Extracellular Superoxide Dismutase Expression in Papillary Thyroid Cancer Mesenchymal Stem/Stromal Cells Modulates Cancer Cell Growth and Migration

Tumor stroma-secreted growth factors, cytokines, and reactive oxygen species (ROS) influence tumor development from early stages to the metastasis phase. Previous studies have demonstrated downregulation of ROS-producing extracellular superoxide dismutase (SOD3) in thyroid cancer cell lines although according to recent data, the expression of SOD3 at physiological levels stimulates normal and cancer cell proliferation. Therefore, to analyze the expression of SOD3 in tumor stroma, we characterized stromal cells from the thyroid. We report mutually exclusive desmoplasia and inflammation in papillary and follicular thyroid cancers and the presence of multipotent mesenchymal stem/stromal cells (MSCs) in non-carcinogenic thyroids and papillary thyroid cancer (PTC). The phenotypic and differentiation characteristics of Thyroid MSCs and PTC MSCs were comparable with bone marrow MSCs. A molecular level analysis showed increased FIBROBLAST ACTIVATING PROTEIN, COLLAGEN 1 TYPE A1, TENASCIN, and SOD3 expression in PTC MSCs compared to Thyroid MSCs, suggesting the presence of MSCs with a fibrotic fingerprint in papillary thyroid cancer tumors and the autocrine-paracrine conversion of SOD3 expression, which was enhanced by cancer cells. Stromal SOD3 had a stimulatory effect on cancer cell growth and an inhibitory effect on cancer cell migration, thus indicating that SOD3 might be a novel player in thyroid tumor stroma.

In solid tumors, paracrine factors secreted from the stroma regulate cancer cell growth and migration [1][2][3][4][5][6][7][8][9] . Reactive oxygen species (ROS), a well-known paracrine factor, contribute to stromal myofibroblast maturation 10 , thus emphasizing the effect of ROS in tumorigenesis. Extracellular superoxide dismutase (SOD3) has anti-oxidative, anti-inflammatory, anti-apoptotic, and growth promoting characteristics, exhibiting the most potent therapeutic responses and growth regulatory characteristics in cardiovascular and cancer models [11][12][13][14][15][16][17][18][19][20][21][22] . The expression of SOD3 is increased in a benign thyroid tumor goiter model and gradually downregulated in cell lines that model advanced papillary and anaplastic thyroid cancers correlating with the level of oncogene activation 23,24 . Of note, downregulation of growth stimulating SOD3 in epithelial cancer cells is controversial, particularly in light of recent data demonstrating SOD3-driven immortalization and even the transformation of murine primary cells, hence suggesting abrogation of the growth advantage in cancer cells 23-28 . In the current study, we describe mesenchymal stem cells (MSCs) isolated from non-carcinogenic thyroids (Thyroid MSCs) and papillary thyroid cancer (PTC MSCs), the latter showing desmoplastic characteristics. Importantly, a redox gene expression analysis showed downregulation of SOD3 in papillary thyroid cancer TPC1 cells compared to Nthy control cells and upregulation in PTC MCS compared to Thyroid MSCs, hence suggesting autocrine-paracrine conversion of SOD3 mRNA expression. A functional analysis of stromal secreted SOD3 corroborated previously published data 20,26 showing increased cancer cell proliferation and decreased cell migration in co-culture. Therefore, our data suggest that the growth-promoting characteristics of SOD3 are not limited to the initial benign growth phase of tumorigenesis but are sustained to the end phase of tumor development.

Histological analysis of papillary thyroid cancer and follicular thyroid cancer stroma sections.
In thyroid cancers, desmoplastic stromal reactions, which correlate to lymph node metastasis, are a relatively common early phenomenon present in up to 80% of medullary thyroid cancers 29 . Characterization of papillary (PTC) and follicular (FTC) thyroid cancers 12 out of 20 cases (60%) demonstrated fibrosis or mononuclear cell infiltration. In PTC 40% of tumors showed desmoplastic regions and 30% inflammatory regions, whereas 40% of PTC tumors showed no detectable changes in stroma. In one case (10%) the stroma contained both desmoplastic and inflammatory regions. Interestingly, 50% of the cases suggested mutual exclusion between fibrosis and inflammation ( Fig. 1A-D and F-I). The analysis of FTC showed desmoplasia or mononuclear cell infiltration in 8 out of 10 cases (80%). In seven cases (70%) there was mutual exclusion between fibrosis and inflammation: in five cases (50%) there was desmoplasia without inflammation and in two cases (20%) there was increased inflammation without fibrosis. In two cases (20%) there was no desmoplasia or inflammation, and in one case (10%) FTC stroma showed both desmoplasia and increased mononuclear cell content (Fig. 1E,J).

