The noninvasive breast cancer cell line, MCF7 (ER⁺, PR⁺, HER2^–, Research Resource Identifiers #RRID: CVCL_0031, ATCC Number: HTB-22) were purchased from The Global Bioresource Center $|$ American Type Culture Collection (ATCC, Manassas, VA, USA) and cultured according to the recommended growth conditions. In detail, MCF7 were grown at 37 °C in Dulbecco’s Modified Eagle Medium (DMEM, HiMedia Leading BioSciences Company, Mumbai, Maharashtra, India, Catalog number: AL007-500ML) supplemented with Fetal Bovine Serum to a final concentration of 10% (HiMedia Leading BioSciences Company, Mumbai, Maharashtra, India, Catalog number: RM10432), L-Glutamine (Gibco, Thermo Fisher Scientific, Waltham, MA, USA, Catalog number: 25030024) and Penicillin/Streptomycin (Gibco, Thermo Fisher Scientific, Waltham, MA, USA, Catalog number: 15070063) both to a final concentration of 1%, in a humidified atmosphere with 5% CO₂. In all the experiments in which TH was applied to cells, we used a hormonal concentration that resembles the physiological exposure of cells to the active hormone T3 (30.0 nM T3, in a time-course treatment of 24–48 hours, Sigma-Aldrich St. Louis, MO, USA, Catalog number: T6397). In all the experiments in which type II deiodinase was inhibited, we used a specific D2 inhibitor, 3,3^′,5^′-Triiodo-L-thyronine, commonly known as reverse-T3, rT3 (30.0 nM rT3, in a time-course treatment of 24–48 hours, Cayman Chemical Company, Ann Arbor, MI, USA, Catalog number: Cay17598). The transfection protocol was performed using Lipofectamine-2000 (Invitrogen™, Carlsbad, CA, USA, Catalog number: 11668019), according to the manufacturer’s instructions. All experiments carried out in the current study meet all relevant ethical regulations. Furthermore, all cell lines used have been authenticated and validated by (1) morphology check by microscope, (2) identity verification by Short Tandem Repeat (STR) profiling and (3) mycoplasma detection. Cell lines were thus tested negative for mycoplasma contamination.

The generation of the MCF7 CRISPR/Cas9 control and DIO2 knock out (D2KO) clones is reported elsewhere [27]. In brief, D2KO MCF7 cells were obtained following targeted mutagenesis of the DIO2 gene in MCF7 cells by using the CRISPR/Cas9 technology from Santa Cruz Biotechnology (Dallas, TX, USA) (DIO2 CRISPR/Cas9 KO Plasmid (h), Catalog number: sc-402262) and control D2WT MCF7 cells were stably transfected with the CRISPR/Cas9 control plasmid (Control CRISPR/Cas9 Plasmid, Catalog number: sc-418922). 72 hours after transfection, the cells were sorted using Fluorescence Activated Cell Sorting (FACS) for Green Fluorescent Protein expression (FACS gating/sorting strategy is reported in Supplementary Fig. 1a). To avoid off-target effects, three different single D2KO clones were selected and analyzed by PCR to identify alterations in coding regions (Supplementary Fig. 1b). Furthermore, DIO2 depletion was confirmed by DNA sequencing (Supplementary Fig. 1c).

The intracellular T3 and T4 levels were measured as previously described [28, 29]. Briefly, frozen D2KO and wild type (WT) MCF7 cell samples were homogenized on ice in phosphate buffer containing 0.25 M sucrose, 1 mM EDTA, 0.1 M NaPO4, and 10 mM 1,4-Dithiothreitol (DTT, Sigma-Aldrich St. Louis, MO, USA, Catalog number: D062), and sonicated. Then, 100.0 µg of protein extract was incubated for 1 hour at 37 °C in 0.3 mL PE buffer with 10 mM DTT. Then, 20 nM T4-13C6 (Sigma-Aldrich St. Louis, MO, USA, Catalog number: T-076) was added to the reactions and incubated at 37 °C for 6 hours. Proteins were precipitated in acetonitrile (Sigma-Aldrich St. Louis, MO, USA, Catalog number: AX0156) and supernatant evaporated to dryness at 37 °C in a rotavapor. The dried extract was reconstituted with 30.0 µL of a 95:5 methanol:NH4-OH solution (Methanol with 0.1% Ammonium Hydroxide, Honeywell Research Chemicals, Germany, Catalog number: LC2183) and centrifuged at 13.000 rpm for 10 minutes. Samples (5.0 µL) were injected on a Raptor Biphenyl (Restek Srl. Milan, Italy, Catalog number: 9309A12) 2.7 µm, 100 mm $\to$| 2.1 mm using as Mobile Phase A 0.1% Formic acid in water and as Mobile phase B 0.1% Formic acid in methanol. Calibration curves (1.0 to 40.0 nM) were prepared dissolving pure T4, T3, and T4-13C6 in 95:5 methanol: NH4-OH solution. The analyte MRMs were for T3-13C6 (658.07 $>$ 612.1 $>$ 514.1); T3 (652.07 $>$ 606.1 $>$ 508.1); T4 (778.0 $>$ 731.9 $>$ 323.9); T4-13C6 (784.1 $>$738.04). Measurements were conducted in triplicate.

In this study, primary MAT-MSCs were extracted from the mammary adipose tissue of healthy females undergoing surgical mammary reduction, free of neoplastic, metabolic or endocrine diseases [Age: 40.7 $\pm$ 9.0; Weight (kg): 68.7 $\pm$ 9.0; Height (cm): 166.0 $\pm$ 11.5; BMI: 24.8 $\pm$ 0.2]. Samples were obtained from the Surgery Department of the University Hospital Center “Federico II” of Naples, with patients and the Institution Ethical Committee (Protocol Number 138/16). Written informed consents were collected before surgical procedures. Briefly, 1.0 mg/mL of Dispase (Sigma-Aldrich St. Louis, MO, USA, Catalog number: C2139) was utilized to enzymatically digest the biopsies to isolate MAT-MSCs [30, 31]. Digestion solution was mixed and incubated for 1 hour at 37 °C. The reaction was finally stopped by adding the same volume (1% w/v) of High Glucose Dulbecco’s Modified Eagle Medium (DMEM, HiMedia Leading BioSciences Company, Mumbai, Maharashtra, India, cod. AL007) supplemented with 10% Fetal Bovine Serum (FBS, HiMedia Leading BioSciences Company, Mumbai, Maharashtra, India, cod. RM10432) , enriched in L-Glutamine (Gibco, Thermo Fisher Scientific, Waltham, MA, USA, cod. 25030024) , Penicillin/Streptomycin (Gibco, Thermo Fisher Scientific, Waltham, MA, USA, cod. 15070063), Antibiotic-Antimycotic (Gibco, Thermo Fisher Scientific, Waltham, MA, USA, Catalog number: 15240062), all of them to a final concentration of 1%. The cell pellet was centrifuged at 1200 g $\times$ 5 minutes at 16 °C and then resuspended in full DMEM/F-12 cell culture medium (Dulbecco’s Modified Eagle Medium/Nutrient Mixture F-12, Thermo Fisher Scientific, Waltham, MA, USA, Catalog number: 11320033), passed through a 70-µm Cell Strainer (Corning, Fisher Scientific, Hampton, New Hampshire, USA, Catalog number: 431751) to remove the remaining tissue debris and seeded in cell culture flasks at a density of 5000–10,000 cells/cm². Fresh culture medium was added after 2 days post-seeding. Finally, MAT-MSCs were grown in full DMEM/F-12 cell culture medium at 37 °C and 5% CO₂, and medium was changed every 48–72 hours. All tests were performed on cells from passages 4 to 10, and cells were routinely tested for Mycoplasma. The cells were tested to be Mycoplasma-free.

