To knock-down EMMPRIN expression, we used the procedure described before (36). Briefly, the lentivirus vector with four mouse EMMPRIN siRNA sequences (CT26-KD) or the empty vector (CT26-NC, both from Applied Biological Materials, Richmond, BC, Canada) were used to infect the parental CT26 (CT26-WT) cell line at a multiplicity of infection (MOI) 135 with polybrene (8 μg/ml, Merck, Rathway, NJ, USA). The four siRNA sequences are given in Supplementary Table S1 . After infection, cells were grown for a total of 3 weeks in full medium with puromycin (10 μg/ml) to select for positive cells, and the medium was replaced every 3-4 days. Because the vector did not contain the GFP marker that is known to be immunogenic (37), we used the limiting dilution approach to isolate sub-clones that were similar in respect to EMMPRIN expression. These selected sub-clones were characterized for their EMMPRIN expression at the mRNA levels, membranal and secreted protein expression ( Figure 1 ). The 4T1-KD and LLC-KD cells were generated in the same manner.

The murine colon carcinoma cell line CT26 (ATCC CRL-2638) was cultured in RPMI-1640 medium with 10% fetal calf serum (FCS), 1% penicillin/streptomycin, 1% non-essential amino acids, 1% amphotericin B, 1% pyruvate and 1% L-Glutamine (full medium, all reagents from Biological industries, Beit Ha’emek, Israel). The mouse brain endothelial cell line bEND3 (ATCC CRL-2299) was cultured in high glucose DMEM, with 10% FCS, 1% penicillin/streptomycin, and 1% glutamine. The murine mammary carcinoma 4T1 (ATCC CRL-2539) and the Lewis lung carcinoma (LLC, ATCC CRL-1642) cell lines were grown in Dulbecco’s modified Eagle’s medium (DMEM) with 10% FCS, 1% penicillin/streptomycin, 1% amphotericin B, and 1% L-Glutamine. All cells were split every 3–4 days at a ratio of 1:4 using trypsin-EDTA, and were used at passages 3–15. All cells were routinely checked for the presence of mycoplasma.

To form a monolayer, CT26-WT, CT26-KD or CT26-NC cells (8x10⁴ cells/well) were incubated overnight in 300µl full medium in a 24 well plate to allow their adherence. Then the medium was replaced with 300µl serum starvation medium to avoid the masking of signals, and the cells were incubated for 48 hours. To form 3D monoculture spheroids, we used the scaffold-free liquid overlay method, where CT26-WT, CT26-KD or CT26-NC cells (5x10³ cells/100 μl) were seeded in full medium in a U-shaped 96-well plate that was previously coated with 1% agarose (50 μl/well in PBS). Cells were incubated for 72 hours in full medium, to allow them to mature and form tight spheroids.

To assess viability, CT26-WT, CT26-KD and CT26-NC cells were seeded as monolayers in 96-well plates (10⁴ cells/well/100μl of serum starvation medium) for 48 hrs. Viability of the cells was assessed by adding 10µL of the cell counting kit 8 reagent (CCK-8, Abcam, Cambridge, United Kingdom) and incubating for additional 2 hrs. Alternatively, cells were seeded as spheroids (5x10³ cells/100μl full medium) and incubated for 72 hrs. Then the spheroids in 100μl medium were moved to another well and 10μl CCK-8 were added. Cell viability was determined according to the absorbance measured at 450 nm with 620 nm reference, resulting from the reduction of the tetrazolium salt WST-8 in the mitochondria of viable cells, and normalized to the CT26-WT cells.

Additionally, to measure cell proliferation we used the BrdU Cell Proliferation Assay kit (Biovision/Abcam, Waltham, MA) according to the manufacturer’s instructions. This kit is based on the detection of the DNA-incorporated pyrimidine analogue using a primary mouse anti-BrdU antibody and an anti-mouse HRP-linked secondary antibody. The color developed by the HRP substrate and read at 450nm is proportional to the BrdU incorporated and to number of proliferating cells.

