Matrigel^® was obtained from Corning, Inc. The ROCK1 inhibitor Y-27632 was purchased from Merck & Co., Inc.

The human EC cell line, ECC-1 was purchased from American Type Culture Collection (cat. no. CRL-2923) and was cultured in medium consisting of RPMI-1640 (Thermo Fisher Scientific, Inc.) supplemented with 10% FBS and 1% penicillin/streptomycin. The ECC-1 cell line was authenticated using STR profiling. The immortalized human normal endometrial fibroblast cell line, T-HESC (ATCC^® CRL-4003™), was purchased from the American Type Culture Collection and cultured with Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin.

The sources of CAFs and BAFs and patient age are shown in Table I. All samples were collected from women undergoing surgery at University of Malaya Medical Centre. The study was approved by The Ethical Committee of University of Malaya Medical Centre (approval no. 865.19). Written informed consent was obtained from all participants. The fibroblast lines were labelled using an arbitrary number assigned to each recruited patient. CAF and BAF cultures were established as previously described (16). First, ~1 g of tissue samples was transported to the laboratory in RPMI-1640 supplemented with 10% FBS and 1% penicillin/streptomycin. The samples were washed twice with PBS and minced to 1-mm³ size, then digested with 2 mg/ml collagenase (Worthington Biochemical Corporation) in a rotator for 1 h at 37°C. After digestion, the tissue samples were washed with PBS and cultured in RPMI-1640 supplemented with 10% FBS and 1% penicillin/streptomycin at 37°C. The medium was changed every 72 h, and the cells were subcultured when they reached confluence.

Second, CAFs or BAFS were further isolated from the primary cell culture using human anti-fibroblast magnetic microbeads (cat. no. 130-050-601; Miltenyi Biotec GmbH). Briefly, 1×10⁶ cells were centrifuged at 300 × g for 10 min, at room temperature. The cell pellets were resuspended in 100 µl calcium- and magnesium-free PBS (pH 7.2) containing a final concentration of 0.5% BSA and 2 mM EDTA dissolved, then incubated with 20 µl human anti-fibroblast microbeads for 1 h. The cells were then separated using MiniMACS™ cell separator (Miltenyi Biotec, Inc.). The isolated cells were cultured in the aforementioned medium and maintained below passage 10, to maintain a phenotype close to that of the primary tissues. These fibroblasts were characterized by high expression of fibroblast markers (vimentin and α-SMA) and low expression of epithelial markers (EpCAM and E-cadherin) (Fig. S1).

Total RNA was extracted from cultured cells using TRIzol^® (Invitrogen; Thermo Fisher Scientific, Inc.). A total of 1 µg RNA was converted into cDNA using Revert Aid RT Kit (Thermo Fisher Scientific, Inc.), according to the manufacturer's instructions. The primer sequences used to quantify the mRNA expression levels of Rho GTPases, fibroblast markers and epithelial markers are listed in Table II. GAPDH was used as the internal control gene. The specificity of the primer sequences was validated using Primer-BLAST (National Center for Biotechnology Information). The qPCR was performed using ABI StepOne Plus (Applied Biosystems) in 40 cycles using 5X HOT FIREPol EvaGreen qPCR Mix (Solis Biodyne), 10 pmol/µl forward and reverse primer, 10 ng/µl cDNA template and PCR-grade water. The following thermocycling conditions were used for qPCR: Initial denaturation for 10 min at 95°C, followed by 40 cycles of denaturation at 95°C for 15 sec and annealing/extension at 55°C for 45 sec. The relative expression levels were calculated using the 2^−∆∆Cq method (17). For ECC-1 and CAF monoculture, the relative fold-changes in gene expression were normalized to the non-Matrigel cultures (2D cultures), whereas for the ECC-1 and CAF co-cultures, the relative fold changes in gene expression were normalized to those of the ECC-1 monoculture to obtain the expression levels of the CAFs. The relative fold changes from three different CAF lines were averaged (EC48Fib, EC49Fib, and EC50Fib).

