Briefly, aftersexual maturity, male mice have higher trabecular bone mass[20,21], a lower level of bone remodelling [20,22], greater perios-teal diameter and ultimate bone strength [23,24] than females.For this reason, phenotypes of low bone remodelling are morereadily detectable in female mice, since the level of bone remodel-ling is already low in the male [25,26], and phenotypes demon-strating high levels of bone remodelling are more readilydetectable in male mice [27].

In mice, as in humans [28], periosteal expansion (modelling)continues throughout life at the mid-diaphysis [20] while trabecu-lar bone volume continues to reduce [20], and bone remodellinglevels reduce with age on trabecular surfaces in rodents [21]. Tra-becular bone volume peaks at 3 months of age, and at this age,periosteal apposition is still occurring in the tibiae and femora,making this an ideal time for analysis of genetically altered mice[20].

Although both humans and mice demonstrate increased perios-teal apposition with age, cortical structure is vastly different inthese two species. Mice lack the characteristic Haversian systemsthat give cortical bone its structure in larger mammals. These sys-tems comprise multiple concentric rings surrounding central bloodvessels, but the smaller thickness of cortical bone in the mousemeans that such organisation is not required. Although mice donot possess Haversian systems, they are a useful model of perios-teal growth, and of growth-related remodelling of the cortical bonefrom the rapidly deposited woven bone, that contains disorganisedcollagen fibres, to the more mechanically competent lamellar bone,in which collagen fibres are oriented to maximise bone strength.

Osteoclast-derived coupling factors in bone: roles in modellingand remodelling

The osteoclast lineage, including mature, resorbing osteoclastsand their precursors, provide coupling factors that match bone for-mation to the level of bone resorption [4,7]. These include somereleased from the bone matrix during bone resorption, such asIGF-1 and TGFb [29,30], and an increasing number of factorssecreted by both inactive and active osteoclasts, including cardio-trophin-1 [31], sphingosine-1-phosphate [32], BMP6 and Wnt10b[32], collagen triple helix repeat containing 1 (CTHRC1) [33], and

Sema4D [34]. Since none of these factors are expressed exclusivelyby osteoclasts, in vivo studies to determine the role of osteoclasticrelease of them will require the generation of cell specific knock-outs of each factor.

Cardiotrophin-1 (CT-1) is a member of the IL-6 family of cyto-kines. While it is expressed in a number of different organs, withinthe skeleton its expression is restricted to the osteoclast [31].When osteoblasts are exposed to it, either in vitro or in vivo, theirdifferentiation and bone forming activity are increased [31]. Fur-thermore, expression of the osteocyte-derived bone formationinhibitor, sclerostin is reduced by CT-1 [25]. When mice with glo-bal deletion of CT-1 were studied, although osteoclast numberswere increased, osteoclast activity was reduced. This low level ofresorption resulted in a high bone mass phenotype, termed osteo-petrosis. Furthermore, in neonate and adult male mice, bone for-mation on trabecular surfaces was lower in the absence of CT-1pointing to a lack of coupling factor activity in bone remodelling[31].

In studies of other osteopetrotic mutants, Karsdal et al. [35]noted that trabecular bone formation was frequently maintainedin mice with osteoclast-rich osteopetrosis (such as c-src deficiency)[36,37], in which osteoclasts were present but showed impairedfunction. In contrast, osteoclast-poor osteopetrosis, in which boneresorption was low because of a lower number of osteoclasts, dis-played impaired bone formation on trabecular surfaces [38–40].This suggested that, although both matrix-derived and osteo-clast-derived factors play a part, it is the osteoclast itself, not nec-essarily its resorptive action, which is the key source of couplingfactors for trabecular osteoblasts.

Local release of soluble factors from osteoclasts provides a sim-ple mechanism to control bone formation in the context of boneremodelling, where osteoblasts and their precursors are in closeproximity, but the question of whether osteoclast-derived couplingfactors also influence cortical bone remains puzzling because thisis a surface on which osteoblasts and osteoclasts are not in directapposition. This particular question has not been studied at lengthin mouse models. However, when RANKL was provided by lym-phocytes in a rescue of the osteoclast-deficient osteopetroticRANKL null mouse, cortical remodelling, and periosteal growthwas restored even though osteoclasts were only restored to theendocortical surfaces [41]. This indicates that coupling factors fromendocortical osteoclasts are able to signal to the osteoblast lineageon the periosteal surface (Fig. 1). The coupling factors that are mostimportant for this set of signals are not fully described. Further-more, in most models in which putative coupling factors have beendeleted, there is a lack of data describing cortical dimensions, thewoven or lamellar nature of the cortical bone, or the level of boneformation on the periosteal surface. In only two coupling factor-null mice have cortical dimensions been reported: CT-1 null miceshow no defect in periosteal modelling [31], but mice that lackcoupling factor BMP6 show impaired periosteal growth [42], sug-gesting that the latter may be a coupling factor that stimulatesperiosteal growth. However, the BMP6 null mice also have reducedlongitudinal growth due to loss of BMP6 at the growth plate, andthis would also limit periosteal growth.