To observe changes in morphology, cells (3x10³ cells/well) were seeded in a 96-well plate in 100μL of full medium as monolayers. After 48 h of incubation, cells were fixed with 50μL of cold methanol for 5 min, dried in air for 5 min, and washed three times with PBS. Then cells were stained with 50μL of 0.1% of crystal violet solution (Merck) and incubated for 20 min. in room temperature. Excess of the crystal violet stain was removed with three washes with double distilled water (DDW). Morphological changes were investigated under light microscopy, and the aspect ratio (the ratio between the length and the width of the cells) indicating the extent of cell elongation was determined using the ImagePro Plus 4.5 software (Media Cybernetics, Inc., Rockville, MD, USA). The morphology of cells seeded in the 3D system was determined by taking images of the spheroids and determining the aspect ratio of the spheroid using the ImagePro Plus 4.5 software.

Total RNA was extracted from cells cultured in 2D or in 3D using the Total RNA Purification Kit (Norgen Biotek Corp, Thorold, ON, Canada) according to the manufacturer’s instructions. RNA concentrations were determined using the NanoDrop-One 1000 spectrophotometer (Thermo Scientific, Waltham, MA, USA). Three μg of total RNA were reverse transcribed using the FIREScript RT cDNA synthesis Mix with oligo (dT) and random primers kit (Solis BioDyne, Tartu, Estonia) according to manufacturer’s instructions. Eighty ng of the resulting cDNA were amplified in the StepOne system (Applied Biosystems, Foster City, CA, USA) in triplicates, using the 5X HOT FIREPol EvaGreen qPCR Mix Plus (Solis BioDyne) and 200nM of the primers ( Table 1 ). We used initial activation at 95°C for 12 min followed by 40 cycles of 95°C for 15 sec, 60°C for 20 sec, and 72°C for 20 sec., to determine the mRNA expression level of the different genes or their endogenous reference gene PBGD. Alternatively, 1μg of total RNA were transcribed using the microRNA cDNA synthesis kit (QuantaBio, Beverly, MA, USA) and then the universal primer and the miR-146a-5p or the RNU6B (U6) small RNA that served as endogenous control were used as primers ( Table 1 ) and amplified in triplicates using the Syber green super mix (QuantaBio) kit. The relative levels of gene expression were calculated by the comparative ^ΔΔCT method, using the non-treated cells as calibrators in each experiment, to compare the relative quantity (RQ) between the samples.

Concentrations of the mouse EMMPRIN in cell supernatants (diluted 1:100) were determined using the matched antibody pair kit (Abcam), and those of MMP-9 and VEGF (supernatants diluted 1:100) were evaluated with the DuoSet ELISA kits (R&D Systems, Minneapolis, MN). Phosphorylation of the ERK1/2 and the p38 MAPKs in cell lysates (R&D systems) as well as of EMMPRIN (Abcam) in cell lysates were assessed using their DuoSet ELISA kits and normalized to the total protein. All kits were carried out according to the instructions of their manufacturers.

The bEND3 mouse microvessel endothelial cells were plated in a 96-well plate (4x10⁴ cells/well/100 μl) in full medium and allowed to reach confluency overnight. A scratch was made with a pipette tip, and detached cells were washed away. Then, conditioned media (CM) obtained from previous experiments with CT26 cells cultured in 2D or in 3D were diluted 1:2 with the full endothelial cell medium and applied onto the scratched bEND3 cells. Images of the scratch at the beginning of the experiment (time 0h) and at the end of the experiment (time 24h) were obtained by an inverted microscope. The length to which cells migrated to was measured using the ImagePro Plus 4.5 software.

