FAM83B is involved in thyroid cancer cell differentiation and migration

FAM83B has been recently identified as an oncogene, but its role in thyroid cancers (TC) is still unclear. We examined the expression of FAM83B and its possible involvement in cell migration and differentiation, in neoplastic/normal thyroid tissues and in TC human cell lines. FAM83B expression in TC varies according to the tumor histotype, being significantly downregulated in more aggressive and metastatic tissues. FAM83B levels in cell lines recapitulate patients’ samples variations, and its total and cytoplasmic levels decrease upon the induction of migration, together with an increase in its nuclear localization. Similar variations were detected in the primary tumor and in the metastatic tissues from a follicular TC. FAM83B knock down experiments confirmed its role in thyroid differentiation and cell migration, as demonstrated by the reduction of markers of thyroid differentiation and the increase of the mesenchymal marker vimentin. Moreover, the silencing of FAM83B significantly increased cells migration abilities, while not affecting the oncogenic RAS/MAPK/PI3K pathways. Our data indicate for the first time a role for FAM83B in TC cell differentiation and migration. Its expression is reduced in dedifferentiated tumors and its nuclear re-localization could favour distant migration, suggesting that FAM83B should be considered a possible diagnostic and prognostic biomarker.

www.nature.com/scientificreports/ Consistently, patients with low FAM83B levels tended to have a shorter disease-free survival (DFS) rate than those with high FAM83B levels (Log rank 0.151) (Fig. 1C). When considering PTCs separately (n = 20), DFS was not significantly different between cases with low or high FAM83B expression (P = 0.224), possibly due to the low statistical power of the sample.
FAM83B mRNA expression profile in normal thyroid tissue, primary thyroid cancers and metastases. Since a huge FAM83B expression variability was observed in TC group, we decided to analyze separately the 34 tumors with different tumor histotypes as follows: 5 follicular thyroid cancers (FTC), 20 papillary thyroid cancers (PTC), and 9 poorly differentiated/anaplastic thyroid cancers (PDTC/ATC). FAM83B mRNA was higher in FTCs compared to either NTs, or PTCs, or PDTC/ATCs (p = ns, p < 0.05 and p < 0.001, respectively). On the other hand, FAM83B levels were constantly lower in PDTC/ATCs with respect to NTs (p < 0.001). PTCs were downregulated with respect to NTs, when considering the median levels of FAM83B expression, but a huge variability was recorded ( Fig. 2A). Immunohistochemistry experiments showed that FAM83B protein levels parallel mRNA levels, with a positive cytoplasmatic staining being detected in normal thyroid tissue and in FTC samples, but not in PTC and PDTC cases ( Supplementary Fig. 1). Interestingly, among PTCs, we found that the 2 RET/PTC1 mutated cases and 1 out of 2 RET/PTC3 cases had expression levels higher than the median, while BRAF mutated and wild type cases were at the median or below the median levels (Fig. 2B). For some patients we had the opportunity to evaluate the FAM83B expression in paired PTC/ NT tissues and we found that FAM83B mRNA is significantly lower in tumors than in contralateral normal thyroid tissues (p = 0.0046) (Fig. 2C), consistent with data extracted from the TCGA THCA database (Fig. 2D). Considering FTC and PTC tissues and their unpaired metastases (FTC-and PTC-derived MTS), we observed that FAM83B mRNA level were significantly lower in FTC-derived MTS compared to primary FTCs (p < 0.05), whereas the decrease observed between PTCs and PTC-derived MTS was not significant (Fig. 2E,F).

FAM83B levels in human thyroid cell lines recapitulate patient samples variations and significantly change upon cell migration.