Recent experiments using cathepsin K inhibitors to reduceosteoclast activity in preclinical models have shown that bone for-mation on both trabecular and cortical surfaces is retained whenosteoclast activity is reduced (for review see [43]). This suggeststhat coupling factors released by the cathepsin K-deficient osteo-clast may continue to be available to periosteal bone-forming oste-oblasts. Mice were generated that lack cathepsin K specifically inosteoclasts to investigate whether this increase in bone formationrelated to a change in osteoclast activity or signalling. These mice(both males and females) exhibited significantly greater trabecularbone formation and increased femoral cross-sectional area,

N.A. Sims, C. Vrahnas / Archives of Biochemistry and Biophysics 561 (2014) 22–28 25

implying a higher level of periosteal bone formation [44]. Osteoclastsisolated from these mice demonstrated increased support forosteoblast differentiation; this effect was inhibited by a spingo-sine-1-phosphate antagonist. This suggests that even when non-functional, osteoclasts may release coupling factors, includingsphingosine-1-phosphate, and thereby stimulate both periostealand trabecular bone formation. Other coupling factors may alsobe involved, since it has been reported that osteoclast-mediatedrelease of matrix-derived IGF-1 and BMP-2 also occurs whenosteoclasts are treated with cathepsin K inhibitors in vitro [45].These findings suggest potential mechanisms by which osteoclastactivity can be inhibited by cathepsin K inhibitors for osteoporosistreatment while bone formation is simultaneously increased.

How are periosteal osteoblasts exposed to coupling factors pro-duced by distant osteoclasts on the endocortical surface? Thisquestion remains unanswered, but an appealing model is that sta-ble coupling factors required for periosteal bone formation may betransported through the osteocyte lacuna-canalicular network.Alternatively, the primary cellular target of coupling factors maybe the osteocyte; an example of this is the way that CT-1 inhibitsosteocytic sclerostin production [31]. This could then increasebone formation on the adjacent bone surface. The question ofhow sclerostin is directed to the appropriate bone surface on whichmodelling is required also remains unanswered.

Osteoclast-derived coupling factors may also contribute to boneformation when it is stimulated by parathyroid hormone (PTH).PTH is classically regarded as a hormone that stimulates boneresorption. This action is mediated by its ability to stimulate pro-duction of RANKL by osteoblast-lineage cells [9], including osteo-cytes [46]. In contrast to this catabolic action, daily injection ofPTH can be used therapeutically to increase bone formation andbone mass in postmenopausal women [47]. This action dependsalso on direct PTH action on the osteoblast lineage where it alsopromotes differentiation of committed osteoblast precursors [48],inhibits osteoblast apoptosis [49,50] and reduces osteocytic pro-duction of the bone formation inhibitor sclerostin [51,52]. SincePTH stimulates both bone formation and resorption, and the activ-ities of both osteoclasts and osteoblasts are linked through the nor-mal process of bone remodelling [53,54], it has been proposed thatanabolic action of PTH depends, at least in part, on its ability tostimulate osteoclast activity [4]. This was first suggested manyyears ago, when rapid morphological changes in osteoclasts wereobserved in rats injected with PTH [55].

A role for osteoclasts in the anabolic action of PTH was furtherindicated by the reduction in the anabolic effect with concomitantbisphosphonate treatment [56], a lack of anabolic response inyoung mice with osteoclast-poor osteopetrosis [57], but retentionof anabolic response in young mice with osteoclast-rich osteope-trosis [58]. This suggested that the osteoclast contribution to PTHanabolic action may not require bone resorption, just the presenceof osteoclasts and their coupling factors. In addition, this suggeststhat osteoclasts of patients treated with cathepsin K inhibitorsmight still provide sufficient signal for PTH anabolic action to beretained. In addition to the effects of long-term osteoclast inhibi-tion described above, we found that transient inhibition of osteo-clasts by co-administration of low dose sCT blocked PTH anabolicresponse in normal young female rats [59]. This seemed to accordwith the hypothesis that active osteoclasts are required for theanabolic effect of PTH. However, in addition to inhibiting cou-pling-mediated actions, sCT treatment also increased osteocyticexpression of the bone formation inhibitor sclerostin [59]. Thissuggested an alternative mechanism by which sCT may inhibitthe PTH anabolic effect and by which endogenous CT may stimu-late bone formation. Surprisingly, when the same experimentwas carried out in 6 month old sham-operated and ovariectomizedrats, with a lower level of bone remodelling, only the stimulation of