The upper chamber of inserts (8μM pores size), was coated with 50μL of 0.15 mg/ml liquefied basement membrane extract (Coultrex^® BME, R&D systems), and incubated at 37°C for 4 hours to solidify the BME. Then, excess fluid was discarded, and the ECM was equilibrated with 40μL of medium for 1 hour. CT26-WT, CT26-KD or CT26-NC cells (3x10⁴ cells/insert) were seeded as monolayers in the upper chamber in 100μL serum starvation medium. Alternatively, formed spheroids were transferred to the upper chamber of the inserts. The inserts were then transferred to a 24-well plate, where the lower chamber was filled with 500μL of medium with 10% FCS as a chemoattractant. Cells were allowed to migrate to the other side of the insert membrane for 24 h, and were then fixed in 4% formaldehyde for 15 minutes. Cells or spheroids in the upper side of the insert were thoroughly removed using a cotton swab, and cells in the lower part of the membrane were stained with 0.05% Crystal Violet for 20 min. at room temperature. Images of three separate fields were taken of each membrane (Magnification x10). The stained area of the cells that migrated to the other side of the membrane was evaluated using the ImagePro Plus 4.5 software and the area of cells migrating towards serum starvation medium was subtracted from it.

CT26-WT, CT26-KD or CT26-NC cells (3x10⁴ cells/100μl full medium) were seeded on sterile cover slips and allowed to adhere overnight, after which the medium was replaced with serum starvation medium, and cells were incubated for 48 h. Then, cells were fixed with 300μL of 4% formaldehyde for 10 min following by three washes with PBS. Cells were then blocked in buffer (2% donkey normal serum, 0.1% Triton-X100 in PBS) for 1 hour at room temperature. The primary antibodies rabbit anti-Vimentin (diluted 1:700, Abcam) or rat anti-E-cadherin (diluted 1:200, BioLegend, San Diego, CA, USA) were added and incubated in 120μL of the blocking buffer overnight at 4°C. Next, cells were washed with PBS three times, and the secondary antibodies donkey anti-rabbit Alexa Fluor^® 555 (diluted 1:1000) or donkey anti-rat Alexa Fluor^® 488 (diluted 1:1000) were incubated for 1 h in the dark at room temperature. Cells were washed again once in PBS, and once with PBS with 10nM DAPI for 5 min. Coverslips were mounted with Fluoromount G on carrying glass, and sealed with a nail polish.

Spheroids were formalin fixed and paraffin embedded (FFPE), and then cut into 4μm thick sections. The sections were deparaffinized, with K-clear plus (Kaltek, Saonara, Italy), and rehydrated with decreasing ethanol concentrations. Antigen retrieval was carried out for 15 min in a microwave in citrate buffer pH 6.0, and endogenous peroxidase activity was blocked with 10 min incubation with 3% hydrogen peroxide. Slides were then blocked with 4% BSA with 0.02% triton X100 and incubated at 4°C overnight with the rabbit anti-vimentin antibody diluted 1:1400 (Abcam). After three washes (0.05% Tween 20 in PBS), the HRP-polymer anti-rabbit (Zytomed, Berlin, Germany) was added and incubated for 1 h at room temperature, following three washes and incubation for 40 min with the DAB substrate kit (Zytomed). Sections were counterstained with hematoxylin Gil III (Leica Biosystems, Deer Park, IL) and were imaged in the bright field trinocular microscope (Olympus BX-60, Tokyo, Japan) and acquired with the MS60 camera and the MShot Image Analysis System V1 (MSHOT, Guangzhou Micro-shot Technology Co., Guangzhou, China).

CT26-WT, CT26-KD and CT26-NC cells (10⁶ cells each) were centrifuged and re-suspended in PBS with 1% FCS, and then incubated with 0.5µg of rat anti-mouse FITC-conjugated anti-EMMPRIN or with its isotype control (BioLegend) for 30 min at 4°C. After washing, the cells were fixed in 0.1% formaldehyde and analyzed by flow cytometer, (LSRFortessa, BD Biosciences, San Jose, CA). Dead cells were excluded from the analysis by their forward and sideway light-scattering properties.