In order to get more insights into FAM83B mRNA levels variations found in the different thyroid cancer subtypes, we investigated its expression in cell lines derived from healthy and neoplastic thyroid tissues. Preliminary western blot experiment performed on 8 different cell lines derived from healthy tissue (NTHY-ORI), PTCs (TPC1 and K1), PDTC/ATC (SW579 and SW1736) and metastasis (HTCC3 and FTC133) showed that FAM83B levels in the different cell lines replicate those observed in patients of the present series and in the TCGA cohort ( Supplementary Fig. 2). In this context, paralleling the above reported findings in tissues, FAM83B expression was lower in K1 cells, which harbor a BRAF mutation, than in TPC1 cells, which carry a RET/PTC fusion. We selected one representative cell line for each group for the following experiments. Real-time qRT-PCR experiments showed a variability in FAM83B mRNA levels among the different cell lines. In particular, NTHY-ORI (derived from normal tissue) and TPC1 (derived from PTC with RET/ PTC1 fusion) cells expressed high levels of FAM83B, while SW1736 (derived from ATC) and FTC133 (derived from an FTC metastasis) expressed low levels (Fig. 3A). Western blot experiments revealed that FAM83B protein levels vary accordingly to the mRNA ones (Fig. 3B). These cells lines are known to harbor different migration ability and sensitivity to contact inhibition. Namely, NTHY-ORI have high motility in a non-confluent state and a low motility when confluent, TPC1 have a low motility either in non-confluent or confluent state, whereas SW1736 and FTC133 always harbor high motility consistent with their aggressive features [23][24][25][26][27][28][29] . Thus, we investigated if the differences detected in FAM83B expression were due to differences in cell movement and migration. Scratch assays demonstrated that all cell lines were able to migrate with similar abilities, with the exception of TPC1 cells, that showed low motility (Fig. 3C,D). Moreover, Western blot experiments showed that FAM83B protein levels decreased in all migration induced cells, again with the exception of TPC1 cells (Fig. 3E).
FAM83B subcellular localization varies during cell migration. We then investigated the subcellular localization of FAM83B in quiescent and migrating cells by direct immunofluorescence. All cell lines had a diffuse cytoplasmic signal and, in agreement with the above reported Western blot data, the signal was stronger in quiescent TPC1 and NTHY-ORI, while it was extremely low in SW1736 and FTC133 cells (Fig. 4A,B, Supplementary Fig. 3). Interestingly, the induction of cell migration is accompanied by a reduction in the FAM83B cytoplasmic signal (Fig. 4A,C, Supplementary Fig. 3), strictly correlated to the ability to move. In particular, the mobility and the lack of contact inhibition associated to a significant decrease in FAM83B positivity in NTHY-ORI and in SW1736 cells. On the other hand, not significant reductions were seen either in TPC1 cells (due to the above-mentioned low motility) or in and FTC133 (which already displayed very low positivity due to their high motility). The decrease in cytoplasmic FAM83B was accompanied by an increase in the nuclear localization in all migrating cells (Fig. 4D,E).
To confirm at the tissue level the results on migration and subcellular localization, we analyzed FFPE samples from a patient affected with metastatic and aggressive FTC. In particular, we obtained the primary tumor and two neck muscle metastases obtained from different surgeries at 6.5 months and 1.5 year after thyroidectomy. In the primary tissue, we observed a positive cytoplasmic FAM83B staining which was progressively lost moving towards the more undifferentiated area (Fig. 5). The first local muscle metastasis showed a positive cytoplasmic staining in the center, and a nuclear expression in the peripheral area, while the second local muscle metastasis showed a homogeneous positive nuclear staining for FAM83B (Fig. 5).
FAM83B silencing reveals its role in thyroid cancer cell differentiation and migration. To better elucidate the role of FAM83B in the maintenance of thyroid cell differentiation and in cell migration, we performed silencing experiments using two different siRNAs 11  www.nature.com/scientificreports/ which expressed high FAM83B levels. We found that the knock down of FAM83B significantly affect normal and tumor thyroid cell differentiation, inducing in NTHY-ORI a reduction in the expression of the thyroid marker genes PAX8 and NIS (Fig. 6A,B), and in TPC1 cells a significant increase in the expression of the mesenchimal marker vimentin (Fig. 6C,D). On the other hand, the silencing of FAM83B did not affect the oncogenic RAS/ MAPK/PI3K pathways, as the phosphorylation levels of two of the main downstream effectors, ERK and AKT, were unchanged (Fig. 6E). Finally, the FAM83B silencing significantly increased the migration abilities of TPC1 cells, that reached levels similar to those observed in SW1736 and FTC133 cells (Fig. 6F).

