Cell competition between anaplastic thyroid cancer and normal thyroid follicular cells exerts reciprocal stress response defining tumor suppressive effects of normal epithelial tissue

The microenvironment of an early-stage tumor, in which a small number of cancer cells is surrounded by a normal counterpart milieu, plays a crucial role in determining the fate of initiated cells. Here, we examined cell competition between anaplastic thyroid cancer cells and normal thyroid follicular cells using co-culture method. Cancer cells were grown until they formed small clusters, to which normal cells were added to create high-density co-culture condition. We found that co-culture with normal cells significantly suppressed the growth of cancer cell clusters through the activation of Akt-Skp2 pathway. In turn, cancer cells triggered apoptosis in the neighboring normal cells through local activation of ERK1/2. A bi-directional cell competition provides a suppressive mechanism of anaplastic thyroid cancer progression. Since the competitive effect was negated by terminal growth arrest caused by radiation exposure to normal cells, modulation of reciprocal stress response in vivo could be an intrinsic mechanism associated with tumor initiation, propagation, and metastasis.


Introduction
At earlier stages of carcinogenesis, malignant transformation starts from a single cell that grows within an epithelial monolayer. The initiated cell continues to proliferate and accumulates genetic alterations, which give rise to malignant cancer cells. However, tissue microenvironment, in which multiple types of cells coexist, affects malignant propagation of initiated cells [1][2][3][4][5][6][7]. In fact, in a tissue microenvironment, the initiated cells are likely to interact with normal counterparts in a dynamic fashion over time. Such interactions may lead to a balance  [57]. Cells were cultured in DMEM (Wako, Tokyo) supplemented with 10% fetal bovine serum (FBS) (TRACE Bio) under standard conditions in a humidified incubator at 37˚with 5% CO 2 . NTECs and ACT1 cells were checked their human origin by chromosome analysis, and mycoplasma contamination was routinely inspected by DAPI staining.

Irradiation
NTECs cultured in a culture flask were exposed to 10 Gy of γ-rays from a γ-ray irradiator equipped with a 137 Cs source (Pony Industry Co., Ltd, Osaka) at a dose rate of 1 Gy/min.

Cell labelling
ACT1 was labelled with the Qtracker 655 cell labeling reagent (ThermoFisher Scientific, Tokyo, Japan). Cells were cultured onto sterilized 22 x 22 mm glass slips (Matsunami, Tokyo, Japan) placed in the 35-mm culture dishes. For labelling, a 10 nM solution was prepared according to the manufacturer's protocol and added to the dishes with ACT1 cells for 1 hr at 37˚followed by two washes with fresh growth medium.
NTECs were transfected with the pHOS-H2B-GFP plasmid (BD Bioscience, Tokyo) from which the green fluorescent protein (GFP)-tagged histone H2B is produced. Introduction of the plasmid into NTEC was performed by electroporation (Electric Cell Fuser, ECF2001, Wakenyaku, Tokyo). Exponentially growing cells were collected by trypsinization, suspended in PBS, and 200 μl of cell suspension was added to a electroporation cuvette (0.2 cm) with plasmid DNA (1 μg). Three pulses with an intensity of 400V/cm with a constant pulse duration of 1 msec were used. Immediate after the pulse, cell suspension was transfer to 100-mm dishes with 10 ml of DMEM medium and cultured for three days. Cells were then collected by trypsinization, replated onto three 100-mm dishes with 10 ml of DMEM medium containing 400 μg/ml G418 (WAKO Pure Chemicals, Osaka), and cultured for further 7 day before G418-resistant colonies were formed. Colonies were randomly isolated, and independent clones were cultured in T25 flasks. A part of each clone was replated onto glass-bottomed dishes (AGC Techno Glass Co., Ltd, Tokyo), and GFP expression was checked under a fluorescent microscope (BZ-9000, KEYENCE, Osaka) to select GFP-H2B-positive clones. The clone, which expressed the strongest GFP fluorescence, was named GFP-NTECs.

Co-culture of ACT1 and NTEC cells
Exponentially growing ACT1 cells were plated on glass coverslips at low cell density (400 cells/ slip) and incubated for about 5 days until they formed small cell clusters. Then, NTECs were added to the culture (10 5 cells/slip), and incubated for further 3-5 days.

