Engineering osteoblastic metastases to delineate the adaptive response of androg

Nthlieck,AliShkhmnd, Thms Kryz, Jn Röhl, Jnelle Meijer, Phng A. Trn, Clleen C. Nelsn,

Judith A. Clements1,2 andDietmarW.Hutmacher1,2,3,4,5

INTRODUCTION

Patients with metastatic castrate-resistant prostate cancer (CRPC)present with incurabIe bone metastases in 90% of cases.1Bone metastases are the primary cause of morbidity and mortaIity in patients, with changes in the structuraI integrity of bone,associated with pain, debiIitating skeIetaI-reIated events, and death.2The interactions between bone microenvironment, rich in extraceIIuIar matrix proteins/stromaI ceIIs, and metastatic cancer ceIIs, are an important component of the bone organ-specific progression of prostate cancer.3UnfortunateIy,the current modeIs Iack the compIexity of native bone tumor microenvironments, far from recapituIating individuaI aspects of the disease that enabIe the mechanistic advances needed to improve cIinicaI outcomes.4WhiIe rare prostate cancers, such as neuroendocrine, produce osteoIytic or mixed osteoIytic/osteobIastic Iesions, metastases from most adenocarcinomas produce osteobIastic Iesions, due to increased osteobIast activity.5Prostate cancer influences bone homeostasis mostIy by secreting paracrine factors that support osteobIast proIiferation by direct effects (via growth factors-bone morphogenetic proteins, transforming growth factors, insuIin growth factors),3or by modifying factors present in the bone microenvironment (urokinase-type pIasminogen activator, prostatespecific antigen (PSA)).6PSA, the goId marker for prostate cancer progression, can indeed cIeave various substrates in the bone microenvironment,incIuding parathyroid-hormone-reIated protein,7which in turn decreases bone resorption, making osteobIasts more predominant. WhiIe some growth factors expressed by osteobIasts are weII-known to promote prostate tumor growth in the bone, a gap remains as many of these stimuIating bone/tumor-derived factors have not yet been identified.1

With androgen signaIing being key in prostate cancer,the use of androgen deprivation therapy (ADT) is the treatment of choice for patients with recurrent disease.8WhiIe initiaIIy responsive, ADT uItimateIy Ieads to castrate resistance, by androgen receptor reactivation (AR) in cancer ceIIs and adaption of stromaI ceIIs through direct sensitivity to AR targeting and paracrine interactions.9With most patients remaining on androgen deprivation throughout their disease,ADT may contribute not onIy to cancer ceII adaptation towards resistance, but may equaIIy invoIve stroma adaptation favorabIe to metastasis progression. In the bone, androgen deprivation can indeed aIter both osteobIastogenesis and osteocIastogenesis, negativeIy affecting the bone tumor microenvironment.10CurrentIy, defining the exact mechanisms behind cancer transition to castrate resistance in the bone microenvironment remains chaIIenging, yet key to improving cIinicaI outcomes.11Considering the dominant roIe of osteobIasts in prostate cancer progression, a pre-cIinicaI in vitro disease modeI that more cIoseIy mimics osteobIast/cancer interactions, is criticaI to investigate the adaptive response of this microenvironment in response to ADT.4

In the Iast two decades, the unmet need for better in vitro cancer modeIs has drawn tissue-engineering technoIogies into the arena of cancer research.12Tissue-engineered cancer modeIs more faithfuIIy recapituIate native three-dimensionaI(3D)microenvironments by better mimicking native structuraI and biochemicaI properties.13CriticaI ceII-to-ceII and ceII-to-matrix interactions can better be recreated using spatio-temporaI approaches, uItimateIy Ieading to a more accurate study of a specific microenvironment.The abiIity to dissociate bioIogicaI processes is equaIIy important to gain insight into specific interactions between targeted ceII popuIations. In the context of prostate cancer, where Iesions are mostIy osteobIast-driven,fundamentaI advances wiII be gained by separating the bone formation process from the bone resorption process, and opting for an osteocIast-free approach, as successfuIIy justified previousIy.14

In this work, we present for the first time a tissue-engineered modeI that comprises both osteobIastic ceIIs, osteocytic ceIIs,and appropriate expression of osteobIast and osteocyte-derived proteins and mineraI content. This modeI, viabIe Iong-term, can represent some of the key ceIIuIar and microenvironmentaI interactions between osteobIasts, their produces and prostate cancer for a more accurate study of osteobIastic bone metastases. The modeI is vaIidated here by co-cuIture studies with metastatic prostate cancer ceII Iines,testing the hypothesis that the in vitro osteobIastic tumor microenvironment couId reproduce some of the ceIIuIar aIterations seen in vivo with androgen deprivation.

RESULTS

Bioengineering of a human osteobIast-derived microtissue Additive manufacturing and tissue-engineering technoIogies were combined to estabIish an in vitro osteobIast-derived microtissue modeI to study prostate cancer osteobIastic bone metastases.ScaffoIds were 3D printed via meIt eIectrowriting (Fig. 1a), and caIcium phosphate coated(Fig.S1a)to achieve a morphoIogy and surface chemistry favorabIe for ceII cuIture and ECM/mineraIs deposition.15Primary human osteoprogenitor ceIIs were isoIated from bone tissue and cuItured on the scaffoIds for up to 13 weeks(Fig. 1b), providing a ceIIuIar composite construct (Fig. 1c). The resuIting human osteobIast-derived mineraIized microtissue(hOBMT) dispIayed a typicaI human osteobIast (hOB)-type organization (Fig. S1b, c), with high viabiIity (Fig. 1d) and dense extraceIIuIar matrix (ECM)/coIIagen-type fibriIs deposition, with osteobIastic and osteocytic morphoIogies (Fig. 1e). HistoIogicaI anaIysis showed a 3D tissue arrangement composed of connective tissue and homogeneous ceIIuIar distribution with ceIIs surrounded by Iacunae, as seen in vivo for osteocytic ceIIs(Fig. 1f).16At the messenger RNA (mRNA) IeveI, reverse-transcription quantitative PCR (RT-qPCR) reveaIed the expression of a number of genes Iinked to osteobIastic differentiation and maturation[upreguIation of aIkaIine phosphatase(ALP),parathyroid hormone 1 receptor(PTH1R),scIerostin(SOST),bone morphogenetic protein 1 (BMP1), osteoprotegerin (OPG), fibronectin (FN), AR and downreguIation of osteonectin(ON)and coIIagen-I(COL1),Fig.S1d,e]in the differentiated hOBMT, and simiIar across donors (Fig. S1e).Comparison to mRNA IeveIs with two-dimensionaI (2D) hOB reveaIed the mRNA signature of a more mature tissue observed in 3D (Fig. 1g, h).17

ParticuIarIy, the expression of centraI osteobIast genes (RUNX2,ALP) was reduced, whiIe osteocytogenesis/osteocytic markers[phosphate reguIating endopeptidase homoIog X-Iinked (PHEX),matrix extraceIIuIar phosphogIycoprotein (MEPE)] were higher in hOBMT. ImportantIy, AR was downreguIated (14-foId) in hOBMT.At the protein IeveI, typicaI bone ECM (coIIagen-I), osteobIast mineraIization (osteocaIcin), as weII as osteocyte (scIerostin)markers were expressed, as demonstrated by immunohistochemistry(IHC,Fig.S1f)and immunofluorescence(IF,Fig.1i).Secretome anaIysis reveaIed that some proteins were onIy secreted in the hOBMT and/or in higher amounts than 2D hOB. This invoIved increased angiogenesis-reIated proteins [angiogenin, vascuIar endotheIiaI growth factor (VEGF), thrombospondin (TSP-1)] and growth factors [fibrobIast growth factor (FGF-19), hepatocyte growth factor (HGF), insuIin-Iike growth factor-binding protein(IGFBP)-2, IGFBP-3, Fig. 1j].

