Skeletal loading regulatesbreast cancer-associated osteolysisin a loading intens

时间：2025-01-07

Yao Fan,Aydin Jalali,Andy Chen ,Xinyu Zhao,3,Shengzhi Liu,Meghana Teli,Yunxia Guo,Fangjia Li,Junrui Li,Amanda Siegel,Lianxiang Yang,Jing Liu,Sungsoo Na,Mangilal Agarwal,Alexander G.Robling,Harikrishna Nakshatri ,Bai-Yan Li and Hiroki Yokota,,,

INTRODUCTION

Bone is a mechanosensitive organ,and it constantly remodels itself using mechanical loading as one of its major cues.1Osteocytes are the most abundant type of cells in the bone matrix,and they act as mechanosensors and induce load-driven bone remodeling.2While mechanical loading activatesmany loadsensitive genes via a wide spectrum of signaling pathways,little is known about the potential effects of mechanical stimulation on tumor-osteocyte communication.Osteocyte-mediated loading effects are potentially different from those observed in tumor cells that directly receive mechanical stimulation.3–5In this study,we examined the interactions between osteocytes and migratory breast cancer cells in the presence and absence of mechanical stimulation in vitro and in vivo.

Bone is the most frequent site of metastasis of breast cancer.6Interactions between tumor cells and bone-resorbing osteoclasts are known to induce a vicious feed-forward cycle in which osteolytic responsesare ampli fied by a loop of interactionsamong PTHrP and TGFβ.7In addition to interactions with osteoclasts,tumor cells are also attracted by chemoattractants in the bone marrow and ECM,8and some ECM proteins,such as collagen and proteoglycans,are reported to act as attractants.9For instance,type Icollagen is the most abundant protein in the ECM of the bone matrix,and receptors such as integrin and discoidin domain proteins have been shown to interact with collagen and alter the expression of genes that are involved in the epithelial-tomesenchymal transition(EMT).10–11The major aim of this study was to identify the effect of mechanical stimulation on tumorosteocyte communication.

We used conditioned media derived from osteocytes with and without mechanical stimulation and monolayer and 3D culture assays to evaluate whether mechanical stimulation induces differential effects on migratory breast cancer cells regarding their viability,migration,and EMT-linked gene expression.Furthermore,in a mouse model of breast cancer-associated osteolysis,we evaluated whether skeletal loading of tumorloaded osteolytic bone inhibits or promotes tumor progression.In tumor-metastasized bone,tumor cells and osteocytes are considered the seed and soil.The chief focus in this study was the effect of mechanical stimulation on the soil(osteocytes and bone)and the subsequent impact of the modi fied soil on the seed(tumor-inoculated bone).Fluid flow-driven shear stress(0.25 Pa and 1 Pa)was applied to osteocytes,and tibia loading(1–5 N)was applied to tumor-inoculated bone.

In biophysical therapy to suppress in flammation and tissue degradation in osteoarthritic joints,we and others observed that low-intensity loads of the knee(0.5–1.0 N)provided bene ficial outcomes,but medium loads(2N or above)promoted the expression and activity of matrix metalloproteinases(MMPs),12and excessive mechanical loading was shown to promote osteoarthritis.13Based on these observations,our hypothesis is that medium mechanical stimulation in the form of oscillatory fluid flow(1 Pa shear stress)or skeletal loading(5 Nloads)induces EMT-like responses in tumor cells,leading to increased migration and invasiveness.In contrast,we hypothesize that low mechanical stimulation(0.25 Pa shear stress or 1 N loads)reverses this effect and induces MET-like responses.

In this study,we used the breast cancer cell line MDA-MB-231 and its TMD cell(mouse xenografted clone)variant,human primary breast cancer cells,14and the mouse mammary tumor cell lines 4T1,4T1.2,and EO771.To induce mechanical stimulation in vitro,oscillatory fluid flow between a pair of parallel plates was applied to MLO-A5 osteocytes15at 1 Hz with a peak shear stress of 0.25 Pa or 1 Pa.Tumor cells were then treated with MLO-A5-derived conditioned medium(CM)or MLO-A5 fluid flow-treated conditioned medium(FFCM).Using a mouse model of tibial osteolysis,tibia loading(oscillatory compressive loading)was applied to two mouse strains,C57BL/6 and BALB/c,after tibial injection of EO771 and 4T1.2 mouse mammary tumor cells,respectively.The daily loading condition was sinusoidal loads at 2Hz with peak-to-peak loads at 1 N,2 N,and 5N for 5min.

