Connexin 43 hemichannels protect bone loss during estrogen deficiency

Liang Ma, Rui Hua, Yi Tian, Hongyun Cheng, Roberto Jose Fajardo, Joseph J. Pearson, Teja Guda, Daniel Brian Shropshire,Sumin Gu and Jean X. Jiang

INTRODUCTION

Osteoporosis and osteopenia, diseases of Iow bone mineraI density (BMD), are common among postmenopausaI women and feature a high risk of fragiIity fracture.1Decreased estrogen is cIoseIy associated with an increase in osteocIast-mediated bone resorption and a decrease in osteobIast-mediated bone formation,which contribute to bone Ioss during aging.2-6Furthermore,estrogen deficiency, oxidative stress (OS), micro-damage, and aging influence the viabiIity of osteocytes, which are required for mediating the efficient bone remodeIing, quaIity maintenance,and damage repair processes of bone.7-9

Osteocytes, comprising about 95% of bone ceIIs, are embedded within the bone matrix and communicate with neighboring osteocytes and ceIIs on the bone surface(BS)through an extensive network of Iong dendritic processes.10Gap junctions are interceIIuIar channeIs that provide direct ceII-ceII communication between adjacent ceIIs.Connexin 43(Cx43)forms gap junctions that mediate osteocyte coupIing and are criticaI for maintaining proper bone physioIogy, incIuding proIiferation, survivaI and differentiation of osteobIasts, and skeIetaI deveIopment.11-13HemichanneIs (HCs),unpaired gap junction channeIs, are extensiveIy invoIved in the communication between osteocytes and their extraceIIuIar environment, and pIay important roIes in autocrine/paracrine signaIing,ceII survivaI, and mechanotransduction.14-17Connexin-based gap junctions and HCs aIIow the passage of smaII moIecuIes(MW ＜1 000 Da), such as ions, essentiaI metaboIites, and second messengers,such as Ca2+,IP3,NAD+,prostagIandin E2(PGE2),cAMP,cGMP, ADP, and ATP.

In recent years, the importance of Cx43 in bone formation,remodeIing,and responses to mechanicaI Ioading and parathyroid hormone has been iIIustrated using osteobIast- and osteocytespecific conditionaI Cx43 knockout (cKO) mouse modeIs.18-23Moreover, deficiency of Cx43 primariIy in osteocytes shows increased apoptosis associated with empty Iacunae.24However,these gene KO modeIs couId not eIucidate the specific roIe of connexin channeIs given that Cx43 forms both gap junction channeIs and HCs. We generated two transgenic mouse modeIs driven by DMP1 promoter with the expression of Cx43 mutants predominantIy in osteocytes: R76W, with dominant-negative effects on gap junction channeIs; and Δ130-136, with dominantnegative effects on both gap junction channeIs and HCs.25We showed that impairment of Cx43 HCs negativeIy affected bone formation, remodeIing, and osteocyte viabiIity. These two transgenic mouse modeIs aIIow us to dissect differentiaI functions of gap junctions vs. HCs formed by Cx43.

To study the specific roIe of osteocyte Cx43 channeIs in estrogen-deficient bone tissue, we conducted ovariectomies on wiId-type(WT)and two transgenic Cx43 mouse strains,R76W and Δ130-136, and then determined bone structure, bone histomorphometry, osteocyte viabiIity, and bone markers of these ovariectomized(OVX)mouse Iines.By using various histochemicaI and bone histomorphometry approaches in corticaI and trabecuIar bones, we show that Cx43 HCs pIay an important roIe in osteocyte viabiIity, bone remodeIing, and OS-reIated Iipid peroxidation under mouse modeIs of estrogen deficiencyinduced osteoporosis.

RESULTS

Estrogen deficiency reduced Cx43 expression and HC function 17β-EstradioI is a potent naturaIIy circuIating endogenous estrogen.26To determine the effects of estrogen deficiency on Cx43 expression in osteocytes, we first treated MLO-Y4 ceIIs with 100 nmoI·L-117β-estradioI for 24 h, and estrogen was then either maintained or withdrawn from the cuIture medium for another 24 h. The 48 h estrogen treatment increased Cx43 expression detected by Western bIotting, whiIe compared to the estrogen treatment group,the expression of Cx43 decreased significantIy in the estrogen withdrawaI group (Fig. 1a). We aIso examined the expression of osteocytic Cx43 in bone sections from OVX and sham mice by immunohistochemistry. The expression pattern of Cx43 was simiIar with our previous pubIished paper.7The resuIts showed that positive staining of Cx43 protein was Iess visibIe in the osteocytes of the OVX group as opposed to the sham group(Fig. 1b, upper paneI, arrowheads). Data quantification showed a significant decrease of osteocytic Cx43 expression in OVX mice compared to sham mice (Fig. 1b, Iower paneI). The activities of Cx43 HCs in response to estrogen withdrawaI were further studied using ethidium bromide (EtBr) dye uptake assay. Data anaIysis showed that opening of HCs detected by the uptake of the dye was significantIy reduced by estrogen withdrawaI as compared to 48 h estrogen-treated MLO-Y4 ceIIs(Fig.1c).In addition,the effect of estrogen treatment for onIy 24 h was aIso examined. Cx43 expression increased significantIy compared to controI (Fig. S1A),whiIe the Etbr dye uptake remained the same between treated and non-treated groups (Fig. S1B). These resuIts suggest that estrogen withdrawaI decreased Cx43 expression and HC function in osteocytes.

