Dulbecco’s modified Eagle’s medium (DMEM), minimum essential medium (MEMα), and other chemicals, unless otherwise stated, were obtained from Sigma (St. Louis, MO, USA). HyClone fetal bovine serum (FBS) was from ThermoFisher (Pittsburg, PA, USA). Trypsin–EDTA, Alexa Fluor 488-conjugated goat anti-rat and anti-rabbit IgG, MitoTracker Deep Red FM, CellTracker Orange (CTO), DQ collagen type IV, BODIPY (493/503), Hoechst Dye, and Gentamicin (G418) were from Invitrogen (Carlsbad, CA, USA). StemXVivo Adipogenic Supplement was from R&D Systems (Minneapolis, MN, USA). Rosiglitazone was from Cayman Chemical Company (Ann Arbor, MI, USA). PureCol collagen type I was from Advanced Biomatrix (San Diego, CA, USA). Cultrex™ (rBM; reduced growth factor) was from Trevigen (Gaithersburg, MD, USA). Rat monoclonal F4/80 was from Abcam (Cambridge, MA, USA). Rabbit monoclonal carbonic anhydrase 9 (CA9) and hexokinase II (HK2) were from Cell Signaling Technology (Danvers, MA, USA). Transwell systems (Costar™ Transwell™ Permeable Supports with 0.4-μm pore size) were from Corning (Corning, NY, USA). FABP4 inhibitor (BMS309403) was from Calbiochem (San Diego, CA, USA). Atglistatin was from Axon Medchem (Groningen, Netherlands).

PC3, an androgen independent cell line derived from a bone metastasis of a high-grade prostate adenocarcinoma, was purchased from American Type Culture Collection (ATCC; Manassas, VA, USA). The PC3-DsRed cell line was established by stable transfection with pDsRed2-N1 vector (Clontech Laboratories, Palo Alto, CA, USA), containing the neomycin-resistant gene. Transfection was performed using Lipofectamine 2000 and pooled populations of stable cells were selected, expanded, and maintained in medium supplemented with 400 mg/ml of G418 (29). L929 cells (source of M-CSF for macrophages, purchased from ATCC) were cultured until confluent and conditioned medium was collected, centrifuged, and stored at −80°C until ready for use. PC3 and L929 cells were cultured in DMEM supplemented with 10% FBS, 10 mM HEPES, and 100 U/ml penicillin–streptomycin. All cells were maintained in a 37°C humidified incubator ventilated with 5% CO₂.

Collection of murine cells was performed in accordance with the protocol approved by the institutional Animal Investigational Committee of Wayne State University and NIH guidelines (Protocol # 15-12-025; IP, PI). Primary mouse bone marrow stromal cells (mBMSC) were isolated from femurs and tibiae of 6- to 8-week-old FVB/N mice and induced to become bone marrow adipocytes according to our previously published protocols (5). Specifically, bone marrow from each tibia and femur was flushed with of DMEM containing 20% FBS (4 ml/bone) using a 26-gauge needle. Marrow suspension was mixed and broken apart using a 20-gauge needle and seeded into a 6-well cell culture plate (~3 ml/well). After 24 h, non-adherent cells were removed by replacing the medium. Cells were cultured to confluency by changing the medium every 2–3 days and then expanded to larger dishes as needed. For adipogenic differentiation, mBMSCs (~600,000/well) were mixed with bovine collagen I and seeded in six-well plates (500 μl/well) for Transwell coculture or 60-mm dishes (500 μl, ~1,300,000/dish) for 3D invasion assays. Approximately 48–72 h later, upon reaching confluency, cells were treated with adipogenic cocktail (30% StemXVivo Adipogenic Supplement, 1 μM insulin, 2 μM Rosiglitazone; DMEM and 10% FBS) for 8–10 days (5).

For the preparation of bone marrow macrophages (BMMs), bone marrow was flushed from femurs and tibiae of 10- to 12-week-old FVB/N male mice with BMM growth medium (MEMα containing 20% FBS and 30% L929-conditioned media as the source of M-CSF; 4 ml/bone) using a 26-gauge needle (30). The cell suspension was mixed with an additional 20 ml of BMM medium using a 20-gauge needle and plated onto three 100-mm Petri dishes (12 ml/dish). Cells were cultured for 4–5 days in a 37°C humidified incubator ventilated with 5% CO₂ to obtain differentiated BMMs.