Colony Formation Assay is reported elsewhere [13]. In brief, D2KO and WT MCF7 cells (5.0 $\times$ 10³ cells) were seeded in cell culture plates to form colonies. 1 week after plating, cells were washed with Phosphate-Buffered Solution (PBS) and stained with 1% Crystal Violet (Sigma-Aldrich St. Louis, MO, USA, Catalog number: C6158) in 20% ethanol (Carlo Erba Reagents S.r.l., Cornaredo MI, Italy, Catalog number: 4146052) for 30 minutes at room temperature (RT). Colonies were manually counted and imaged using a Leica DMi8 microscope digital camera system and the Leica Application Suite LAS X Imaging Software – Version 3.6.0.20104 2018 (Leica Microsystems GmbH, Wetzlar, Germany). The counted colonies were plotted as a graph of the number of colonies per field.

Cell-cycle analysis was performed as previously described [13]. Cell-cycle profile of D2KO and WT MCF7 cells was determined by FACS analysis (FACS Canto2, Becton Dickinson, Rowa, Italy). After cells synchronization, consisting in 24 hours of serum starvation, at least 10,000 cells were firstly fixed in ice-cold 70% ethanol at –20 °C, and then stained for 10 minutes at room temperature with a mixed solution of 5.0 µg/mL Propidium Iodide (PI) (Sigma-Aldrich St. Louis, Missouri (MO), USA, Catalog number: P4864) and 0.25 mg/mL Ribonuclease I (RNase I, 10 U/µL, Thermo Fisher Scientific, Waltham, MA, USA, Catalog number: EN0601). Obtained data were analyzed with the MODFIT Lt 3.0 Software – Version 3.0 (BD Biosciences, Rome, Italy).

Multicellular 3D-mammosphere were generated from subconfluent MCF7 cells using Ultra-Low Attachment surface plates. In brief, D2KO and WT MCF7 cells were resuspended in complete DMEM growth medium and plated in an Ultra-Low Attachment Polystyrene 96-well (Corning™-Costar®, Fisher Scientific, Hampton, NH, USA, Catalog number: 7007) at a final concentration of 5 cells/well, with or without MAT-MSCs Conditioned Media (MAT-MSCs CM, ratio 1:1). Mammosphere diameters were quantified after 2, 5 and 7 days after seeding by Software associated to the Leica DMi8 microscope digital camera system and the Leica Application Suite LAS X Imaging Software (Version 5.0.2, Leica Microsystems GmbH, Wetzlar, Germany). DAPI Nucleic Acid staining (4^′,6-Diamidino-2-Phenylindole, Dihydrochloride, Invitrogen™, Carlsbad, CA, USA, Catalog number: D1306, Staining Condition: 1:1000 DAPI stock solution 2.5 mg/mL, 10 minutes directly in cell culture medium) and Fluorescence Laser-Scanning Microscopy (Leica Microsystems GmbH, Wetzlar, Germany) for in vivo imaging were performed on not-fixed 7 days after seeding mammospheres.

The bidirectional interaction between MAT-MSCs and MCF-7 cells was analyzed by performing 2D-MAT-MSCs/MCF7 coculture assay, according to the manufacturer’s protocol. In detail, MAT-MSCs (2.5 $\times$ 10⁴ cells) were seeded in the bottom chamber of a transwell-6 culture system (Corning™-Costar®, Fisher Scientific, Hampton, NH, USA, Catalog number: 353090, Permeable Support for 6-well plate with 0.4 µm Transparent PET Membrane), while D2KO or WT MCF7 cells (3.5 $\times$ 10⁴ cells) were plated in the upper chamber. MAT-MSCs and MCF-7 cells were cocultured, each in own cell growth medium, for 72 hours. In parallel, MAT-MSCs and D2KO and WT MCF7 cells were monocultured, each in its own cell growth medium, as control. After 72 hours, Conditioned Media (CM) were collected, and cells were harvested for RNA extraction analysis.

MAT-MSCs and MCF7 migration assays were performed using Transwell migration chambers (Corning™-Costar®, Fisher Scientific, Hampton, NH, USA, Catalog number: 353182, Permeable Support for 12-well plate with 8.0 µm Transparent PET Membrane). In brief, MAT-MSCs or D2KO and WT MCF7 cells (1.0 $\times$ 10⁵ cells) were added to the upper chamber of a transwell insert, while the lower chamber, pre-coated for 30 minutes with a thin Matrigel layer (BD Matrigel Matrix, BD Biosciences Discovery Labware, Europe, Catalog number: 356234), was supplemented with 50% Conditioned Medium (CM) collected from MAT-MSCs or either D2KO and WT MCF7 cells. After 5 days of incubation at 37 °C, cells that remained on the top of the filter were scrubbed off, and cells that had migrated to the underside of the filter were fixed and stained with 1% Crystal Violet in 20% ethanol for 30 minutes at room temperature (RT). Migrated cells were manually counted and imaged using a Leica DMi8 microscope digital camera system and the Leica Application Suite LAS X Imaging Software (Version 5.0.2, Leica Microsystems GmbH, Wetzlar, Germany). The counted cells were plotted as a graph of cells migrated per field.