Time-lapse microscopy
Time-lapse imaging was performed by BioStation-ID (GE Helthcare Bioscience, Tokyo, Japan) with a x10 objective lens. Images were captured in every 5 min for up to 72 h. In each experiment, at least ten fields were imaged in GFP fluorescence channel along with phase contrast.

Live-cell imaging of apoptosis
For the detection of apoptosis in living cells, CellEvent Caspase-3/7 Green Detection Reagent (ThermoFisher Scientific, Tokyo, Japan) was used. The reagent (2 μl) was diluted in growth medium and added to the dishes containing ATC1 clusters and NTECs. The cultures were incubated for 30 minutes according to the manufacturer's instruction, and apoptotic cells with activated caspase-3/7, which showed bright green fluorescence, were analyzed by time-lapse microscopy using Biostation-ID.

Image analysis
Images were taken under a fluorescent microscope (BZ-9000) and the fluorescence area corresponding to a cell cluster size was determined using Image J software [58]. Briefly, the selection tool was used to set the area of interest, and the image was converted to Grayscale. The pixels were highlighted using the threshold dialog, and the total number of pixels was defined by selecting the menu command "measurements". For analyzing phosphorylation of ERK1/2, the distance from ACT1 cell cluster was marked with the straight selection tool. Next, the colour images were splitted using the menu command "Image-Type-Colour-Split channels", and using the blue channel, which is for the DAPI staining, the total number of NTEC cells per distance was calculated. The number of NTEC cells with phosphorylated ERK1/2 was quantified with the merged image between the blue and red channels using the menu command "Image-Colour-Merge channels", and cells positive for both channels were counted.

Western blotting
Cells were treated with lysed in RIPA buffer (50 mM Tris-HCl (pH 7.2), 150 mM NaCl, 1% NP-40, 1% sodium deoxycholate, and 0.1% SDS). The cell lysate was cleared by centrifugation at 15,000 rpm for 10 min at 4˚C, and the supernatant was used as the total cellular protein.