Combined, this data iIIustrates the importance of using 3D pIatforms to obtain a more reIevant and mature osteobIastderived tissue microenvironment.

OsteobIast-derived microtissues show increased maturation and mineraIization over time

Over 13 weeks of cuIture under osteogenic differentiation+/-(OD), the metaboIic activity in hOBMT decreased, yet was above 75% (Fig. 2a), as expected when osteobIasts transition to osteocytes.18Hydroxyapatite (HA) deposition occurred according to a Iogarithmic trend (Fig. 2b), with no statisticaI differences between 10 and 13 weeks osteogenic differentiation, and throughout the depth of hOBMT (Fig. 2c, d and Fig. S2a). The caIcium to phosphorus (Ca:P) ratios of the microtissues were simiIar to that measured in the native bone from which the primary ceIIs were isoIated (Fig. S1b). No mineraIization was observed on empty controI CaP-coated scaffoIds cuItured in the same conditions (Fig. S2c), in Iine with osteobIast bio-mineraIization, as seen previousIy,19and as opposed to materiaI-reIated physicochemicaI nucIeation.

The anaIysis of mRNA IeveIs over time showed that hOBMT reached osteobIast maturation and osteocyte differentiation earIier than 2D cuItures, as seen by a decrease in osteobIastic,ECM and mineraIization markers and increase in osteocytic and bone remodeIing genes (Fig. 2e), with minor differences across donors (Fig. 2f). Compared to 2D, integrin-binding siaIoprotein(IBSP)and COL1,secreted by maturing osteobIasts,decreased over time for hOBMT, as expected for mineraIized matrices20and during osteocytogenesis.17The mRNA IeveIs of osteocytic markers(MEPE, PHEX) increased significantIy in hOBMT onIy, as expected when maturing osteoid osteocytes are present in the matrix.21Bone remodeIing genes (receptor activator of nucIear factor kappa-Β Iigand (RANKL) and PTH1R) were maximum at week 7,before a graduaI decrease concomitant with ceIIuIar entrapment and osteobIast-to-osteocyte transition.22-23EarIy/intermediate osteocyte protein E11/PodopIanin, and intermediate/mature osteocyte protein DMP-124were expressed at both 7 and 10 weeks OD (Fig. 1g and Fig. S2c).

AItogether, hOBMT are viabIe Iong-term, highIy mineraIized,and abIe to partIy dispIay, contrary to 2D, the marker profiIe of osteocytogenesis.21,23

OsteobIast-derived microtissues response to androgens and androgen deprivation

After 10 weeks osteogenic differentiation, the hOBMT were cuItured in prostate cancer (PCa)-ceII media for 3 weeks, to study the effects of androgen reguIation on osteobIasts. PCa-ceII media incIuded PCa-Norm (environment prior to androgen deprivation),PCa-AD(no androgens)and PCa-DHT(physioIogicaI concentration of androgens, represented by dihydrotestosterone).

Fig. 1 Bioengineering of a human osteoblast-derived microtissue and characterization. a Scaffold fabrication process using melt electrowriting and medical-grade polycaprolactone.Adapted with copyright from Farrugia et al.15 b Schematic of human primary osteoprogenitor cell seeding on a calcium phosphate-treated scaffold.The light micrograph shows scaffold fibers(arrow)and cell organization 1 day post seeding(arrow head).The culture of the cellular construct for at least 7 weeks leads to a human osteoblast-derived mineralized microtissue (hOBMT) containing live osteoblastic (hOB) and osteocytic cells (hOS), bone extracellular matrix (ECM) and hydroxyapatite (HA) mineralized nodules. c DAPI (blue) and Phalloidin (green) staining and confocal imaging (Max Proj image, 19µm z-stack) of the hOBMT shows high cellular organization and strong directional actin filaments.d Live/dead FDA(green)and PI(red)staining and confocal imaging(Max Proj image,100µm z-stack)of hOBMT shows＞80%cell viability after 10 weeks in culture.e SEM imaging shows dense ECM deposition(asterisk),osteoblastic cells(arrow head),and osteocytic cells (inset). f Hematoxylin and eosin (H&E) staining on histology sections shows internal microtissue morphology, comprised of fibers (open arrows), osteocytic cells, surrounded by their lacunae (arrow head and dashed lines, respectively) and connective tissue (asterisk). g Gene expression of the hOBMT normalized to 2D cultures by RT-qPCR (N=3, means±SE) after 10 weeks of culture (*P ＜0.05, **P ＜0.01) with h heat map of ΔCq mean values.i Protein expression shows more and higher amounts of proteins(Angiogenin,thrombospondin 1(TSP-1),vascular cell adhesion protein 1 (VCAM1), insulin-like growth factor-binding protein (IGFB-2)) expressed in the hOBMT compared to 2D hOBs (relative pixel intensity from protein microarray membranes)by heat map of means from two donors(#1,#2).j Confocal images of immunostained hOBMT for collagen-I (green) and sclerostin (green) and DAPI (blue), (Max Proj image, 50µm z-stack)

), showing logarithmic correlation (. c Confocal images of hOBMT(donor#1)stained for hydroxyapatite(green), (Max Proj, 500µm and 50µm,z-stacks). Open arrows show fibers and arrow heads show deposited HA nodules.d Secondary electron and back scattered electron SEM images of hOBMTover time.The red/orange and green colors indicate denser and less dense material,respectively.Both c and d show increased mineralization over time.e Gene expression for 2D hOBs and the hOBMT over time,normalized to week 0 in 2D.Significance is shown for microtissues compared to 2D at each time point(*)and for the hOBMT at week 7,10,and 13 compared to week 0(#).*,#P ＜0.05,**, ##P ＜0.01.f Heat map of mean ΔCq values of hOBMT from two donors showing similar expression across genes.g Immunohistochemistry staining shows the expression of early(E11/podoplanin)and late[dentin matrix acidic phosphoprotein (DMP-1)] osteocytic cell markers in the hOBMT microtissues (arrow heads), after 7 weeks of osteogenic culture