The total amount of messenger RNAs (mRNAs) was extracted using TRIzol reagent (Life Technologies Ltd, Carlsbad, CA, USA, Catalog number: 15596018). Complementary DNAs (cDNAs) were synthesized using SuperScript™ VILO™ MasterMix (Life Technologies Ltd, Carlsbad, CA, USA, cod. 11755-050) starting from 1.0 µg of mRNA, according to the manufacturer’s instructions. On CFX Connect Real-Time PCR Detection System (BioRad, Hercules, CA, USA, cod. 1855201), Real-Time PCR was directed using the fluorescent double-stranded DNA-binding dye SYBR Green (BioRad, Hercules, CA, USA, cod. 1708882). All samples were analyzed in technical triplicate and standardized to an endogenous control, Cyclophilin-A (CYPA). The Relative Quantification (RQ) and the expression of each mRNA were determined utilizing the comparative Ct methodology and expressed as N-fold differences in target gene expression [N*target = 2^{(Δ⁢Ct sample-Δ⁢Ct calibrator)}]. Specific primers for each gene were designed to work under the same cycling conditions [95 °C for 10 minutes $>$40 cycles at 95 °C for 15 seconds $>$60 °C for 1 minute], thus generating amplicons of comparable sizes (200–300 bp). Primer combinations were placed whenever possible to span an exon-exon junction and the total RNAs were digested with DNAse to avoid genomic DNA contamination. Primer sequences are indicated in the Supplementary Data 1.

To estimate the total protein expression, MCF7 cells were lysed with ice-cold Protein Lysis Buffer [0.5% TRITON X-100; 25.0 mM TRIS-HCl pH 7.4; 300.0 mM NaCl; 1.0 mM CaCl₂], supplemented with sodium fluoride (NaF, Sigma-Aldrich St. Louis, MO, USA, Catalog number: 201154), sodium orthovanadate (Na3VO4, Sigma-Aldrich St. Louis, MO, USA, Catalog number: S6508), $\beta$-glicerophosphate (Sigma-Aldrich St. Louis, MO, USA, Catalog number: G5422), Phenyl-Methyl-Sulfonyl-Fluoride (PMSF, Sigma-Aldrich St. Louis, MO, USA, Catalog number: P7626) and Protease Inhibitor Cocktail (Sigma-Aldrich St. Louis, MO, USA, Catalog number: P8340), for 30 minutes on ice from whole cell lysates. Protein concentration was measured by spectrophotometer absorbance using Bio-Rad Protein Assay Dye Reagent (BioRad, Hercules, CA, USA, Catalog number: 5000006). After protein estimation, each sample was admixed with 2$\times$ Protein Loading Buffer [150.0 mM TRIS-HCl pH 6.8; 6.0% SDS; 30% glycerol, Bromophenol Blue] and boiled for 7 minutes at 99 °C for denaturation. Finally, cell proteins were separated by 8–10 or 12% SDS-PAGE followed by Western Blot. The membrane was then blocked with 5% Bovine Albumin Serum (BSA, Sigma-Aldrich St. Louis, MO, USA, Catalog number: A9647) in Tris-Buffered Saline (TBS)-0.2% Tween and incubated over-night at 4 °C with the following primary antibodies: anti-Cyclin-D1 (Santa Cruz Biotechnology, Catalog number: sc-246), anti-E-Cadherin (BD Biosciences, Catalog number: 610181), anti-ERK1/2 (K-23) (Santa Cruz Biotechnology, Catalog number: sc-94), anti-phospho-p44/42 MAPK (ERK1/2) (Thr202/Tyr204) (Cell Signaling, Danvers, MA, USA, Catalog number: 4370), anti-N-Cadherin (Elabscience, Catalog number: E-AB-32170), anti-p21 [Waf1] (Santa Cruz Biotechnology, Catalog number: sc-397), anti-p27 [Kip1] (BD Transduction Laboratories, Catalog number: 610242), anti-p38 MAPK (Cell Signaling, Catalog number: 9212), anti-phospho-p38 MAPK (Thr180/Tyr182) (Cell Signaling, Catalog number: 9211S), anti-Slug (Cell Signaling, Catalog number: 9585S), anti-Snail (Cell Signaling, Catalog number: 3895S), anti-ZEB1 (abcam, Catalog number: ab-155249) and anti-Tubulin (Santa Cruz Biotechnology, Catalog number: sc-5546), this last used as loading control. Detection was established by using peroxidase-conjugated secondary specific antibodies (anti-mouse IgG-HRP, BioRad, Hercules, CA, USA, Catalog number: 1706516 and anti-rabbit IgG-HRP, BioRad, Hercules, CA, USA, Catalog number: 1706515). The signals were detected by chemiluminescence using an ECL kit (Millipore, Burlington, MA, USA, Catalog number: WBKLS0500) on ChemiDoc XRS+ System/Images Lab Software (BioRad, Hercules, CA, USA, Catalog number: 1708265). All Western Blots were run in triplicate and bands were quantified with ImageJ Software (Version 1.8.0, NIH Image, Bethesda, MD, USA). Dilution for each antibody is reported in Supplementary Data 2.

For immunofluorescence staining of MCF7 cell-cultured coverslips, attached cells were first washed in cold PBS and then fixed with 100% cold buffered formaldehyde solution 4%, pH 6.9 (Sigma-Aldrich St. Louis, MO, USA, Catalog number: 1004968350) on ice for 10 minutes. Cells were then permeabilized with 0.2% Triton X-100 (Sigma-Aldrich St. Louis, MO, USA, Catalog number: T8787) for 15 minutes, washed with PBS and then blocked for 30 minutes with 2% BSA in PBS. Next, cells were incubated over-night at 4 °C with primary anti-E-Cadherin (BD Biosciences, Catalog number: 610181), anti-N-Cadherin (Elabscience, Catalog number: E-AB-32170) and anti-ZEB1 (abcam, Catalog number: ab-155249) antibodies (dilution for each antibody is reported in Supplementary Data 2), appropriately diluted into blocking solution. Next day, cells were washed with PBS and then incubated with Alexa Fluor Cross-Adsorbed Secondary Antibody [anti-mouse IgG (H+L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor 594, Invitrogen™, Carlsbad, CA, USA, Catalog number: A-21203; anti-Mouse IgG (H+L) Cross-Adsorbed Secondary Antibody, Alexa Fluor 488, Invitrogen™, Carlsbad, CA, USA, Catalog number: A-11001; anti-Rabbit IgG (H+L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor 594, Invitrogen™, Carlsbad, CA, USA, Catalog number: A-21207; anti-Rabbit IgG (H+L) Cross-Adsorbed Secondary Antibody, Alexa Fluor 488, Invitrogen™, Carlsbad, CA, USA, Catalog number: A-11070], in the dark, at room temperature (RT), for 1 hour. Finally, after the DAPI Nucleic Acid staining, coverslips were mounted using 80% glycerol. The images were captured using a Leica DMi8 microscope and the Leica Application Suite LAS X Imaging Software (Version 5.0.2, Leica Microsystems GmbH, Wetzlar, Germany) and quantified using ImageJ Fiji Software (Version 2.14.0, ImageJ Wiki, Madison, WI, USA).