Bone marrow adipocytes were differentiated in six-well plates as described above. On the day of the Transwell setup, tumor cells were plated on acid-washed glass coverslips (12 mm in diameter) coated with reconstituted basement membrane (rBM; Cultrex™; Trevigen). Briefly, 100 μl of 1 mg/ml rBM was added per coverslip and allowed to polymerize for 45 min at RT, then the excess was removed via suction, and the coverslips were allowed to dry for 15 min at RT before cells were plated. PC3 cells (30,000–50,000) were plated in 500 μl of normal growth medium and allowed to settle for a minimum of 4 h at 37°C. Adipocytes were prepared for Transwell coculture by washing 3× with PBS and adding 2 ml of DMEM supplemented with 5% FBS. Transwell membranes were placed above the adipocyte culture according to manufacturer’s instructions and coverslips with attached tumor cells (2× coverslips per transwell) were placed on each membrane with medium gently added on top (2 ml). The control coverslips were cultured separately in DMEM supplemented with 5% FBS. All cells were allowed to grow for 48 h at 37°C.

Adipocytes were prepared as described above for the 2D system. The experimental approach to prepare 3D cultures of prostate carcinoma cells has been adapted from the original protocols for breast cancer progression models developed by the Brugge laboratory (31). Non-diluted rBM (15.35 mg/ml) was used for the preparation of all cultures. In brief, acid-washed coverslips were placed in 35 mm dishes, and 45 μl of rBM was carefully added on top of each coverslip, making a continuous surface without going over the edge of the coverslip. The dishes were placed in 37°C for 15 min to polymerize, then 60 μl of PC3 cell suspension was placed on top of each rBM-coated coverslip (30,000–35,000 cells/coverslip). The dishes were placed in a 37°C incubator for 45 min–1 h to allow the cells to settle and attach to the matrix. After cells were attached, 2–3 ml of media (5% FBS, 2% rBM overlay) was added to the 35-mm dish control coverslips. For 3D Transwell cultures, one 3D coverslip (covering only ~24% of the membrane and thus allowing free flow of factors and nutrients) was placed on each Transwell membrane positioned above the differentiated adipocyte culture, and 2 ml of medium was gently added (Figure 1). Cells were allowed to grow for 48–120 h depending on experimental design.

For the analysis of the expression and localization of carbonic anhydrase 9 (CA9) (Figure 2), control and Transwell coverslips from 2D and 3D cultures were washed with PBS, fixed with cold methanol, and incubated with rabbit monoclonal anti-CA9 antibody (1:50) at 4°C overnight. To label the mitochondria, coverslips were washed with PBS, and incubated with 200 nM MitoTracker Deep Red for 30–45 min at 37°C prior to fixation. For immunodetection of hexokinase 2 (HK2) (Figure 3), cells were washed with PBS, fixed with cold methanol, and stained with rabbit monoclonal anti-HK2 antibody (1:100) at 4°C overnight. Alexa Fluor 488-conjugated goat anti-rabbit IgG (1:1000) was used as a secondary antibody for both CA9 and HK2 immunostaining. For 2D cultures, DAPI was used as a nuclear stain, and coverslips were mounted using Vectashield mounting medium (Vector Laboratories) before imaging with Zeiss LSM 510 META NLO confocal microscope using 40× oil immersion lens. Coverslips from 3D cultures were left in PBS in 35-mm dishes, Hoechst Dye (1:1000) was added for labeling of nuclei, and imaging was performed at an extended depth of focus with a Zeiss LSM 510 META NLO confocal microscope using a 40× dipping lens.