To analyze the EMT,we examined the expression of Snail and EMT-linked genes.To measure the activation of tumor-induced osteoclastic bone resorption,TGFβ,an ECM-enriched growth factor,was measured.16The expression and activity of Src is known to correlate with advanced malignancy and poor prognosis in a variety of cancers.17–18Osteopontin(OPN)is known to play a role in cell adhesion and migration and tumorigenesis.19Mass spectrometry-based protein analysis predicted the involvement of OPN in the mechanosensitive migratory behaviors of tumor cells.OPN involvement in osteocyte-tumor interactions was further probed by analyzing the action of OPN on Src activity and molecular tension force using two fluorescence resonance energy transfer(FRET)-based techniques,as well as RNAinterference.In a FRETexperiment using a Src biosensor,the activity level of Src in response to varying amounts of OPNwas evaluated based on the conformational change in the Src biosensor.20In a second FRET experiment using a vinculin force biosensor,the migratory behavior of individual tumor cells was evaluated in which the FRET lifetime increased by stretching a molecular linker in the force sensor.21

The response to mechanical stimulation was further measured using proteins and volatile organic compounds(VOCs)that were designed to determine general levels of osteolysis and tumor progression.More speci fically,we analyzed the expression of PPARγ,a transcription factor that promotes lipid metabolism and is generally associated with the pathogenesis and development of tumors.22Cancer-associated VOCs are a generalized biomarker of disease.23We therefore also measured VOClevels in the urine of differentially treated mice.The results of this study indicate that tumor-osteocyte interactions are affected by mechanical stimulation in vitro and in vivo,suggesting an important role for biophysical signals on metastatic tumor cells in bone.

RESULTS

Differential effects of A5 osteocyte conditioned medium(CM)and fluid flow-treated CM(FFCM)

We first conducted in vitro analysis using osteocyte-derived conditioned media with and without fluid flow treatment(Supplementary Fig.S1).In this study,we advanced the differentiation stage of parental MLO-A5 early-stage osteocytes by incubating them with ascorbic acid.A5 cells expressed a higher level of Sclerostin and DMP1,two osteocyte markers,compared with that of Y4 osteocyte-like cells(Fig.1a).Using TMD breast cancer cells(a variant of MDA-MB-231 cells),we examined the effects of A5 CM and FFCM.A5 CM inhibited the healing of wounds in the scratch assay(Fig.1b),stimulated cellular proliferation in an Ed U assay(Fig.1c),and compacted tumor spheroids(Fig.1d).A5 CM also reduced Snail,Vimentin,MMP9 and Slug and elevated E-cadherin and p-Akt(Fig.1e).Notably,A5 FFCM(1 Pa flow application)reversed the responses to A5 CM.FFCMat 1Pa promoted cellular migration(Fig.1f),reduced cellular proliferation(Fig.1g),and expanded tumor spheroids(Fig.1h).FFCM also upregulated Snail,Vimentin,MMP9 and Slug and downregulated E-cadherin and p-Akt(Fig.1i).

Migratory behaviors in response to A5 CM and FFCM

To further evaluate the differential effectsof CMand FFCM at 1Pa,we examined the activity level of Src using a FRET-based biosensor.Collagen treatment was used as a control to suppress Src based on our previous work.9Src activity was downregulated by A5 CM as well as by incubation with 10 μg·mL-1collagen,but the observed downregulation was reversed by FFCM(Fig.2a).The migratory behaviors of three cell lines(TMD,MDA-MB-231,and 4T1 breast cancer cells)were also suppressed by A5 CM and Y4 CM(Supplementary Fig.S2a–d).The degree of suppression was stronger in the presence of A5 CM than Y4 CM.

Tumor spheroids presented high af finity to A5 spheroids in the control medium(Fig.2b).A5 CM inhibited the migration of tumor spheroids(green)toward A5 spheroids(red),and this inhibition was suppressed by FFCM at 1 Pa.Live cell imaging revealed that individual tumor cellsin FFCMappeared to be more slender with a higher aspect ratio(the ratio between the major and minor axes)than those in A5 CM(Fig.2c).Furthermore,the trajectory of individual cells in the FFCM suggested a random walk in which the mean-square displacement was proportional to time,while the trajectory in the A5 CM showed a restricted random walk(Fig.2d).Taken together,these cellular images also support the notion that FFCM at 1 Pa activated Src and induced tumor cell migration compared with that of A5 CM.

Proteins differentially expressed in A5 CM and FFCM

We next evaluated the expression of TGFβas a potential regulator of the action of A5 CM and FFCM,since TGFβis known to play a vital role in osteolysis.Compared with the level in A5 CM,Western blot analysis and ELISA showed that TGFβin FFCM was increased(Fig.3a,b).Snail and p-Src were increased in response to 10 and 100 ng·mL-1TGFβ1 and β3,but PPARγwas decreased(Fig.3c).To further investigate proteins that are potentially responsible for the actions of A5 CM and FFCM,we conducted mass spectrometrybased analysis.In addition to TGFβ,we identi fied five proteins that were differentially expressed(Fig.3d).These proteins(Fibronectin,Nucleolin,Osteopontin,Vimentin,and Pro filin 1)were more highly expressed in A5 CM than in FFCM.As predicted,treatment with each of these proteins,except for Pro filin 1,downregulated Snail in TMD breast cancer cells and 4T1 mammary tumor cells(Supplementary Fig.S3).The level of Sclerostin was not altered(data not shown).