D2KO and WT MCF7 cells were normally cultured and 72 hours post-plating Conditioned Media were collected and screened for the concentration of a panel of different cytokines and chemokines (i.e., Interleukin-1β (IL-1$\beta$), IL-6, IL-8, IL-10, Interleukin-1 Receptor Antagonist (IL-1RA), Interleukin-1 Receptor alpha (IL-6 R alpha), B-Cell Activating Factor (BAFF), Vascular-Endothelial Growth Factor (VEGF), Chemokine C-C motif Ligand-3 (CCL-3), CCL-4, CCL-11, C-X-C motif chemokine ligand 10 (CXCL-10), FURIN, CRIPTO, alpha Synuclein, urokinase (u-PA), Interferon-$\gamma$ (IFN-$\gamma$), Tumor Necrosis Factor-alpha (TNF-alpha) and Total Inhibin) using a custom Human Magnetic Luminex Assay (CUSTOM-LXSA-H-21, R&D System), according to the manufacturer’s protocol. The magnetic bead-based multiplex assay was performed on a Luminex® 200™ System (Bio-Techne, R&D Systems, Austin, TX, USA, Catalog number: LX200-XPON-RUO) and the results were analyzed with xPONENT® Software 4.3. Concentration (expressed as pg/mL) of detected molecules are reported in Table 1 and Supplementary Data 3.

All animal procedures were approved by the Institutional Animal Care and Use Committee (IACUC, nos. 167/2015-PR and 354/2019-PR) and conducted in accordance with the guidelines of the “Department of Human Health, Animal Health and Ecosystem and International Relations” from Ministero della Salute (MSAL). 2-months-old male C57BL/6 mice, purchased from Charles River Laboratories (Wilmington, MA, USA), were intraductally xenografted in absence of any surgical manipulation. In detail, mice were anesthetized by 0.01 mL/g intraperitoneal (IP) injection with ketamine (9.0 mg/mL), and the hair surrounding the the fourth pair of abdominal mammary glands were trimmed with a shaver and wiped with 70% ethanol. 2.0 $\times$ 10⁵ D2KO and WT MCF7 cells, suspended in 10.0 µL of sterile PBS, were backloaded into the Insulin Syringes U-100 30G 1.0 mL/cc 5/16” (8 mm). The needle was gently inserted directly into the mammary duct of mice and tumor cells were slowly ejected. Mammary glands were collected 8 weeks post-tumor cell injection. At the end of the experimental procedures, animals were euthanized by exposure to a mixture of isoflurane and oxygen (4% isoflurane e 0,8 Lt/min oxygen) by inhalation and then killed by cervical dislocation.

For histology and immunostaining, engrafted mammary glands were harvested, fixed in 10% neutral buffered formalin (Bio-Optica, Milano, Italy, Catalog number: 05-01005Q) and paraffin-embedded. Hematoxylin/Eosin (H/E) and immunostaining were performed on 15.0 µm thick tissue sections. Embedded-murine mammary gland sections were dewaxed in xylene, subsequently passed through decreasing ethanol concentrations for rehydration and then permeabilized in 0.2% Triton X-100/PBS. Slides were stained with Hematoxylin Solution (Sigma-Aldrich St. Louis, MO, USA, Catalog number: GHS116) and counterstained with Eosin Y Solution (Sigma-Aldrich St. Louis, MO, USA, Catalog number: HT110216) prior to dehydrating and mounting with Eukitt mounting Solution (BioOptica S.p.A, Milan, Italy, Catalog number: 09-00250). For immunostaining analysis, sections were blocked in 1% Bovine Serum Albumin (BSA, Sigma-Aldrich St. Louis, MO, USA, Catalog number: A9647)/0.2% Tween (Sigma-Aldrich St. Louis, MO, USA, Catalog number: P2287) /PBS for 1 hour at room temperature (RT). Primary antibodies were incubated over-night at 4 °C in the blocking buffer and then washed in 0.2% Tween/PBS. Secondary antibodies were incubated at room temperature (RT) for 1 hour and washed in 0.2% Tween/PBS. Images were acquired with a Leica DMi8 microscope and the Leica Application Suite LAS X Imaging Software.

The isolation and Analysis of CTCs from Mice Blood Samples have been reported elsewhere [32]. In brief, mice blood samples were collected in RNA Protect Animal Blood Tube and stored at 4 °C. When necessary, the tubes were incubated at room temperature (RT) for 2 hours to achieve complete lysis of blood cells and total RNA was purified using the RNeasy Protect Animal Blood Kit (QIAGEN, Hilden, Germany, Catalog number: 73224). Complementary DNAs (cDNAs) were synthesized using SuperScript™ VILO™ MasterMix (Invitrogen™, Carlsbad, California (CA), USA, Catalog number: 11755-050) and the samples were tested by Real-Time PCR for the expression of typical clinical biomarkers for breast cancer patients, i.e., Krt8, Krt18, Krt19 and c-Met. For primer sequences see Supplementary Data 1.

The data are expressed as means $\pm$ Standard Deviation (SD) of 3 independent experiments. Statistical differences and significance between samples were determined using the Student’s two-tailed t-test or Ordinary one-way ANOVA. A p-value 0.05 was considered statistically significant. The reproducibility, sample sizes and, where appropriate, statistical analyses are described in the Figure Legends.

To explore the DIO2 gene expression, prognosis, and clinicopathological characteristics in human breast cancer, we collected data from The Cancer Genome Atlas-BReast CAncer (TCGA-BRCA, https://portal.gdc.cancer.gov/projects/TCGA-BRCA) Data Portal and performed a bioinformatic analysis of the differential gene expression in normal and primary breast cancer samples, stratified or not in function of specimens and different histopathological tumor stage. We observed that DIO2 was significantly overexpressed in BRCA tumor tissues (Fig. 1a) and enhanced independently of patient’s gender (Fig. 1b), age (Fig. 1c), individual cancer stage (Fig. 1d), BRCA subclass (Fig. 1e), BRCA histologic subtype (Fig. 1f) and presence and/or absence of nodal metastasis status (Fig. 1g). Additionally, to evaluate the prognostic value of DIO2 in breast cancer, we carried out Kaplan-Meier analysis. By dividing the cancer cases into low- and high-DIO2 expression groups according to the DIO2 expression levels, we found that high-DIO2 expression breast cancer group had lower Overall Survival (OS) (HR = 0.97, Log Rank p = 0.86) (Fig. 1h) and Disease-Free Survival (DFS) (HR = 0.79, Log Rank p = 0.57) than those with low-DIO2 levels (Fig. 1i). Conversely, DIO3 gene expression was significantly downregulated in breast cancer tissues and reduced in each above-mentioned condition (Supplementary Fig. 2). These data suggest that the TH modulating enzymes are inversely regulated in breast cancer, advising that TH status might be fine regulated in human breast cancer cells.

Fig. 1.