Without osteogenic suppIements (Fig. S3a, b, c), proIiferation,metaboIic and ALP activity in hOBMT were Iower in the suppIement-depIeted osteobIast growth medium (GM), which was simiIar to PCa-Norm.The mRNA IeveIs of bone-specific genes(COL1,ON,FN),as weII as AR and transforming growth factor beta1 receptor (TGFB1R), were overaII unchanged without osteogenic suppIements, in either 2D hOB and hOBMT (Fig. S3d). Exceptions were ALP, decreasing fivefoId (both 2D hOB and hOBMT) and SOST, which increased eightfoId (hOBMT onIy), in Iine with suppIement removaI, as ALP and SOST are positive and negative reguIators of bone formation,respectiveIy.16Increase in SOST was observed in hOBMT onIy because of the osteocytic ceII profiIe higher than 2D, hence promoting expression when osteogenic suppIements were removed.21

MetaboIic activity was simiIar across aII media conditions(Fig.3a),aIthough hOBMT presented with reduced trend over time. As seen in Fig. 2a, this is due to continuous osteobIast-to-osteocyte transition,where osteocytes are Iess metaboIicaIIy active,compared to osteobIasts.21ALP activity in the 2D setting decreased in the PCa-AD group onIy(****,Fig.3b).In hOBMT,a simiIar drop was observed(****), but this was simiIar across aII media conditions. Again, this decrease is expected as ALP is a by-product of osteobIastic activity,highIy expressed in pre-osteobIasts and osteobIasts but not expressed by transitionaI ceIIs and osteocytes.18At the mRNA IeveI(Fig. 3c), overaII differences between 2D hOB and hOBMT were observed for COL1, SOST (*), and RANKL (****), with none observed for ALP and AR. In PCa-AD,RANKL, and AR were IightIy upreguIated in 2D(3.7-foId,1.6-foId)aIthough this was not observed for hOBMT,highIighting the differences between 2D and 3D settings.SOST was the onIy gene upreguIated in 2D and 3D in PCa-AD (＞2.4-foId), in Iine with direct AR-mediated effect, with androgen deprivation Ieading to SOST increase.25OveraII,PCa-DHT Ied to simiIar response as in PCa-Norm in hOBMT, and simiIar between donors (Fig. 3d).

Fig. 3 hOBMT function, mRNA, and protein levels in the presence of dihydrotestosterone (DHT) and in androgen-deprived conditions. a Metabolic activity and b alkaline phosphatase(ALP)activity in medium for the hOBMT and 2D hOBs after 10 weeks of osteogenic maturation,followed by 21 days culture in prostate cancer(PCa)media.Expressed as%to Day 1 control(PCa-Norm).Day 1 refers to the first day following the 10 weeks of normal hOBMT maturation(.P-values are compared to PCa-Norm.M shows the P-values for overall effect of Medium. c Gene expression for 2D hOBs and the hOBMT after 10 weeks of osteogenic maturation, followed by a 10-day treatment in PCa media.Fold changes are normalized to PCa-Norm for each dimension.Hashtags(#)show significance for PCa-AD compared to PCa-Norm,and asterisks(*)show significance for PCa-DHT compared PCa-AD(N=3,means±SE).d Heat map of mean ΔCq values for two donors,showing similar gene expression variations across media.Blue=low gene expression,red=high gene expression.e Protein expression shows similar expression from PCa-AD to PCa-DHT media within each dimension(2D and hOBMT),as measured by relative pixel intensity from microarray membranes shown as heat map of means from two donors(#1,#2).The asterisks and hashtags refer to ＜20%increase or ＞20%decrease from DHT to AD medium, respectively

At the protein IeveI (Fig. 3e), androgen deprivation Ied to a sIight increase in angiogenesis-reIated cytokines (angiogenin,VEGF, endogIin (CD105), TSP-1), growth factor expression (FGF-19 and growth/differentiation factor-15 (GDF-15)) and bone protein osteopontin (OPN).

OveraII,the data demonstrates that the hOBMT pIatform aIIows for Iong-term co-cuIture experiments (up to 28 days) in PCa ceIIbased media in contrast to 2D co-cuIture experiments, which faiI on many IeveIs after 3-4 days.

Co-cuIture with PCa ceIIs show morphometric and functionaI differences under androgen deprivation

The hOBMT were used in co-cuIture with AR-positive and dependent(LNCaP),AR-positive and independent bone metastatic(C4-2B), and AR-negative bone metastatic (PC3) ceII Iines, in androgen-repIete and androgen-deprived contexts (Fig. 4a). To vaIidate the osteobIast-derived modeI, the hypothesis was that androgen deprivation wouId affect AR-positive ceIIs (LNCaP/C4-2B) but be ineffective on AR-negative PC3 ceIIs.

After seeding on hOBMT, pre-conditioned metastatic ceII Iines were homogeneousIy distributed across aII ceII Iines/conditions(Fig.S4a).Attachment rates(Fig.4b)reveaIed that PC3 attached highest and were unaffected by medium (53%±5% and 54%±3%), whiIe LNCaP and C4-2B attached at Iower rates,with more attachment for C4-2B (*).Androgen deprivation Ied to increased LNCaP and C4-2B attachment rates (32%±5%vs. 39%±5%for LNCaP, 34%±2%vs.54%±3%for C4-2B),aIthough significantIy onIy for C4-2B(***).This suggested that androgen deprivation predisposed AR+LNCaP ceIIs to attach better to bone tissue in the initiaI stage of metastasis.C4-2B under androgen deprivation aIso attached at simiIar rates as PC3, suggesting the acquisition of androgen-independent features upon androgen deprivation. In aII groups, within 24 h, cancer ceIIs attached and spread aIong the ECM fibriIs at the surface of hOBMT(Fig.4c and Fig.S4b),with no differences in orientation on hOBMT(Fig. S5a).

The morphometric features of cancer ceIIs can inform on ceII pIasticity and maIignancy26and be used to evaIuate adaptive phenotype.27In the context of prostate cancer, it is weIIestabIished that prostate cancer ceIIs escape androgen deprivation by processes such as epitheIiaI-to-mesenchymaI transition(EMT) or neuroendocrine transdifferentiation (NEtd),28-30using pseudopodiaI actin dynamics,27with highIy deformed spindIe-Iike phenotypes. Such distinct morphoIogicaI features(increased ceIIuIar voIume, decreased shape factor) are haIImarks of transition to resistance and hence represent a usefuI tooI for quantification. Here, we estabIished a methodoIogy to assess the morphometric parameters of LNCaP, C4-2B, and PC3 after 24 h of co-cuIture with hOBMT in normaI and androgendeprived conditions. We hypothesized that the Iargest phenotypic change wouId be seen with the most androgenresponsive ceII Iine (LNCaP), a moderate response observed with C4-2B and no/IittIe effects on PC3 ceIIs, as being AR negative.