Effects of OPN on cellular migration

Among the five predicted proteins,in vitro analysis showed that OPN signi ficantly altered PPARγ,p-Src and Snail in a dosedependent fashion(Fig.3e,f).Notably,lower doses of OPN(0.1 μg·mL-1)upregulated PPARγ,p-Src and Snail,while higher doses(2 μg·mL-1)reduced them(Fig.3e,f).Consistent with the suppression of p-Src and Snail,2 μg·mL-1OPN inhibited tumor cell wound healing.To further evaluate the role of OPN,we employed a FRET-based vinculin sensor and evaluated the molecular force of cell migration.24This sensor was designed to alter the fluorescence lifetime under tensile force.We observed that a low dose of OPN(0.05 μg·mL-1)increased the fluorescence lifetime,indicating elevated tensile force,while a high dose of OPN(2 μg·mL-1)decreased it(Fig.3g,h).Furthermore,partial silencing of OPN by siRNA in A5 osteocytes increased TGFβ,and CM downregulated p-Src and Snail in tumor cells(Fig.3i).Collectively,these results were consistent with the dosedependent role of OPN,in which fluid flow reduced OPN in A5 cells,induced tensile force on tumor cells,and stimulated migratory behaviors.

Responses to fluid flow at 1 Pa and 0.25 Pa in tumor cell lines and primary breast cancer cells

Thus far,the responses to CM and FFCM at 1 Pa were evaluated using a variant of MDA-MB-231 breast cancer cells.To further evaluate the responses to CM and FFCM,we employed other cell lines and primary human breast cancer cells.FFCM at 1 Pa upregulated Snail in 4T1 mammary tumor cells as well as in two types of primary breast cancer cells(Fig.4a).EO771 and 4T1.2 mammary tumor cells had reduced wound healing in response to A5 CM(Fig.4b,c).In these cell lines,Snail was downregulated by A5 CM and upregulated by FFCM at 1Pa(Fig.4d).Taken together,the responses to CM and FFCM were consistent among the six types of mammary tumor and breast cancer cells that we examined.

Using two levels of flow-induced shear stress(0.25Pa and 1 Pa),we next examined whether fluid flow intensity alters the migratory behaviors or expression levels of flow-sensitive genes such as p-Src and Snail in TMD cells,4T1.2 cells,and EO771 cells(Fig.4e–g).When A5 osteocytes received low-level fluid flow(0.25 Pa),the tumor cell wound healing was suppressed,and p-Src and Snail were downregulated.In contrast,with 1 Pa shear stress,FFCM stimulated wound healing and elevated the levels of p-Src and Snail.In A5 osteocytes,the expression of TGFβand OPN was also altered depending on the flow intensity(Fig.4h).OPN was increased by 0.25 Pa and reduced by 1Pa shear stress.

Strain measurement

In vitro analysis identi fied an antitumor action of osteocytes that is strengthened by FFCM at 0.25 Pa and reversed by FFCM at 1 Pa.To further evaluate the effect of mechanical stimulation,we employed a mouse model and conducted dynamic tibia loading(axial compression to the tibia)using different(low and medium)loads(Fig.5a).

Prior to evaluating load-driven effects on tumor-induced osteolysis,we conducted strain measurements and evaluated the mechanical microenvironment in the proximal tibia in which tumor cells were inoculated.The first method,using a pair of strain gauges immobilized in the proximal tibia(Fig.5b),detected load magnitude-dependent strain in C57BL/6 mice(Fig.5c;N=5).In response to 1N loads at 2Hz,for instance,the induced strain was 98μstrains(peak compression at Channel 1)and 46μstrains(peak tension at Channel 2).The second method using 3D digital correlation imaging(DCI)also showed that cyclic compression(2 Hz with 1,2,and 5 N loads)induced load-dependent bone deformation and strain(Fig.5d).The maximum compressive strain in the proximal tibia with DCIwas 260μstrains(1 N),480μstrains(2 N),and 1 060μstrains(5 N)(Fig.5e;N=3).The observed variations in the strain values indicate a complex strain pattern in the proximal tibia in response to tibia loading.Using normal control BALB/c mice without tumor inoculation,we also evaluated the effect of daily tibia loading for 2 weeks on trabecular bone in the proximal tibia(Fig.5f–h).The results showed that tibia loading at 5 Nelevated BV/TV(bone volume ratio)and BMD(bone mineral density)and reduced Tb.s(trabecular separation).However,tibial loading at 1 N did not alter these parameters.