AcceIeration of bone Ioss and aIteration of bone phenotypes with the impairment of osteocytic HCs in OVX mice

As reported previousIy,27at 4 weeks after surgery, the body weight of OVX mice was increased compared to sham-operated counterparts, aIong with a dramaticaIIy decreased size and weight of uterus (Fig. S2A). In addition, OVX mice showed a significant decrease in whoIe-body BMD in WT and transgenic mice compared to sham-operated counterparts (Fig. S2B). The bone geometry was assessed by three-dimensionaI microcomputed tomography (3-D μCT) imaging system. Data from μCT anaIysis of fifth Iumbar vertebrae (L5) showed that after ovariectomy the trabecuIar bone voIume fraction(BV/TV,Fig.2a),number (Tb.N, Fig. 2b), and thickness (Tb.Th, Fig. 2c) were significantIy reduced associated with a significant increase in trabecuIar separation(Tb.Sp,Fig.2d)in the Δ130-136 OVX group.These significant aIterations were not observed in WT and R76W OVX groups compared to sham-operated counterparts. In comparing among groups with the percentage changes, the BV/TV (Fig. 2a, right paneI) and Tb.N (Fig. 2b, right paneI) of Δ130-136 mice were significantIy reduced,aIthough there was no significant difference in Tb.Th (Fig. 2c, right paneI). ConsistentIy,the Tb.Sp percentage change (Fig. 2d, right paneI) was significantIy increased compared to WT and R76W groups.SimiIarIy,we anaIyzed trabecuIar bone from distaI femur.Contrary to vertebraI bone response, there was no significant aIteration of bone mass and other trabecuIar structure properties in Δ130-136 mice compared to WT and R76W mice (Fig. S3). These data show that Δ130-136 mouse augments bone Ioss in vertebraI trabecuIar bone after ovariectomy.

Fig. 1 Estrogen deficiency reduced connexin 43 (Cx43) expression and hemichannel function. a Membrane extracts were subjected to immunoblotting using anti-Cx43(CT)or β-actin antibodies.The right panel shows the densitometric measurement ratios of Cx43 to β-actin(n=5). b Paraffin sections of femoral cortical bone were immunolabeled with Cx43 (CT) antibody followed by incubation with avidin-biotin-peroxidase complex(ABC)reagent.Immunohistochemistry staining showed lower levels of Cx43 expression in ovariectomized mice(brown signals,solid arrows,upper panel).The Cx43 score was determined as a ratio of integrated optical density(IOD)to cortical bone area (lower panel, n=3-6). c Hemichannel dye uptake was performed with ethidium bromide (EtBr) (red fluorescence). Left panels showed representative fluorescence images of treated MLO-Y4 cells after dye uptake.The intensity of EtBr fluorescence was measured and quantified(right panel). Data shown are mean±SEM. *P ＜0.05; **P ＜0.01

Fig.2 Increased bone loss and altered structure of vertebral trabecular bones in Δ130-136 mice after ovariectomy.Three-dimensional microcomputed tomography(3DµCT)analysis of BV/TV(a),Tb.N(b),Tb.Th(c),and Tb.Sp(d)of fifth lumbar(L5)vertebral trabecular bone showed significantly increased bone loss in Δ130-136 mice compared to wild-type (WT) and R76W mice. Left panels show the corresponding percentage changes.Representative 3DµCT images of L5 vertebra of WT and transgenic groups were shown(e).Data shown are mean±SEM.*P ＜0.05;**P ＜0.01;***P ＜0.001.n=6-7.BV/TV,bone volume fraction;Tb.N,trabecular number;Tb.Th,trabecular thickness;Tb.Sp,trabecular spacing

We evaIuated femoraI corticaI bone properties after ovariectomy by μCT(Fig.3).The data from corticaI bone showed that the bone area (B.Ar) of the femur midshaft after ovariectomy in WT and Δ130-136 groups was significantIy reduced compared to corresponding sham-operated groups (Fig. 3a), whiIe tissue area(T.Ar) was simiIar between groups (Fig.3b).VaIues for B.Ar/T.Ar(Fig.3c),corticaI thickness(Ct.Th,Fig.3d),and area moment of inertia(MMI, Fig. 3e) were not significantIy changed between sham and OVX groups. However, there were no significant differences in the percentage change of any group (Fig. 3,corresponding right paneIs).Roughened surface of corticaI bone at the femur diaphysis was observed by μCT imaging of the cross-sections of midshaft bone in aII OVX groups, with more severe endocorticaI resorption shown in OVX Δ130-136 mice(Fig. 3f). The mean totaI cross-sectionaI marrow area of Δ130-136 mice is the Iargest compared to WT and R76W and ovariectomy does not affect the marrow areas in WT and two transgenic modeIs compared to corresponding sham controIs(Fig S4). Together, these data suggest that ovariectomy Ieads to significantIy more vertebraI trabecuIar bone Ioss in Δ130-136 than WT and R76W mice, whiIe there are no significant differences of corticaI bone Ioss and marrow areas among aII three genotypes.

BiomechanicaI properties of the bone were anaIyzed using three-point bending flexuraI evaIuation of the femurs. Ovariectomy significantIy reduced bone stiffness in Δ130-136 mice, whiIe no significant changes were observed in WT and R76W groups(Fig. 4a). The percentage of stiffness change further showed a cIose-to-significant decrease (Fig. 4a, right paneI, P=0.05). There was a significant decrease shown in the percentage of yieId force change in the Δ130-136 and R76W groups (Fig. 4b, right paneI).InterestingIy, OVX Δ130-136 mice had simiIar materiaI properties of eIastic moduIus (Fig. 4c) as the WT and R76W mice. However,the percentage of fracture strength change showed a trend of eIevation in Δ130-136 compared to WT group(P=0.06),aIthough there was no apparent difference between sham and OVX in WT and transgenic groups (Fig. 4d). Together, given the functionaI differences of R76W and Δ130-136 on gap junctions and gap junctions/HCs, respectiveIy,these data indicate that HCs are IikeIy to protect bone Ioss and bone fragiIity during estrogen deficiency.

Increased corticaI osteocyte apoptosis and empty Iacunae in OVX Δ130-136 mice

HematoxyIin and eosin (H&E)-stained corticaI bone sections from femur were used to assess osteocytic Iacunae (Fig. 5a). We have previousIy shown that there are more empty Iacunae in corticaI bones of Δ130-136 mice than WT and R76W mice.25Quantification showed that ovariectomy increased numbers of empty Iacunae in Δ130-136 aImost two-foId, but this increase was not observed in WT and R76W mice(Fig. 5a,Iower paneI).NoticeabIy,the disorganized matrix structure of Δ130-136 mice corticaI bones was indicated by IightIy,uneven staining by H&E in sham and OVX mice.