The spheroids formed at day 1 and day 14 were digested with 2 mg/ml collagenase (Worthington Biochemical Corporation) in a rotator for 1 h at 37°C. After digestion, the cells were centrifuged at 300 × g for 10 min, at room temperature. The cell pellets were resuspended in 100 µl calcium- and magnesium-free PBS (pH 7.2) containing a final concentration of 0.5% BSA and 2 mM EDTA dissolved, then incubated with 20 µl human anti-fibroblast microbeads and 20 µl human CD326 (EpCAM) microbeads (cat. no. 130-061-101, Miltenyi Biotec, Inc.) for 1 h at 4°C. The cells were then separated using MiniMACS™ cell separator (Miltenyi Biotec, Inc.).

ECC-1 cells and fibroblasts were transduced with pre-made lentiviral particles for red fluorescence protein (RFP; cat. no. LVP429) and green fluorescence protein (GFP; cat. no. LVP426) obtained from GenTarget Inc., respectively. Briefly, the cells were seeded at a density of 5×10⁴ cells/well in a 24-well plate. After 24 h, 15 µl pre-made lentiviral particles (1×10⁷ IFU/ml) were added, together with 235 µl of serum-free medium. A 250 µl volume of complete medium was added after 6 h. Selection was carried out by supplementing the cultures with a final concentration of puromycin of 1 µg/ml (MilliporeSigma) for 2 weeks.

RFP-expressing ECC-1 cells were co-cultured with GFP-expressing fibroblasts at 1:1 ratio (2×10⁴ cells per dish) each on Matrigel^® (Corning, Inc.)-precoated glass bottom cell culture dish (WillCo Wells B.V.). The cells were cultured using DMEM F12 (Thermo Fisher Scientific, Inc.) supplemented with 10% FBS and 1% penicillin/streptomycin. The culture medium was changed every 3–4 days over a period of 14 days.

Immunofluorescence staining was performed on ECC-1 and CAF monocultures, as well as ECC-1 and CAF co-cultures on day 1 and day 14. The 3D spheroids were fixed with 4% paraformaldehyde at 4°C for 20 min, followed by permeabilization with 0.01% Triton-X at 4°C for 20 min. The spheroids were stained with anti-phosphorylated (p-) MLC primary antibody at 4°C, overnight (cat. no. 3671S; diluted 1:500; Cell Signaling Technology, Inc.), and then with Alexa Fluor^® 647-conjugated secondary antibody at 4°C for 1 h (cat. no. A32733; diluted 1:1,000, Thermo Fisher Scientific, Inc.). The primary and secondary antibodies were diluted with ImmunoDetector Protein Blocker/Antibody Diluent (BIOTnA), which included bovine serum albumin as the blocking reagent. The slides were mounted with Vectashield Antifade Mounting Medium with DAPI (Vector Laboratories, Inc.). The images were analysed using a DeltaVersion deconvolution microscope (GE Healthcare).

Fluorescence images of the 3D spheroids were analyzed using a Leica TCS SP5 II laser confocal microscope (Leica Microsystems GmbH). The Helium-Neon (543 nm) and Argon (488 nm) lasers were used to detect RFP and GFP signals, respectively. The images were acquired sequentially to avoid crosstalk between fluorescence channels. A stack of optical sections was acquired by changing the position of the focal plane in the z-direction with a step size of 2.98 µm using a 10X air objective. Images taken in each z-plane were combined into one single picture using the maximum intensity projection function. The 3D reconstruction of z-planes was then analyzed using LAS X 3D Visualization version 3.1.0.15537 (Leica Microsystems GmbH). The 3D reconstruction of the x, y and z planes of the 3D sphere were adjusted to the center of the plane in the cross-section analysis. The distance between the outermost edges of the z-plane was measured as the diameter of spheroids.

Live cell imaging was performed using a Nikon Eclipse Ti fluorescence microscope (Nikon Corporation). The hardware was controlled using NIS Element version 4.30. To maintain cell viability during imaging, an environmental chamber at 37°C with 5% CO₂ supply was mounted on the microscope. A z-stack of images with a step size of 8 µm was taken at 1-h intervals using a 10X S plane Fluor dry objective. The entire imaging process was performed over 16 h. To create videos illustrating of cell motility, a sharp and focused single z-plane image was selected at each time point. The velocity (µm/h) of the cells was calculated as the distance travelled by the cells (µm) divided by time (h) (18).