Load-dependent effects of tibia loading

After intratibial injection of EO771 or 4T1.2 cells,we applied tibia loading to C57BL/6 or BALB/c mice at three intensities(1,2,and 5N loads).We observed that daily tibia loading of C57BL/6 mice with 1 Nloads reduced the degradation of trabecular bone in the proximal tibia(Fig.6a).More speci fically,a signi ficant increase in BV/TV,Tb.n(trabecular number),and BMDwas observed(Fig.6b).In contrast,osteolysis of trabecular bone was stimulated by tibia loading of C57BL/6 mice with 5 N loads(Fig.6c,d),as well as BALB/c mice with intratibial injection of 4T1.2 cells and 5 N loads(Fig.6e,f).No signi ficant effect wasobserved in C57BL/6 mice with 2 Nloads(Fig.6c,d),and the administration of 1 Nloadsto BALB/c mice prevented bone loss in the tibia(Supplementary Fig.S4a,b).However,administration of 5 N loads to C57BL/6 mice damaged not only the proximal tibia but also the distal femur(Supplementary Fig.S4c,d).Of note,5 N loads did not signi ficantly damage the distal femur of BALB/c mice(data not shown).Histological images with H&Estaining showed that the sample with 5 Nloads presented signi ficant bone damage distal to the growth plate in the proximal tibia in BALB/c mice compared with that of the sham loading sample(Fig.7a,b).In the proximal tibia,loading at 5 N reduced the bone area ratio and increased the tumor area ratio(Fig.7c,d).Fuchsin staining for microdamages indicated an increased number of microcracks in the proximal cortical bone in the tibia of the tumor-induced osteolytic samples with 5 N loads(Fig.7e).

Involvement of the mevalonate pathway in urine-derived VOCs We previously determined that terpenes and terpenoids in the mevalonate pathway are some of the VOCs that are signi ficantly altered in mice with breast cancer tumors.23To evaluate metabolic changes caused by loading,we analyzed urine-derived VOCs(Supplementary Table S1).Hierarchical clustering and principal component analysis(PCA)were conducted on samples from C57BL/6 and BALB/c mice(Fig.8a–f).In the PCAplane,the normal control and the placebo control were separately positioned along the first PCAaxis for both mouse breeds(Fig.8b,e).The 2 N-loaded samples were positioned between the normal and placebo controls,while the 5N-loaded sampleswere located away from the normalcontrol.PCA of VOCs showed clustering of compounds along the first PCA axis,with the right cluster corresponding to the tumor-enriched VOCs(Fig.8c,f).Among the tumor-enriched compounds,four VOCs(nerol,acetophenone,dolichol,and bisabolol)were linked to the mevalonate pathway(Fig.8g).Because this pathway regulates lipid metabolism,we analyzed the expression levels of PPARγ,a transcription factor that is involved in lipid metabolism,in three tumor cell lines(TMD,4T1.2,and EO771).In all cell lines,PPARγwas reduced by FFCM at 0.25Pa and increased by FFCMat 1 Pa(Fig.8h).Since the mevalonate pathway leads to cholesterol synthesis,we evaluated the effects of cholesterol in TMD and 4T1.2 cells.The results revealed that in both cell lines,cholesterol increased MMT-based cellular viability and migratory capability and upregulated MMP9,p-Src,and Snail(Supplementary Fig.S5).

Load-dependent regulation of serum TGFβ,D-dimer,and cholesterol and gene expression

In an attempt to associate fluid flow-driven gene regulation in vitro with load-driven responses in vivo,we analyzed the levels of selected markers in serum and the expression of metabolic and EMT-linked genes in bone marrow.ELISAresults showed that 5 N loads increased TGFβand D-dimer in the serum(Fig.9a,b).Of note,TGFβwas elevated by FFCMat 1 Pa,and D-dimer isa marker for thrombus,indicating pathological loading of osteolytic bone.Because the mevalonate pathway leads to cholesterol production,we analyzed serum cholesterol levels in C57BL/6 mice.Loads at 1N reduced the cholesterol level,but 5N loads increased it(Fig.9c).Western blot analysis of bone marrow-derived cells revealed that the level of OPN was increased by 1 N and reduced by 5 N loads.Conversely,the levels of PPARγ,TGFβ,Snail,and p-Src were reduced by 1N and increased by 5 N loads(Fig.9d).Collectively,these results suggest that medium-level mechanical stimulation(1 Pa shear stress in vitro and 5 N loads in vivo)promotes EMT-linked genes such as TGFβ,Snail,and p-Src and low-level stimulation(0.25 Pa shear stress in vitro and 1 N loads in vivo)inhibits them.