serum P1NP levels was reduced by sCT co-administration [60]. Thetransient inhibition of osteoclast activity therefore only partiallyinhibited the anabolic effect of PTH in older rats, suggesting thatthe role of osteoclast-derived coupling factors in the anabolicaction of PTH may be most important in states of high bone remod-elling, as exists in the young animal compared to the adult. It isalso possible that while transient osteoclast inhibition is sufficientto block PTH anabolic action in young animals, any osteoclasticcontribution to PTH-induced bone formation in older animalsmay be prevented only by a constant blockade of osteoclast activ-ity, as occurs in humans treated with therapeutic osteoclast inhib-itors, such as bisphosphonates or anti-RANKL.

Clinical studies in postmenopausal women that have combinedosteoclast inhibitors with anabolic PTH therapy provide conflictingresults, and are complicated by the limitation that bone biopsiesare rarely available in such studies so data is limited to BMD andbiochemical markers. Early work showed that the combination ofa bisphosphonate and PTH limited the effect of PTH on BMD[61,62]. A recent study indicates that the combination of anti-RANKL treatment with PTH has a cumulative effect on BMD, butthe PTH-induced increase in bone formation itself was blocked,as indicated by significant reductions in the serum markers P1NPand osteocalcin [63].

Thus, the osteoclast provides coupling factors to both trabecularand periosteal surfaces to stimulate bone formation in both phys-iological modelling and remodelling. They also contribute to theanabolic action of PTH, and possibly its paracrine analog PTHrP[14], most significantly in states of high remodelling, as in theyoung animal.

EphrinB2:EphB4: osteoblast: osteoclast communication or aninter-osteoblast-lineage signal?

EphrinB2 is a membrane-bound receptor tyrosine kinase that,in bone, is expressed at all stages of osteoblast differentiationand in osteoclasts [64,65]. To induce signalling above baseline lev-els of phosphorylation, EphrinB2 must interact with a membranebound receptor, such as EphB4, which is expressed by the osteo-blast lineage, but not by osteoclasts [64,65]. Two distinct featuresof membrane-bound Ephs and ephrins are: (1) their requirementfor direct cell-to-cell interaction, and (2) their ability to generatebidirectional signals where forward signalling through the Ephreceptor and reverse signalling through the ephrin ligand occurat the same time [66].

In vitro data has indicated that when osteoclasts and osteoblastscome into contact, subsequent signalling through EphrinB2 in theosteoclast lineage restricts their differentiation, while signallingin the osteoblast through EphB4 stimulates bone formation [64].This suggested that ephrinB2 may act as a membrane-boundosteoclastic coupling factor that stimulates bone formation in oste-oblasts via EphB4 intracellular signalling. However, osteoclast line-age-specific deletion of EphrinB2 presented no detectable bonephenotype [64]. This suggests that any inhibitory role of ephrinB2signalling in the osteoblast lineage is redundant with other path-ways, or may indicate that contact between osteoclasts and osteo-blasts, as exists in cell culture conditions is a rare event inphysiological conditions. For example, in bone remodelling osteo-clasts and osteoblasts act on the same surface but at different times[67], while in bone modelling, they act on different surfaces.

The osteoblast lineage, including osteoblast precursors andosteocytes, expresses both ephrinB2 and EphB4 [65]. In addition,there is extensive contact among these cells that is required forbone formation [68–70]. Although ligand-independent phosphory-lation is possible, the high degree of contact within the osteoblastlineage is likely to explain the endogenous phosphorylation of both

26 N.A. Sims, C. Vrahnas / Archives of Biochemistry and Biophysics 561 (2014) 22–28

tyrosine kinases in these cells in vitro [65]. A key role for osteoblas-tic ephrinB2 signalling in bone formation was suggested by itsstrong and specific upregulation by parathyroid hormone (PTH)and its related paracrine protein PTHrP [65]. The importance ofthe ephrinB2:EphB4 interaction was shown by inhibition of latestage osteoblast differentiation and mineralisation in vitro[65,71,72], a finding that was reproduced in vivo, both in the pres-ence and absence of PTH [71]. In vivo, this inhibition of osteoblastdifferentiation resulted in an accumulation of early-stage osteo-blasts, indicated by high mRNA levels of Runx2, Osx, Alpl, Col1a1and PTH1R [71]. These data suggested that ephrinB2:EphB4 inter-actions between bone-surface osteoblasts, and possibly osteocytes,are required for the progression of osteoblasts to full maturity, andfor full expression of late osteoblast/osteocyte markers.