In-depth the cancer genome atlas program (TCGA) DIO2 gene expression analyses in BReast CAncer (BRCA). (a–g) Differential DIO2 gene expression analysis between normal and BRCA samples (a), stratified in function of patient’s gender (b), patient’s age (c), individual cancer stages (d), BRCA subclasses (e), BRCA histologic subtypes (f) and nodal metastasis status (g) through the online UALCAN Database (https://ualcan.path.uab.edu). DIO2 levels, expressed as transcripts per million (TPMs), are shown using box plots. (h,i) Kaplan-Meier curves for patient’s overall survival (OS) (h) and disease-free survival (DFS) (i) classified by low and high DIO2 gene expression levels in BRCA (OS and DFS median, hazar ration (HR) with 95% confidence interval), performed using the survival plot module of GEPIA2 tool (http://gepia2.cancer-pku.cn). Statistical analyses were computed by the Wilcoxon test. TPM, transcripts per million; TNBC, triple negative breast cancer; IDC, infiltrating ductal carcinoma; ILC, infiltrating lobular carcinoma; INOS, infiltrating carcinoma NOS; N0-N1-N2-N3 are pathologic_N descriptions in UALCAN. In detail: N0: No regional lymph node metastasis; N1: Metastases in 1 to 3 axillary lymph nodes; N2: Metastases in 4 to 9 axillary lymph nodes; N3: Metastases in 10 or more axillary lymph nodes.

To address the functional relevance of DIO2 expression in breast cancer, we genetically depleted the DIO2 gene using CRISPR/Cas9 technology in a noninvasive breast cancer cell line, MCF7. Effective DIO2 depletion in D2KO cells was verified by measuring D2 mRNA levels and enzymatic activity (Fig. 2a,b). Moreover, the intracellular T3 and T3/T4 ratio was potently reduced in the D2KO versus WT cells, confirming that D2KO cells have a lower availability of T3 consequent to DIO2 depletion (Fig. 2c–e). We then analyzed the functional consequences of DIO2 targeted mutagenesis on cell proliferation. A significant enhancement of cell growth was observed after DIO2 genetic depletion in MCF7 cell line compared to control cells (Fig. 2f). Accordingly, colony formation assay confirmed a growth advantage for the DIO2-depleted cells (Fig. 2g,h). Moreover, we found that also the anchorage-independent growth of MCF7 cells was affected after DIO2 depletion, as shown by the enhanced sphere diameters compared to control cells (Fig. 2i and Supplementary Fig. 3a). Analysis of the cell-cycle progression showed that DIO2-depletion altered the cell-cycle phase distribution in asynchronized cells, increasing the population of cells in S-phase and reducing those in G1- and G2-phases. In addition, analysis of cell-cycle progression in synchronized cells confirmed that loss of D2 results in a higher percentage of duplicating cells already 12 h after the release from the arrest of the cell cycle compared to control cells (Fig. 2j and Supplementary Fig. 3b). To evaluate if enhanced cell proliferation was linked to the reduction of TH signal activation, we measured the protein expression of known positive (p21 [Waf1], p27 [Kip1]) and negative (Cyclin-D1) TH target genes involved in the control of the cell-cycle [33, 34, 35]. Compared to control cells, Cyclin-D1 and phospho-ERK expression was mildly increased in D2KO cells, while the expression of cell-cycle inhibitors p21 [Waf1] and p27 [Kip1] and the anti-proliferative phospho-p38 MAPK was reduced, confirming that DIO2 depletion also in the BRCA context increases the expression of a negative TH target gene, Cyclin-D1, and reduces the expression of positive TH target genes, p21 [Waf1], p27 [Kip1] (Fig. 2k and Supplementary Fig. 4a). To further confirm that the D2KO-dependent increase in cell growth was due to the blockage of the D2-mediated T4-to-T3 conversion, we treated D2KO cells with physiological T3 concentrations and observed that T3 treatment rescued the fast-cycling profile of D2KO cells (Supplementary Fig. 4b). Analogously, the treatment of MCF7 control cells with a specific D2 inhibitor, reverse-T3 (rT3) [36], enhanced the proliferation rate of D2WT cells to almost that of D2KO cells (Supplementary Fig. 4c). These findings indicate that DIO2 genetic ablation, reducing the nuclear T4-to-T3 conversion, enhances cell proliferation of MCF7 cells by reducing the expression of genes involved in cell-cycle arrest and fostering the expression of proliferative genes.

Fig. 2.

Lowering of THs intracellular levels through the DIO2 genetic depletion enhances breast cancer cell proliferation. (a) mRNA expression of D2 assessed by real-time PCR in DIO2 knock out (D2KO) and wild type (WT) MCF7 cells. (b) D2 deiodinase assay was performed in D2KO and WT MCF7 total protein cells lysates and expressed as amount of heavy T3 produced per hour using heavy T4 as substrate. (c–e) T3 levels [pmol] (c), T4 levels [pmol] (d) and T3/T4 ratio levels [pmol] (e) produced by the deiodination assay performed in D2KO and WT MCF7 total protein cells lysates. (f) Growth curves of D2KO and WT MCF7 cells. (g,h) Clonogenicity of D2KO and WT MCF7 cells determined by colony formation assay by seeding cells (5.0 $\times$ 10³ cells per well in 6-well plates) and growing for 1 week. Magnification 4$\times$. (i) Clonogenicity of D2KO and WT MCF7 cells determined by 3D-Mammosphere Formation Assay. Scale bars represent 50 µm. Quantification of the relative sphere diameters are shown in Supplementary Fig. 3a. (j) Flow cytometry cell-cycle distribution in asynchronized and synchronized D2KO and WT MCF7 cells at 0, 12, 24, 36 and 48 hours after exit from arrest. (k) Western blot analysis of cell-cycle and regulatory proteins (Cyclin-D1, p21 [Waf1], p27 [Kip1], p-ERK1/2, p-p38 MAPK) in D2KO and WT MCF7 cells. Tubulin was used as loading control. The immunoblot quantifications (Densitometry Analysis) are shown in Supplementary Fig. 4a and the uncutted/unprocessed gels were provided in the Supplementary Source Data file. All the results are shown as means $\pm$ SD from at least 3 independent experiments. p-values were determined by the two-tailed Student’s t-test. THs, thyroid hormones; mRNA, messenger RNA; D2, type II deiodinase; T3, 3,3^′,5-Triiodo-L-thyronine; T4,l-thyroxine; ERK, extracellular signal-regulated kinase; MAPK, mitogen-activated protein kinase; PCR, polymerase chain reaction; MW, molecular weight; ns, not significant. Source data are provided as a Supplementary Source Data file.