DISCUSSION

The main goal of this study was to determine whether osteocytes cultured under static conditions or mechanical stimulation express and/or release factors that alter the metastatic behaviors of breast cancer cells.Both in vitro and in vivo results demonstrated that osteocyte action on cancer cells differs with and without mechanical stimulation in an intensity-dependent manner.In vitro analysisrevealed that A5 CM without FFinhibited cellular migration by downregulating Src activity and Snail expression.FFCM at 0.25Pa further downregulated the Src-Snail axis.With FFat 1 Pa,however,the responses were altered.FFCM at 1Pa upregulated Src and Snail and promoted cellular migration.Our findings on cellular migration at 1 Pa were consistent with previous work,25and we identi fied further dependence of this effect on mechanical intensity.The FF-driven responses at 1Pa were largely consistent with the EMT.Collectively,the present in vitro analysis indicates that FFat 1 Pa may alter the bone microenvironment from a pro-METto pro-EMT milieu,while FF at 0.25 Pa changes the microenvironment to a further pro-METmilieu.

Metastatic dependence on loading intensity was also observed in the mouse model of tibial osteolysis.Previous loading work used SCID immunocompromised mice with the MDA-MB-231 cell line,26but we employed two immunocompetent strains(BALB/c and C57BL/6)with two tumor cell lines(4T1.2 and EO771).While~4 N loads at 4 Hz with~600μstrains for 5 min are reported to be bene ficial,27we observed load intensity-dependent responses with 1,2,and 5 N loads at 2 Hz for 5 min.Our findings showed that 1 Nloads with~300μstrains protected bone from tumor-induced osteolysis and 5 N loads with~1 000μstrains increased microcracks and bone degradation.No clear difference was observed with 2 N loads with~500 μstrains.While intensity-dependent responses were observed in both mouse strains,the protective effects with 1 N loads were more signi ficant in C57BL/6 than in BALB/c mice.Furthermore,the observed dependence on load intensity in vivo was consistent with in vitro FFresponses,in which shear stress at 0.25 Pa inhibited the migratory behaviors of tumor cells and stimulated them at 1 Pa.The observed dependence on loading intensities was consistent with the serum concentrations of TGFβ and cholesterol,as well as the protein levels of PPARγ,OPN,and TGFβin bone marrow,which presented loading intensity-dependent expression pro files.

The involvement of mechanical factors in tumor progression has been previously reported.27–29The stiffness of the ECM is known to alter the proliferation and survival of not only tumor cells but also stromal cells in the tumor microenvironment.28Tumor-induced pressure is reported to promote tumor growth in metastatic prostate cancer.29In laryngeal carcinoma cells,FF is reported to induce the EMTby enhancing cellular migration.4In this study,we applied FFto osteocytes in vitro and compressive skeletal loading to the tibia in vivo.Osteocytes in vivo receive oscillatory shear stress by load-driven deformation of the bone matrix and deformation-induced FF in the lacuno-canalicular network.30In the present study,mechanical stimulation with 0.25 Pa FF and 1 N loads were bene ficial,while stimulation with 1 Pa FF and 5 N loads were detrimental.These mechanical conditions heavily depend on culture conditions such as the serum concentration in the culture media as well as loading modalities and mouse strains used for skeletal loading.Further analysis is recommended to evaluate loading intensity-dependent tumor-osteocyte communication.

Mass spectrometry-based protein analysis and Western blotting revealed that Fibronectin,Nucleolin,Vimentin,and OPN were potential molecular contributors to the observed differences between A5 CM and FFCM.Fibronectin is a high molecular weight glycoprotein that binds to integrins.31Nucleolin is known to be involved in cell differentiation,adhesion,in flammation and tumor development,32while Vimentin isan intermediate filament protein that is used as a marker of mesenchymal-derived cells.33In this study,we focused on the role of OPN in the responses to mechanical stimulation,since OPN is known to be involved in tumor progression.19

Aschematic diagram of the load-driven action of OPNand TGFβ in osteocytes is illustrated,together with the subsequent expression of p-Src and Snail in tumor cells(Fig.9e).The level of TGFβin osteocytes was altered by OPN in a dose-dependent fashion.Tumor cells also responded to OPN in a manner that was dependent on its concentration in osteocytes.We observed that the level of OPN was differentially regulated depending on the intensities of FF on osteocytes and tibia loading.A low dose of OPN in osteocytes elevated p-Src and Snail in tumor cells,while high doses of OPN inhibited these factors.Of note,PPARγwas downregulated under 0.25 Pa FFand upregulated under 1 Pa FF.TGFβ suppresses PPARγin human fetal lung fibroblasts,34and our data also showed downregulation of PPARγby TGFβ.Thus,unlike the regulation of p-Src and Snail,PPARγwas mainly regulated by OPN but not TGFβ.