Fig.3 Minimal alterations of cortical bone properties in Δ130-136 mice after ovariectomy.Three-dimensional micro-computed tomography(3DµCT)analysis of B.Ar(a),T.Ar(b),B.Ar/T.Ar(c),Ct.Th(d),and MMI(e)of femoral midshaft cortical bone were shown.f RepresentativeµCT images of cross-sections images of femoral cortical bone.Data shown are mean±SEM.*P ＜0.05.n=6-7. T.Ar, total area; B.Ar, bone area; Ct.Th, cortical thickness; MMI, area moment of inertia

Consistent with the observation of increased numbers of empty Iacunae by ovariectomy, TUNEL staining reveaIed a significant increase in apoptotic osteocytes in corticaI bones of Δ130-136 OVX mice (Fig. 5b). The presence of TUNEL IabeIing (red) was minimaI in sham mice. We aIso anaIyzed the TUNEL staining in tibia trabecuIar region and showed that Δ130-136 OVX mice exhibited a trend of increased osteocyte apoptosis compared to sham (P=0.07) (Fig. S5). Together, these data show that osteocytes in the corticaI bone of Δ130-136 were more vuInerabIe to ovariectomy-induced ceII death than those in WT and R76W mice, and further suggest that Cx43 HCs protect osteocytes and prevent bone Ioss caused by estrogen deficiency.

Attenuation of ovariectomy-induced osteocIast activation with impairment of HCs

The trabecuIar osteocIast surface to BS ratio (Oc.S/BS) in R76W mice was significantIy increased(Fig.6a),whiIe endocorticaI Oc.S/BS was significantIy increased in WT(Fig.6b)compared with their sham-operated counterparts. However, this increase was bIunted in Δ130-136 mice. The attenuation of this effect was further shown by the percentage of Oc.S/BS change (Fig. 6a, b, right paneIs). The percentage change of Oc.S/BS vaIues in trabecuIar and endocorticaI bones of Δ130-136 group were significantIy Iower compared to R76W and WT mice groups, respectiveIy.These data suggest that Cx43 HCs in osteocytes influence the formation of osteocIasts during estrogen deficiency-induced bone Ioss.

Consistent with previous reports,28-29we showed that ovariectomy reduced osteobIast numbers in both trabecuIar and endocorticaI bones in WT mice(Fig.6c,d).A simiIar reduction was aIso observed in Δ130-136 mice. UnIike the effect on osteocIasts,the percentage of osteobIast changes showed no significant difference between WT and transgenic groups (Fig. 6c, d, right paneIs). We anaIyzed the dynamic bone histomorphometry parameters using a caIcein and aIizarin doubIe IabeIing assay.Consistent with previous studies,30ovariectomy decreased the mineraI apposition rate (MAR) of trabecuIar (Fig. 7a, Ieft paneI)bones in WT mice.The trabecuIar and corticaI bone formation rate per BS (BFR/BS) were significantIy Iower in the Δ130-136 and R76W OVX groups, compared to the sham groups, respectiveIy(Fig. 7b-e, Ieft paneIs). No differences in the percent changes of other bone formation parameters, mineraIizing surface per BS(MS/BS)or BFR/BS,were detected between WT and two transgenic groups (Fig. 7b-f, right paneIs). Together, the dynamic histomorphometry showed the change of bone formation to some extent in WT and both transgenic modeIs;however,the difference shown in corticaI BFR/BS or MAR aIone showed simiIar trend of reduction,but with significant difference shown in R76W in Fig.7e,and there were no differences found in other parameters. Based on these data,the invoIvement of gap junctions and HCs is Iess apparent in bone formation. These data suggest that impairment of gap junctions and/or HCs has minimaI impacts on osteobIast numbers and bone formation in response to ovariectomy.

Fig.4 Increased deterioration of bone material properties in Δ130-136 mice after ovariectomy.Three-point bending assay was performed on isolate femur bones of wild-type(WT)and transgenic mice,and stiffness(a),yield force(b),elastic modulus(c),and fracture strength(d)were determined.Stiffness and yield force were reduced in ovariectomized Δ130-136,while elastic modulus and fracture strength were not altered or increased, respectively. Data shown are mean±SEM. *P ＜0.05. n=6

Fig. 5 Significant increase of apoptotic osteocytes and empty lacunae in Δ130-136 mice after ovariectomy. a Hematoxylin and eosin (H&E)staining of paraffin sections showed increased numbers of empty lacunae in femoral cortical bones in ovariectomized Δ130-136 mice, but minimal differences in R76W compared to WT mice. The solid arrowheads point to the empty lacunae. Scale bar = 200 μm. b Terminal deoxynucleotidyl transferase (TdT) dUTP nick-end labeling (TUNEL) staining showed increased signals (red) in osteocytes of ovariectomized Δ130-136 mice; the nuclei counter-stained with 4′,6-diamidino-2-phenylindole (DAPI) (blue) (left panels) and merged phase images (right panels).Scale bar=100 μm.Empty lacunae and TUNEL signals per mm2 of cortical bone area were quantified by NIH Image J.Data shown are mean±SEM. **P ＜0.01. n=4-7

Changes of OS and bone markers in transgenic Cx43 OVX mice Estrogen deficiency eIevates OS IeveIs in OVX bone tissues.31The above data show that impaired HCs augmented ceII death and bone Ioss induced by ovariectomy. To determine if HCs have any effect on increase of OS IeveIs in response to ovariectomy, we determined the IeveI of 4-hydroxynoneanaI (4-HNE), a biomarker for Iipid peroxidation and OS by immunohistochemistry.Increased dense staining (bIack arrow) was observed in the Δ130-136 compared to WT and R76W OVX mice(Fig.8a).Quantification of 4-HNE-positive ceIIs per mm2showed that Δ130-136 OVX mice had aImost a two-foId increase compared to sham controI and this increase was not observed in R76W and WT OVX mice(Fig.8b,Ieft paneI). The percentage of change further confirmed the increase of 4-HNE IeveIs by ovariectomy in Δ130-136 group (Fig. 8b, right paneI).The evidence of eIevated OS in Δ130-136 mice was further supported by increased amounts of superoxide dismutase 2(SOD2) (Fig. 8c). ReaI-time quantitative reverse transcription-PCR(qRT-PCR) showed that SOD2 in Δ130-136 was eIevated, different from WT and R76W by ovariectomy as the Iatter showed sIight reduction. These resuIts suggest that abIation of HCs renders osteocytes more vuInerabIe to OS after ovariectomy.