Given the enhanced DIO2 expression in human breast cancer progression and the correlation with a poor prognosis in breast cancer patients, we investigated whether the D2-mediated TH activation could be responsible for the invasive conversion of breast cancer cells. Therefore, we evaluated the contribution of the TH signal in promoting the epithelial-to-mesenchymal transition (EMT) in MCF7 breast cancer cells. As shown in Fig. 3a, DIO2depletion reduced the expression of known EMT markers (i.e., N-Cadherin, Vimentin, Zeb-1, Snail, Slug, and Twist), in association with a remarkable induction of the epithelial marker E-Cadherin, thereby resulting in a clear attenuation of the E-Cadherin/N-Cadherin ratio and E-Cadherin/Vimentin ratio at mRNA levels. The same results were obtained by analyzing the protein expression of epithelial and mesenchymal markers by Western Blot and immunofluorescence staining (Fig. 3b,c). To verify whether the activation of the TH signal is a mandatory requisite of the EMT induction in breast cancer cells, we treated D2KO cells with T3 in a time-course and observed that TH treatment rescued the loss of mesenchymal phenotype acquired as a consequence of the genetic DIO2 ablation (Supplementary Fig. 5). As an additional control of TH-dependent EMT induction, we treated MCF7 WT cells with rT3 in a time-course and we observed that the expression of mesenchymal markers was reduced, while the expression of the epithelial marker E-Cadherin was enhanced at both mRNA and protein levels (Supplementary Fig. 6). Overall, these data demonstrate that the genetic DIO2 depletion attenuates the breast cancer EMT, lowering the propensity to acquire an invasive phenotype.

Fig. 3.

Loss of D2 attenuates the THs-dependent epithelial-to-mesenchymal transition (EMT) program. (a) mRNA expression of epithelial (E-Cadherin) and mesenchymal (N-Cadherin, Vimentin, Zeb-1, Snail, Twist) markers and their ratio assessed by real-time PCR in DIO2 knock out (D2KO) and wild type (WT) MCF7 cells. (b) Western blot analysis of EMT markers in D2KO and WT MCF7 cells. Tubulin was used as loading control. The immunoblot quantifications (densitometry analysis) are shown in Supplementary Fig. 5a and the uncutted/unprocessed gels were provided in the Supplementary Source Data file. (c) Confocal images of E-Cadherin, Vimentin and Zeb-1 immunofluorescent staining in D2KO and WT MCF7 cells. Data are presented as higher magnification overviews (representative of 3 images per sample). Magnification 40$\times$. Scale bars represent 10 µm. Relative quantification of E-Cadherin, Vimentin and Zeb-1 immunofluorescent signals (Integrated Density) are represented by histograms. All the results are shown as means $\pm$ SD from at least 3 separate experiments. p-values were determined by two-tailed Student’s t-test. Source data are provided as a Supplementary Source Data file.

To investigate if the D2-mediated TH activation influences the tumor-stroma crosstalk in the breast cancer context, we cultured the D2KO and WT MCF7 cells with Conditioned Media (CM) from Mammary Adipose Tissue-derived Mesenchymal Stromal/Stem Cells (MAT-MSCs) and, in addition, we studied the reverse influence of D2KO and WT MCF7 cells on MAT-MSCs (Fig. 4a). CM from three different patients-derived MAT-MSCs (MAT-MSCs CM) were collected and used to define the effects of stromal secreted soluble factors on the proliferation and migration of D2KO and WT MCF7 cells. MAT-MSCs CM increased cell proliferation of WT MCF7 cells compared to MCF7 cultured alone as control and this effect was enhanced in D2KO MCF7 cells (Fig. 4b and Supplementary Fig. 7a,b). In contrast, the genetic ablation of DIO2 in MCF7 cells did not change the morphology and the proliferation rate of MAT-MSCs, as indicated by the unaltered cell density (Supplementary Fig. 7c). These results indicate that not only the D2-mediated TH activation influences the MCF7 behavior in term of proliferation, but also their responsiveness to the Mesenchymal Stromal/Stem Cells.

Fig. 4.

Attenuation of TH signaling via DIO2-depletion impacts on MCF7/MAT-MSCs crosstalk. (a) Schematic illustrating experimental design. Conditioned Media (CM) from (1) MAT-MSCs and (2) D2KO and WT MCF7 cells were collected and used to study their reverse/reciprocal influence in proliferation and migration assays. (b) Clonogenicity of D2KO and WT MCF7 cells, treated or not with MAT-MSCs CM, was determined by 3D-mammosphere formation assay. Quantification of relative sphere diameters is shown in Supplementary Fig. 7a. Magnification 10$\times$. Scale bars represent 50 or 200 µm. (c) Migration ability of D2KO and WT MCF7 cells, treated or not with MAT-MSCs CM, was determined by transwell migration assay. Magnification 4$\times$. Scale bars represent 50 µm. The overall motility, expressed as the number of migrated cells/fields, is shown using scatter dot plots. All the results are shown as means $\pm$ SD from at least 3 independent experiments. p-values were determined by the two-tailed student’s t-test. Source data are provided as a Supplementary Source Data file.

Next, we evaluated the migration ability of D2KO and WT MCF7 cells conditioned or not with the growth media of MAT-MSCs. Treatment with MAT-MSCs CM increased the migration ability of MCF7 cells compared to own monoculture cells as control and this effect was appreciably attenuated in D2KO MCF7 cells (MCF7 D2WT alone: 121.8 $\pm$ 10.2; MCF7 D2KO alone: 73.5 $\pm$ 10.3; MCF7 D2WT + CM MAT-MSCs: 170.3 $\pm$ 13.6; MCF7 D2KO + CM MAT-MSCs: 114.8 $\pm$ 5.7) (Fig. 4c). When we performed the inverse experiment, i.e. the reciprocal influence of D2KO and WT MCF7 CM on the migration of MAT-MSCs, we observed that the migration of MAT-MSCs was enhanced by the CM of both D2KO and WT MCF7 cells compared to MAT-MSCs alone, with no differences between CM of D2KO and WT MCF7 cells (Supplementary Fig. 7d).

Then, we evaluated the EMT transcriptomic signature from D2KO and WT MCF7 cocultured or conditioned with the medium of MAT-MSCs versus matched D2KO and WT MCF7 monocultured as control (Fig. 5a). Similar to what was observed in D2KO and WT MCF7 comparison (Fig. 3), mRNA expression profile of EMT genes in the MCF7 cocultured with MAT-MSCs indicated that MAT-MSCs enhanced the overall MCF7 cell EMT propriety in dependence of the thyroid status of breast cancer cells. In detail, the expression of the epithelial marker E‑Cadherin in D2KO and WT MCF7 cells cocultured with MAT-MSCs was reduced compared to their corresponding monocultured control group, and concurrently the expression of mesenchymal markers N‑Cadherin and Vimentin and EMT‑associated transcription factors, including Zeb-1, Snail and Twist, had the opposite trend (Fig. 5b). Accordingly, MAT-MSCs CM experiments paralleled the results of coculture assays (Fig. 5c). These results suggest that MAT-MSCs are involved in the promotion of MCF7 cell EMT process in a TH-dependent manner since the thyroid status of breast cancer cells is important to prompt their mesenchymal conversion to facilitate the migration and invasion.