Because of the nature of bidirectional signalling, sEphB4 blocksboth EphB4 (reverse) and ephrinB2 (forward) signalling, making itimpossible to discern from these pharmacological experimentswhether forward or reverse signalling is most important for osteo-blast differentiation. Early data using a mouse that overexpressesEphB4 in the osteoblast lineage demonstrated a significantincrease in osteoblast activity in female mice, suggesting that for-ward signalling may be most important [64]. However, in our anal-ysis of these mice, backcrossed to C57BL/6, there was no differencein their bone mass or bone formation activity compared to litter-mate wild type males or females (F.M. Takyar, H.J. Brennan, N.A.Sims, unpublished observations), suggesting that promoting EphB4forward signalling does not stimulate bone formation in remodel-ling in all strains of mice.

Therefore, ephrinB2:EphB4 signalling within the osteoblastlineage provides a check-point through which osteoblasts mustpass to reach late stages of osteoblast differentiation. The key ques-tion of whether it is forward or reverse signalling that is mostimportant during bone remodelling, and relative roles in trabecularvs. cortical bone remain unresolved.

What are the respective roles of preosteoblast- and osteocyte-derived RANKL?

Inhibition of the ephrinB2:EphB4 interaction in vivo and in vitroalso promoted RANKL mRNA levels in cultured osteoblast lineagecells, and enhanced their support of osteoclast formation[65,71,72]. Consistent with this finding, in the context of fracturehealing, EphB4 overexpression in the osteoblast lineage was asso-ciated with a lower number of osteoclasts at the fracture site [73].The greater number of early-stage osteoblasts and increased sup-port of osteoclast differentiation in vitro and in vivo without anincrease in their bone forming activity suggests that inhibition ofthe ephrinB2:EphB4 interaction in the osteoblast lineage has sepa-rated the osteoblast’s bone forming effect from its support ofosteoclastogenesis. Further, this suggests that late stage osteo-blasts are less supportive of osteoclast function. This is consistentwith early in vitro studies that showed greater RANKL expressionin less differentiated osteoblasts cultured in vitro [74], however,more recent studies have suggested that late-stage osteoblast line-age cells, particularly osteocytes, are also a physiologically signifi-cant source of RANKL.

The most highly studied mechanism by which the osteoblastlineage controls osteoclast formation is their production of RANKL.RANKL expression in the osteoblast lineage is directly stimulatedby a wide range of factors, including PTH [9], 1,25-dihydroxyvita-min-D3 [75] and IL-6 family cytokines [76]. It has long beenthought that the major contributor of RANKL to bone remodellingis the osteoblast lineage, most likely osteoblast precursors [2],although RANKL has been known for many years to be expressedby newly embedded osteocytes as well [77].

Recent in vivo studies suggest that fully differentiated andmatrix-embedded osteocytes are also a significant source ofRANKL. Cell-specific knockouts demonstrated osteopetrosis inmice that lack RANKL throughout the osteoblast lineage, and a sig-nificant, but more mild, osteopetrotic phenotype, includingincreased trabecular bone mass, in mice with RANKL deletion tar-getted to late osteoblasts and osteocytes only [46,78]. Althoughcortical bone was not a focus of that work, the reduction in corticalthickness associated with hindlimb unloading, suggesting thatosteocytic RANKL may be of importance in endocortical resorptionof disuse-associated bone loss [77]. Cortical thickness may also begreater in osteocyte-specific RANKL null mice, but statistical anal-ysis of this was not reported, nor the change in endocorticalresorption.

When RANKL deletion was induced in the osteoblast lineage in26 week old mice, no detectable change in trabecular bone masswas detected, suggesting that the contribution of early osteo-blast-derived RANKL to trabecular osteoclastogenesis may be mostimportant during in trabecular bone during growth [46]. Notably,this difference in trabecular bone phenotypes between inducedosteoblast- and osteocyte-specific deletion of RANKL after develop-ment was not reproduced in a recent manuscript that achieved asimilar level of RANKL knockdown, albeit in younger mice [79].In the latter work, data was reported only for male mice, and theearlier work did not state which sex was analysed. Additionally,in direct contrast to the findings of Nakashima et al. [78] RANKLmRNA levels were higher in osteoblast-rich cell preparations ofthe later paper compared to osteocyte-rich preparations, a findingconsistent with early studies of differentiating osteoblasts in vitro[74].

Some insights into the paracrine and endocrine pathways thatdepend on osteocyte-specific RANKL may be gained from micewith osteocyte-specific deletion of pathways known to stimulateRANKL. In three examples, osteocyte-specific deletion of the vita-min D receptor (VDR), the PTH receptor (PTHR1) or the IL-6 familyco-receptor subunit gp130 revealed no osteopetrosis, nor anyalteration in RANKL expression in bone [80–82]. Furthermore,and in the case of the VDR null mice, a robust RANKL response to1a,25-dihydroxyvitamin-D3 administration was observed. Thissuggests that the main source of RANKL produced in response tothese particular stimulatory factors in bone remodelling of theadult trabecular bone is not the osteocyte.