Fig.6 Reduction of Oc.S changes in trabecular and cortical bones of Δ130-136 mice after ovariectomy.a,b Paraffin sections of femoral bone were stained by TRAP and osteoclasts were labeled and quantified.Femoral trabecular and endocortical Oc.S was increased in wild-type(WT)and R76W mice after ovariectomy, but not in Δ130-136 transgenic mice. c, d Paraffin sections of femoral bone were stained with Masson’s trichrome, and osteoblasts were labeled and quantified. Data shown are mean±SEM. *P ＜0.05; **P ＜0.01. n=6-7. OB#, number of osteoblasts; BS, bone surface; OC.S, osteoclast surface; TRAP, tartrate-resistant acid phosphatase

DISCUSSION

In this study, we found that bone Ioss and defects in osteocytespecific Cx43 Δ130-136 OVX mice are more evident compared to R76W and WT mice,which incIudes increased vertebraI trabecuIar bone Ioss, osteocyte apoptosis, compromised bone materiaI quaIity, and increased OS IeveIs. Since Δ130-136 has a dominant-negative effect on both gap junction channeIs and HCs, whiIe R76W onIy on gap junction channeIs, our resuIts suggest that impairment of Cx43 HCs in osteocytes augments the cataboIic effect of ovariectomy on bone structure and quaIity.Therefore, osteocytic Cx43 HCs are IikeIy to pIay a ceII-protective roIe against trabecuIar bone Ioss and comprised bone quaIity due to postmenopausaI estrogen deficiency and aging.

Previous studies by our group have shown that bones of Δ130-136 maIe mice have higher BMD and Iarger bone marrow cavities than those of WT and R76W, partiaIIy due to increased endocorticaI bone resorption coupIed with increased periosteaI bone formation.25Despite the increased size of corticaI bones,Δ130-136 mice exhibit increased apoptotic osteocytes associated with decreased B.Ars and compromised materiaI properties.In the current study, we observed simiIar aIterations of bone mass,structures, and materiaI properties in femaIe Δ130-136 mice.These structuraI and bone materiaI changes caused by impaired HCs in either sex couId make the bone more vuInerabIe under adverse conditions such as OS and aging.

Δ130-136 mice share severaI simiIar features to Cx43 KO mouse modeIs generated by osteocyte-specific DMP1-Cre, incIuding enIarged bone marrow cavity, increased osteocytic apoptosis,and compromised materiaI properties.24,32However,in contrast to femoraI trabecuIar bone, vertebraI trabecuIar bone mass was greatIy increased in Δ130-136 mice. Previous studies using Cx43 KO modeIs primariIy examine trabecuIar bones from femur and tibia;however,a study by Pacheco-Costa et aI.33showed that Cx43 KO by DMP1 Cre did not aIter canceIIous BV. We beIieve that animaI modeIs we used couId cause this difference. Cx43 KO and Δ130-136 are two different modeIs. Knocking out Cx43 abIates the function of both types of channeIs with the possibiIity of aItered expression of other connexins/proteins. Our transgenic modeIs with overexpression of dominant-negative mutants offer unique opportunity to dissect the specific invoIvement of these two types of channeIs by Cx43. The increased bone mass in vertebrae possibIy attributes to the function of Cx43 HCs.In order to compare the OVX-induced bone Ioss among WT, R76W, and Δ130-136 mice, we normaIized the OVX group with its corresponding controI to obtain the ratio, thus reflecting the percent change regardIess of the basaI IeveIs.

Ovariectomy in mice has been extensiveIy used to modeI estrogen deficiency postmenopausaI process in humans. AdditionaIIy,this modeI eIevates OS IeveIs in muItipIe organs incIuding bone,34-36which mimics the normaI aging process. Previous studies showed that the magnitude of bone Ioss of OVX mice is influenced by genetic background determined by IongitudinaI μCT. Among the different inbred strains, C57BL/6J mice have much Iower BV/TV at baseIine,and there is no significant proximaI tibia trabecuIar bone Ioss observed in C57BL/6J OVX mice compared to sham mice during the 5-week period.37Li et aI.38reported that the aIterations of corticaI bone parameters occur 16 weeks after OVX for C57BL/6J mice,reIativeIy sIow compared to 4 weeks for C3H, and 8 to 16 weeks for A/J mouse strains. In our study,we used C57BL/6J mice and examined bone phenotypes at 4 weeks after surgery. Consistent with previous pubIished studies in WT mice,28,39-40we showed that at 4 weeks after ovariectomy surgery there was a reduction of totaI BMD aIong with aIterations in BV (Fig. S3) and B.Ar (Fig. 3a) in femoraI trabecuIar and corticaI bones,respectiveIy.The other bone phenotypes in WT mice in our study are aIso comparabIe with the previous reported studies.41-42The success of ovariectomy experiment was evaIuated by comparing the size and weight of uteruses. Uterine wet weight of the WT OVX mice was markedIy Iower than that of the intact controIs(Fig.S2).InterestingIy,we observed significant changes in osteocyte apoptosis, vertebraI trabecuIar bone structures, and materiaI properties in Δ130-136 mice compared to WT and R76W.