Fig. 5.

Thyroid status of MCF7 cells influences their MAT-MSCs-dependent EMT stimulation. (a) Schematic illustrating experimental design. (1) Co-culture experiments of MAT-MSCs and D2KO or WT MCF7 cells; (2) Conditioned media (CM) from MAT-MSCs was collected and used to treat D2KO and WT MCF7 cells. Both experimental conditions have been used to assay MAT-MSCs influence on breast cancer cell EMT properties. (b,c) mRNA expression of epithelial (E-Cadherin) and mesenchymal (N-Cadherin, Vimentin, Zeb-1, Snail, Twist) markers and their ratio assessed by real-time PCR in D2KO and WT MCF7 cells, co-cultured with MAT-MSCs (b) or treated with MAT-MSCs CM (c). All the results are shown as means $\pm$ SD from at least 3 separate experiments. p-values were determined by ordinary one-way ANOVA. ns, not significant; MAT-MSCs, mammary adipose tissue-derived mesenchymal stromal/stem cells. Source data are provided as a Supplementary Source Data file.

Since MAT-MSCs have emerged for their immunosuppressive properties [37], we assessed how the alteration of DIO2 in MCF7 breast cancer cells might modulate their functional immunomodulatory properties. MAT-MSCs were cocultured or conditioned with the medium of D2KO and WT MCF7 cells, with matched MAT-MSCs monocultured as control, and then the expression of a plethora of immunomodulatory genes, including IL-1$\alpha$, IL-1$\beta$, IL-6, IL-8, CXCR4, BAFF, BAFFR, CCL19, CCL21, SDF1 and VEGFA was evaluated (Fig. 6a). While the CM from WT MCF7 cells efficiently enhanced the expression of the above-mentioned inflammatory genes in comparison to MAT-MSCs alone, both in coculture and CM experiments (Fig. 6b,c), the CM from D2KO MCF7 cells failed to stimulate MAT-MSCs in the expression of these genes. To reinforce the concept that TH acts as a potent endocrine signal driving pro-inflammatory effects, we treated MAT-MSCs with the conditioned medium of D2KO MCF7 cells in presence of physiological T3 concentration in a time-course and observed that T3 treatment significantly rescued the expression of the above-mentioned genes, whose levels were enhanced to almost that of MAT-MSCs conditioned with the medium of MCF7 control cells (Supplementary Fig. 8a). As an additional control of TH-dependent immune-induction of MAT-MSCs, we treated these latter with the conditioned medium of MCF7 control cells in presence of rT3 in a time-course and observed that the rT3-mediated D2 inhibition, blocking the T4-to-T3 conversion, paralleled the results of MAT-MSCs conditioned with the medium of D2KO MCF7 cells (Supplementary Fig. 8b), as indicated by the reduced levels of all the immunomodulatory genes.

Fig. 6.

The raising of THs levels in breast cancer cells acts as a pro-inflammatory mediator in MAT-MSCs immune activation. (a) Schematic illustrating experimental design. (1) Co-culture experiments of MAT-MSCs and D2KO or WT MCF7 cells; (2) Conditioned media (CM) from D2KO or WT MCF7 cells were collected and used to treat MAT-MSCs. Both experimental conditions have been used to assay the influence of the TH status of breast cancer cells on the immunomodulatory properties of MAT-MSCs. (b,c) mRNA expression of inflammatory genes (IL-1$\alpha$, IL-1$\beta$, IL-6, IL-8, CXCR4, BAFF, BAFFR, CCL19, CCL21, SDF1 and VEGFA) assessed by real-time PCR in MAT-MSCs, co-cultured with D2KO or WT MCF7 cells (b) or treated with D2KO or WT MCF7 cells CM (c). All the results are shown as means $\pm$ SD from at least 3 separate experiments. p-values were determined by ordinary one-way ANOVA. ns, not significant. Source data are provided as a Supplementary Source Data file.

In addition, we profiled a panel of cytokines and chemokines released in the CM of D2KO and WT MCF7 cells. Notably, we found a drastic reduction in the secretory profile of D2KO MCF7 cells compared to WT MCF7 cells (Table 1 and Supplementary Fig. 8c). Indeed, D2KO MCF7 cells secreted a lower amount of pro-inflammatory cytokines such as IL-1$\beta$, IL-6, IL-8, IL-10 and, conversely a significantly higher amount of anti-inflammatory antagonists such as IL-1RA and IL-6 R alpha (Table 1 and Supplementary Fig. 8c). Together, this evidence poses the TH to act as a pro-inflammatory mediator involved in the immune induction of MAT-MSCs.

To shed light on the TH influence on the early-stage BRCA malignancy, we used a surgery-free intraductal delivery model, namely the Mouse-INtraDuctal (MIND) injection, as a tool for in vivo characterization of tumor formation and progression. D2KO and WT MCF7 cells (2.0 $\times$ 10⁵ cells in 10.0 µL) were intraductal injected into the fourth abdominal mammary gland of male C57BL/6 mice and cells were allowed to naturally grow (Fig. 7a). Mammary glands were collected 8 weeks post-tumor cell injection, before the development of overt lesions as assessed by palpation, but concurrently with microlesions as confirmed by histological assessment. Hematoxylin/Eosin (H/E) staining showed that both intraductal injections of D2KO and WT MCF7 cells resulted in the formation of solid tumors with necrotic cores, with no evidence of tumor cell leakage into the adjacent adipose tissue (Fig. 7b). Specifically, both D2KO and WT MCF7 MIND tumors were not characterized by an infiltrating fat phenotype but showed a histological profile more comparable to human breast cancers. To confirm whether these tumors were MCF7-derived in situ lesions, we tested them for the expression of E-Cadherin and Vimentin markers by immunofluorescence staining. As shown in Fig. 7c, D2KO and WT MCF7 MIND cells, when xenografted to the mouse milk ducts, recapitulated own in vitro EMT features. Differently from WT MCF7-derived in situ MIND, D2KO MCF7-derived in situ MIND was characterized by higher expression of the epithelial marker E-Cadherin and a reduced expression of the mesenchymal marker Vimentin. Thus, the in vivo MIND model confirmed what was observed in in vitro data, that DIO2 depletion and the concurrent T4-to-T3 reduction represent a crucial barrier to counteract breast progression toward invasiveness and metastasis.

Fig. 7.