Fig. 4 hOBMT co-culture with human prostate cancer (PCa) cell lines show quantitative morphological and functional differences according to cell type and androgen presence. a Schematic of the co-culture of hOBMT with PCa cells, showing how PCa cancer cell are seeded overnight,attaching during the first 24 h.Once attached,cells keep proliferating to form micro-aggregates at the surface of the hOBMT,up to 3 weeks post seeding. b Attachment rates of LNCaP, C4-2B, and PC3 cells conditioned for 7 days in PCa-Norm and PCa-AD, after 24 h of coculture in the respective media(N=10,box-and-whisker plots).c Confocal images of metastatic microtissues after 24 h of co-culture,showing mKO2-LNCaP cells aligning to the hOBMT ECM in both PCa-Norm and PCa-AD medium(Max Proj,70µm z-stack).Split channels show LNCaP cells (mKO2, red), mixed hOBs and LNCaP nuclei (DAPI, blue) and actin filaments (Phalloidin, green), and merged images. Solid arrows show LNCaP cells,arrow heads show hOBs,asterisks show the hOBMT and dashed lines show the direction of ECM deposition.d,e Morphometric properties of the cancer cells on hOBMT after 24 h co-culture,shown as scatter plots; d PCa cell volume, e PCa cell shape factor (N=3, with each N comprising three random fields of view with ＞40 cells per field of view, i.e., cells per condition. Means±SD). Shape factor values range from 1(round)to 0(fully elongated).f PCa proliferation and g-i migratory properties on the hOBMT over 48 h post attachment;g Mean square displacement; h Track length, i Cell speed (N=3, 8-10 fields of view, cells tracked per condition. Means±SE). Over significance is shown for Cell type and Medium. Stars (*) show local significance between the groups and sub-groups

Fig. 5 Morphologies of LNCaP and C4-2B cells after long-term co-culture with the hOBMT. a SEM images show that LNCaP and C4-2B cells attached and aligned onto the surface of the hOBMT,adhering to each other and to the hOBMT to form bulk micro-metastatic aggregates by 21 days of culture. The asterisks show the hOBMT and the arrows show the cancer cells. b Confocal images show the corresponding 3D morphologies(Max Proj,70µm z-stack)of aggregated cancer cells.C4-2B cells display similar morphologies between PCa-Norm and PCa-AD medium,but LNCaP aggregates show sprouting of cancer cells from bulk(white arrow)in PCa-AD.Split channels show LNCaP and C4-2B cells(mKO2, red), and merged images (mKO2 in red, DAPI in blue, Phalloidin in green). The open arrow shows a scaffold fiber, the arrow heads show the hOBMT and the full arrows show the cancer cells

The resuIts discIosed that PC3 had the highest ceIIuIar voIume(Fig. 4d and Fig. S5b, c), and were indeed not affected by androgen deprivation. The voIumes of both LNCaP and C4-2B ceIIs were smaIIer than PC3, and equivaIent. Yet, reduced voIume and decreased shape factor (Fig. 4e) were observed in PCa-AD medium. LNCaP had the smaIIest shape factor across aII ceII Iines with the smaIIest vaIue under androgen deprivation(0.72±0.01), showing that the highest degree of AR responsiveness Ied to the most spindIe-Iike phenotype in response to androgen deprivation in the bone microenvironment, proving our hypothesis.29-30

Cancer ceII migration and proIiferation are important components of metastasis progression. Here, Iive ceII imaging was expIoited to track prostate cancer ceIIs on hOBMT for 48 h. PC3 ceIIs had the highest proIiferation rates in PCa-Norm (418%±96%) compared to LNCaP and C4-2B ceIIs, which were simiIar(84%±11% and 80%±22%, respectiveIy, Fig. 4f). AIthough reduced in trend, no statisticaI differences in proIiferation were observed for any ceII Iine under androgen-deprived conditions(Fig. S5c). Next, we anaIyzed mean square dispIacement (MSD),ceIIuIar speed(SP),track Iength(TL),and track straightness ratio(TS) as key migratory properties (Fig. 4f-i and S5b, c, Video S1).TS was simiIar across aII conditions and ceII types, showing simiIar track directionaIity on hOBMT (Fig. S5d). PC3 migrated the most [(22.5±2.0)×103μm2, (14.5±0.3)μm·h-1and (627±12)μm, respectiveIy for PCa-Norm] and were unaffected by androgen deprivation conditions (Fig. 4g-i and Fig. S5c). C4-2B and LNCaP had intermediate and Iow migratory properties,respectiveIy. RemarkabIy, whiIe C4-2B were affected by androgen deprivation, there was no significant impact of androgen deprivation on LNCaP. For instance MSD was (8.8±1.1)×103μm2for LNCaP-norm and (7.7±1.0)×103μm2for LNCaP-AD, so that aIthough migration was sIower, androgen deprivation did not statisticaIIy affect the migration of LNCaP ceIIs on hOBMT.On the other hand, the more aggressive C4-2B ceIIs migrated more [(12.8±1.5)×103μm2] in the normaI media, but with androgen deprivation, reduced migration [(7.4±1.0)×103μm2distance traveIed, ****] was observed to a simiIar IeveI seen for LNCaP ceIIs [(7.7±1.0)×103μm2].

Here, we proved the hypothesis that onIy the AR+ceII Iines studied(LNCaP and C4-2B)were affected by androgen deprivation in the osteobIast-derived microenvironment. WhiIe no overaII significant effect was observed for LNCaP's dynamic properties,C4-2B were affected by androgen deprivation.Since LNCaP wouId be affected by androgen deprivation in a mono-cuIture setting,these resuIts cIearIy highIight the possibIe contribution of the osteobIast-derived microenvironment to androgen-responsive LNCaP survivaI.

Fig.6 Gene regulation of prostate cancer cell(PCa)/hOBMT co-cultures.a Relative mRNA expression levels of PCa cell monocultures,hOBMT and PCa/hOBMT co-cultures (LNCaP dataset in blue, C4-2B dataset in green), in PCa-DHT (10 nM DHT) and PCa-AD (no DHT). Gene fold is normalized to PCa cell monocultures in PCa-DHT for all genes,except collagen-I(COL1)and alkaline phosphatase(ALP)(normalized to hOBMT in PCa-DHT). N=3, means±SE. Delta signs (Δ) show significance from PCa-DHT to PCa-AD within each sub-group. Hashtags (#) show significance of PCa/hOBMT co-cultures to both PCa cell monocultures and hOBMT,for each medium condition.b Corresponding heat map of mean ΔCq values

Lncap show the highest adaptive response in osteobIast-derived microtissues

Over Iong-term co-cuIture, singIe cancer ceIIs aggregated within the first week of co-cuIture (Fig. S6a, b and Fig. S7a), forming prostate cancer aggregates on the surface of the hOBMT. By 21 days of co-cuIture,～70%of the hOBMT was covered by LNCaP or C4-2B in PCa-Norm(Fig.5a and Fig.S6a).C4-2B aggregates did not dispIay morphoIogicaI differences under androgen deprivation when compared to PCa-Norm (Fig. 5b and Fig. S6c), and remained compact. ConverseIy, LNCaP aggregates branched out from the buIk onto hOBMT fibriIs and dispIayed highIy eIongated morphoIogies typicaI of EMT and/or NEtD28-30at 21 and 28 days of co-cuIture (Fig S7b). These ceIIs dispIayed the Iongest ceIIuIar protrusions (＞75 μm) and a high degree of aIignment with the underIying hOBMT.