Equal amounts of cellular lysates (20 μg/lane) that were extracted in RIPA buffer were loaded on a gradient 4-20% SDS-PAGE, and proteins were separated by electrophoresis and transferred onto a nitrocellulose membrane (Advansta, San Jose, CA, USA). Block-Chemi (Advansta) reagent served to block the membranes for 1 hour. Then the primary antibodies goat anti-EMMPRIN (R&D systems, diluted 1:1000), mouse anti-TRAF6 antibody (Santa Cruz Biotechnology, Dallas, TX, USA, diluted 1:500) or mouse anti-β-actin (ProteinTech, Rosemont, IL, USA, diluted 1:10,000) in blocking solution were added for overnight incubation at 4°C, following 3 washes in TBST buffer (1xTris-buffered saline with 0.1% Tween 20). Then the HRP-conjugated secondary antibodies donkey anti-goat IgG or goat-anti-mouse IgG (Jackson ImmunoResearch Labs, West Grove, PA, USA, diluted 1:5,000) was incubated for 1 hour, and after three more washes, the membranes were incubated with the WesternBright ECL HRP substrate (Advansta). The protein bands were visualized using the Omega Lum G imaging system (Aplegen, Pleasanton, CA, USA).

The Lipofectamine RNAi MAX (Thermo Fisher Scientific/Ambion, Austin, TX, USA) was diluted 1:25 in Opti-MEM medium and combined with an Opti-MEM medium containing 150 nM of miRNA-146a-5p mimic (final concentration was 30 nM), or its negative controls (Thermo Fisher Scientific/Ambion). The generated lipid complexes were spread on a well in a 24-well plate for 20 minutes. Then the CT26-WT cells (8x10⁴ cells) were added in full medium without antibiotics. After 24 hours of incubation, cells in monolayers were washed with PBS and starvation medium (300 μl) was added for 48 hours. To generate transfected spheroids, cells were trypsinized and 5x10³ cells were plated in a well of 96-well plates coated with 1% agarose in full medium (70μl) for 48 hours. At the end of the incubation cellular lysates, RNA extracts or supernatants were collected for analysis.

To simulate metastatic outbreak that relies on the proliferation of metastatic cells in a 3D matrix, we followed the protocol described in (38). Briefly, we coated 96-well plates with 40μl of ice cold Coultrex™, that was solidified by placing the plates in a humidified incubator for 30 min. Cells were then seeded onto the BME-coated plates (10³ cells/well/60μl) in CT26 serum-starvation medium supplemented with 2% FCS + 2% BME (assay media). Every 4 days, cells were re-fed with the assay media. Images were taken at different time points. The mean area of the resulting cell structures was quantified using ImagePro Plus 4.5 software.

All experiments were independently repeated at least four times in triplicates, and results are represented as mean ± standard error of mean (SEM). Statistical significance between three groups was determined using one-way ANOVA followed by Bonferroni’s post-hoc multiple comparison test. Statistical significance was determined at P-values less than 0.05 (α<0.05). All statistical analyses were performed using Prism 10.10 software (GraphPad, La Jolla, CA, USA).

To study the involvement of EMMPRIN in the formation of spheroids and in their properties, we stably knocked-down its expression in the mouse colon carcinoma cell line CT26, as described in the methods section and in (36). The expression of EMMPRIN was then validated at the RNA and protein levels, and as a secreted protein. Although the levels of EMMPRIN mRNA were not different between the CT26-WT, CT26-KD and CT26-NC cells ( Figure 1A ), the protein levels at the membrane ( Figure 1B ), in the cells lysates ( Figure 1C ) and the secreted levels ( Figure 1D ) were significantly reduced by about 50-60%. This reduction was also confirmed in western blot analyses ( Figures 1E, F ). This reduced level of expression in the CT26-KD cells still allowed their proliferation and survival. The lack of change in the EMMPRIN mRNA levels suggests a post-translational regulation of EMMPRIN, and we have previously implicated miR-146a-5p in such a regulation (39).