The mechanism by which osteoblast-lineage cells make RANKLavailable to osteoclast precursors is not yet known, particularlysince early in vitro studies indicated that cell–cell contact betweencultured osteoblast-lineage cells and osteoclast precursors wasrequired for RANKL to support osteoclast formation [83]. Osteo-blast precursors are not tethered to the bone surface, so their con-tact with osteoclast progenitors can occur directly in the marrowspace. For matrix-embedded osteocytes, it has been suggested thatsoluble RANKL, or membrane-bound RANKL on exosomes may pro-vide a mechanism [84]. While this may be quite tightly controlledduring bone remodelling under the canopy of the BMU, it remainsto be established how RANKL may be provided to sites distant fromosteoblast lineage cells, where osteoclasts contribute to bone mod-elling, such as the metaphyseal periosteum.

Are signals to osteoclasts maintained when osteoblasts orosteocytes are reduced?

If signals from the osteoblast are key to maintaining osteoclastnumbers, what happens when osteoblasts are depleted? A numberof genetic and pharmacological models provide clues. While someosteoblast-lineage specific deletions that lead to low osteoblastnumbers, such as PTHrP [85], also cause low levels of osteoclasts,

N.A. Sims, C. Vrahnas / Archives of Biochemistry and Biophysics 561 (2014) 22–28 27

there are others, such as deletion of b-catenin, where osteoblastnumbers are reduced and osteoclast formation is increased [86].There are still other instances where osteoblast numbers are low,but osteoclast numbers are retained at normal levels, as whengp130 or osterix were deleted in osteoblasts [82,87] (see below).These differences may relate to a shift in the numbers of osteo-blasts at different stages of differentiation, and therefore RANKLexpression, as observed with sEphB4 treatment (see above). Alter-natively, they may indicate a loss of paracrine stimulation ofRANKL by each of these factors.

In more extreme examples, such as when osteoblasts are com-pletely ablated in adult mice by stem cell mobilisation therapies[88,89] the surprising finding was that there was no reduction inosteoclast numbers. Similar observations were made when genet-ically altered mice were generated with a thymidine kinase-medi-ated inducible deletion of osteocalcin-expressing osteoblasts [90].The latter model suggests that the cells in the osteoblast lineagerequired to support osteoclast formation are not the matrix-pro-ducing osteoblasts. Rather, it is osteoblast precursors and/ornon-dividing cells, such as osteocytes that support trabecular boneremodelling.

In contrast to the osteoblast ablation studies, diphtheria toxin-induced deletion of osteocytes [91] led to a high level of resorption.This suggests either that osteocytes inhibit osteoclast formation(suggesting that their expression of OPG is more significant thantheir expression of RANKL), or supports the concept that osteocyteapoptosis, induced in this case, by diphtheria toxin, stimulatesosteoclast formation, as occurs with microdamage [92].

Role of osteocytic IL-6 family cytokines in trabecular andcortical bone

Another example of low bone formation in the presence of nor-mal osteoclastogenesis is mice in which gp130, the IL-6 cytokinefamily transducing receptor subunit, was specifically deleted inthe osteoblast lineage [82]. The IL-6 family is a large family of cyto-kines, including IL-6, interleukin 11 (IL-11), leukemia inhibitoryfactor (LIF), cardiotrophin-1 (CT-1) and oncostatin M (OSM). Eachof these cytokines acts by forming a complex that includes a com-mon transmembrane receptor subunit, glycoprotein 130 (gp130)[93].

Most members of the IL-6 family of cytokines have been shownto robustly stimulate osteoclast formation in vitro [94,95], anaction mediated by stimulation of RANKL expression by culturedosteoblasts [11,96]. These findings led to the understandingthat these cytokines are largely pro-osteoclastic. However, manyof these same cytokines also stimulate bone formation[25,31,93,97–100].

The anabolic action of IL-6 cytokines appears to rely on two keymechanisms, firstly stimulating osteoblast commitment at theexpense of adipogenesis [25,31] while also inhibiting expressionof the osteocyte-specific Wnt signalling antagonist sclerostin[25]. However, it seems that the osteocytic pathway of action ismost important for bone formation during bone remodelling, sincemice lacking gp130 in the full osteoblast lineage and mice lackinggp130 only in osteocytes demonstrate essentially the same pheno-type [82]. In both models, trabecular bone mass was low due to alow level of trabecular bone formation, while periosteal diameterwas increased, indicating region-specific roles for gp130 cytokinesin bone formation.