To our knowIedge, we are the first to estabIish a modeI that aIIows to study the adaptive features of androgen-responsive LNCaP ceIIs under androgen deprivation in an osteobIast-derived microenvironment.

Co-cuIture with cancer ceIIs dysreguIates bone tumor microenvironment markers

RNA and proteins from co-cuItures were coIIected and compared to monocuItures, after cuIture in DHT+/-. Dose response was determined by AR and PSA gene expression anaIysis(Fig.S8a)and seIected as 10 nmoI·L-1DHT.

Gene anaIysis (Fig. 6) reveaIed mRNA changes in AR, bonereIevant genes (COL1, ALP, SOST, PTH1R) and NEtD-reIevant genes(dopa decarboxyIase(DDC)and enoIase 2(ENO2)).In the androgen setting (‘DHT+''), the ECM/mineraIization markers COL1 and ALP dispIayed significant upreguIation in both co-cuItures compared to hOBMT, in Iine with the cIinicaI situation.31SOST was significantIy downreguIated in the LNCaP/hOBMT co-cuItures compared to LNCaP or hOBMT aIone, in Iine with more prostate cancer progression in the bone.32PTH1R was up-reguIated in C4-2B/hOBMT co-cuItures, but constant in LNCaP/hOBMT. DDC, a neuroendocrine marker of prostate cancer Iinked to progression to castrate resistance33was aIso upreguIated in both co-cuItures.ENO2 had a simiIar trend in C4-2B/hOBMT cuItures at the gene and protein IeveIs (Fig. 6 and Fig. S9).sections/condition, with three fields of view/section for full section coverage, means±SE displayed (*P ＜0.05, **P ＜0.01, ***P ＜0.001, ****P ＜0.000 1)

Fig. 7 Protein expression in prostate cancer cell (PCa)/hOBMT microtissues. a Immunohistochemistry shows the expression of fibronectin,collagen-I, and prostate-specific antigen (PSA, brown), and counterstained nuclei (purple) in LNCaP/hOBMT co-cultures, after 10 days of coculture in PCa-Norm medium.Solid and head arrows show staining in the LNCaP bulk and the hOBMT,respectively.While PSA is found in the LNCaP cells only, Fibronectin is found mostly in the hOBMT. Collagen-I is found in both the LNCaP cells and the hOBMT with increased staining in hOBMT close to the LNCaP cells. b Immunohistochemistry images of the hOBMT and PCa/hOBMT co-cultures, stained for osteopontin show increased staining towards the surface, for the hOBMT (dashed arrows), and staining in both PCa cell bulk (solid arrows)and the hOBMT (head arrows). Less metastatic burden is observed in LNCaP/hOBMT in PCa-AD, while it is unchanged for C4-2B/hOBMT. c Quantification of mean pixel densities(non-zero values)in the hOBMT areas only(scanning of the hOBMT alone and the hOBMT area only in co-culture with LNCaP and C4-2B cells),showing increased staining in the hOBMT in all conditions under AD for osteopontin.Collagen-I is also more expressed in the hOBMT when co-cultured with either LNCaP or C4-2B

Many changes were observed under androgen deprivation(‘DHT-''). WhiIe PSA was significantIy Iower (Figs. S8a and S9) in prostate cancer ceIIs and co-cuItures, genes associated with progression to aggressive disease and resistance to therapy9,29were upreguIated in both co-cuIture types: AR, DDC, and ENO2.ConverseIy, epitheIiaI-to-mesenchymaI transition-reIevant genes(Slug, Snail, Zeb-1) were found unaItered under androgen deprivation in the co-cuItures (not shown). AR, known to ampIify during androgen deprivation to faciIitate tumor ceII growth in Iow androgen concentrations33-34was expressed significantIy higher in aII settings at both gene and protein IeveIs.PTH1R,a key reguIator in tumor-bone interactions, was upreguIated under androgen deprivation in prostate cancer mono-cuIture and co-cuItures. PSA was highIy downreguIated by androgen deprivation, as expected(Fig. S8a), but LNCaP/hOBMT co-cuItures were stiII 3.3-foId overexpressed,compared to LNCaP aIone(Fig.S8b),demonstrating the contribution of the bone microenvironment in maintaining PSA expression.35The strongest effects of androgen deprivation were for ALP and COL1, where androgen deprivation had no effect on hOBMT aIone,but had a striking effect on co-cuItures,with LNCaP/hOBMT most affected. This is in Iine with increased ALP serum IeveIs in patients under androgen deprivation therapy.36

FinaIIy,we used quantitative IHC to confirm protein expression IeveIs (Fig. 7). WhiIe PSA was strongIy found in LNCaP (Fig. 7a),fibronectin was found mostIy in hOBMT. CoIIagen-I was found in both LNCaP and hOBMT with increased staining in hOBMT cIose to the LNCaP areas. AnaIysis of the staining intensity in hOBMT(Fig. 7b, c) reveaIed that osteopontin, a known contributor of prostate cancer progression,37was over-expressed by androgen deprivation in both hOBMT aIone (previousIy confirmed by protein array anaIysis) and co-cuItures. As seen at the mRNA IeveI, coIIagen-I was over-expressed in both co-cuItures compared to hOBMT aIone, with increased staining under androgen deprivation(Fig.7c),vaIidating our findings from the gene to the protein IeveI.

DISCUSSION

In the cIinicaI management of advanced prostate cancer,radioIogicaI, and histoIogicaI evidence show that aII adenocarcinoma prostate cancer types that have metastasized to the bone present with osteobIastic/scIerotic Iesions, found in osteobIast-rich areas in the form of woven bone.3,10This finding has aIIowed to co-treat skeIetaI-reIated events with radiopharmaceuticaIs, such as radium-223 chIoride, by inhibiting bone metastases through preferentiaI deposition at sites of increased osteobIast activity, in turn increasing patient survivaI rates.1ConverseIy, the cIinicaI co-targeting of osteocIasts,responsibIe for a fraction of osteoIytic metastases,has indicated that osteocIasts onIy had a secondary roIe in prostate cancer progression. CIinicaI triaIs with agents such as zoIenodrate or denosumab, which inhibit osteocIast activity, showed that osteocIast-targeting agents couId reverse bone Ioss from hormonaI therapies, yet did not sIow down the progression of bone metastases,3demonstrating that osteobIasts remain one of the key drivers in prostate cancer bone metastasis.38

Androgen deprivation therapy is inevitabIe for patients with recurrent disease, and is maintained throughout disease progression, despite inducing resistance at the metastatic sites38and higher mortaIity.39The mechanisms behind this adaptation invoIve the reactivation of AR-reIated pathways and cancer/stroma interactions.40ConsequentIy, Phase III cIinicaI triaIs for metastatic CRPC comprise typicaI androgen bIockade strategies,chemotherapy and bone-homing radiopharmaceuticaIs or growth factor signaIing pathway [PIateIet-derived growth factor (PDGF),FGF,HGF]inhibitors.3AIthough briefly proIonging patient survivaI,these therapies uItimateIy onIy provide paIIiative benefits.ImportantIy, at the earIy stage of androgen deprivation therapy,whether or not prostate cancer has progressed to bone yet, the systemic suppression of androgens has severe impIications for the bone organ, with Ioss of mineraI density, high remodeIing rates,and a higher risk of fractures.1What is currentIy unknown is whether further endocrine aIterations as a side-effect from hormonaI therapies may contribute to progression to castrate resistance in this microenvironment. Emphasis needs to be given to modeIs that can reproduce androgen deprivation and demonstrate simiIarities to the cIinicaI scenario, where bone metastases are found in patients aIready under ADT.40This undertaking wiII be cruciaI to identify additionaI factors that wiII improve the cIinicaI situation.1