The morphology of the cells changed both in monolayers and in spheroids. CT26-WT and CT26-NC monolayers presented a network of elongated cells, indicating their mesenchymal state, whereas the CT26-KD cells were a little more spread on the culture dish ( Figure 2A , upper panel). The change in the morphology was reflected by the high aspect ratio (4.04 ± 0.5) that describes the spindle-like morphology of the CT26-WT and CT26-NC cells, which was reduced by 42% in the CT26-KD cells ( Figure 2B ). Similar results were obtained for the mouse mammary carcinoma 4T1 cells, where the aspect ratio was decreased in the 4T1-KD cells by 18.5%, and for the Lewis lung carcinoma (LLC) cells where the aspect ratio was decreased by 27.5% in the LLC-KD cells ( Supplementary Figures S1A, B ).

In the 3D model, the CT26-WT and CT26-NC cells formed round, almost perfect spheres when plated on wells coated with 1% agarose, which was manifested by the aspect ratio of 1.18 ± 0.016. These spheroids were large, with a mean diameter of 573 ± 12.56μm ( Figure 2A , lower panel). Similarly, the 4T1-WT and the LLC-WT cells also generated round spheroids with an aspect ratio of 1.054 ± 0.005 and 1.046 ± 0.004μm, respectively ( Supplementary Figures S1A, B ). In contrast, the CT26-KD cells formed non-spherical and more elliptic aggregates (mean of long diameter 706 ± 22, mean of short diameter 416 ± 10.1), resulting in an increase of their aspect ratio by 37%. However, despite the different morphology, the CT26-WT, CT26-KD and CT26-NC cells did not present differences in their area (253,000 ± 11,325 μm), suggesting that their cellular mass was not different. Likewise the 4T1-KD cells and the LLC-KD cells generated non-spherical elliptic aggregates, with an aspect ratio of 1.78 ± 0.17 and 1.42 ± 0.004, respectively ( Supplementary Figures S1A, B ).

To explore the effects of the spatial organization on the EMT status of CT26 cells that express EMMPRIN in high (CT26-WT, CT26-NC cells) or low levels (CT26-KD cells), we evaluated some of the properties that are associated with a mesenchymal phenotype. These include their proliferative capacity, drug resistance, invasiveness, angiogenic potential, and ability to express genes related to the EMT process and to dormancy. However, because monolayers and spheroids were cultured in different conditions, we compared separately the magnitude of the difference between the CT26-WT/CT26-NC and CT26-KD cells in each of the different spatial organizations.

The CT26-KD cells cultured as monolayers and as spheroids presented reduced rates of proliferation relative to their respective CT26-WT and CT26-NC cells, as assessed both by the cell counting kit (CCK-8) and by the DNA incorporation of BrdU ( Figures 3A, B ). In the monolayers, CT26-KD cells showed a 21% or 27% reduction relative to the CT26-WT cells in the CCK-8 and BrdU assays, respectively, whereas in the spheroids the proliferation was inhibited by 35% and 48%, respectively. Thus, in cells with reduced EMMPRIN expression, proliferation in spheroids is attenuated more than in monolayers.

To assess the ability of cells to develop drug resistance, we exposed them to the chemotherapeutic drug etoposide, an inhibitor of topoisomerase II that is highly expressed in CT26 cells (40). Only about 50% of both the CT26-WT and CT26-NC cells cultured as monolayers survived upon exposure to low concentrations of etoposide, reaching 18% and 13% survival in the high concentration of 15μM etoposide, respectively ( Figure 3C ). In contrast, the CT26-KD cells in monolayers exhibited greater resistance to the drug, reaching 55% survival at 15μM etoposide. The 3D organization in spheroids conferred resistance to the drug even in the CT26-WT and CT26-NC cells, which exhibited 52% and 46% survival at 15μM etoposide, respectively, consistent with previous reports of increased drug resistance in spheroids relative to monolayers (41). The CT26-KD cells in spheroids exhibited 77.3% survival at that concentration ( Figure 3D ). Thus, the reduced EMMPRIN expression together with the spheroid structure conveyed the best protection against etoposide-induced cell death.