Statistics
For each experiment, at least three independent repeats were carried out. The non-parametric Mann-Whitney test was used for colony size comparison. Weighted linear regression was used to analyze region-specific phosphorylation of ERK1/2. Statistical calculations were performed using JMP 15 Pro software. The p-value < 0.05 were considered indicative of statistical significance.
The time-lapse analysis demonstrated that NTECs added to the culture are gradually accumulated around ACT1 clusters even at very low cell density, and crowding around ACT1 clusters becomes obvious approximately 6-8 hrs later (S1 Fig). After 3-5 days incubation, NTECs fully occupied the space between the ACT1 clusters (S2 Fig), and the growth of ACT1 clusters seemed to be suppressed, as we observed little change in the sizes of the clusters among NTECs compared with ATC1 clusters in monocluture. To confirm the suppressive effects GFP-tagged NTECs, which enable to discriminate normal cells in co-culture, were used (Fig 2). ACT1 cluster sizes in monocultures (Fig 2A) are obviously larger than those in co-cultures (Fig 2B), and the difference is statistically significant (Fig 2C), demonstrating that cell competition between neighbouring NTECs and ACT1 clusters retarded ACT1 cluster growth. According to the results of immunofluorescence using anti-Ki-67 antibody, ACT1 clusters lose proliferating potential. Particularly, ACT1 cells in the center of clusters became negative for Ki-67, while all of ACT1 cells are positive for Ki-67 in monoculture (S3 Fig). This confirms that cell competition between NTECs and ACT1 clusters retarded cell cycle progression of ACT1 cells.
In contrast to the growth arrest of ACT1 clusters, the time-lapse analysis also exhibited regional death of NTECs in the areas close to ACT1 clusters (Fig 3 and S1 Movie). During the first approximately 48 hours after inoculating NTECs, growth of NTECs was more than that of ACT1, so that dividing cells were observed more in NTECs, and then, morphologically distinct cells, which were small round but not the mitotic cells, were emerging in proximity to ACT1 clusters (Fig 3B and 3C).
Detailed time-lapse analysis revealed that cell blebbing, which is one of the defined features of apoptosis, preceded the cells being round-shaped (S4 Fig). It was found that small roundshaped cells were subsequently ruptured. The time-lapse analysis also confirmed that they were derived from Qtracker-negative cells (S4 Fig). Small round-shaped cells were also GFPtagged and positive for vimentin, which is a marker for NTECs (S5 Fig), confirming that cell death was induced in NTECs. Temporal analysis of time-lapse images shows that dead cells were appearing continuously over 60 hours (Fig 3D), while none of them was detectable during the first approximately 12 hours. Judging from their morphology, NTECs was likely dead by apoptosis, which was confirmed by live-imaging using cell-permeable fluorescent substrates for Caspase-3 and 7. As shown in Fig 4, NTECs in the areas close to ACT1 clusters are marked by green fluorescence (Fig 4B and 4C), and those cells become round-shaped cells afterwards (Fig 4C and 4D), indicating that Caspase-3/7-mediated apoptosis is involved in NTEC death To uncover the mechanisms underlying position-specific cell death, we performed an immunofluorescent analysis using antibodies against phosphorylated forms of MAP kinases. Among ERK1/2, p38 and JNK1/2, only ERK1/2 was found to be phosphorylated in NTECs. whereas augmented phosphorylation is common in ACT1 clusters (Fig 5).
Importantly, NTECs with phosphorylated ERK1/2 are localized to the areas close to ACT clusters, and notably, round-shaped dead NTECs are phosphorylated ERK1/2-positive (S7    Fig 5, phospho-ERK1/2 positivity was observed more in NTECs closer to the clusters, so that the frequency of phospho-ERK1/2-positive NTECs is determined in the areas every 100 pixels from the ACT1 cluster border (Figs 6 and S8). Analysis of distance-dependent ERK1/2 phosphorylation revealed the highest frequency is observed in NTECs closest to the ACT1 clusters (Fig 6E), however, even NTECs within the area more than 500 pixels apart from the ACT1 cluster border, which are equivalent to more than 300 μm apart from the clusters, still show phosphorylation of ERK1/2.

Molecular pathways involved in cell competition
Since the growth of ACT1 clusters was significantly retarded in co-cultures with NTECs, we further analysed molecular changes in ACT1 cells as well as NTECs using discriminatory cell collection technique (S9 Fig). As the substrate attachment of ACT1 cells is much stronger than that of NTECs, we briefly trypsinized co-cultures without PBS wash and collected NTECs. Then, cells were trypsinized again to collect cancer cells remaining in the dish. Purity of the NTEC and ACT1 cell population was checked by using the GFP positivity determined under the fluorescence microscope, and more than 99.9% of the first collected cells and less than 0.5% of the second collected cells were positive for GFP, respectively.
Western blot analysis demonstrated that multiple phosphorylations of RB were significantly reduced in ACT1 cells that competed with NTECs, while total RB protein level was not changed (Fig 7A). As RB phosphorylations are targeted by several Cyclin/Cdk kinases, we examined the levels of their inhibitors. Among p21 WAF1/Cip1 , p16 INK4A , and p27 Kip1 , the p27 Kip1 protein, which are expressed in ACT1 cells at lower level as compared with NTECs, was profoundly upregulated in competed ACT1 cells (Fig 7B).
We also observed the decreased expression of Cdk2 and Cyclin D in competed ACT1 cells (Fig 8A), indicating that upregulation of p27 Kip1 and downregulation of Cyclin D/ Cdk2 could be involved in reduced RB phosphorylation. Competed ACT1 cells also displayed lower level of phosphorylated Skp2 at serine 64 and increased level of phosphorylated forms of Akt (Fig 8B).
In agreement with the immunofluorescence results, phosphorylation of ERK1/2 was higher in ACT1 cells compared with NTECs, however, the level is not changed by cell competition (Fig 9A). As we observed an increased ERK1/2 phosphorylation in NTECs which were close to ACT1 clusters, we examined phosphorylation of ERK1/2 in NTECs from bulk co-culture. Despite regional phosphorylation was detected on immunofluorescence analysis, we did not detect changes in ERK1/2 phosphorylation under such conditions (Fig 9A). We also observed increased phosphorylation of JNK1/2 in both types of competed cells (Fig 9B).