Primary BMMs were differentiated from murine bone marrow cells, as described above. To examine their ability to infiltrate the pre-formed tumor spheroids in the absence or presence of adipocytes, the 3D tumor-adipocyte Transwell cocultures were set up and cultured for 3 days, as described above. To distinguish tumor cells from macrophages, PC3 cells were pre-labeled with CTO (1:1000 in serum-free media for 1 h at 37°C) prior to plating on rBM-covered coverslips. After 3-day culture, medium with rBM overlay was removed from the Transwell or control dishes, 60 μl of cell suspension containing 150,000 BMMs was plated on top of each coverslip and allowed to settle for 45min-1 h at 37°C, then new medium (containing 2% rBM overlay) was reapplied. Cells were allowed to grow for additional 2 days (Figure 4) before cultures were fixed in 3.7% formaldehyde and immunostained for murine macrophage marker F4/80 using rat anti-mouse F4/80 antibody (1:50). Alexa Fluor 488 conjugated goat anti-rat IgG (1:1000) was used as a secondary antibody. Hoechst Dye (1:1000) was added for labeling of nuclei, and imaging was performed at an extended depth of focus with Zeiss LSM 510 META NLO confocal microscope using a 40× dipping lens.

Cleavage of DQ-collagen IV substrate by live PC3 cells grown in a control or Transwell coculture with adipocytes in the presence or absence of BMMs was assayed in real time and quantified based on published protocols (32–34). Briefly, single cell suspensions of tumor cells (30,000–35,000) were plated on top of coverslips coated with rBM containing DQ-collagen IV (1:30) and overlayed with 2% rBM. Cells were grown alone or in Transwell coculture with adipocytes for 3 days prior to addition of pre-labeled BMMs (Figure 5). After an additional 2-day culture, Hoechst dye was added as nuclear marker and DQ-IV proteolysis under all culture conditions was imaged live at an extended depth of focus with Zeiss LSM 510 META NLO confocal microscope using a 40× dipping lens.

To examine the effects of BMMs on 3D growth and proteolysis by prostate tumor cells cocultured with marrow adipocytes, we pre-labeled differentiated BMMs by incubating them with a 1:1000 dilution of with CellTracker Orange (CTO) in serum-free medium for 1 h at 37°C. Labeled BMMs were then washed 3× with PBS and maintained in normal BMM medium overnight until use. On the day of 3D Transwell setup, adipocytes and prostate tumor cells were prepared as described above. Coverslips were coated with rBM and 75,000 BMMs mixed with 30,000 PC3 cells were plated together on top of a coated coverslip, as described above. As before, one 3D coverslip was placed on each Transwell membrane positioned above differentiated adipocyte culture, and 2 ml of medium was gently added (Figure 6). Cells were allowed to grow for 3 days, and DQ-IV proteolysis was imaged live at an extended depth of focus with a Zeiss LSM 510 META NLO confocal microscope using a 40× dipping lens.

Bone marrow-derived mBMSC cells were mixed with collagen I matrix and plated in 60-mm dishes (500 μl/dish; ~1,300,000/500 μl). Cells were differentiated into adipocytes over the course of 8–10 days as described above. Post-differentiation, each plate was washed 3× with PBS, with all PBS carefully removed before 1 ml of collagen I was added on top and allowed to polymerize at 37°C for 45 min–1 h. A total of 4 ml of medium, containing 400,000 dsRed-expressing PC3 cells, was gently added on top. Cultures were maintained in a 37°C humidified incubator ventilated with 5% CO₂ for up to 96 h. At designated time-points, plates were washed with PBS, fixed for 45 min with 3.7% formaldehyde at RT, stained with BODIPY (493/503) (1:1000, 1 h at RT), nuclei were marked with Hoechst dye, and imaged at an extended depth of focus with a Zeiss LSM 510 META NLO confocal microscope using a 20× dipping lens (Figure 7). For inhibitor studies (Figure 8), FABP4 inhibitor (BMS309403; 1 μM) and Atglistatin (10 μM) were added to both the collagen I layer between the adipocytes and the tumor cells as well as the medium, and were replenished in the medium daily.

All images were captured using a Zeiss LSM510 META NLO confocal microscope (Carl Zeiss AG, Göttingen, Germany). For 2D cultures, a 40× oil immersion lens was used, and a single slice was captured. 3D cultures were imaged in 35 mm dishes with a 40× water immersion lens, and z-stacks of 2–4 μm optical slices were captured. 3D reconstruction of optical slices was performed using Volocity software (PerkinElmer, Waltham, MA, USA). For quantification of DQ-IV proteolysis, Volocity software was used to determine integrated intensity per image by dividing the total sum of green fluorescence signal (thresholded using intensity) by the number of nuclei in the image. At least three replicates were measured per condition. The totals in each group were averaged and shown as mean ± SEM.