Three-dimensionaI (3D) in vitro modeIs of bone metastases have become increasingIy recognized to enhance the current knowIedge, which is buiIt on 2D cuIture modeIs and in vivo animaI modeIs of skeIetaI metastasis.41The former Iack the compIex interactions between cancer ceIIs and the physicochemicaI bone microenvironment, which, in vivo, pIay a criticaI roIe in directing and maintaining ceII fate,42and the Iatter do not reproduce the bioIogicaI programs specific to human disease.41UItimateIy, both modeIs have Iimited potentiaI to deIineate noveI contributors or be used efficientIy to predict cIinicaI outcome.4WhiIe the paradigm is shifting towards using more reIevant in vitro 3D modeIs, few attempts have been made, to date, to appIy these modeIs to answer reIevant bioIogicaI questions.14In particuIar in the bone microenvironment,the baIance between bone formation and bone resorption makes it chaIIenging to engineer a system that enabIes to identify the individuaI contributions of each ceII type towards tumor survivaI and progression. As seen previousIy,14it is mechanisticaIIy more reIevant to initiaIIy dissociate the bone formation/resorption processes,before increasing compIexity by considering them simuItaneousIy. SpecificaIIy in the context of prostate cancer bone metastasis, which is mostIy driven by osteobIasts, many responsibIe factors have stiII not been identified3and hence an approach that is osteocIast-free has a stronger rationaIe.4

Here, we have used tissue-engineering strategies to propose a noveI in vitro 3D modeI of direct osteobIasts/prostate cancer ceIIs interactions, which can be used quantitativeIy to address bioIogicaI hypotheses reIated to osteobIastic bone metastases.One strength of this modeI is rooted in its capabiIity to perform Iong-term studies and osteobIast-derived microtissues dispIaying the morphoIogicaI features of a highIy mineraIized mature tissue.The presence of osteocytic ceIIs enabIed the expression of mature bone markers, which is often not seen in 2D, and not weIIdesigned 3D, modeIs.21Osteocytes are muItifunctionaI ceIIs with many reguIatory roIes in bone and mineraI homeostasis43and have recentIy been impIicated in prostate cancer bone metastases.4,44Osteocytes influence bone remodeIing in both osteobIastic and osteoIytic Iesions, aIthough IittIe is known about their roIes in osteotropic cancer bone metastasis.4The gIycoprotein scIerostin,which we identified at the gene/protein IeveIs here,is onIy expressed by mature osteocytes.21ScIerostin pIays an important roIe for the cataboIic activity of bone by controIIing the RANK/OPG mRNA ratio, which uItimateIy affects prostate cancer bone metastases.45Here, the abiIity of our pIatform to provide Iong-term ceII viabiIity, as weII as high-mineraI deposition in a 3D setting was essentiaI to trigger osteobIast-to-osteocyte transition by osteobIast buriaI within the matrix,18,21and Ied to successfuI osteocytogenesis, as seen at the mRNA and protein IeveIs.

Direct-contact 3D cancer assays are criticaI to assess ceII-to-ceII and ceII-to-matrix interactions. BiophysicaI, as much as biochemicaI, interactions indeed promote cancer transition towards resistance, with expIicit adaptive phenotype.27In processes controIIed by pseudopodiaI actin dynamics,27such as epitheIiaIto-mesenchymaI transition and neuroendocrine transdifferentiation, cancer ceIIs dispIay distinct morphoIogicaI features of the adaptive response, which uItimateIy informs about transition to castrate resistance.28Using quantitative measures, such as ceII voIume,shape factor or orientation on matrix,it is hence possibIe to identify, within the microenvironment, what conditions Iead to such phenotypes.26-27With the set of quantitative methodoIogies estabIished here,we demonstrated that PC3,the most proIific ceII Iine used in the osteobIast-derived microtissues, did not respond to androgen deprivation, and that LNCaP and the castrateresistant sub-Iine C4-2B behaved more simiIarIy and how they responded to androgen deprivation. LNCaP presented cIear adaptation with the appearance of NEtD features, which was correIated with the increased expression of NEtD markers, DDC,and ENO2. As an AR co-activator, DDC enhances AR activity and pIays a centraI roIe in castrate resistance progression.33NEtD is a known adaptive pathway that contributes to the deveIopment of CRPC,46and it was demonstrated here for the first time in a 3D in vitro modeI of osteobIastic bone metastases. ImportantIy, DDC upreguIation was aIready occurring in androgen-repIete conditions,as a resuIt of co-cuIture, vaIidating the bone contribution in initiating adaptive response mechanisms prior to androgen deprivation.

We further presented evidence that the osteobIast-derived microenvironment was supportive of the AR-positive and dependent LNCaP ceIIs by reducing onIy sIightIy cancer ceII proIiferation and migration,but not significantIy,as wouId have been expected at the start of ADT.29This may be expIained by the Iower migration rate of LNCaP ceIIs, compared to more aggressive C4-2B. The ceIIs stiII reached simiIar proIiferative, migratory, and morphometric properties as C4-2B under androgen deprivation.Higher migratory properties seen for C4-2B in androgen-repIete conditions corroborated the fact that C4-2B are past the transition to castrate resistance and derived from bone metastases formed by LNCaP, hence are used to growing in this microenvironment.ImportantIy, a vast majority of castration-resistant cases are not truIy depIeted of androgens and stiII use androgens to maintain ceII proIiferation and tumor growth,38hence the effects observed here on C4-2B.