Discussion
FAM83B has been recently identified as an oncogene involved in the development and progression of several human cancers 1,9,10,12,13,30 , and is one of eight members of a protein family (FAM83A-H) characterized by a highly conserved domain necessary and sufficient to drive transformation 2 . To date, the role of FAM83B in thyroid cancers is unclear, though Cipriano and coauthors reported its overexpression in 17 thyroid tumors from a series of different human cancers 1,22 . In the present study, we found a positive correlation between low FAM83B mRNA levels and thyroid cell dedifferentiation. Indeed, FAM83B was downregulated in TCs and metastatic samples compared to normal thyroid tissues, and concordant data were obtained by the analysis of the TCGA THCA database (https:// www. cancer. gov/ tcga accessed on May 2020). Consistently, a decrease in FAM83B expression was also found in the metastatic tissue with respect to the corresponding primary tumor, which was more evident and significant in distant metastases from FTCs, than in locoregional recurrences from PTCs. Moreover, patients with low FAM83B levels tended to have a higher frequency of distant metastatization and a shorter DFS rate than those with high levels, further highlighting the correlation between dedifferentiation and low FAM83B mRNA levels, and consistent with data found in ovarian cancer 8 , in lung squamous cell carcinoma-SCC 9 , and triple-negative breast cancer 20 . Interestingly, we found that FAM83B expression levels in TC vary according to the tumor histotype, suggesting the possibility of different roles of this protein in the maintenance and tumorigenicity of the cancer cells. Compared with normal tissues, FAM83B resulted constantly and significantly downregulated in PDTCs and ATCs cases. In PTCs a FAM83B expression lower than NTs was observed, which was significant comparing paired tissues either from our series or from TCGA. On the other hand, FAM83B expression in FTCs was higher than that found in NTs, suggesting that this protein could have a different role in different histotypes, as observed in lung carcinoma 11 .
In order to get more insights into FAM83B levels variations in the different thyroid cancer subtypes, we investigated its expression in cell lines derived from healthy and pathological thyroid tissues and we found that they nicely recapitulate the findings obtained in tissue samples. In particular, NTHY-ORI (derived from normal tissue) expressed high levels of FAM83B mRNA and protein. Although a variable FAM83B expression was found in PTCs, tumors with RET/PTC1 fusions had levels higher than the median, and this finding was consistent with the high levels recorded in TPC1 cells, which derive from a PTC with a RET/PTC1 fusion. On the other hand, both SW1736 (derived from ATC) and FTC133 (derived from an FTC metastasis) express low levels, as the corresponding tumor samples. Although the limitations of 2D cell cultures are well known, the variations in FAM83B observed in the different cell lines paralleled those observed both in patients and in the TCGA cohort, thus prompting us to speculate that this model may be useful in the initial understanding of cellular biology of FAM83B in thyroid cancer.
Given that FAM83B levels are low in FTC metastatic tissues and in metastasis-derived FTC-133 cell line, we hypothesized a role in cell migration. Several FAM83 family members have been reported to be involved in the regulation of cell migration in different tumors 12,[15][16][17][18][19]31 . Interestingly, in the tested cell lines FAM83B protein levels decreased in migrating cells. In particular, NTHY-ORI, which still have contact inhibition but retain migration abilities 23,26 , had high FAM83B levels at confluency, but expression levels promptly decreased when migration was induced by scratch assays. On the other hand, TPC1 showed very low migration abilities, as already reported 24,28 . Consistently, FAM83B expression levels did not vary upon migration induction. Finally, SW1736 and FTC133 cell lines, for which the loss of contact inhibition and the reduction of cell-cell adhesiveness is known 25,27,29 , displayed a further reduction in FAM83B expression during migration. Interestingly, upon    www.nature.com/scientificreports/ migration induction we observed not only a reduction in FAM83B positivity, but also its re-localization from the cytoplasm to the nucleus. Similar variations were also detected in samples obtained from a FTC patient, with the cytoplasmic FAM83B staining stronger in more differentiated areas and progressively lower in the undifferentiated parts of the tumor. At the metastatic level, the nuclear localization of FAM83B was detected at the invasion front in the firstly developed metastasis and was exclusive and homogeneous in the metastasis developed several years after first treatment. This finding is consistent with what reported for FAM83H in colorectal cancer tissues, where a nuclear staining was observed in a small population of cancer cells located at the edge of or detached from the tumor mass 15 , and indicates that the FAM83 family members, and likely their subcellular localization, might have a role in cell migration as previously suggest in other cancer types 12,[15][16][17][18][19]31 .
Finally, results obtained by FAM83B silencing on the two cell lines with higher FAM83B mRNA levels, NTHY-ORI and TPC1, may support its role in thyroid cancer cell migration. In particular, upon FAM83B silencing, migration abilities significantly increased, while the oncogenic RAS/MAPK/PI3K pathways were found not to be affected.