OPN is a negatively charged,secreted phosphoprotein that is rich in the ECM of bone matrix.It has been reported that a secretory isoform of OPNinducesthe EMT,while itsintracellular/nuclear isoform induces the MET.35It has also been reported that cancer cells express different splice variants.36While OPN may facilitate cellular adhesion and contribute to the migration and invasion of tumor cells,37–38it is also possible that an excessive amount of OPN inhibits cellular migration because of strong tumor cell attachment to the substrate surface.The observed dose-dependent effect of OPN and the intensitydependent responses to mechanical stimulation present a complex interplay of mechanosensitive ECM proteins in tumorosteocyte communication.

In addition to characterizing loading effects at the loaded site,we analyzed load-driven metabolic responses in mouse urine.23The bene ficial effects of 1 N loads on the tibia at the loading site,as well as the detrimental in situ effects of 5 N loads,were associated with systemic changes in metabolites.Principal component analysis revealed that the tumor samples and 5 N-loaded samples were enriched for several volatile metabolites in the mevalonate pathway in urine-derived VOCs,while 2 N-loaded samples had reduced levels of these metabolites.The level of PPARγin the tibial bone marrow was consistent with the changes in VOCs associated with lipid metabolism.The results of this study are consistent with the potential involvement of cholesterol in tumor progression,22but the available data are preliminary.Further analysis is needed to examine the possibility of using VOCsto evaluate the ef ficacy of therapeutic interventions.

In this study,we mainly examined the regulation of the Snail-Src axis in EMT/MET transitions.Src is a nonreceptor tyrosine kinase,and many lines of evidence show its critical role in the progression of breast cancer.39Src is activated by mechanical stimulation through ECM-integrin interactions.40We observed that FFCM altered the expression and activity of Src in tumor cells that did not receive direct mechanical stimulation.FF at 1 Pa activated Src,while FFat 0.25 Pa inactivated Src.Growth factors can activate Src,41and we observed that TGFβ1 was upregulated by FFCM at 1 Pa and downregulated by FFCM at 0.25 Pa.Snail expression was also altered by FFCM in a shear stress intensitydependent manner.

This work demonstrates the involvement of mechanical stimulation in the regulation of the bone microenvironment and how tumor cells interact with the microenvironment.While the presented results reveal a novel feature of tumor-osteocyte interactions,the study has a few limitations.Our experiments used mouse bone cells and human cancer cells,and the potential effects of cross-species interactions should be taken into consideration.Interactions might be age-dependent.Furthermore,the bone microenvironment includes many other types of cells,such as osteoclasts,osteoblasts,and immune cells,which may respond differently to FF.The role of Wnt signaling and the potential involvement of Lrp5/Lrp6 coreceptors should be investigated in connection with the mechanotransduction of bone.42

In summary,this study revealed that osteocyte-tumor communication in the presence or absence of mechanical stimulation induced a substantially different effect on tumor cell behaviors.OPN,identi fied by mass spectrometry,was differentially expressed in A5 CM and FFCM,as well as in bone marrow from loaded and nonloaded tibiae,and was at least in part responsible for the observed responses in the Src-Snail regulatory axis in tumor cells.The results of tumor-osteocyte signaling suggest that strategies to prevent bone metastasis associated with breast cancer might bene fit from the inclusion of mechanical loading of the skeleton,but the loading intensity should be carefully monitored.

MATERIALSAND METHODS

Cell culture

Human breast cancer cell-derived MDA-MB-231 cells,TMD cells(CH3 BioSystems,Amherst,NY,USA),4T1 cells,4T1.2 mammary tumor cells(obtained from Dr R.Anderson at the Peter MacCallum Cancer Institute,Australia),and EO771 mammary tumor cells were grown in DMEM(Corning,Inc.,Corning,NY,USA).MLO-A5 and MLO-Y4 osteocyte-like cells(obtained from Dr L.Bonewald at Indiana University,USA)were grown inαMEM(Gibco,Carlsbad,CA,USA).For the tumor cells,the culture media were supplemented with 10%fetal bovine serum(FBS)and antibiotics.For MLO-A5 cells,the media contained 5%FBS and 5%fetal calf serum.Primary breast cancer cells(1604-27 T and 1304-37 T)were cultured in a 3:1 v/v mixture of F-12 and DMEM supplemented with 5% FBS,0.4 μg·mL-1hydrocortisone,5 μg·mL-1insulin,8.4 ng·mL-1cholera toxin,10 ng·mL-1epidermal growth factor,24 μg·mL-1adenine,and 5 μmol·L-1Y-27632.The cells were maintained at 37°C with 5%CO2.Conditioned media(A5 CM and Y4 CM)were collected from cells at~80%con fluence after 24 h incubation in media consisting of 1%FBS and antibiotics.