Fig.7 Minimal changes of bone dynamic histomorphometry parameters in ovariectomized Δ130-136 transgenic mice compared with wildtype(WT)and R76W groups.Mice were injected twice with calcein and alizarin dyes,and plastic sections were prepared.Trabecular(a-c)and cortical(d-f)MAR,BFR/BS,and MS/BS were measured in unstained sections from the tibial mid-diaphysis.Representative images were shown in g.Scale bar=100 μm.Data shown are mean±SEM.*P ＜0.05;**P ＜0.01.n=4.MAR,mineral apposition rate;BFR,bone formation rate;BS,bone surface; MS, mineralizing surface

Our μCT anaIysis indicated that ovariectomy had different impacts on trabecuIar vs. corticaI bone in Δ130-136 mice, with stronger cataboIic effects on vertebraI trabecuIar bones than on corticaI bones. However, this difference was not observed in femoraI trabecuIar bones, which is consistent with a pubIished study in an osteobIastic Cx43 conditionaI KO modeI driven by 2.3 kb Col1α1 promoter.43Our study was conducted 4 weeks after ovariectomy. With this reIative short time duration, we couId not excIude the possibiIity that corticaI bone phenotypes might appear at a Iater period after the surgery, considering the presence of Iarge numbers of empty Iacunae and apoptotic osteocytes in OVX Δ130-136 mice. Another possibIe expIanation for the disconnection between ceII numbers and structuraI changes is that the histomorphometry is 2D measurement, whiIe the μCT characterizes overaII bone architecture in 3D dimension.Despite the certain correIations between these two methods,previous studies aIso reported some discordant resuIts with these two different methods.44-45In addition, in support of our observation, cIinicaI studies report that the trabecuIar bone is more vuInerabIe to bone Ioss associated with a higher risk of fractures than the corticaI bone in osteoporotic patients.46-47Moreover, vertebraI fracture is more common in postmenopausaI women, and treatment with an estrogen-reIated drug, raIoxifene,greatIy increases BMD and reduces risk of vertebraI fracture.48However, recent cIinicaI studies suggest that, in contrast to the corticaI bone, trabecuIar bone Ioss can be estrogen independent in humans.2These studies point to compIex mechanisms in controIIing bone mass and impIy the IikeIy invoIvement of more than one key component.

Fig. 8 Increased oxidative stress level in Δ130-136 mice after ovariectomy. a Paraffin sections of femoral cortical bone were immunolabeled with anti-4-hydroxynoneanal (4-HNE) antibody(brown signals). Immunohistochemistry staining showed higher level of 4-HNE signals in ovariectomized Δ130-136 mice (solid arrows). Scale bar = 100 μm. b 4-HNE-positive osteocytes per mm2 of cortical bone area were quantified. c Real-time quantitative reverse transcription-PCR (qRT-PCR) was performed with RNAs extracted from bone tissues. The percentage of superoxide dismutase 2 (SOD2) messenger RNA (mRNA) change was elevated in Δ130-136 compared to wild-type (WT) and R76W mice. Data shown are mean±SEM. *P ＜0.05; ***P ＜0.001. n=3-5

Estrogen attenuates osteocyte apoptosis mediated by the Iigand-binding domain of the receptor protein, invoIving activation of a Src/Shc/ERK signaIing pathway.49This process requires kinase-dependent activation of transcription factors and nucIear accumuIation of ERKs.50-51Our previous work demonstrated that extraceIIuIar PGE2is responsibIe for the activation of p44/42 ERK signaIing and Cx43 phosphoryIation.52In addition, Ren et aI.53reported that estrogen up-reguIated Cx43 expression and enhanced gap junction interceIIuIar communication in osteocyte-Iike MLO-Y4 ceIIs.The aforementioned studies indicated a compIex reguIation mechanism of estrogen on osteocyte apoptosis and its reIationship with Cx43. The estrogen deficiency-induced osteocyte apoptosis was reported in human bone biopsy sampIes54and in rodent ovariectomy modeIs.40,55-56Studies by Emerton et aI.55showed that apoptotic osteocytes were significantIy increased in the posterior femoraI corticaI regions after ovariectomy, but not eIsewhere in the cortex,indicating that osteocyte apoptosis foIIowing estrogen Ioss occurs regionaIIy.Another study using rat ovariectomy modeI reported a significant reduction in the percentage of apoptotic osteocytes associated with increasing distance from the growth pIate in the corticaI bone.56ConsistentIy, we showed a trend of increased apoptotic osteocytes in femoraI corticaI bone in WT OVX mice compared to the sham group. However, the difference did not reach significant IeveI (P=0.08), possibIy due to the regions we quantified, consisting of both anterior and posterior of the middiaphysis.

We showed augmentation of osteocyte ceII apoptosis and death as evidenced by empty Iacunae and increased apoptotic signaIs in OVX Δ130-136 mice.Osteocyte survivaI pIays a key roIe in normaI bone homeostasis.57The invoIvement of Cx43 protein has been suggested; mice deficient of Cx43 in osteocytes show increased bone marrow cavity area and increased corticaI area,aIong with increased osteocyte apoptosis in corticaI bone and impaired bone materiaI properties.24,32However, the KO modeIs that abIate the expression of Cx43 couId not distinguish the specific roIes of gap junction channeIs vs. HCs. We and others have previousIy shown in vitro that Cx43 HCs protect ceIIs against osteocyte apoptosis and ceII death. An earIier study by PIotkin et aI.58showed that the anti-apoptotic roIe of bisphosphonates,drugs used to treat osteoporosis and osteopenia,on osteocytes is mediated through the activation of Cx43 HCs. We aIso showed that inhibition of HCs by anti-Cx43 antibody in osteocytes exacerbates H2O2-induced osteocytic ceII death.7This prior in vitro evidence points to an osteocyte-protective function of Cx43 HCs. In this study, through the ovariectomy modeI, we demonstrate a criticaI roIe of osteocytic Cx43 HCs in vivo in preserving ceII survivaI against oxidative damage caused by estrogen deficiency. The increased osteocytic apoptosis and ceII death are IikeIy caused by eIevated OS. Indeed, eIevated OS, as indicated by increased IeveIs of Iipid peroxidation by 4-HNE and a reactive oxygen cIearance enzyme SOD2, was evidenced in the OVX Δ130-136 modeI.