In vivo characterization of THs-dependent BRCA tumor formation and progression. (a) Schematic representation of Mouse-INtraDuctal (MIND) injection. The experimental protocol is depicted in the panel. (b,c) Hematoxylin/Eosin (H/E) (b) and immunofluorescent staining with epithelial (E-Cadherin) and mesenchymal (Vimentin) markers (c) of D2KO and WT MCF7 MIND tumors (representative of 3 images per sample, n = 6 mice/group). Magnification 20$\times$. Scale bars represent 50 µm. (d) mRNA expression of four metastatic biomarkers (Krt8, Krt18, Krt19 and c-Met) was measured by real-time PCR in circulating tumor cells isolated from blood samples of D2KO and WT MCF7 xenografted MIND mice (n = 8 mice/group). All the results are shown as means $\pm$ SD from at least 3 separate experiments. p-values were determined by the two-tailed Student’s t-test. Source data are provided as a Supplementary Source Data file.

To keep track of the expansion of breast cancer to distant sites after intraductal engraftment, we processed the peripheral blood of D2KO and WT MCF7-MIND mice in order to detect and separate the fraction of cells sloughing off the main tumor and extravasing into the blood, namely the Circulating Tumor Cells (CTCs). After collecting blood samples, the expression of four specific markers, i.e. Krt8, Krt18, Krt19, and c-Met, whose transcriptomic levels are a typical signature of BRCA progression in the blood of early breast cancer patients [38, 39], was analyzed. Interestingly, the above-mentioned breast cancer-related markers were expressed at higher levels in WT MCF7-MIND mice than D2KO MCF7-MIND mice (Fig. 7d).

In conclusion, these data, not only provide new insights into the complex role of TH in tumor biology but represent evidence that TH critically regulates BRCA progression and its dialogue with the surrounding milieu. Consequently, interfering with the TH signal might be a valid strategy for anti-cancer therapies.

Although different data in literature have described that the TH-activating enzyme, D2, is up-regulated in advanced human cancers, still few data are available on the expression of D2 in BRCA, and its association with the clinicopathological characteristics of this tumor. By performing a wide in silico analysis of different datasets, we described a potent correlation of high DIO2 expression with poor survival rate for the patients. These data suggest that D2 plays a significant role in aggressive tumors and in the promotion of metastatic potential in BRCA. Moreover, they indicate D2 as a potential useful prognostic marker for patients. Our in vitro and in vivo experiments confirm the in silico data, indicating that D2 inactivation in MCF7 cells drastically attenuates their EMT potential while boosting cell proliferation. The central role of TH in boosting MCF7 invasiveness was further proved by the rescued invasive phenotype of D2KO MCF7 cells treated with T3, as well as by the drastically reduced expression of EMT marker genes induced by treatment of WT MCF7 cells with rT3. Noteworthy, the in vivo experiments, based on the intraductal injection of D2KO and WT MCF7 cells, reinforced such a role for D2, since lesions formed by injection of D2KO cells were characterized by an epithelial phenotype, while EMT genes were drastically down-modulated in comparison with lesions formed by the injected WT MCF7 cells.

In BRCA, chemotactic factors released by cancer cells, which include matrix metalloproteinases, growth factors, and inflammatory cytokines, attract MSCs to the site of tumorigenesis [42]. Furthermore, various tumors, BRCA included, are stimulated by MSCs to grow and progress [43]. In the BRCA context, endogenous MSCs exert powerful pro-tumorigenic activities within the tumor stroma enhancing fibrovascular desmoplasia, tumor growth and metastasis [44, 45]. In detail, breast cancer cells stimulate the MSC to secrete different chemokines thereby accelerating the conversion of cancer cells toward invasion and malignancy. Although the stimulation of MSCs has been investigated, the molecules regulating the mechanism by which they crosstalk with the tumor cells have yet been poorly elucidated. Here, we observed that a functional TH activation mediated by D2 is strictly required not only for the intrinsic cancer biology properties of MCF7 cells, but also for their sensitivity to the MSCs-mediated pro-invasive conversion. Noteworthy, our previous works have indicated that the Deiodinases-mediated regulation of T3 availability deeply impacts cancer formation and progression with dynamic and opposite effect on tumor formation and invasion. Indeed, while the D3-mediated THs inactivation occurs in the early phases of tumorigenesis and in the amplification phase of Cancer Stem Cells (CSCs) [46, 47, 48], the subsequent D3 down-regulation and D2 up-regulation enhance the intracellular activation of T4-to-T3 and thus boosts tumor invasiveness and metastatic conversion [15, 27, 49]. In the light of these evidence, the observed opposite effects of D2 inactivation on MCF7 proliferation and EMT can be ascribed to the ability of THs to reduce cell proliferation while promoting tumor progression. Differently from the above studies, the present work addresses the intricate role of THs in the Tumor Micro-Environment (TME) and in the interplay of cancer cells with the surrounding milieu. Interestingly, we observed that D2 is produced by MCF7 and catalyzes the T4-to-T3 conversion within cancer cells. This in turn amplifies the ability of MSCs to stimulate MCF7 cells and also enhances the sensitivity of MCF7 cells to MSCs conditioning. Blocking D2 interferes with such a dialogue and is thus a valid tool to counteract the crosstalk within the TME.

Malignant evolution of tumors is often triggered by a strong inflammatory response, in which the TME plays a critical role in fostering EMT of cancer cells and secreting a variety of pro-tumorigenic cytokines. While it is generally acknowledged that increased THs are important modulators of inflammation and immune activities [50], still controversial functions have been described on the role of THs in inflammation, probably due to the pleiotropic properties of these hormones and their context-dependent differential effects. Moreover, even less is known about the role of THs as diffusible molecules in the inflammatory properties of the TME. Importantly, we observed that the MSCs-dependent production of a plethora of pro-inflammatory genes, exacerbated by the MCF7-MSCs interplay, was drastically reduced when D2 was inhibited in MCF7 cells. Indeed, our data show that MSCs cultured with MCF7 amplified the inflammatory gene expression, but such effect was lost when MSCs were cultured with D2KO MCF7, and this impairment was completely rescued by the addition of exogenous T3 to the co-cultures.

In conclusion, the current work adds different novel insights into the complex role of THs and its intracellular activation in BRCA. All the data point to a potent effect of THs in exacerbating the tumor invasiveness. These observations are in line with clinical data showing that BRCA is increased in hyperthyroidism [51, 52], and that stage I ER⁺ BRCA patients treated with THs for the management of hypothyroidism have worse outcomes [53]. Furthermore, it has been also shown that, acting via non-genomic mode, THs foster BRCA formation and invasion [54, 55]. In light of the above considerations, we conclude that the intratumoral-activated TH can increase the aggressiveness of BRCA and that the inhibition of the D2 enzyme can be viewed as a valid tool to attenuate BRCA progression. Moreover, our data strengthen the concept that elevated THs levels in hyperthyroid patients can worsen the cancer risk, suggesting that treatment with THs in patients with both hypothyroidism and BRCA should be finely monitored to ascertain the THs levels in the physiological range.