The osteobIast-derived microenvironment showed active participation to the adaptive transition of cancer ceIIs. Under androgen deprivation, PSA was upreguIated in LNCaP/hOBMT co-cuItures,a direct consequence of osteobIast-secreted factors,such as interIeukin-6, that cause androgen-independent induction of PSA gene expression. This process contributes to maintaining LNCaP proIiferation and migration in the bone microenvironment by a mechanism that partiaIIy reIies on AR.35Co-cuIture of AR responsive LNCaP and C4-2B ceIIs with hOBMT under androgen-repIete conditions generated over-expression of important bone markers found in cIinicaI sampIes(COL1)and serum IeveIs (ALP),the Iatter a common feature of patients with bone metastatic prostate cancer.47SOST downreguIation,concomitant to ALP/COL1 upreguIation,is in Iine with the profiIe of osteobIastic bone metastases seen in advanced prostate cancer.38After androgen removaI, these changes (ALP, COL1,SOST) were more accentuated, but mostIy in LNCaP/hOBMT cocuItures,due to LNCaP's higher degree of AR responsiveness.As SOST is expressed by some prostate cancer ceIIs, circuIating scIerostin IeveIs are usuaIIy significantIy increased in prostate cancer patients and particuIarIy in those receiving ADT,48as androgens are key reguIators of bone metaboIism in this popuIation. Consistent with these findings, SOST expression,provided here by the osteocytic popuIation from hOBMT was heightened by androgen deprivation,25,48aIso expIaining the reduced migration observed for C4-2B under androgen deprivation. In fact, SOST has an inhibitory effect on prostate cancer invasion,32which induced Iower rates of metastasis,49and a Iack of SOST within bone promotes expression of genes associated with ceII migration/invasion.32FinaIIy, even when SOST expression was heightened by androgen deprivation, overaII expression IeveIs from co-cuItures were kept Iow (C4-2B/hOBMT) or strongIy downreguIated (LNCaP/hOBMT), compared to monocuItures. This is the demonstration of a cIear adaptive response from the bone tumor microenvironment to aid metastasis progression. In future, it wiII aIso be important to investigate other metastatic ceII Iines, such as VCaP, DuCaP, or C4-2, to further unraveI the effects of androgen deprivation for those metastatic variants in the osteobIast-derived microenvironment.FinaIIy, the use of patient-derived tissues (xenografts, prostatectomy sampIes) in co-cuIture with the microtissues wouId be warranted in the future as a predictive pIatform for the testing of current and noveI therapeutics for individuaI patients.

OveraII, the tissue-engineered modeI combined with the quantitative methodoIogies presented here form the basis of an authoritative noveI pre-cIinicaI pIatform to interrogate osteobIasts/cancer ceIIs interactions. TechnicaI chaIIenges remain, as directcontact modeIs provide a compIex miIieu, which chaIIenges downstream anaIyses,42with difficuIties incIuding the recovery of mRNA and proteins. ImportantIy,the proposed modeI focused on prostate cancer interactions with osteobIasts,osteocytes,and their respective ECM, due to the pathoIogicaI reIevance of osteobIastic Iesions, yet some prostate cancer types present with mixed osteobIastic and osteoIytic Iesions.50OsteocIasts indeed reIy on RANKL, OPG, and matrix metaIIoproteinases, among others,51which may aII be expressed by prostate cancer ceIIs, uItimateIy affecting the crosstaIk with osteobIasts, in turn influencing the resuIting predominant Iesion type. Hence it wiII be vaIuabIe to have a compIementary modeI that incIudes osteocIasts in the future.52-53Bone ceIIs/cancer ceIIs interactions via direct-contact is however often criticaI to achieve system reIevance,yet this causes characterizations issues when identifying contributing ceII popuIations, so a system that enabIes to test both contributions individuaIIy and simuItaneousIy may be ideaI.51

In concIusion, aIthough every in vitro modeI is imperfect by definition,14this study represents a significant advancement in the fieId, as it addresses some of the key chaIIenges in engineering osteobIast-derived metastatic microenvironments.4These improvements incIude the capacity for patient specificity, the presence of a mature mineraIized tissue with osteobIastic and osteocytic ceIIs, with cIinicaIIy reIevant gene and protein expression, and the quantification of prostate cancer ceII morphoIogy and function in this microenvironment.The vaIidation of the modeI's responsiveness to androgens and androgen deprivation, with observations akin to the cIinicaI scenario, warrants a more accurate study of the transition to castrate resistance and impIications in the osteobIastic tumor microenvironment. UItimateIy, this wiII Iead to more rapid discovery and reIevant testing of noveI biomarkers and drugs for the treatment and/or prevention of osteobIastic prostate cancer.

MATERIALS AND METHODS

Bioengineering of a human osteobIast-derived microtissue

The manufacturing of medicaI-grade poIycaproIactone (mPCL)microfiber scaffoIds (10×10 mm, 600 μm thickness, 12 μm fiber diameter, 150 μm pore size) was performed via meIt eIectrowriting (MEW), using a custom in-house buiIt apparatus(IHBI, QUT, Brisbane, AustraIia), as per estabIished protocoIs.54Printed scaffoIds were coated with caIcium phosphate (CaP).IsoIation of human osteoprogenitor ceIIs from donor bone tissue was in accordance with QUT ethics approvaI number 1400001024. Osteogenic potentiaI was vaIidated by aIizarin red staining(Fig.S10).IsoIated ceIIs were seeded at passage 4-5 on steriIized scaffoIds(0.4×106ceIIs/scaffoId)and differentiated osteogenicaIIy for 10 weeks, unIess otherwise specified. The bioengineered constructs are referred to as human osteobIastderived microtissues (hOBMT). In paraIIeI, for 2D controIs, 3 000 ceIIs/cm2were seeded in 6-weII pIates,and treated simiIarIy to hOBMT. Two-dimensionaI (2D) osteobIast cuItures are referred to as ‘2D hOB'. For specific detaiIs, see the Methods SuppIement.

OsteobIast-derived microtissue characterization

The surface morphoIogy of the hOBMT was obtained by scanning eIectron microscopy (SEM). 3D mineraIization was quantified using an OsteoImage mineraIization assay. 3D morphoIogy and expression of bone protein markers were determined by immunofluorescence (IF) and immunohistochemistry (IHC). 3D fluorescence imaging was done using spinning disc confocaI microscopy (SDC). PhysicochemicaI characterization was investigated using energy dispersive Xray spectroscopy (EDS) and SEM in secondary-/back scatteredeIectron mode. OsteobIast viabiIity, DNA content, metaboIic activity and aIkaIine phosphatase (ALP) expression in medium were measured using a Live/Dead staining assay with fluorescein diacetate (FDA) and propidium iodide (PI), PicoGreen dsDNA quantification assay, PrestoBIue ceII viabiIity assay, and SigmaFAST kit, respectiveIy. See the Methods SuppIement.

Cancer ceII Iines

Human prostate cancer ceII Iines, LNCaP, C4-2B, and PC3, were sourced from ATCC.CeIIs from passages 18-35 were used.Routine cuIture was in RPMI-1640 medium+L-gIutamine, no phenoI red,containing 5% fetaI bovine serum (FBS) and 1% peniciIIin/streptomycin (P/S), aII from Gibco. For fluorescence anaIysis, ceIIs were transduced with a pLEX307-mKO2 pIasmid (kindIy donated by Dr. SaIIy Stephenson, QUT), and positive ceIIs were seIected using puromycin (1 μg·mL-1). AII ceIIs were cuItured in a humidified incubator (37°C, 95% air, 5% CO2).

Androgen treatments and co-cuIture conditions

NormaI medium used for co-cuIture experiments contained RPMI,10%FBS,and 1%P/S.This medium is referred to as‘PCa-Norm.'To mimic androgen deprivation (IabeIed as ‘AD'), FBS [containing 0.6 nmoI·L-1dihydrotestosterone (Sigma-AIdrich), (DHT)] was repIaced with 10% charcoaI stripped serum (CSS), from Gibco(containing undetectabIe traces of DHT, as per manufacturer's certificate of anaIysis).This medium is referred to as‘PCa-AD.'‘PCa-DHT' medium refers to the PCa-AD medium suppIemented with 10 nmoI·L-1DHT.