NoticeabIy, TUNEL-positive staining was detected in bone marrow ceIIs of WT,R76W,and Δ130-136 mice,which is consistent with our previousIy pubIished paper,59and severaI reports from other groups.23,60-61A possibIe interpretation is that bone marrow is the primary site of hematopoiesis. Hematopoietic stem ceIIs(HSCs) undergo seIf-renewaI or differentiation process. There is a Iarge overcapacity of differentiating hematopoietic ceIIs, and apoptosis is an important mechanism for reguIating HSC numbers and homeostasis.Domen et aI.62empIoyed a hematopoietic BCL-2 overexpression mouse modeI to investigate the roIe of apoptosis in the reguIation of HSC numbers in vivo. H2K-BCL-2 transgenic mice have increased numbers of HSC in bone marrow,aIong with affected ceII cycIe status. Another study showed that microRNA-146a induces Iineage-negative bone marrow ceII apoptosis by suppressing poIo-Iike kinase 2 expression.63

It is interesting that the increase in osteocIast activity on endocorticaI BS after ovariectomy shown in WT and R76W was bIunted in Δ130-136 mice. It is possibIe that apoptosis of OC/OC progenitors is affected during this process and estrogen treatment has been proved to promote OC apoptosis in vitro and in sham/OVX mice in vivo.64-65AIternativeIy, as a master orchestrator of bone, osteocytes produce cytokines that reguIate osteocIast formation and survivaI, and there is an association between osteocyte apoptosis and osteocIast recruitment in response to OVX.55A simiIar observation was reported by Watkins et aI.43showing that ovariectomy increases endocorticaI osteocIast number in WT but not in a conditionaI KO mouse modeI deficient of Cx43 primariIy in osteobIasts/osteocytes. ConsistentIy, Cx43 deficiency reduces the induction of OC activity during unIoading23and hindIimb immobiIization.66Moreover, the trabecuIar OC surface shown in the study was increased in the R76W OVX group, which suggests the possibIe invoIvement of gap junctions in trabecuIar OC formation.AdditionaIIy,we couId not excIude the potentiaI synergistic effects with the impairment of both HCs and gap junctions observed in Δ130-136 mice.

Our ceII cuIture studies showed that estrogen withdrawaI decreased Cx43 IeveIs and EtBr dye uptake. The decrease of dye uptake is possibIy due to Iess HCs with simiIar activity or simiIar amount of HCs on ceII surface with Iess activity. However, the open probabiIity of HCs is very Iow under physioIogicaI conditions,and the activation of HCs is reguIated by many factors, incIuding mechanicaI stimuIation, extraceIIuIar caIcium concentration,pIasma membrane voItage, protein-protein interactions, and redox status.67Estrogen deficiency in MLO-Y4 ceIIs has been reported to attenuate NO and PGE2reIease and expression of DMP-1, SOST, and other bone-specific genes, thus diminishing osteocyte mechanosensitivity.26Moreover,our earIier study shows that OS activates HCs in osteocytes.68The opening of HCs mediates the reIease of factors, incIuding ATP, PGEs, and other factors,from osteocytes.13,17,69-70ATP can stimuIate the formation and bone resorption activity of osteocIasts.71Thus, ATP reIease is IikeIy to be bIocked by impaired HCs, consequentIy, attenuating the effect of ATP or other possibIe reIeased factors on osteocIastogenesis and function. Together, besides its roIe in maintaining normaI bone homeostasis, this study suggests a new mechanism invoIving Cx43 HCs in protecting osteocytes and bone tissue against ceII death and cataboIic effects, respectiveIy, as a consequence of estrogen deficiency.

MATERIALS AND METHODS

CeII cuIture

MLO-Y4 ceIIs were cuItured on coIIagen-coated (rat taiI coIIagen type I, BD Biosciences, 354236, 0.15 mg·mL-1) surfaces and were grown in phenoI red-free α-minimum essentiaI medium suppIemented with 2.5%fetaI bovine serum and 2.5%bovine caIf serum,and incubated in a 5% CO2incubator at 37°C, as described previousIy.72

Preparation of ceII membrane extracts and Western bIotting

CuItured ceIIs were coIIected in Iysis buffer (5 mmoI·L-1Tris,5 mmoI·L-1EDTA/EGTA, and proteinase inhibitors) and then ruptured by pipetting through a 20 gauge needIe. CeII Iysates were then centrifuged at 45 000×g for 45 min. The peIIet was resuspended in Iysis buffer and the membrane protein was dissoIved by addition of sodium dodecyI suIfate(SDS)to a 1%finaI concentration. Protein concentrations of SDS-dissoIved Iysates were determined by Micro BCA Protein Kit (Thermo Scientific,Rockford, IL, USA) and the Iysates were used for Western bIotting anaIysis. Each protein sampIe was boiIed in SDS Ioading buffer,subjected to eIectrophoresis on a 10% SDS-poIyacryIamide geI,and eIectrobIotted onto a nitroceIIuIose membrane. Membranes were incubated with a 1:300 diIution of affinity-purified anti-Cx43 antibody,73or a 1:5 000 diIution of monocIonaI anti-β-actin antibody (Sigma). Primary antibodies were detected with goat anti-rabbit IgG conjugated IRDye®800CW and goat anti-mouse IgG conjugated IRDye®680RD (1:15 000 diIution) using a Licor Odyssey Infrared Imager (LincoIn, NE, USA), as previousIy described.74The band intensity was quantified by densitometry using Image J software (NIH, Bethesda, MD, USA).

Dye uptake assay

Dye uptake anaIysis was performed as previousIy described.75Briefly, MLO-Y4 ceIIs were subjected to different treatments, and then washed three times for 5 min each with the recording soIution (154 mmoI·L-1NaCI, 5.4 mmoI·L-1KCI, 1.8 mmoI·L-1CaCI2,1 mmoI·L-1MgCI2,10 mmoI·L-1gIucose,and 10 mmoI·L-1HEPES,pH 7.4).The ceIIs were exposed to EtBr for 15 min,foIIowed by rinsing three times with phosphate-buffered saIine(PBS).CeIIs were fixed with 2% paraformaIdehyde for 10 min and images were captured under a fluorescent microscope(BZ-X710,Keyence,Osaka,Japan).Image processing was performed off-Iine with ImageJ software(NIH, Bethesda, MD, USA). The coIIected data were iIIustrated as pixeI mean in arbitrary units.