Mesenchymal stem cells from thyroid and papillary thyroid cancer.
Most of the tissues have been suggested to contain multipotent mesenchymal stem/progenitor cells that support tissue renewal and function as a source of cytokines and growth factors 30 . To study the presence of MSCs in papillary thyroid cancer and a non-carcinogenic thyroid tissue counterpart, we isolated plastic adherent mesenchymal cells and characterized their phenotype. To test the stemness of the isolated cells, adipocyte, chondrocyte, and osteocyte lineage differentiation assays were performed to define the multipotency of the isolated cells. We observed similar multipotency among the bone marrow derived MSCs (BM MSC), the Thyroid MSCs, and the PTC MSCs, indicating the presence of mesenchymal stem cells in the adult thyroid tissues and papillary thyroid cancer tumor tissues ( Fig. 2A-C). The analysis of the expression of cell surface cluster of differentiation (CD) molecules used to identify MSCs by flow cytometer suggested no differences among the BM MSCs, Thyroid MSCs, and PTC MSCs  PTC MSCs have fibrotic characteristics. Because our data showed the presence of fibrosis in papillary thyroid cancer, we studied desmoplastic molecular marker expression from Thyroid MSCs and PTC MSCs. Strikingly, in all five patients, PTC MSCs had higher FIBROBLAST ACTIVATING PROTEIN (FAP) mRNA expression compared to Thyroid MSCs. Similarly, PTC MSCs had significantly higher COLLAGEN 1 TYPE A1 (Col1 A1) and TENASCIN mRNA expression levels as compared to Thyroid MSCs in four out of five and in three out of five patients, respectively. Although we observed variation in fibrotic marker expression in the patient 1 (TENASCIN expression was at the level of controls), in the patient 2 (TENASCIN expression lower than in the controls), and in the patient 3 (Col1 A1 expression was at the level of controls) our data could suggest that mesenchymal stem cells may function as an origin for myofibroblasts ( Fig. 3A-O).
SOD3 expression is reduced in epithelial cancer cells and is increased in MSCs with fibrotic characteristics. Thyroid cancer progression and fibrosis correlate with increased reactive oxygen species (ROS) production and increased redox enzyme expression 23,31 . Hence, we investigated the gene expression of the superoxide dismutase SOD1-3 family and the NOX1-5 family from human immortalized Nthy thyroid cells modeling a normal thyroid and from papillary thyroid cancer patient-derived TPC1 cells. Comparable to our previous data 26 , we observed increased SOD2, NOX1, NOX2, and NOX5 mRNA expression in papillary thyroid cancer cells compared to Nthy cell line. Importantly, we found decreased expression of SOD3 recurrently with recent works 23,26 and an absence of NOX4 expression (Fig. 4A).
The redox gene expression patterns of Thyroid MSCs compared to PTC MSCs were similar to the expressions observed in thyroid epithelial Nthy and TPC1 cells, with the exceptions of SOD3 and NOX4. The Paracrine MSC secreted SOD3 increases epithelial cancer cell growth. According to previous data, BM MSCs could reduce cancer cell proliferation via the paracrine system, thereby inhibiting tumorigenesis 32 . On the contrary, tumor-associated stromal cells, such as tumor-associated fibroblasts and mesenchymal stem cells, have been shown to promote cancer cell proliferation and tumor growth 33,34 . Therefore, we tested the growth stimulating properties of Thyroid MSCs and PTC MSCs on the papillary thyroid cancer TPC1 cell line. In line with previous reports, a BrdU DNA incorporation assay for TPC1 papillary thyroid cancer cells cultured together with Thyroid MSCs suggested decreased DNA replication, whereas cancer cells grown in the presence of PTC MSCs showed significantly increased BrdU incorporation (Fig. 5A,B).
To probe the paracrine effect of SOD3 expression on cancer cell growth, we over-expressed SOD3 in Thyroid MSCs (Thyroid MSC SOD3) and silenced the gene expression by SOD3 RNAi in PTC MSCs (PTC MSC SOD3 RNAi) (Fig. 5C,D). Co-culture of TPC1 cells with Thyroid MSC SOD3 suggested significantly (p < 0.01) increased DNA replication of TPC1 cells compared to the parental Thyroid MSCs (Fig. 5E,F). Correspondingly, RNAi SOD3 treatment of PTC MSCs decreased (p < 0.001) TPC1 DNA replication compared to parental PTC MSCs (Fig. 5G,H), thus demonstrating that PTC MSCs may support cancer cell growth by secreting SOD3.
SOD3 has been demonstrated to regulate the activation of a large number of growth-related signaling molecules 14,27 . We then studied the growth factor gene expression in Thyroid MSC SOD3 and PTC MSC SOD3 RNAi cells. The analysis revealed correlations among SOD3, FIBROBLAST GROWTH FACTOR 9 (FGF9), and BONE MORPHOGENIC PROTEIN 2 (BMP2) expression (Fig. 5I,J).
Paracrine effect of SOD3 on epithelial cancer cell migration. Desmoplastic thyroid tumors frequently metastasize to the lymph nodes 29 , suggesting increased intratumoral cancer cell migration toward the lymphatic vasculature. In the current work, the Matrigel migration analysis demonstrated significantly (p < 0.001) increased TPC1 migration levels towards Thyroid MSCs compared to PTC MSCs (Fig. 6A), which may indicate a lower affinity of TPC1 cells to tumor stroma. The analysis of the effect of SOD3 on TPC1 cell migration showing significantly (p < 0.01) reduced migration toward SOD3 over-expressing Thyroid MSCs (Thyroid MSC SOD3) (Fig. 6B) could suggest regulation of local cancer cell migration by paracrine secretion of the enzyme. However, silencing of SOD3 in PTC MSCs (PTC MSC SOD3 RNAi) failed to show increased cancer cell migration (Fig. 6C), indicating that SOD3 might have a different function in fibrotic MSCs or that the activation process of PTC MSCs modifies the ROS balance, hence altering the effect of single redox genes.
Previous publications have proposed that inflammatory and cancer cell migration are regulated by the same chemoattractants 35,36 . We recently demonstrated SOD3-derived decreased expression of TNFα, IL1α, IL6, MIP2, and MCP-1 inflammatory cytokines in cardiovascular models with consequent inhibition of monocyte/macrophage migration 17,20,37 . The expression analysis of IL1α, MCP-1, and IL8 showed reduced cytokine expression in PTC MSCs compared to Thyroid MSCs (Fig. 6D-J). The data, however, suggest that only IL1α and MCP-1 are downregulated by SOD3 in stromal cells (Fig. 6D,E), whereas IL8 expression is independent of the enzyme (Fig. 6F). Western blot analysis for MCP-1 (Fig. 6G) and flow cytometer analysis for IL8 (Fig. 6H,I)