ECC-1 cells were cultured in RPMI1640 supplemented with 10% FBS and 1% penicillin/streptomycin in a 24-well plate until they reached 100% confluence. A small linear scratch was created in the middle of the well using a 10 µl pipette tip. The ECC-1 cells were treated with 2% FBS medium or conditioned medium collected from the CAFs over 3 days. The wound width was measured using a Nikon Eclipse Ti fluorescence microscope (Nikon Corporation).

ECC-1, CAFs and BAFs were seeded at 5×10⁴ cells/well on fibronectin pre-coated coverslips in 24-well plates. After overnight incubation, the cells were treated with Y-27632 at 50, 75 or 100 µM for 24 h. The cells were washed with PBS and fixed for 20 min in 3.7% formaldehyde solution in PBS at 4°C. The cells were further permeabilized with 0.1% Triton X-100 in PBS. The cells were stained with 50 µg/ml phalloidin-TRITC (MilliporeSigma) in PBS for 40 min at room temperature. The cells were washed with PBS three times. The slides were mounted with Vectashield Antifade Mounting Medium with DAPI (Vector Laboratories). The cells were analyzed using a Nikon Eclipse Ti fluorescence microscope (Nikon Corporation).

Cell viability was assessed using an MTT assay. ECC-1, CAFs and BAFs were seeded at 5×10³ cells/well in 96-well plate overnight. The cells were treated with Y-27632, at concentrations ranging from 0.2 to 100 µM or vehicle (0.01% DMSO) for 24 h. At the end of treatment, 20 µl MTT solution (5 mg/ml) was added to each well. Following 4 h of incubation at 37°C, 100 µl 10% sodium dodecyl sulfate was added to dissolve the formazan crystals in an additional 4 h of incubation at 37°C. Absorbance was measured using a spectrometer at 575 nm, with a reference of 650 nm.

In cell seeding, RFP-expressing ECC-1 cells were co-cultured with GFP-expressing fibroblasts at a 1:1 ratio (2×10⁴ cells per dish) on a Matrigel^® (Corning, Inc.)-precoated glass bottom cell culture dish (WillCo Wells B.V.). A total of 100 µM Y-27632 and control (10% FBS) was added during cell seeding. Live cell imaging was performed as aforementioned.

Statistical analysis was carried out to assess the differences between the means of the control and test groups using unpaired Student's t-test. One-way ANOVA followed by Tukey's post hoc test or Bonferroni's correction was performed when there were >2 comparison groups. All statistical analyses were performed using GraphPad Prism version 6 (GraphPad Software, Inc.). P<0.05 was considered to indicate a statistically significant difference.

The isolated fibroblast cells were first characterized for the expression of fibroblast and epithelial markers. Both isolated CAFs (EC48Fib, EC49Fib and EC50Fib) and BAFs (EF2Fib, EH6Fib and NE14Fib) showed high expression of vimentin and α-SMA, with low expression of EpCAM and E-cadherin (Fig. S1). The effects of these fibroblasts on EC cell motility were examined using wound healing assays. ECC-1 tumor cells were treated with conditioned medium collected from CAFs. As shown in Fig. S2, there was no significant ECC-1 cell migration following treatment with conditioned medium treatment, except that the conditioned medium collected from EC48Fib led to a small but significant increase in migration at day 3. Thus, the effects of CAFs on EC cell motility were further investigated using live cell imaging of the ECC-1 and CAF co-cultures in Matrigel^®-coated dishes over 16 h.

In order to visualize the individual movement of EC cells and fibroblasts in 3D co-culture, the cells were transduced with RFP and GFP lentivirus, respectively. Selection was performed by supplementing the cultures with puromycin with a final concentration of 1 µg/ml for 2–8 weeks. Fig. 1A demonstrates that the EC cells and fibroblasts expressed RFP and GFP, respectively, following puromycin selection. Approximately 80% of ECC-1 cells and fibroblasts expressed RFP and GFP, respectively.