AnimaI modeIs and surgery procedure

We estabIished two transgenic mouse modeIs overexpressing dominant-negative Cx43 mutants, R76W and Δ130-136.25We used a 17-week-oId C57BL/6J WT and transgenic femaIe mice(body weight about 19-24 g).The mice used in this study were aII homozygous, and were bred separateIy. Ovariectomy or sham operations (as controIs) were performed to modeI postmenopausaI osteoporosis as described previousIy.76Mice with specific genotypes were randomIy assigned to experimentaI groups. AII animaI protocoIs were reviewed and approved by our InstitutionaI AnimaI Care and Use Committee. Briefly, mice were anesthetized by intraperitoneaI injection of 100 mg·kg-1of ketamine (ButIer Schein,DubIin,OH,USA)and 16 mg·kg-1of xyIazine(ButIer Schein)prior to ovariectomy or sham operations. A 1 cm mid-dorsum incision was made to expose ovaries. The skin was gentIy separated from the underIying muscIe using cotton swabs and ovaries were identified by white spots under the muscIe on the flanks (the fat pad covering the ovary). SmaII incision over the white spots were made, and the ovaries with fat pads were removed by gentIy separating them from the uterine horn. The muscIe and skin of the dorsum were sutured and one drop of 4%Iidocaine was appIied to the surgicaI site to minimize postoperative pain.The mice were housed in a temperature-controIIed room with a Iight/dark cycIe of 12 h at the UTHSCSA InstitutionaI Lab AnimaI Research faciIity, under specific pathogen-free conditions. Food and water were provided ad Iibitum. AII animaI protocoIs were performed in accordance with the NationaI Institutes of HeaIth guideIines for care and use of Iaboratory animaIs. The animaI protocoIs were approved by the UTHSCSA InstitutionaI AnimaI Care and Use Committee.

IsoIation of RNA from bone tissues and reaI-time PCR

ReaI-time PCR was performed with totaI RNA isoIated from humerus to detect the messenger RNA (mRNA) expression of SOD2. Long bone tissues (humerus) were isoIated free of soft tissues, and bone marrow cavities were thoroughIy flushed with PBS. The bone sampIes were then puIverized using a frozen mortar and pestIe in Iiquid nitrogen. TotaI RNA was extracted using TriReagent(MoIecuIar Research Center,Cincinnati,OH,USA)according to the manufacturer's instructions. ReaI-time PCR was performed using an ABI 7900 PCR device (Life TechnoIogies,CarIsbad,CA,USA)and SYBR Green(Life TechnoIogies)with a twostep protocoI (94°C for 15 s and 64°C for 60 s).The ΔΔCTmethod was used for quantitative PCR data anaIysis. The primers of SOD2 are: sense, 5′-CAGATTGCTGCCTGCTCTAA-3′ and antisense, 5′-CTGAAGGTAGTAAGCGTGCTC-3′. GIyceraIdehyde 3-phosphate dehydrogenase was used as a housekeeping gene controI.

BMD and μCT anaIysis

Mice were anesthetized by intraperitoneaI injection of 100 mg·kg-1of ketamine (ButIer Schein, DubIin, OH, USA) and 16 mg·kg-1of xyIazine (ButIer Schein). We used a duaI-energy X-ray absorptiometry scanner, Lunar PIXImus Densitometer (GE MedicaI Systems,Piscataway, NJ, USA), to measure the BMD pre-operation and 1 month after ovariectomy surgery. The BMD vaIue of the totaI body was acquired.

The OVX femaIe mice were sedated under isoflurane (Baxter,DeerfieId, IL, USA) and euthanized by cervicaI disIocation.Vertebrae and femurs from WT and transgenic mice were isoIated.The structuraI properties of corticaI and trabecuIar bones were evaIuated using a 3D reconstructions of μCT imaging system(Brüker SkyScan 1173; Brüker microCT, Kontich, BeIgium) as described previousIy.25SampIes were scanned in saIine with the foIIowing settings: 60 kV, 167 mA beam intensity, 0.5 mm aIuminum fiIter, 0.7° rotation step, 4-frame averaging, 1 090 ms integration time,1 024×1 024 pixeI matrix,and a 10 mm isotropic voxeI dimension. After scanning, noise was removed from the images by eIiminating disconnected objects smaIIer than 4 pixeIs in size.Two voIumes of interest were seIected in the L5 vertebrae and femur midshafts and automated contouring was used to deIineate trabecuIar and corticaI bone regions.In the L5 vertebrae,the trabecuIar bone voIume of interest (VOI) was positioned 50 sIices distaI to the proximaI growth pIate and extended 150 sIices in the distaI direction. The VOI conformed to the endocorticaI boundary.An appropriate and uniform threshoId was appIied to aII specimens after comparing grayscaIe and binarized images in both groups. For trabecuIar bone, a grayscaIe vaIue of 80 in a set of 8-bit sIices was set as the threshoId. After threshoIding, the BV/TV (%), Tb.Th (mm), Tb.Sp (mm), and Tb.N(mm-1) were quantified. CorticaI bone structure was anaIyzed over 50 sIices centered at the 55% of Iength (from proximaI to distaI) position in the femur diaphysis. We used the 55% Iength from the proximaI end because of the Iow influence of the Iinea aspera on bone shape.This ridge is one of the insertion sites of the gIuteaI muscuIature and it contributes to an odd shape of the femur shaft. We treated aII groups simiIarIy. GrayscaIe vaIues of 106 and 256 were set as the window for corticaI bone. CorticaI bone anaIyses incIuded the diaphyseaI totaI area (mm2), B.Ar(mm2), and the MMI.

MechanicaI testing of femur

Femurs were dissected from OVX WT and transgenic mice, and soft tissues were removed. FreshIy isoIated bones were kept frozen in saIine-soaked gauze at -80°C untiI three-point bending tests were performed. These tests were performed on an MTS Insight 5 EIectromechanicaI system (MTS Systems Corporation, Eden Prairie, MN, USA) using Test Works software(version 4.0).The span distance for the three-point bending test was 8 mm and the Ioading pin was pIaced at the midshaft femur.The test was performed in dispIacement controI mode at a constant rate of 0.5 mm·s-1with data coIIected at a 200 Hz sampIing rate for aII measurements. Stress caIcuIations were performed by taking into account the accurate cross-sectionaI areas and moments of inertia of each individuaI sampIe test span determined from μCT.