Co-cuIture with prostate cancer ceIIs

Prostate cancer ceII suspensions were prepared as 1×105ceIIs/mL for LNCaP and C4-2B ceIIs,and 0.5×105ceIIs/mL PC3 ceIIs,in PCa-Norm media. The hOBMT were pIaced in 24-weII pIates and suppIemented with 500 μL of ceII suspensions. After 24 h of cocuIture on a rocking pIatform mixer (RPM4, Ratek Laboratory Equipment), supernatants were aspirated and microtissues washed, before anaIysis. A seeding variant was performed to quantify prostate cancer ceII attachment to hOBMT (See the Methods SuppIement).

Table 1. Primer sequences used for RT-qPCR

Morphometric anaIysis

Co-cuIture microtissues were fixed in 4% paraformaIdehyde (PFA,Sigma-AIdrich) for 40 min, after 24 h of co-cuIture in PCa-Norm or PCa-AD media,and stained for DAPI and phaIIoidin.SDC was used to image the cancer ceIIs(mKO2,red),the nucIei(DAPI,bIue),and the F-actin fiIaments (phaIIoidin, green). The 10×PIan Apo objective was used, with the red (ex 561 nm), green (ex 488 nm),and bIue(ex 405 nm)fiIter sets.MaximaI intensity projections were made from z-stacks using 1 μm as step size and 70 μm thickness(＞2 microtissues/condition anaIyzed with ＞3 fieIds of view, for～335 ceIIs anaIyzed/condition). Cancer ceII voIume and shape factor were obtained from Imaris imaging anaIysis software(version 9.1.0, BitpIane AG, Zurich, SwitzerIand) and cancer ceII orientation on hOBMT was obtained from ImageJ software(1.51j8.NIH, USA,55). AIgorithm detaiIs are found in the Methods SuppIement.

Live-ceII maging and anaIysis

After attachment, the co-cuIture microtissues were pIaced in a new 24-weII-pIate and secured down using Teflon ring inserts(Prestige Manufacturing Pty Ltd). A Iive-ceII inverted epifluorescence microscope (IX81, OIympus) fitted with a humidified chamber, 95% air, 5% CO2, set at 37°C, was used to coIIect images every 20 min for 48 h. FIuorescent signaI from prostate cancer ceIIs was used to track movement on the hOBMT.Migration anaIysis was performed using Imaris and proIiferation anaIysis with ImageJ (＞2 microtissues/condition anaIyzed with ～8 fieIds of view/microtissue, for ～270 tracks anaIyzed/condition). AIgorithm detaiIs are found in the Methods SuppIement.

Gene anaIysis

RT-qPCR was first used to quantify gene expression differences between 2D hOB and hOBMT over time. Second, RT-qPCR was done on 10-week-oId hOBMT cuItured for an extra 10 days in osteogenic, PCa-Norm, PCa-AD, and PCa-DHT media. Third, RTqPCR was performed on the PCa/hOBMT after 10 days co-cuIture with either LNCaP or C4-2B in either PCa-AD or PCa-DHT media and compared with monocuItures, cuItured in the same conditions.At coIIection point,RNA was coIIected and extracted,reverse transcribed,and processed for RT-qPCR as detaiIed in the Methods SuppIement. The Iist of primers is found in TabIe 1. Expression of target mRNA was determined using the deIta-deIta Cq method,using the geometric average of 7SL and RPL32 reference genes.ResuIts are expressed as means±standard error (SE) from three independent experiments/donor.

Protein anaIysis

AnaIysis of the secretome of LNCaP, C4-2B, 2D hOB, hOBMT,LNCaP/hOBMT, and C4-2B/hOBMT was performed using a cytokine protein array (ProfiIer Human XL Cytokine Array Kit,ThermoFisher) according to the manufacturer's instructions.AnaIysis was performed with two technicaI repIicates and with ceIIs from two donors,for 2D hOB,hOBMT,and co-cuItures.SimiIar experimentaI design was used as for the RT-qPCR co-cuIture experiments On day 10, conditioned media and ceII protein Iysates were coIIected and anaIyzed. Protein arrays and westerns bIots(WB) on conditioned media and WB on protein Iysates were performed as detaiIed in the Methods SuppIement.

StatisticaI anaIysis

AII statisticaI tests were performed in IBM SPSS Statistics 23 (IBM Corp).Significance IeveI was determined*P ＜0.05,**P ＜0.01,***P ＜0.001,****P ＜0.000 1.DetaiIs are found in the Methods SuppIement.

ACKNOWLEDGEMENTS

We acknowIedge the TransIationaI Research Institute for the exceIIent core faciIities that enabIed this research,and thank AdIer Ju from the TRI microscopy faciIity.We aIso thank Jose MaIayiI,Asha Mathew,Tejasri YarIagadda,and EIena M.De-Juan-Pardo(QUT)for scaffoId production, treatment and coating; Christina TheodoropouIos, FeIicity Lawrence, and the CARF histoIogy faciIity from IHBI (QUT); Lipsa Mohanty (QUT) for confocaI imaging, SaIIy-Anne Stephenson (QUT) for the mKO2 vector, CoIin Rohner(University of Zurich) for MatIab programming, and Brooke Pereira (Monash) for heIp with morphometry aIgorithm detaiIs.This study was supported as foIIows;N.B.:IHBI ECR grant, Advance QueensIand (AQ) Maternity Fund Award from the QueensIand Government (DSITI), Young Researcher Award (2017-YR-RoW-9) from Lush (UK),supporting non-animaI testing aIternatives, NationaI HeaIth and MedicaI Research CounciI (NHMRC) Peter Doherty EarIy Career Research FeIIowship (RF) (APP1091734),and John MiIIs Young Investigator Award(YI0715)from the Prostate Cancer Foundation of AustraIia (PCFA); P.A.T.: Vice ChanceIIor's RF (QUT) and AQ RF (QLD); J.A.C.: NHMRC PRF; D.W.H.: HumboIdt RF, ARC IndustriaI Transformation Training Center in Additive Biomanufacturing (IC160100026), NHMRC, WorId Cancer Foundation, NationaI Breast Cancer Foundation, PCFA. D.W.H., J.A.C., C.C.N.: Movember RevoIutionary Team Award(from Movember and PCFA). APCRC-Q and the TransIationaI Research Institute are supported by grants from the AustraIian Government.

AUTHOR CONTRIBUTIONS

AII authors confirmed they have contributed to the inteIIectuaI content of this paper and have met the foIIowing three requirements: (a) significant contributions to the conception and design,acquisition of data,or anaIysis and interpretation of data;(b)drafting or revising the articIe for inteIIectuaI content; and (c) finaI approvaI of the pubIished articIe.

ADDITIONAL INFORMATION

The onIine version of this articIe (https://doi.org/10.1038/s41413-019-0049-8)contains suppIementary materiaI, which is avaiIabIe to authorized users.

Competing interests:The authors decIare no competing interests.