In the monoculture, the ECC-1 tumor cells migrated in an amoeboid mode without forming cell projections, with an average velocity of 2.68±0.72 µm/h (Fig. 1B and O; Table III; Data S1). The CAFs formed long projections, and ECC-1 cells could move on top of these projections. The ability of CAFs to form long cell projections was not influenced by the tumor cells (Fig. 1C, E and G), with an average velocity of 9.39±1.30 µm/h (Fig. S3A; Table III; Data S2, S3 and S4). However, in the presence of CAFs, the velocity of ECC-1 increased 3-fold (7.86±0.6 µm/h, P<0.0001) (Fig. 1D, F, H and O; Table III; Data S5, S6 and S7). The pro-migratory effects were not observed with BAFs. The tumor cells could not move on top of the BAFs, likely due to the lack of cell projections (Fig. 1I, K and M). There were no significant changes observed in tumor cell velocity when co-cultured with BAFs (4.06±1.01 µm/h, P=0.28) (Fig. 1J, L, N and O; Table III; Data S8, S9 and S10). To examine whether the tumor cells affect fibroblast motility, the velocity of CAFs and BAFs was examined in the presence of ECC-1 cells. The motility of the CAFs was not affected by the tumor cells. Indeed, there was no significant change in CAF velocity when co-cultured with ECC-1 (7.14±0.97, P=0.71) compared with monocultures (9.39±1.30 µm/h) (Fig. S3A; Table III; Data S2, S3 and S4). Similarly, the velocity of the BAFs did not change significantly when co-cultured with tumor cells (2.95±0.61 µm/h, P=0.47) compared with monocultures (3.56±0.51 µm/h), except that the EH6Fib line showed an increase in velocity from 2.26±0.52 in monoculture to 4.53±0.85 µm/h (P=0.0349) in the presence of tumor cells (Fig. S3B; Table III; Data S11, S12 and S13). Altogether, these findings indicated that the ability of CAFs to mobilize EC cells through cell projections was likely a CAF-specific property and not the result of reciprocal interactions with the tumor cells.

To examine the effects of enhanced EC cell motility in the presence of CAFs, the 3D tumor-fibroblast co-culture duration was prolonged. Tumor spheroid formation was observed 24 h post co-culture, and the cells within the spheroids began to dissociate after 14 days of co-culture. Thus, the diameter and number of tumor spheroids were examined between day 1 and 14 in the presence or absence of CAFs.

In ECC-1 monocultures, tumor spheroids were observed on day 1 with a mean diameter of 66.67±4.17 µm, which gradually increased from day 7 to 14 (191.7±18.16 µm) (Fig. 2A and D). The number of tumor spheroids in this monoculture decreased from 49.00±3.21 to 26.00±4.50 from day 1 to 14 (Fig. 2F). According to the 3D image reconstruction, from day 7 to 14, the ECC-1 spheroids formed were compact (Fig. 2A). This may explain the consistent RFP signal observed from ECC-1 spheroid throughout this time period. However, in CAF (EC48Fib) monocultures, the cells grouped and formed clusters. The GFP signal from CAFs gradually decreased, and the cells were loosely packed as seen from the 3D reconstruction analysis (Fig. 2B).