Invasion cultures in 60 mm dishes were imaged using a 20× water immersion lens, and z-stacks of 4 μm increments were captured. 3D reconstruction of optical slices was performed using Volocity software. For quantification of distance traveled by the tumor cells, images were opened in Zeiss LSM Image Browser. After removing the DIC channel, the Ortho tool was used to select and measure the distance from the middle of PC3 (red) layer to the middle of the adipocyte (green) layer. Distances between PC3 and adipocyte layers were recorded from at least three biological triplicates. For inhibitor studies, distances were shown as percent of distance at baseline (4-h control).

All data analyses were performed using GraphPad Prism Software version 6.05. Data were presented as mean ± SD and statistically analyzed using unpaired Student’s t-test. For three or more groups, one-way analysis of variance was used.

Adipocytes are metabolically active cells that have been shown to promote growth and aggressiveness of several cancers, particularly those that grow in adipocyte-rich microenvironments such as the breast or ovary (10, 35, 36). The fat cells residing specifically in the bone marrow have been increasingly credited with the ability to support and promote metastatic growth in bone (4–9). We have shown previously that exposure of prostate tumor cells to marrow adipocyte-supplied factors increases their proliferation and invasiveness (5). Here, to better mimic physiological conditions and to allow exploration of molecular mechanisms of adipocyte involvement in metastatic tumor growth in bone, we developed a cell culture model that allows 3D growth of prostate cancer cells exposed to adipocytes.

Seeding of the single-cell suspensions of PC3 prostate carcinoma cells on rBM-coated coverslips with 2% rBM overlay results in a formation of tumor spheroids (Figure 1). Cells cultured in their normal growth medium (CONTROL) form compacted, rounded spheroids after 3-day culture (Figure 1B, top panels). In contrast, tumor cells grown in a Transwell coculture with marrow adipocytes (TRANSWELL) form much larger and more disorganized clusters (Figure 1C, top panels). Longer, 5-day culture of PC3 cells results in further growth under either of the conditions, with Transwell cultures exhibiting significantly more robust proliferation and acquiring much more disordered morphology compared with control cultures (Figures 1B,C, bottom panels). Quantification of 3D volumes of the spheroids confirms the growth-promoting effects of adipocytes on tumor cells and indicates potential utility of this experimental system for interrogation of adipocyte–tumor cell interactions (Figure 1D).

One way that adipocytes can affect tumor cell behavior is by regulating cancer cell metabolism (37). Adipocyte-supplied lipids have been shown to feed into the glycolytic pathway and (38–40) and induce the Warburg effect in tumor cells (41–45). One of the master regulators of metabolic reprograming is hypoxia inducible factor (HIF), a key driver of hypoxic stress (46). Bone marrow adipocytes have a capability of inducing HIF-1α signaling in prostate tumor cells grown in 2D monolayer culture (Podgorski et al., unpublished results¹). It is, however, well-recognized that tissue dimensionality and associated oxygen status have a potential to profoundly affect hypoxic response in tumor cells (19, 47). Therefore, we examined the utility of our 3D Transwell model to measure hypoxia response to adipocytes. Our 3D cultures established on glass coverslips are easily fixable and suitable for immunocytochemical analyses. We used carbonic anhydrase 9 (CA9), a HIF-1α target gene (48), as a measure of hypoxic response in our system. The immunofluorescence analysis of CA9 protein revealed a significant increase and typical membrane localization in 2D tumor cell cultures exposed to adipocytes (Figures 2A,B), indicating activation of HIF-1α signaling. Interestingly, cells grown in 3D culture showed robust CA9 expression even under control conditions (Figure 2D). However, although tumor cells grown in transwell coculture with adipocytes expectedly formed significantly larger, disorganized structures (Figures 2E,H,I,L,M), there was no visible increase in CA9 expression compared with spheroids grown under control conditions (Figures 2D,F,G,J,K). This was further confirmed by the quantification of CA9 fluorescence/nuclei in each of the 3D structures (data not shown). 3D growth alone appears to activate HIF-1α signaling, which is consistent with the reports of heterogeneous oxygen distribution in 3D cultures (47). This speaks to the importance of considering tissue architecture while examining metabolic responses within the tumor microenvironment. Further studies are underway to validate these findings.