Discussion
Local vascular permeability with consequent extravasation of fibrinogen and plasminogen initiate the formation of fibrin gel deposits, the early form of tumor stroma, which attracts migrating fibroblasts, epithelial cancer cells and inflammatory cells. Importantly, continuous tumor growth creates a demand for continuous stroma expansion. Thus, tumor stroma has different developmental stages that affect cancer cells, regulating their proliferation and local migration. According to recent data, the development of stroma is connected to the progression of carcinogenesis, such as lymph node metastasis 39 , indicating that stroma responds to the microenvironmental needs of epithelial cancer cells through paracrine growth factor secretion and direct cell-to-cell contact [40][41][42] . Additionally, activated desmoplastic stroma, which is frequently observed in medullary thyroid cancers, is used as a clinical intraoperative diagnostic marker to characterize tumor progression and to predict lymph node metastasis 29,39,43,44 . In the present work, we observed a marked inverse correlation between desmoplasia and inflammatory cell infiltration in PTC and FTC: 12 out of 20 patients had increased desmoplasia or increased mononuclear cell infiltration, whereas only two out of 20 patients had simultaneous desmoplasia and inflammation (Fig. 1). Previous articles have suggested that aggressive cancers, such as pancreatic ductal adenocarcinoma characterized by an excessive stromal reaction, may have reduced intratumoral vascularization 45 that could cause reduced mononuclear cell migration into desmoplastic tumors. However, in a recent work the immunohistochemical analysis of fibroblast activating protein α -positive medullary thyroid carcinomas showed no correlation between new vessel formation and stromal fibrosis 29 , thus indicating other mechanisms than the physiological fibrotic barrier in the inhibition of cell migration. Inflammation has also been suggested to precede fibrosis development: infiltrated inflammatory cells secrete profibrogenic cytokines that affect fibroblast activation, perivascular fibrosis development, and cellular migration 46,47 .
An interesting, although rare, characteristic of thyroid tumors is intratumoral heterotopic ossification [48][49][50][51][52][53][54] , which may suggest the presence of multipotent mesenchymal stem/stromal cells in thyroid tissue. Indeed, tumor-associated MSCs recently characterized in ovarian carcinoma and prostate cancer have similar phenotype and functionality as BM MSC but altered growth factor expression profile 55,56 . The current data, which demonstrate the presence of MSCs in non-carcinogenic thyroid and papillary thyroid cancer with a similar differentiation capacity and CD marker expression as BM MSCs (Fig. 2) strengthening the previous observations, could further indicate that MSCs maintain their stemness even in conditions where they are under constant external carcinogenic stimuli. However, a more accurate gene expression analysis of fibrotic markers in Thyroid MSCs and PTC MSCs (Fig. 3) showed differentiation stage-related differences that affected functional properties, such as proliferation (Fig. 5A) and migration (Fig. 6A), in neighboring cells. We demonstrated that PTC MSCs had increased expression of fibrotic markers (Fig. 3). Although MSC cultures in general are heterogeneous containing cells with various degrees of differentiation, the current data may suggest that tumor stroma MSCs exposed to ROS have been demonstrated to contribute to carcinogenesis and myofibroblast differentiation. Interestingly, NOX4 expression, which is absent in thyroid cancer cells 26 (Fig. 4), was observed in MSCs, which may suggest a paracrine role for the protein in the thyroid. Furthermore, SOD3 expression suggested autocrine-paracrine conversion from epithelial cancer cells to mesenchymal stromal cells (Fig. 4), which could shed light on the function of SOD3 in tumorigenesis. Previous data have shown increased SOD3 expression in an experimental rat thyroid goiter model 23 with corresponding growth-supporting function at physiological expression levels 14,18,25,26,28,57 . A strong connection of SOD3 to growth regulation is caused by positive a feedback loop between the enzyme, small GTPase activation, and RAS-ERK1/2 signaling 24,25,27,28 . Moderately increased small GTPase RAS activation stimulates SOD3 mRNA synthesis, but at aberrant levels of RAS activity (> 10-fold RAS activation level), there is a sudden decrease in SOD3 expression 24 . Therefore, we have hypothesized that SOD3 benefits the initial phase of tumorigenesis at low (< 10-fold) RAS activation levels. Once SOD3 expression reaches non-physiological toxic levels, cancer cells are programmed to downregulate autocrine SOD3 production and instead stimulate stromal paracrine SOD3 secretion simultaneously with activation of stroma, as shown in Figs 3 and 4. Paracrine secretion of SOD3 then continues to support cancer cell growth (Fig. 5) and, correspondingly, decreases the affinity of cancer cells to tumor stroma by decreasing chemotactic cytokine expression (Fig. 6), allowing local intratumoral cancer cell migration.
In conclusion, based on the current data, PTC MSCs with a fibrotic fingerprint might function as a source of myofibroblasts. Importantly, clarification of the mechanisms of fibroblast activation could result in the discovery of novel drug targets that are specific to thyroid cancer 58 or cancer stroma. ROS, although forming a challenging molecular ensemble, control growth-and migration-related signaling routes that potentially contain drug targets. The current results suggest the autocrine-paracrine conversion of SOD3 expression, corroborating the growth supportive nature of the enzyme throughout the carcinogenic process. Thus, the identification of SOD3-coordinated signal transduction in tumor stroma and epithelial cancer cells may reveal small drug target molecules that could be used in combination therapy for the treatment of thyroid cancer.