Interestingly, when co-cultured with EC48Fib, some ECC-1 cells were observed along the protrusions and projections of EC48Fib. ECC-1 and EC48Fib formed at least 10-fold larger spheroids on day 7, with a mean diameter 708.3±110.2 µm, compared with that of ECC-1 monocultures. From the 3D reconstruction analysis, the cells in the co-culture formed irregular spheroids (Fig. 2C). The diameter of the spheroids remained stable, as no apparent change in mean diameter was observed from day 7 to 14 (652±97.25 µm; Fig. 2A), although GFP signals from the EC48Fib within the spheroids was reduced substantially. For ECC-1 and EC48Fib, the mean diameter of the spheroids from the co-cultures was significantly higher than that of ECC-1 monoculture on day 1 (708.33±110.2 µm) and day 14 (652±97.25 µm), respectively. Interestingly, ECC-1 and EC49Fib or EC50Fib co-culture only formed spheroids from day 7 onwards, with a mean diameter of 177.6±38.68 and 500±212 µm, respectively (Fig. 2D and E; Fig. S4A). On day 1, ECC-1 and CAF co-culture did not show an increase the number of spheroids compared with the ECC-1 monoculture (49.00±3.21) (Fig. 2F and G). On day 14, the number of spheroids in ECC-1 and CAF co-cultures slightly increased (4.0±0.58), although there were no significant changes compared with the ECC-1 monoculture (26.00±4.50) (Fig. 2F and G). Similar observations were obtained when the experiment was repeated with other CAF lines (EC49Fib and EC50Fib), as the CAFs promoted EC spheroid formation over 14 days (Fig. S4A). These observations indicated that CAFs may facilitate EC spheroid formation because of their pro-migratory effects on EC cells.

To further determine whether the observed effects on spheroid formation were specific to CAFs, the aforementioned experiments were carried out using BAFs. Similar with the previous section, in ECC-1 monoculture, tumor spheroids were observed on day 1, and gradually increased from day 7 to 14 (Fig. 3A). In BAF monocultures, the EF2Fib protruded from the cellular membrane and formed clusters at day 1 (141.7±8.33 µm), with no significant change in the mean diameter from day 1 to 14 (125.0±0.00 µm) (Fig. 3B). Interestingly, in ECC-1 and EF2Fib co-cultures, some ECC-1 cells were detected along the protrusion and projection of the fibroblasts (Fig. 3C). The spheroid diameters increased only slightly from day 1 to 14 (152.9±27.92 µm). However, there was no significant change compared with the ECC-1 monoculture at day 14 (Fig. 3D and E). The number of spheroids did not significantly change between day 1 and 14. Interestingly, BAFs suppressed the number of spheroids compared with the ECC-1 monoculture (Fig. 3F and G). Similar observations were obtained when the experiments were repeated with the NE14Fib line (Fig. S5A and B).

Co-culture of ECC-1 with another BAF line (EH6Fib) resulted in spheroids with an average diameter of 95.83±15.02 µm on day 1 (Fig. S5D). There was a slight increase in diameter at day 14 (166.7±20.83 µm) (Fig. 3D and E) but no significant change in the number of spheroids compared with day 1 (Fig. 3F and G). However, unlike EF2Fib and NE14Fib, EH6Fib did not form cell protrusions and projections (Fig. S5C). Altogether, our data suggested that BAFs were not able to promote EC spheroid formation.

To delineate the mechanisms underlying CAF-mediated EC tumor cell motility and spheroid formation, the expression of RhoA was examined. The primary function of this protein is to govern actin cytoskeleton reorganization (11). The involvement of these proteins was examined on day 1 when the tumor cells were actively migrating towards each other, and for spheroid formation on day 14 when the tumor cells had formed spheroids.

A previous study showed that the presence of the basement membrane in Matrigel^® culture may affect the expression of RhoA (19). Compared with the non-Matrigel^® cultures (2D cultures), the ECC-1 3D monoculture showed a 1.5-fold upregulation of RhoA mRNA expression on day 1, and that increased to 3-fold on day 14 (Fig. 4A). Moreover, when compared with the non-Matrigel cultures (2D cultures), in the CAF monocultures, RhoA expression was upregulated 4-fold on both day 1 and 14. In the ECC-1 and CAF co-cultures, a 5-fold upregulation of RhoA expression was observed on day 14 compared with the ECC-1 monoculture on day 1 (P<0.05; Fig. 4A).

The expression of ROCK1, a downstream effector of RhoA (11), was then evaluated. ROCK1 was not upregulated in ECC-1 and CAF monocultures on day 1 and 14 (Fig. 4B). Surprisingly, in the presence of CAFs, the expression of ROCK1 increased 2.5-fold on day 1, and this was maintained on day 14 (Fig. 4B). Immunofluorescence staining was then performed to evaluate the expression of p-MLC, a downstream effector of ROCK1 in RhoA/ROCK1 signaling (14,15). The p-MLC protein was highly expressed in ECC-1 and CAF co-cultures at day 14 compared with day 1 (Fig. 4C). However, the expression of p-MLC observed in ECC-1 monoculture and CAF monocultures was low in day 1 and 14 when compared to that in ECC-1 and CAF co-culture, respectively. Collectively, these data suggested that RhoA/ROCK1 signaling in fibroblasts was likely involved in EC motility and spheroid formation in the 3D culture system.