Effect of terminal growth arrest of NTECs on ACT1 cell growth
To determine whether the proliferation of NTECs is required for cell competition, NTECs were exposed to 10 Gy of γ-rays, which induces senescence-like terminal growth arrest as judged by the expression of senescence associated β-galactosidase activity (S10 Fig). NTECs were exposed to γ-rays and incubated for 3 days before adding to ACT1 clusters. The quantity equivalent to that inducing competition (10 6 cells) were added to ACT1 cluster, and ACT1 cluster sizes were measured 3 days later. It is apparent that the cluster sizes are appeared to be significantly greater than those observed in co-cultures with unexposed NTECs (Fig 10). Immunofluoresent analysis reveals that Ki-67 positivity, which was significantly reduced in competed ACT1 clusters, is remarkably recovered under this condition (S11 Fig).

Discussion
This study for the first time outlined the crucial role and possible mechanisms of the cell competition between anaplastic thyroid cancer cells and normal thyroid follicular epithelial cells. Comparison of ACT1 cluster size clearly indicated that the growth of ACT1 cells was significantly suppressed by NTECs (Fig 2). In fact, Ki-67 positivity was significantly reduced in ACT1 clusters, confirming that cell competition was able to suppress cancer cell growth (S3 Fig). Western blot analysis reveals that RB phosphorylation was significantly repressed accompanied by the up-regulation of p27 Kip1 (Figs 7 and 8). Since p27 Kip1 level is regulated through ubiquitin-dependent proteasome activity, we examined phosphorylation of Skp2, an E3 ligase catalyzing p27 Kip1 ubiquitination [59], and found that it was reduced under competed condition (Fig 8B). We also observed up-regulation of Akt phosphorylation which facilitates Akt- dependent phosphorylation of Cdk2 and sequestering of the latter in the cytoplasm [60,61]. These changes together with down-regulation of Cdk2 and Cyclin D could be involved in the retarded growth of competed ACT1 cells (Fig 11).
Besides ACT1 cell growth suppression, cell competition also affected the fate of NTECs manifested by the elimination of cells with contacting ACT1 clusters or located nearby to those. NTECs elimination was taking place through apoptosis accompanied by the region-specific activation of ERK1/2 (Fig 6), although the overall phosphorylation levels of ERK1/2 showed no change in bulk cultures (Fig 9). While activation of ERK1/2 is generally involved in cell proliferation, ERK activity has been also involved the induction of apoptosis [62]. Considering that no apoptosis in NTECs during the first days before the time-lapse analysis was started, this indicates that physical cell contact is required to induce NTECs elimination. Previously, it was demonstrated that mechanical stress through cell-to-cell contact was involved in cell competition [63,64]. In fact, NTECs neighboring ACT1 cluster show increased cell anisotropy (S12D Fig), indicating that an increasing the number of NTECs competed with ACT1 clusters potentiates compression of NTECs at the border. This could cause apoptotic cell death in NTECs.