Methods
Histological analysis. For histological analyses, papillary and follicular thyroid carcinoma samples were fixed in formalin, embedded in paraffin, and cut into 4-5 μ m sections with a microtome. Sections were de-waxed in xylene for 10 minutes, rehydrated by sequential immersion in decreasing ethanol concentrations. The sections were then stained with hematoxylin-eosin and analyzed using an optical microscope (Nikon ECLIPSE Ni, Nikon Instrument Europe B.V. Amsterdam, Netherlands). Two independent observers (F.C. and F.R.) examined the specimens. Isolation and culture of stromal cells. Thyroid mesenchymal stem/stromal cells (MSCs) were isolated from non-malignant sites (Thyroid MSCs) and papillary thyroid cancer sites (PTC MSCs) of patients. De-identified thyroid tissue samples were obtained from patients who underwent thyroid surgery at the University of Wisconsin-Madison. The protocol was approved by the Health Sciences Institutional Review board (IRB) of the University and performed in accordance with the relevant guidelines and regulations 59 . Informed consent was obtained from all subjects according the requirements of University of Wisconsin-Madison protocol permission. Due to differences between individuals all analysis of the study were done comparing thyroid MSCs and PTC-MSCs of the same patient. Tumor tissue was washed several times with PBS, cut into small pieces, and then was plated in alpha MEM (Mediatech, Manassas, VA, USA) supplemented with 10% FBS (Hyclone, Logan, UT), non-essential amino acids (Mediatech) and L-alanine-l-glutamine (Life Technologies, Grand Island, NY, USA). After reaching 70-90% confluency, cells were harvested using TrypLE (Life Technologies) and passaged until reaching passage 4 before characterizing and using them in the study 60 . The MSCs were transduced with SOD3 and RNAi SOD3 lentivirus (MOI 10) (Dharmacon, Lafayette, Colorado, United States).
Gene expression analysis. RNA was isolated using an RNeasy minikit (Qiagen, Hilden, Germany). The cDNA synthesis was performed by QuantiTect Reverse Transcription (Qiagen). SYBR Green PCR master mix (Applied Biosystems, Foster City, CA, USA) was used for qPCR. The primers are listed in the Supplementary  Table T1.