In neither male nor female mice was there a significant changein osteoclastogenesis in either the osteocyte or osteoblast lineageknockouts for gp130 [82]. This suggests that the stimulatoryactions of IL-6 family cytokines on RANKL expression and osteo-clast generation by the osteoblast lineage, while impressive

in vitro, may be of little relevance to physiological bone remodel-ling. Their key role is likely to be one that mediates pathologicalbone resorption, consistent with previous reports that IL-6 nullmice do not show increased osteoclastogenesis in response toovariectomy [101] or inflammatory arthritis [102]. The low osteo-clast numbers observed in mice with global deletion of individualIL-6 family members, OSMR, IL-11R [25,26], are therefore likely torelate to signals from non-osteoblast lineage cells. Surprisinglythough, when osteoclastogenesis was stimulated from OSMR orIL-11R null bone marrow macrophages in vitro, there was noimpairment [25,26]. The key source of pro-osteoclastogenic factorsthat respond to IL-6 family cytokines in normal bone remodellingremains unknown.

The increased periosteal growth in the osteocyte-specific gp130null model [82] indicates that IL-6 family cytokines play an inhib-itory role in modelling periosteal bone. One cytokine that may becapable of this is ciliary neurotrophic factor (CNTF), the only IL-6family cytokine reported to inhibit bone formation [103]. Notably,CNTF is expressed at high levels in skeletal muscle, suggesting thatthe increased periosteal growth in the absence of osteocytic gp130may relate to reduced paracrine signals from the adjacent muscle,a control mechanism specific to the periosteal surface [104].

The findings of a low level of trabecular bone formation andincreased cortical dimensions elicited by gp130 deletion in osteo-cytes contrasts with osteocyte-specific knockouts of other factorsthat stimulate bone formation (b-catenin, Pkd1 and IGF-1); theseall resulted in low bone mass of both cortical and trabecular bonedue to reduced bone formation at both sites [105–107]. It is alsopossible that the increased periosteal growth of the osteocyte-spe-cific gp130 null mouse occurs to compensate for poor bone quality;cortical bone demonstrated reduced yield strength, there was agreater proportion of woven bone, and production of collagen typeI subunits and osteocalcin was lower than controls [82].

Thus signalling of IL-6 family cytokines in osteocytes controlsbone formation in opposite ways on trabecular and periosteal sur-faces; in addition this family of cytokines regulates the collagendistribution in cortical bone through the osteocyte, possibly bylimiting remodelling of cortical bone.

Concluding comments

In conclusion, the field has advanced to the point that we arenow beginning to have a better appreciation of the key intercellu-lar signals that control bone formation and resorption both in thecontext of bone remodelling and modelling. Since the 1950s, ourview of the BMU has extended beyond the immediate interactionsof cells on the surface of bone. Osteoclasts are still understood toproduce coupling factors that stimulate remodelling on trabecularsurfaces; it now seems that they can also signal, through the corti-cal bone, to stimulate bone modelling on the periosteum, althoughtheir method of access still needs to be defined. In addition, thecontributions of RANKL derived from cells at specific stages ofosteoblast differentiation (including osteocytes) to bone remodel-ling and modelling, and their regulation at each stage by endocrineand paracrine factors, remains controversial. Finally, we note thatIL-6 family cytokines can have both a stimulatory effect on boneformation in remodelling, but inhibitory effects on modelling atthe periosteum, a difference likely to be mediated by cells presentin the marrow or on the periosteum that release factors that are yetto be identified.

References

[1] T.J. Martin, E. Seeman, Best Pract. Res. Clin. Endocrinol. Metab. 22 (5) (2008)701–722.

[2] G.A. Rodan, T.J. Martin, Calcif. Tissue Int. 33 (4) (1981) 349–351.

28 N.A. Sims, C. Vrahnas / Archives of Biochemistry and Biophysics 561 (2014) 22–28

[3] H.M. Frost, Bone Biodyn. (1964) 315–333.[4] T.J. Martin, N.A. Sims, Trends Mol. Med. 11 (2) (2005) 76–81.[5] A.M. Parfitt, Bone 30 (1) (2002) 5–7.[6] A.M. Parfitt, in: R. Marcus et al. (Eds.), Fundamentals of Osteoporosis,

Academic Press, London, 2010, pp. 35–54.[7] N.A. Sims, T.J. Martin, BoneKEy Rep. 3 (2014) 481.[8] Y.Y. Kong et al., Nature 397 (6717) (1999) 315–323.[9] Q. Fu et al., J. Biol. Chem. 277 (50) (2002) 48868–48875.