One important consequence of HIF-1α activation is the induction of glycolytic phenotype in tumor cells. We performed immunofluorescence analysis of expression and localization of hexokinase-2 (HK2), a critical enzyme in the first step of glycolysis, that elicits its functions by binding to the voltage-dependent anion channels (VDAC) in the outer mitochondrial membrane (49, 50). Low HK2 expression was detected in monolayer cultures of PC3 cells (Figure 3A), with robust increases in HK2 fluorescence upon exposure to adipocytes (Figure 3E). This is consistent with adipocyte-induced Warburg phenotype in these cells (see footnote text 1). Mitotracker labeling (Figures 3B–D,F–H) showed no observable differences in the number of mitochondria between the control and Transwell conditions. The majority of the HK2 appeared to co-localize with mitochondria (Figures 3C,G), which is consistent with the literature evidence demonstrating that approximately 80% of total HK2 is bound to the mitochondrial VDAC (51).

In stark contrast to the monolayer cultures, a significant HK2 expression was observed in PC3 3D spheroids even under control conditions (Figures 3I,K–N). This indicates a potential enhancement of glycolytic phenotype, consistent with activation of HIF-1α signaling by 3D culture (Figure 2). High HK2 fluorescence closely mirroring mitochondrial pattern was also observed in 3D cultures grown under Transwell conditions (Figures 3J,P–S). However, HK2 fluorescence intensity was not significantly higher than in control cultures, indicating no significant enhancement of HK2 levels by coculture with adipocytes (data not shown).

Bone represents a complex, dynamic microenvironment and establishment of metastatic lesions in the skeleton requires tumor cell interactions with a number of cell types in the marrow. To examine the suitability of our cell culture system to study multi-cell interactions, we incorporated BMMs into 3D tumor spheroids growing under control and Transwell conditions (Figure 4A). Pre-labeling of PC3 tumor cells with CTO allowed visualization of the tumor growing in 3D culture (Figures 4B,C). We tracked the incorporation of BMMs into the tumor spheroid by F4/80 immunofluorescence staining. Tumors growing under Transwell conditions with adipocytes attracted more macrophages than those cultured under control conditions (Figures 4B,C; green fluorescence). Incorporation of BMMs into the spheroid was also visible in the optical slices taken in the middle of the spheroid (Figures 4D–G). These observations suggest that adipocytes might be inducing phenotypic changes in the tumor cells and/or macrophages to promote the BMM affinity for the tumor. Further studies are needed to confirm this hypothesis.

To grow and thrive in their surrounding environments, tumors have to interact with the extracellular matrix (ECM) via biochemical and mechanical processes. Proteolysis of the ECM is one of the key aspects of tumor invasion (52, 53), and we have previously shown that exposure to adipocyte-derived factors increases invasive potential of prostate carcinoma cells (5). One way to effectively visualize ECM proteolysis by the tumor cells and other cell types residing in the tumor microenvironment is through a confocal microscopy assay utilizing quenched-fluorescent (DQ) proteins, an approach developed by the Sloane laboratory over a decade ago (53). We have previously utilized this technique to demonstrate that prostate carcinoma cells are capable of degrading quenched fluorescent derivatives of collagen IV (basement membrane) and collagen I (organic matrix of the bone) (33). Therefore, we used this approach to determine whether exposure to adipocyte-derived factors affects the collagen IV degradation in our model. PC3 cells plated on rBM containing DQ-IV were allowed to establish into spheroids by growing for 3 days under control or Transwell conditions before adding CTO-labeled BMMs and allowing cultures to continue for additional 2 days (Figure 5A). Degradation of the DQ-collagen IV, observed as green fluorescence, was captured and quantified at the end of 5-day culture. In agreement with our previous findings (33), fluorescent cleavage products of DQ-collagen IV were found both intracellularly and pericellularly (Figures 5B–E). Similar levels of fluorescence were detected in PC3 cells cultured alone (without BMMs) regardless if they were under control or Transwell conditions (Figures 5B,D,F,H,J), suggesting no measurable contribution of adipocytes to tumor-driven proteolysis.