Invasiveness is one of the hallmarks of EMT and a mesenchymal phenotype. To explore the invasive abilities of cells, CT26-WT, CT-KD and CT26-NC cells were seeded as monolayers or as spheroids in inserts coated with basement membrane extract (BME), and the area of the cells that migrated and invaded through the extract to the lower side of the insert membrane was evaluated. Relative to their parental CT26-WT or to CT26-NC cells, CT26-KD cells that were seeded as monolayers showed a 2.4-fold increase in their ability to invade the Coultrex^®-coated membrane. CT26-KD cells that were seeded as spheroids exhibited a 3.5-fold increase in their invasive abilities compared to their CT26-WT or CT26-NC counterparts ( Figures 4A, B ).

The reduced expression of EMMPRIN also affected different aspects of angiogenesis. The wound assay was carried out using conditioned media (CM) derived from either the monolayers or the spheroids that were diluted (1:2) with full medium, and applied onto a scratched monolayer of the bEND3 mouse endothelial cells. Factors secreted from the monolayer CT26-KD cells resulted in a 23% attenuation in the rate of gap closure relative to CM derived from CT26-WT or CT26-NC cells ( Figures 5A, C ), suggesting that EMMPRIN participates in determining the angiogenic potential. When cells were seeded as spheroids, the attenuation in bEND3 cell migration based on CT26-KD relative to CT26-WT cells was 36.5% ( Figures 5B, C ). We next examined the levels of pro-angiogenic factors in the CM, and detected that the secretion of VEGF in CT26-KD monolayers and spheroids was reduced by 66% and 57% respectively, relative to their respective CT26-WT counterparts ( Figure 5D ). Interestingly, no changes were observed in the secretion of MMP-9 in the monolayer cells, probably due to the lack of any stimulus that induces MMP-9. In contrast, MMP-9 secretion was enhanced by the CT26-WT and CT26-NC spheroids, and the CT26-KD cells exhibited a 61% reduction relative to the CT26-WT spheroids ( Figure 5E ).

The reduced expression of E-cadherin and the increase in the expression of vimentin is a hallmark of epithelial cells undergoing EMT to acquire a more mesenchymal phenotype. CT26 cells do not express E-cadherin (40), and we have validated this by immunofluorescence (data not shown), so they have already acquired a mesenchymal phenotype, as suggested by their morphology. However, relative to their respective CT26-WT cells, CT26-KD cells enhance their vimentin expression by 1.48-folds in monolayers and by 1.86-folds in spheroids ( Figures 6A, B ). Interestingly, the expression of vimentin in the spheroids is located at the rims in CT26-WT and CT26-NC cells, but penetrates towards the core of the spheroids in CT26-KD cells ( Figure 6A , lower panel). This may suggest that the reduction in EMMPRIN expression drives the cells towards an even more mesenchymal phenotype than their parental CT26-WT cells. The activation of the EMT program in these cells is also evident by the increase in the expression of the EMT-TFs snail, slug, twist1 and zeb1, each by about 2-folds in CT26-KD cells relative to CT26-WT cells seeded as monolayers ( Figures 6G–J ). When the cells were seeded as spheroids, the CT26-KD cells enhanced the expression of snail (2-folds), slug (8-folds), twist1 and zeb1 (3.5-folds) relative to the CT26-WT spheroids ( Figures 6G–J ).

We next asked whether CT26-KD cells are more dormant than their parental CT26-WT cells. The ratio of the phosphorylated kinases ERK1/2 and p38 is often used to assess dormancy, as p38 MAPK can inhibit proliferation and ERK1/2 MAPK can promote proliferation (42). We show that relative to the CT26-WT cells, CT26-KD cells in monolayers exhibit a 69% reduction in the ERK/p38 ratio, and a 76% reduction when seeded as spheroids ( Figure 6C ). Thus, phosphorylation of ERK1/2 is reduced and phosphorylation of p38 is increased, suggesting enhanced dormancy in CT26-KD cells. Likewise, the gene expression of the dormancy markers NR2F1 and p21 in CT26-KD cells was enhanced by about 2-folds in both monolayers and spheroids, and the stemness marker SOX2 mRNA was enhanced by 2-folds in monolayers and by 4.7-folds in spheroids, relative to their respective CT26-WT cells ( Figures 6D–F ).