PLOS ONE
Cell competition between thyroid cancer and normal follicular cells So far, the molecular nature of cell competition has been discussed extensively in Drosophila [21][22][23][24][25], and several driving factors have been identified including c-myc and p53 [26][27][28][29]. In fact, we confirmed augmented expression of c-myc in ACT1 cells (S13 Fig), however, it did not seem to make ACTs cells supercompetitor, since the growth of ACT1 is much slower than NTECs (S2 Fig). Also, differential expression of Scrib does not appear to be involved as there was no difference between NTECs and ACT1 (S13 Fig). The p53 gene is mutated in ACT1 cells, so that the p53 protein level is significantly stabilized (S13 Fig). Together with the mutation in the N-ras gene p53 deregulation could confer the 'winner' phenotype [43], but ACT1 growth was indeed suppressed by the cell competition with NTECs, suggesting that driver mutations may play a different role in different cell context.
Recently, the suppressive effect of cell competition on liver cancer, which was mediated by the YAP induction in peritumoral hepatocytes, was reported [65]. The YAP as well as TAZ transcriptional coactivators are the downstream effectors of the Hippo signaling pathway, which plays critical roles in cell-to-cell contact, cell polarity, and fitness to the neighbors [66][67][68]. While several previous literatures have presented that the YAP1 is overexpressed in tumors including thyroid cancers [69,70], the study clearly demonstrated that the YAP limitedly induced in normal hepatocytes neighboring tumor cells played suppressive role. In fact, the active YAP1 level was augmented in ACT1 cells, which was significantly down-regulated in co-culture (S14 Fig). Since high-density culture did not alter the YAP expression in ACT1 cells, cell competition with NTECs might affect the Hippo signaling pathway. Similar observation was reported by other study, confirming that bidirectional cell competition was indeed involved in cancer growth in vivo [71]. Thus, the down-regulation of YAP, which might be controlled by the Hippo signaling pathway activated through cell competition with NTECs, should be additional pathway that suppress ACT1 cell growth. The exact mechanism of the Hippo signal activation, and an involvement of Akt activation in this pathway need further investigations to confirm. Previously, several reports demonstrated that cell growth rate and fitness per se may not drive cell competition as these are not an absolute common quality and does not determine the outcome of competition [27,36,[72][73][74][75]. Our findings indicated that ATC1 cells have sensitivity to compaction and crowding, and the AKT activation could be a critical event to initiate the downstream cascade lead to RB dephosphorylation (Fig 11). The exact pathway of AKT activation still needs to be determined.
According to the results obtained by live-cell imaging (Fig 3), cell competition was initiated when the numbers of NTECs reached sufficient for cell compression. In order to confirm that continuous cell growth of NTECs is essential for cell competition, NTECs with sufficient cell number but terminated cell growth were examined (Fig 10). As we expected, reduction of ACT1 cluster sizes was significantly negated. We also found no apoptotic cell death in NTECs (S15 Fig). Furthermore, it is demonstrated that overall ERK1/2 phosphorylation is up-regulated upon irradiation (S16 Fig). While locally activated ERK1/2 seemed to be essential for apoptotic cell death of NTECs, the result suggested that not homogeneous activation but regional and accidental activation of ERK1/2 could be involved in region-specific apoptosis induction. Previously, it was reported that downregulation of ERK was involved in Drosophila [76], our results represented that the effect could be cell context-dependent.
It should be emphasized that even if the NTECs close to the ACT1 clusters were forced to die by apoptosis, the presence of NTECs surrounding cancer cells established suppressive cell competition to anaplastic cancer cells. Considering that thyroid follicles are the spheroidal structures formed by a monolayer of follicular cells, it is possible that similar cell competition Cell competition between anaplastic thyroid cancer cells and normal thyroid follicular epithelial cells provoked reciprocal stress response. In competed ACT1 cell clusters, activation of Akt resulted in dephosphorylation of RB, which gave rise to reduced growth. In contrast, competed NTECs, especially those neighboring ACT1 clusters, were forced to die by apoptosis via unscheduled activation of ERK1/2. Our results demonstrate that cell competition is a bi-directional phenomenon, thourgh which the initiation, propagation, and metastasis of tumors are hindered.