Since macrophages are important sources of proteases and key contributors to tumor-associated proteolysis (34, 53), we also measured and quantified the DQ-IV fluorescence following the 2-day coculture of pre-established tumor spheroids exposed to macrophages (Figures 5C,E,G,I,J). Surprisingly, despite the significant infiltration of BMMs into tumor spheroids exposed to adipocytes in a Transwell coculture and visible BMM-associated proteolysis (Figures 5E,I), there was no significant increase in overall tumor-associated DQ-IV degradation compared with cultures grown without macrophages (Figure 5J).

Interestingly, when macrophages were incorporated into the culture at the time of seeding the tumor cells (Figure 6A), significant changes were observed in the morphology of the resulting spheroids and in the tumor-associated proteolysis. Specifically, under control conditions, addition of BMMs led to formation of less compacted and more disorganized tumor spheroids compared with tumor cells cultured alone (Figures 6B,F vs. Figures 6C,G). This change in morphology was also associated with a significant increase in DQ-IV proteolysis (Figure 6J). Even more pronounced effects of BMMs were observed in cultures exposed to adipocytes in the Transwell system (Figure 6D,H vs. Figures 6E,I). Formation of highly disorganized tumor cell/BMM clusters with high levels of DQ-IV fluorescence (Figure 6J) suggested potentially important contribution of adipocytes to tumor- and BMM-driven proteolytic activity in this model.

Our previous work has shown that prostate carcinoma cells exposed to media conditioned by marrow adipocytes take up the fat cell-supplied lipids in a process that results in accelerated tumor growth and invasiveness (5). Prostate tumor cells have been reported to be attracted to adipocyte-rich areas of the bone marrow, where metastases commonly occur (54, 55), and the translocation of adipocyte-stored lipids has been linked to increased tumor cell motility (56). The mechanisms behind the tumor cell attraction to marrow fat cells, however, remain poorly understood. Here, we have designed a simple 3D system in which tumor cell invasion toward adipocytes can be easily visualized, measured, and quantified (Figure 7A). We used collagen I, the predominant component of bone, as the matrix separating BODIPY-labeled bone marrow adipocytes (bottom) and DsRed-expressing PC3 cells (top). Tumor cells invaded through collagen I matrix over the course of 96 h (Figures 7B–E), with complete merging of the two cell types at the 96-h time point.

We next tested two compounds with potential to affect adipocyte–tumor cell interactions: a selective inhibitor of fatty acid-binding protein 4 (FABP4) (BMS309403) (5), and Atglistatin, a selective inhibitor of Adipocyte Triglyceride LIpase (ATGL) (57). We chose to inhibit FABP4 based on our previous studies demonstrating upregulation of this lipid transporter in prostate tumor cells exposed to marrow adipocytes, and the apparent efficacy of BMS309403 in inhibiting prostate tumor cell invasion in the 2D Boyden chamber assay (5). We also used Atglistatin as the means of inhibiting adipocyte-driven lipolysis, which we have shown to be induced by prostate tumor cells (see footnote text 1). As expected, under control conditions, the distance between PC3 cells and marrow adipocytes was significantly reduced after 70 h, compared with the baseline at 4 h (Figures 8A,B). Atglistatin alone did not seem to have any effect on tumor cell invasion toward the adipocytes (Figure 8C). Despite potent inhibition of PC3 invasion in 2D (5), only modest inhibition of tumor invasion was observed with FABP4 inhibitor in a 3D assay (Figure 8D), a result potentially due to differences in matrix and experimental conditions between the two assays. Notably, however, the use of both inhibitors in combination significantly inhibited tumor invasion and restored the distance between the two cell types to the levels measured at 4 h (Figures 8E,F). This suggests a potential link between lipid hydrolysis and transport between adipocytes and tumor cells that needs further investigation.