Y-27632, a selective ROCK1 inhibitor that targets ATP-competitive p160^ROCK and ROCK1, was used to determine whether RhoA/ROCK1 signaling was required for the CAF-mediated increase in EC motility. Cell viability assays and phalloidin-TRITC staining were carried out in ECC-1, CAFs and BAFs in order to determine the optimal concentration of Y-27632 to be used. When Y-27632 was used at concentrations ranging from 0.2 to 50 µM, the viability of ECC-1 cells did not significantly change compared with vehicle-treated cells. However, at 100 µM, ECC-1 cell viability was reduced to 67.20±4.75% (Fig. S6A). CAFs and BAFs treated with concentrations ranging from 0.2 to 100 µM did not show significant changes in cell viability when compared with vehicle-treated cells (Fig. S6B and C). Moreover, ECC-1 cells showed no changes in morphology or in actin structures from concentration of 50 to 100 µM (Fig. S6D). Interestingly, fine and elongated cytoplasmic processes from the cell body were observed in CAFs, possibly due to the inhibition of stress fiber formation following Y-27632 treatment. Such an inhibitory effect was shown in a concentration-dependent manner and stress fiber formation was completely inhibited following treatment with 100 µM of Y-27632 (Fig. S6E-G). Similar observations were made in BAFs (Fig. S6H-J).

ECC-1 cells did not show significant changes in mean velocity following 16-h treatment with 100 µM Y-27632 (0.82±0.07 µm/h) compared with vehicle-treated cells (1.24±0.07 µm/h, P=0.0866) (Fig. 5A). However, this treatment significantly inhibited the mean velocity of the EC48Fib, EC49Fib and EC50Fib CAF lines, resulting in 63.52, 53.33 and 50.42% inhibition, respectively (Fig. 5B). In some of the ECC-1 and CAF co-cultures, Y-27632 significantly inhibited tumor cell velocity with an average 86.36 and 78.3% reduction, in ECC-1 and EC49Fib and ECC-1 and EC50Fib co-cultures, respectively (Fig. 5C). The inhibition was predominantly mediated by CAFs, as motility of ECC-1 cells alone were not affected.

Inhibition of ROCK1 did not alter BAF velocity in monoculture compared with the vehicle-treated cells, except in EH6Fib, resulting in a 33.35% increase (Fig. 5D). There was no significant change observed in ECC-1 velocity in the co-culture with BAFs (Fig. 5E). Taken together, these results suggested that activation of ROCK1 signaling was required for CAF motility, which subsequently also affected EC cell motility in the co-culture model.

The next experiments further assessed whether RhoA/ROCK1 signaling was required for tumor spheroid formation. Compared with the control (10% FBS culture medium), in CAFs treated with 100 µM Y-27632 for 24 h, fine and elongated cytoplasmic processes from the cell body were observed, likely due to inhibition of stress fiber formation (Fig. 6A-C; Data S14, S15 and S16). Although there was no change in ECC-1 monoculture morphology following Y-27632 treatment (Fig. 6D; Data S17), the formation of tumor spheroids in the ECC-1 and CAF co-culture model was markedly inhibited (Fig. 6E-G; Data S18, S19 and S20). The formation of fine cytoplasmic processes from CAFs restricted ECC-1 movement across CAF cell projections, thus inhibiting spheroid formation. Although BAFs also formed cytoplasmic processes upon treatment with Y-27632 (Fig. 6H-J; Data S21, S22 and S23), no spheroid formation was observed in ECC-1 and BAF co-cultures (Fig. 6K-M; Data S24, S25 and S26). Altogether, these data suggested that RhoA/ROCK1 signaling was required for CAF motility and formation of tumor spheroids in 3D cultures.