https://doi.org/10.1371/journal.pone.0249059.g011 could be taking place in vivo. Recently, accumulating evidences have demonstrated that cell competition is an essential component of tissue microenvironment, which is important not only during tissue development and aging but also in preventing cancer development and invasion [1][2][3][4][5][6][7][8]. Cellular heterogeneity of a tissue plays a key role since cancer cells may have different cell stiffness and different proliferation rates than normal cells. Although many of oncogenic mutations have been though to confer supercompetitor phenotype to cancer cells, normal counterparts exert intrinsic tumor-suppressive effects through cell competition activity, which was also reported by others and termed epithelial defence against cancer (EDAC) [77]. While anaplastic cancer is an aggressive form of thyroid cancer, our results indicated that cell competition therapy could be an option. Of importance, since our study showed that termination of growth of NTECs by radiation exposure abrogates suppressive competition properties of NTECs (Fig 10), it should be important to reduce toxicity of radiotherapy and chemotherapy to normal epithelial cells in order to preserve maximum EDAC.
In conclusion, we developed an in vitro cell competition system and demonstrated that the growth of anaplastic thyroid cancer cells was significantly retarded by normal thyroid follicular epithelial cells. Cell competition evoked stress response in cancer cells, which resulted in down-regulation of RB phosphorylation. Reciprocally, it induced stress response in normal cells, which gave rise to position-dependent induction of apoptosis. These results prove that cell competition is obviously a bidirectional phenomenon, in which competed cells are both affected each other. Since ERK1/2 and Akt, well-known pathways involved not only in anaplastic thyroid cancer but also in many other types of tumors, are critical components of cell competition (Fig 11), further studies on identifying target molecules that govern the struggle between cancer and normal cells could provide opportunities for conditioning the situation in favor of normal cells. Finally, it should be noted that our current experimental co-culture system has a potential limitation, as NTECs organize three-dimensional structure in vivo, and cell competition is not on a plastic surface but on the stroma with stromal cells, so that the future studies need to apply organoid culture, in which thyroid cancer cells and normal follicular cells are mixed together. Another limitation is that we only used one anaplastic thyroid cancer cell line, therefore, additional studies using different cancer cell clines should determine whether our observation is common phenomenon or not. ACT1 cells were cultured for 5 days until they formed small clusters before adding GFP-NTECs. Then, ACT1 clusters and GFP-NTECs were co-cultured for further 3 days, fixed with formaldehyde and stained with anti-vimentin antibody (the secondary antibody is labelled with Alexa647, so that the green and red pseudo color were applied) and anti-phosphorylated ERK1/2 antibody (red fluorescence). (A) GFP-NTECs and ACT1 cluster showing ERK1/2 phosphorylation (red). While strong red fluorescence is observed in ACT1 cluster, some groups of NTECs show ERK1/2 phosphorylation. In addition, small round-shaped NTECs are positive for phosphorylated ERK1/2 antibody as indicated by white arrow heads. NTECs. NTECs were exposed to 10 Gy of γ-rays before co-cultured with ACT1 cell clusters. Phase-contrast images of NETCs before (A) and 5 days after 10 Gy of γ-irradiation (B). NTECs were stained according to the protocol described in Materials and Methods. (TIF) S11 Fig. Growth of co-cultureed ACT1 cell clusters with γ-irradiated NTECs. ACT1 cells were cultured for 5 days until they formed small clusters before adding GFP-NTECs. Then, ACT1 clusters and γ-irradiated GFP-NTECs were co-cultured for further 3 days, fixed with formaldehyde and stained with anti-Ki67 antibody (the secondary antibody is labelled with Alexa647, so that the green pseudo color was applied) and anti-53BP1 antibody (red fluorescence). NTECs and ACT1 cells cultured either alone (NTEC or ACT1 only, respectively) or from cocultures (NTEC or ACT1 competed, respectively) and subjected to western blot analysis with indicated antibodies. (TIF) S14 Fig. Down-regulation of the YAP expression through cell competition. ACT1 cells were cultured for 5 days until they formed small clusters before adding NTECs. Then, ACT1 clusters and NTECs were co-cultured for further 3 days, fixed with formaldehyde and stained with anti-vimentin antibody (the secondary antibody is labelled with Alexa647, so that the green pseudo color was applied) and anti-active YAP1 antibody (red fluorescence). Exponentially growing NTECs (A), ACT1 cluster (B), confluent NTECs (C), and confluent ACT1 cells (D) show vimentin (green) and active YAP1 (red) expression. ACT1 cluster co-cultured with NTECs show decreased active YAP1 expression. The bar in (A) indicates 40 μm. (TIF) S15 Fig. No apoptotic cell death and activation of ERK1/2 in 10 Gy-irradiated NTECs cocultured with ACT1 cell clusters. ACT1 cells were cultured for 5 days until they formed small clusters before adding 10 Gy-irradiated GFP-NTECs. Then, ACT1 clusters and GFP-NTECs were co-cultured for further 3 days, fixed with formaldehyde and stained with anti-vimentin antibody (the secondary antibody is labelled with Alexa647, so that the green and red pseudo color were applied) and anti-phosphorylated ERK1/2 antibody (red fluorescence).