[10] N.J. Horwood et al., Endocrinology 139 (11) (1998) 4743–4746.[11] P. Palmqvist et al., J. Immunol. 169 (6) (2002) 3353–3362.[12] K. Henriksen, M.A. Karsdal, T.J. Martin, Calcif. Tissue Int. 94 (1) (2014) 88–97.[13] N.A. Sims, T.J. Martin, Bonekey. Rep. 3 (2014) 481.[14] S. Tonna, N.A. Sims, Calcif. Tissue Int. 94 (1) (2014) 35–45.[15] T.L. Andersen et al., Am. J. Pathol. 174 (1) (2009) 239–247.[16] H.B. Kristensen et al., Am. J. Pathol. 184 (3) (2014) 778–789.[17] W.G. Beamer et al., Bone 18 (5) (1996) 397–403.[18] M.P. Akhter et al., Calcif. Tissue Int. 67 (4) (2000) 337–344.[19] C.H. Turner et al., J. Bone Miner. Res. 16 (2) (2001) 206–213.[20] V. Glatt et al., J. Bone Miner. Res. 22 (8) (2007) 1197–1207.[21] N.A. Sims et al., Bone 30 (1) (2002) 18–25.[22] N.A. Sims et al., J. Clin. Invest. 106 (9) (2000) 1095–1103.[23] B.T. Kim et al., J. Bone Miner. Res. 18 (1) (2003) 150–155.[24] F. Callewaert et al., J. Bone Miner. Res. 25 (3) (2010) 617–626.[25] E.C. Walker et al., J. Clin. Invest. 120 (2) (2010) 582–592.[26] N.A. Sims et al., J. Bone Miner. Res. 20 (7) (2005) 1093–1102.[27] N.A. Sims et al., J. Clin. Invest. 113 (3) (2004) 379–389.[28] C.B. Ruff, W.C. Hayes, J. Orthop. Res. 6 (6) (1988) 886–896.[29] G.A. Howard et al., Proc. Natl. Acad. Sci. U.S.A. 78 (5) (1981) 3204–3208.[30] Y. Tang et al., Nat. Med. 15 (7) (2009) 757–765.[31] E.C. Walker et al., J. Bone Miner. Res. 23 (12) (2008) 2025–2032.[32] L. Pederson et al., Proc. Natl. Acad. Sci. U.S.A. 105 (52) (2008) 20764–20769.[33] S. Takeshita et al., J. Clin. Invest. 123 (9) (2013) 3914–3924.[34] T. Negishi-Koga et al., Nat. Med. 17 (11) (2011) 1473–1480.[35] M.A. Karsdal et al., Biochem. Biophys. Res. Commun. 366 (2) (2008) 483–488.[36] K. Henriksen et al., PLoS One 6 (11) (2011) e27482.[37] M. Marzia et al., J. Cell Biol. 151 (2) (2000) 311–320.[38] X.M. Dai et al., J. Bone Miner. Res. 19 (9) (2004) 1441–1451.[39] N. Sakagami et al., Micron 36 (7–8) (2005) 688–695.[40] J.B. Lian, S.C. Marks Jr., Endocrinology 126 (2) (1990) 955–962.[41] N. Kim et al., Proc. Natl. Acad. Sci. U.S.A. 97 (20) (2000) 10905–10910.[42] M.J. Perry et al., Bone 42 (1) (2008) 216–225.[43] N.A. Sims, K.W. Ng, Curr. Osteoporos. Rep. 12 (1) (2014) 98–106.[44] S. Lotinun et al., J. Clin. Invest. 123 (2) (2013) 666–681.[45] K. Fuller et al., Bone 42 (1) (2008) 200–211.[46] J. Xiong et al., Nat. Med. 17 (10) (2011) 1235–1241.[47] R.M. Neer et al., N. Engl. J. Med. 344 (19) (2001) 1434–1441.[48] H. Dobnig, R.T. Turner, Endocrinology 136 (8) (1995) 3632–3638.[49] R.L. Jilka et al., J. Clin. Invest. 104 (4) (1999) 439–446.[50] D. Stanislaus et al., Bone 27 (2) (2000) 209–218.[51] T. Bellido et al., Endocrinology 146 (11) (2005) 4577–4583.[52] H. Keller, M. Kneissel, Bone 37 (2) (2005) 148–158.[53] A.M. Parfitt, Calcif. Tissue Int. 36 (Suppl 1) (1984) S37–S45.[54] T. Martin, J.H. Gooi, N.A. Sims, Crit. Rev. Eukaryot. Gene Expr. 19 (1) (2009)

73–88.