Ed U and scratch assays

Cellular proliferation was examined using a fluorescence-based cell proliferation kit(Click-iT™ Ed U Alexa Fluor™ 488 Imaging Kit;Thermo-Fisher,Waltham,MA,USA).After fluorescent labeling,we counted the number of fluorescently labeled cellsand determined the ratio to the total number of cells.A wound-healing scratch assay was utilized to evaluate 2-dimensional cell motility.In brief,cells were grown on 12-well plates,and a plastic tip was used to scratch a gap onto the cell layer.After incubation,the areas that were newly occupied with cells in the scratched zone were imaged and measured with ImageJ(National Institutes of Health,Maryland,USA).

Application of fluid flow

Using the previously described procedure,43oscillatory fluid flow at 1 Hz was applied for 1 h to MLO-A5 cells in a parallel plate flow chamber.A5 FFCM was collected with shear stress at 0.25Pa or 1 Pa,and A5 CM was obtained without applying fluid flow.

Western blot analysis and ELISA

Cells were lysed in radioimmunoprecipitation assay(RIPA)buffer,and proteins were fractionated using 10%SDS gels.We used antibodies against Akt,p-Akt,E-cadherin,Slug,Snail,Src,p-Src,TGFβ,Vimentin(Cell Signaling,Danvers,MA,USA),OPN,MMP9,PPARγ(Santa Cruz Biotechnology,Dallas,Texas,USA)and β-actin(Sigma).Protein levels were assayed using a SuperSignal West Femto maximum sensitivity substrate(Thermo Fisher Scienti fic).Using ELISAkits,the levels of TGFβ1(Thermo Fisher),D-dimer and OPN(MyBioSource,San Diego,CA,USA)in A5 CM and FFCM were determined.

Spheroid assay,conditioned media exchange assay,and RNA

interference

The cells were cultured in a U-bottom low-adhesion 96-well plate(S-Bio,Hudson,NH,USA).All spheroid assays were performed in completeαMEM(10%FBS,1%antibiotics).To evaluate the effect of conditioned media,spheroids were formed in separate wells for 48 h,and the media was removed and replaced with conditioned media.RNA interference was conducted using siRNA speci fic to OPN(Cat#AM16708,Life Technologies)with a negative siRNA(Silencer Select#1,Life Technologies)as a nonspeci fic control using the previously described procedure.44

Fluorescence resonance energy transfer(FRET)

To evaluate the role of Src in response to conditioned media,Src activity was quanti fied by FRETimaging as previously described.20ASrc-speci fic biosensor was labeled with cyan fluorescent protein(CFP)and yellow fluorescent protein(YFP).Time-lapse images were acquired at an interval of 5min,and the emission ratio of YFP/CFP for individual cells was computed to determine the activity levels using NIS-Elements software(Nikon).

To evaluate the tension force at a focaladhesion in the presence of varying concentrations of OPN,a plasmid expressing a vinculin tension sensor was transfected,and the fluorescence lifetime images were acquired by a custom-made microscope built on a laser scanning confocal microscope(FluoView 1000,Olympus)using the previously described procedures.21Of note,an increase in the tension force of the vinculin sensor indicated a decrease in FRETef ficiency and an increase in fluorescence lifetime.

Mass spectrometry-based protein identi fication

Three samplesof A5 CMand FFCMeach were harvested and freezedried.Using a previously described procedure,9proteins in the freeze-dried samples were analyzed by reverse-phase HPLC-ESI-MS/MSwith a Dionex-Thermo Fisher Scienti fic UltiMate 3000 RSLCnano System(Thermo Fisher Scienti fic)coupled to a Q-Exactive HFHybrid Quadrupole Orbitrap MS(Thermo Fisher Scienti fic).

Animal model

The experimental procedures were approved by the Indiana University Animal Care and Use Committee and were in compliance with the Guiding Principles in the Care and Use of Animals endorsed by the American Physiological Society.In the mouse model of osteolysis,43–4448 BALB/c and 50 C57BL/6 female mice(~6 weeks old,Envigo)received intratibial injections of 4T1.2 and EO771 cells(2.5×105cells in 20μL PBS)in the left tibia,respectively.After two days of recovery,tibia loading was performed daily on the tumor-inoculated tibiae as previously described.45Using an ElectroForce device(Bose),the left tibiae of the mask-anesthetized mice were given daily loads of 1N,2N,or 5 N(peak-to-peak)at 2 Hz for 5 min.The animals were sacri ficed on day 14.We harvested the tibiae and femurs for histology and μCTimaging,bone marrow-derived cells for Western blot analysis,urine for VOCanalysis,and blood for ELISA.