Growth analysis.
To characterize DNA replication, 10 mM bromodeoxyuridine (BrdU) (Roche, Basel, Switzerland) was added to the growth medium for 15 min. Subsequently, the cells were washed three times with PBS, fixed in 3% paraformaldehyde (Sigma) for 20 minutes at − 20 °C, washed three times with PBS, and stained using anti-BrdU antibody (Roche) primary antibody, FITC-conjugated secondary antibody (Jackson ImmunoResearch Laboratories Inc., West Grove, PA, USA), and Hoechst nuclear stain (Sigma).

Migration analysis.
For the migration assay, 100 μ l of Matrigel (Corning, Corning, NY, USA) at 1 mg/ml was added to the migration chamber (8 microns, BD, San Jose, CA, USA) and allowed to stabilize at room temperature for 30 minutes. The chambers were moved to MSC cultures into 12-well plates, 50 000 TPC1 cells were added to the Matrigel and allowed to migrate for 24 hours at 37 °C. After incubation, the Matrigel was removed from the chamber, and the migrated cells were fixed with 7% paraformaldehyde (Sigma), washed with PBS, and stained with Cristal violet (Sigma). The migrated cells were counted from the high-power microscope fields.

Statistical analyses.
The experiments were repeated at least three times. All results are expressed as the mean ± SD. The p-values (*p < 0.05, **p < 0.01, ***p < 0.001) were determined by two-tail independent samples t-tests.