[55] M.E. Holtrop et al., Calcif. Tissue Int. 27 (2) (1979) 129–135.[56] P.D. Delmas et al., Bone 16 (6) (1995) 603–610.[57] B. Demiralp et al., Endocrinology 143 (10) (2002) 4038–4047.[58] A.J. Koh et al., Endocrinology 146 (11) (2005) 4584–4596.[59] J.H. Gooi et al., Bone 46 (6) (2010) 1486–1497.[60] J.H. Gooi et al., J. Endocrinol. 221 (2) (2014) 181–191.[61] D.M. Black et al., N. Engl. J. Med. 349 (13) (2003) 1207–1215.[62] J.S. Finkelstein et al., J. Clin. Endocrinol. Metab. 95 (4) (2010) 1838–1845.[63] B.Z. Leder et al., J. Clin. Endocrinol. Metab. 88 (1) (2003) 204–201.[64] C. Zhao et al., Cell Metab. 4 (2) (2006) 111–121.[65] E.H. Allan et al., J. Bone Miner. Res. 23 (8) (2008) 1170–1181.[66] K.K. Murai, E.B. Pasquale, J. Cell Sci. 116 (Pt 14) (2003) 2823–2832.[67] E.F. Eriksen et al., Metab. Bone Dis. Relat. Res. 5 (5) (1984) 243–252.[68] B. Ecarot-Charrier et al., J. Cell Biol. 96 (3) (1983) 639–643.[69] Y. Abe et al., Calcif. Tissue Int. 52 (5) (1993) 365–371.[70] I. Gerber, I. ap Gwynn, Eur. Cell Mater. 2 (2001) 10–20.[71] F.M. Takyar et al., J. Bone Miner. Res. 28 (4) (2013) 912–925.[72] T.J. Martin et al., Adv. Exp. Med. Biol. 658 (2010) 51–60.[73] A. Arthur et al., J. Bone Miner. Res. 28 (4) (2013) 926–935.[74] G.P. Thomas et al., J. Endocrinol. 170 (2) (2001) 451–460.[75] S. Kim et al., J. Steroid Biochem. Mol. Biol. 103 (3–5) (2007) 430–434.[76] S. Kim et al., Mol. Endocrinol. 21 (1) (2007) 197–214.[77] V. Kartsogiannis et al., Bone 25 (5) (1999) 525–534.[78] T. Nakashima et al., Nat. Med. 17 (10) (2011) 1231–1234.[79] T. Fumoto et al., J. Bone Miner. Res. 29 (4) (2014) 830–842.[80] L. Lieben et al., J. Clin. Invest. 122 (5) (2012) 1803–1815.[81] W.F. Powell Jr et al., J. Endocrinol. 209 (1) (2011) 21–32.[82] R.W. Johnson et al., J. Bone Miner. Res. 29 (6) (2014) 1492–1505.[83] N. Takahashi et al., Endocrinology 123 (5) (1988) 2600–2602.[84] G. Kogianni, V. Mann, B.S. Noble, J. Bone Miner. Res. 23 (2008) 915–927.[85] D. Miao et al., J. Clin. Invest. 115 (9) (2005) 2402–2411.[86] S.L. Holmen et al., J. Biol. Chem. 280 (22) (2005) 21162–21168.[87] W.-Y. Baek et al., J. Bone Miner. Res. 24 (6) (2009) 1055–1065.[88] I.G. Winkler et al., Leukemia 26 (7) (2012) 1594–1601.[89] I.G. Winkler et al., Blood 116 (23) (2010) 4815–4828.[90] D.A. Corral et al., Proc. Natl. Acad. Sci. U.S.A. 95 (23) (1998) 13835–13840.[91] S. Tatsumi et al., Cell Metab. 5 (6) (2007) 464–475.[92] M. Schaffler et al., Calcif. Tissue Int. 94 (1) (2014) 5–24.[93] N.A. Sims, N.C. Walsh, BMB Rep. 43 (8) (2010) 513–523.[94] T. Tamura et al., Proc. Natl. Acad. Sci. U.S.A. 90 (24) (1993) 11924–11928.[95] C.D. Richards et al., Cytokine 12 (6) (2000) 613–621.[96] C.A. O’Brien et al., J. Biol. Chem. 274 (27) (1999) 19301–19308.[97] D. Metcalf, D.P. Gearing, Proc. Natl. Acad. Sci. U.S.A. 86 (15) (1989) 5948–

5952.[98] J. Cornish et al., Bone 21 (3) (1997) 243–247.[99] L. Malaval, A.K. Gupta, J.E. Aubin, Endocrinology 136 (4) (1995) 1411–1418.

[100] I.J. Poulton et al., J. Bone Miner. Res. 27 (3) (2012) 586–595.[101] V. Poli et al., EMBO J. 13 (5) (1994) 1189–1196.[102] P.K. Wong et al., Arthritis Rheum. 54 (1) (2006) 158–168.[103] N.E. McGregor et al., Calcif. Tissue Int. 86 (3) (2010) 261–270.[104] R.W. Johnson et al., Bone 64 (2014) 47–56.[105] Z. Xiao et al., FASEB J. 25 (7) (2011) 2418–2432.[106] I. Kramer et al., Mol. Cell. Biol. 30 (12) (2010) 3071–3085.[107] M.H. Sheng et al., Bone 52 (1) (2013) 133–144.