Strain measurement

Strain in the tibia in response to axial compressive loading was measured using a strain measurement unit(EDX-14A;Kyowa Americas Inc.),as well as 3D digital image correlation(3D-DIC).In the former method,a pair of strain gauges(SKF-27085,200μm gauge length;Kyowa)was immobilized to the proximal tibia in an orthogonal direction,and the tibia was compressed with 1 N,2 N,or 5 N loads at 2Hz.In the latter method,we employed the previously described procedure,46and the strain was calculated from the dynamic pattern of displacements on the tibial surface.

MicroCTimaging

Microcomputed tomography was performed using Skyscan 1172(Bruker-MicroCT,Kontich Belgium).Scans were performed at a pixel size of 8.99μm.Using manufacturer-provided software,the imageswere reconstructed(nRecon v1.6.9.18)and analyzed(CTan v1.13).We determined BV/TV(bone volume normalized to total volume),Tb.s(trabecular separation),Tb.n(trabecular number),and cortical BMD(bone mineral density)of the proximal tibial segment(2 mm thick)and distal femoral segment(1 mm thick)from the growth plate to the distal section and proximal section,respectively.

Histology

Bone samples were fixed in 4%paraformaldehyde in PBS and decalci fied in a 10%EDTA solution.They were then dehydrated through a series of graded alcohols,cleared in xylene,and embedded in paraf fin.H&Estaining was conducted on the sagittal sections,and the distribution of tumor cells in the tibial bone cavity was determined.Of note,the bone ratio was de fined as the ratio of the bone area to the total area in the proximal tibia(1mm below the growth plate),while the tumor ratio was de fined as the ratio of the tumor area to the total area.Fuchsin staining was also performed to detect microdamage in the bone.The number of Fuchsin-stained microcracks in the cortical bone of the proximal tibia(2 mm from the growth plate)was counted in a blinded fashion in samples from the sham-loaded group and the tibialoaded group at 5 N.

Analysis of volatile organic compounds(VOCs)

Twenty-four urine samples(50μL)were collected from five groups of C57BL/6 mice(10 from the normal control,4 from placebo,4 from 2 N-loaded mice,and 6 from 5 N-loaded mice).For BALB/c mice,37 samples were collected(13 from normal controls,12 from placebo,and 12 from 5 N-loaded mice).The normal control samples were collected on day-1,and the placebo and loaded samples were collected on day 15(C57BL/6 mice)and day 19(BALB/c mice).VOCs were analyzed by solidphase microextraction coupled with gas chromatography-mass spectrometry with quadrupole time-of-fl ight using a previously described procedure.23A matrix with 531 VOCs for C57BL/6 samples and a matrix with 581 VOCs for BALB/c mice were generated.Compounds that were statistically signi ficant in the normal control and placebo groups by Student’sttest were used for hierarchical clustering analysis,and principal component analysis was performed using MATLAB(R2018b;Math Works,Natick,MA,USA).Compounds were identi fied by spectral reference to NIST14,and metabolic pathway analysis was carried out via the KEGG database.

Statistical analysis

For cell-based experiments,three or four independent experiments were conducted,and the data are expressed as the mean±S.D.CM or FFCM from different dishes were collected for each independent experiment.Treatment groups were compared with vehicle-treated controls and no-fl ow control conditions.For experiments using the animal model,the metastasis model was compared with vehicle-injected control,and mechanically loaded metastatic mice were compared with sham-loaded metastatic mice.Statistical signi ficance wasevaluated using one-way analysis of variance(ANOVA).Post hoc statistical comparisons with control groupswere performed using Bonferronicorrection with statistical signi ficance set atP＜0.05.Anonparametric Kolmogorov–Smirnov test was applied to compare cell aspect ratios.The single and double asterisks in the figures indicateP＜0.05 andP＜0.01,respectively.

ACKNOWLEDGEMENTS

The authors appreciate Uma Aryal,Motoki Egi,Jiliang Li,Rahul Prakash,Xun Sun,and Mu Wang for technical support.This study was supported in part by funds from the breast cancer advocacy group 100 Voicesof Hope,as well asby NIHR01 AR52144,R03 CA238555(HY)and R01 AR053237(AGR).NIHR01 AR52144,R03 CA238555(HY)and R01 AR053237(AGR);100 Voices of Hope(HY).

AUTHORCONTRIBUTIONS

Conception and experimental design:L.Y.,J.L.,S.N.,M.A.,A.G.R.,H.N.,B.Y.L.and H.Y.Data collection and interpretation:Y.F.,A.J.,A.C.,X.Z.,S.L.,M.T.,Y.G.,F.L.,J.L.,A.S.,B.Y.L.and H.Y.Drafted the manuscript:Y.F.,X.Z.,A.S.,A.G.R.,B.Y.L.and H.Y.