As CT26-KD cells tend to be more mesenchymal and dormant than their parental CT26-WT cells, we asked whether this could be reflected in a functional assay. We evaluated the potential of the cells to escape dormancy and generate a metastatic-like structure using an in vitro system that was previously demonstrated to distinguish between dormant and metastatic cells (38, 43), and is based on the ability of cells to aggregate and form 3D structures. In contrast to spheroids that get more compacted over time and are devoid of contact with ECM proteins, cells in this assay interact with the basement membrane extract (BME), and proliferate over time to generate an irregular 3D structure that resembles a metastatic lesion and simulates the process of the metastatic outbreak. After 13 days in culture, the CT26-WT and CT26-NC cells developed such 3D structures with typical protrusions and a mean area of 83,090 ± 8,730 μm², thus mimicking the metastatic outbreak. In contrast, the CT26-KD cells remained as very small cellular aggregates, and their mean area was smaller by 81% ( Figure 7 ). Thus, EMMPRIN expression is required for the process of the metastatic outbreak. These results were also confirmed in LLC-WT cells that generated metastatic-like lesions with an average area 143,647 ± 24,810 μm², compared to the LLC-KD cells that did not generate such structures and their average area was reduced by 98% ( Supplementary Figure S1C ).

We have demonstrated that EMMPRIN affects cell morphology, promotes the mesenchymal phenotype and prevents dormancy in spheroids more than in monolayer, raising the question what determines the differences in EMMPRIN expression. Since we have previously linked miR-146a-5p to post-transcriptional regulation on EMMPRIN expression (39), we next evaluated the expression of miR-146a-5p in monolayers and spheroids. Relative to their respective CT26-WT cells, the expression levels of miR-146a-5p were increased in CT26-KD cells by 2-folds in monolayers, and by 5.3-folds in spheroids ( Figure 8A ), whereas the accumulated levels of EMMPRIN in the supernatants were decreased in CT26-KD cells by 55% in the monolayers and by 82% in the spheroids ( Figure 8B ). To reconfirm the link between miR-146a-5p and EMMPRIN and to simulate miR-146a-5p overexpression as observed in the CT26-KD cells, we transfected CT26-WT cells with the miR-146a-5p mimic (CT26-WT-mimic) or its negative control (CT26-WT-NC). Relative to the negative control, the accumulation of EMMPRIN in the supernatants was increased by 1.9-folds in both monolayers and spheroids ( Figure 8C ). Thus, miR-146a-5p positively regulates EMMPRIN and enhances its secretion, whereas EMMPRIN inhibits the levels of miR-146a-5p, generating a negative feedback loop ( Figure 8H ).

We next reasoned that in a negative feedback loop, proteins that are targeted by either miR-146a-5p or EMMPRIN should be affected by the manipulation of any of the regulators. Therefore, we looked first at the secretion of transforming growth factor beta 1 (TGFβ1), that is regulated by EMMPRIN, but not by miR-146a (44, 45). Relative to CT26-WT cells, TGFβ was reduced in CT26-KD cells both in monolayers and in spheroids ( Figure 8D ). Conversely, TGFβ was increased in CT26-WT-mimic cells relative to CT26-WT-NC cells ( Figure 8F ). The tumor necrosis factor receptor−associated factor 6 (TRAF6) is a verified target of miR-146a-5p (46), but not of EMMPRIN. TRAF6 expression was reduced in CT26-KD in monolayers relative to the CT26-WT cells ( Figure 8E ), and in the CT26-WT-mimic cells seeded as monolayers relative to the CT26-WT-NC cells ( Figure 8G ). However, no TRAF6 protein expression could be detected in any of the cells when seeded as spheroids ( Figures 8E, G ).