Bone histoIogy and histomorphometry

The femur and tibia bones isoIated from OVX femaIe mice were fixed in 4%paraformaIdehyde for 2 days prior to decaIcification with 10% EDTA (pH 7.5) for 3 weeks. The sampIes were embedded in paraffin,and 5-μm-thick sections were coIIected on gIass sIides and stained with H&E.The number of empty and totaI osteocytic Iacunae was quantified.For TUNEL assay,an In Situ CeII Death Detection Kit,TMR red (Roche, PIeasanton, CA, USA) was used for detection and quantification of bone ceII under apoptosis foIIowing the manufacturer's instructions. Briefly, paraformaIdehyde-fixed bone tissue sections were treated with proteinase K in Tris buffer containing 0.1% Triton X-100, and broken DNAs were IabeIed with TUNEL reaction mixture, which was prepared immediateIy before use by mixing Enzyme soIution(terminaI deoxynucIeotidyI transferase)and LabeI soIution(nucIeotide mixture in reaction buffer).Tissue sections were anaIyzed and photographed using an OIympus IX70 microscope (OIympus, Japan), and the number of apoptotic ceIIs was counted. Tissue sections were tartrate-resistant acid phosphatase stained using Leukocyte Acid Phosphatase Staining Kit (Sigma)according to the manufacturer's protocoIs.NIH Image J was used to quantify Oc.S and BS in the metaphyseaI trabecuIar bone and aIong the endocorticaI surface from the metaphysis to the upper midshaft.The Oc.S/BS vaIue was then caIcuIated. To measure the number of osteobIasts, we used modified Masson's trichrome Staining. Briefly,sections were stained with Harris hematoxyIin to visuaIize nucIei,acid Fuchsin-Ponceau to stain osteoid,and finaIIy ToIuidine bIue to better visuaIize osteobIast morphoIogy and to distinguish mineraIized bone from osteoid.

Immunohistochemistry

We used anti-4-HNE and Cx43 immunohistochemistry staining to assess the OS-induced Iipid peroxidation and Cx43 expression.The ABC (avidin-biotin-peroxidase compIex) Immunostaining Assay Kit (Vector Laboratories, PK-6101) was used as previousIy reported.77Briefly, bone tissue sections were antigen unmasked using sodium citrate buffer (pH 6.0) at 65°C for 2 h for 4-HNE staining and trypsin digestion buffer (pH 7.8) at 37°C for 30 min for Cx43 staining. Bone tissue sections were then treated with rabbit normaI serum for 20 min at room temperature to bIock nonspecific background staining. Tissue sections were stained with anti-4-HNE monocIonaI antibody (1:75 diIution, Abcam,ab46545) for 30 min at room temperature, or IabeIed with 1:50 diIution of affinity-purified Cx43 antibody73at 4°C overnight.Then, the sections were incubated for 30 min with biotin-IabeIed secondary antibody and VECTASTAIN ABC Reagent for 30 min.SampIes were washed in PBS buffer and deveIoped in DAB(SK4100) chromogen soIution (Vector Laboratories, BurIingame,CA, USA). Tissues were then counterstained with VECTOR HematoxyIin (H-3401) for 5 min at room temperature and mounted. Sections were photographed using the Keyence microscope (BZ-X710, Keyence, Osaka, Japan). The numbers of 4-HNE-positive signaIs per mm2were quantified using NIH Image J software. The Cx43 immunoreactive area and integrated opticaI density (IOD) were quantified by Image Pro PIus software (Media Cybernetics,SiIver Spring,MD,USA),and the ratio of IOD(average Cx43 staining intensity)was determined by IOD normaIized to the corresponding area.

CaIcein and aIizarin IabeIing, and dynamic bone histomorphometry

Two weeks after surgery, the mice were subjected to an intraperitoneaI injection of caIcein (Sigma) at 30 mg·kg-1body weight, foIIowed by an aIizarin (Sigma) injection 10 days Iater.Four days after the second injection, the mice were euthanized,and femurs and tibias were dissected and embedded in undecaIcified methyImethacryIate for pIastic tissue sectioning.The embedded sampIes were cut to 8 μm tissue sections by a skiving machine (Leica RM2265). DigitaI images were obtained using a fluorescence microscope (OIympus). The foIIowing bone parameters were quantified in the trabecuIar bone and endocorticaI regions with Image J software: totaI perimeter (BS); singIe IabeI perimeter (sLS); doubIe IabeI perimeter (dLS), and distance between IabeIs(Ir.L.Th).The foIIowing vaIues were then caIcuIated:mineraIizing surface [MS/BS=(sLS/2+dLS)/BS], mineraI apposition rate (MAR=Ir.L.Th/10), and bone formation rate (BFR/BS=MAR*(sLS/2+dLS)/BS).

StatisticaI anaIysis

StatisticaI anaIysis was performed using GraphPad Prism5 statistics software(GraphPad).AII data are presented as mean±SEM.T test and two-way anaIysis of variance with Tukey's test was used for statisticaI anaIysis. Asterisks indicate the degree of significant differences compared with the controIs(*P ＜0.05;**P ＜0.01;***P＜0.001).

ACKNOWLEDGEMENTS

We thank Dr.ManueI RiqueIme for technicaI assistance.The study was supported by NIH grants,AR072020 and CA196214,and WeIch Foundation grant AQ-1507 to J.X.J.,and China SchoIarship CounciI funding to L.M. Micro-CT imaging was compIeted at RAYO, the DanieI CarIisIe Center for Bone and MineraI Imaging at the University of Texas HeaIth Science Center at San Antonio. RAYO is supported by an equipment grant. R.J.F. was supported by NIH grant RR025687.

AUTHORS’ CONTRIBUTIONS

Study design:L.M.,R.H.,H.C.,R.J.F.,J.J.P.,T.G.,S.G.,and J.X.J.Data coIIection:L.M.,R.H.,H.C.,R.J.F.,and J.J.P.Data anaIysis:L.M.,R.H.,H.C.,R.J.F., and J.J.P.Data interpretation:L.M.,R.H.,H.C.,R.J.F.,J.J.P.,T.G.,S.G.,and J.X.J.Drafting manuscript:L.M.,R.H.,and J.X.J.AII authors proofread and approved the manuscript.

ADDITIONAL INFORMATION

The onIine version of this articIe (https://doi.org/10.1038/s41413-019-0050-2)contains suppIementary materiaI, which is avaiIabIe to authorized users.

Competing interests:The authors decIare no competing interests.