Moreover, we found a decrease of markers of thyroid differentiation and an increase of those associated with acquisition of mesenchymal phenotype, as changes in cell motility are often associated with loss of differentiation or epithelial-to-mesenchymal transition. Although the 2D cell lines model has some intrinsic limitations in the study of cell differentiation, we are tempting to speculate that it can still be considered reliable enough to provide more insight in FAM83B cellular biology. Further studies are needed to confirm FAM83B role in thyroid cell migration and differentiation through 3D multicellular approaches that better resemble patient-derived thyroid tumor. www.nature.com/scientificreports/ In conclusion, the whole of the above reported data indicates for the first time a role for FAM83B in thyroid cell differentiation and migration. Since FAM83B expression is reduced in more dedifferentiated tumors and in metastases, it is tempting to speculate that it could be involved in the maintenance of a differentiated phenotype. In dedifferentiated cells, its decreased expression and its nuclear re-localization could favour distant migration. A low FAM83B expression, even if limited to a part of a tumour sample, could thus indicate the possible existence of circulating tumour cells eventually responsible for distant metastases. For these reasons, FAM83B should be considered a possible diagnostic and prognostic biomarker.

Materials and methods
Patients and samples collect. This is a retrospective cohort study with institutional review board approval (Ethical Committee of the Istituto Auxologico Italiano IRCCS, #2018_09_25_04). Informed consent for the use of thyroid tumor tissues and collection of clinico-pathological information was obtained from all participants. All methods were carried out in accordance with relevant guidelines and regulations (Declaration of Helsinki).
Frozen thyroid tissues from 53 patients undergone total thyroidectomy were used for molecular and expression analysis. In particular, we studied 20 papillary (PTC), 5 follicular thyroid cancers (FTC), 9 poorly differentiated (PDTCs)/anaplastic thyroid cancers (ATC), 16 metastases (MTS), 24 normal thyroid tissues (NT). Formalin-fixed paraffin embedded (FFPE) thyroid tissues were obtained for immunohistochemistry analyses. Patients are followed in a single tertiary care endocrine center and were diagnosed and treated according to the recent guidelines for the management of thyroid cancer during the period 2001-2020 32,33 . Tumors were classified and staged according to the thyroid malignancy World Health Organization classification and the 8th edition of TNM staging 34 . Clinico-pathological features at diagnosis and the final disease outcome, after a mean follow-up of 37 months (range 8-219), were available for all included patients. The molecular characterization of all PTC cases was done by means of a fully validated custom PTC-MA assay, based on matrix-assisted laser desorption/ ionization time-of-flight mass spectrometry 35,36 . Cell culture and siRNA transfection. NTHY-ORI 3-1 (immortalized normal thyroid), TPC1 (derived from human papillary thyroid carcinoma with RET/PTC1 fusion) and K1 (derived from a papillary thyroid cancer with BRAFV600E mutation) cells were grown in RPMI supplemented with Glutamine (Euroclone, Pero, Italy), SW1736 (derived from human anaplastic thyroid carcinoma) cells, SW579 cells (derived from a poorly differentiated carcinoma) and HTCC3 cells (derived from a pleural metastasis of poorly differentiated thyroid carcinoma) were grown in DMEM (Gibco-Thermo-Fisher Scientific), while FTC133 cells (derived from a lymph node metastasis of human follicular thyroid carcinoma) were grown in DMEM/F12 (Gibco-Thermo-Fisher Scientific) in a humidified incubator at 37 °C under 5% CO 2 . All media were supplemented with 10% fetal bovine serum (SigmaAldrich, ST. Louis, Missouri, USA) and penicillin-streptomycin mixture (Sigma-Aldrich). FAM83B expression was downregulated by small interfering RNA (siRNA). Two siRNAs against FAM83B (Hs_ FAM83B_8 and Hs_FAM83B_9 FlexiTube siRNAs, Qiagen GmbH, Hilden, Germany) 11 and a control one (Negative Control FlexiTube siRNA, Qiagen) were transfected in 6-well dishes to a final concentration of 10 nM with HiPerFect (Quiagen) following manufacturer's instruction. All experiments were performed 3 days after transfection. All experiments were performed with cell lines in between the 7th and the 15th passage. All cell lines were authenticated by short tandem repeat (STR) profiling and were routinely screened for mycoplasma contamination with Venor GeM Classic Mycoplasma Detection Kit (Minerva Biolabs GMBH, Berlino, Germany).

Quantitative real-time PCR (qRT-PCR).
Total RNA was isolated from frozen thyroid tissues and thyroid healthy and tumor cell lines using Trizol reagent (Thermo Fisher Scientific, Waltham, Massachusetts, USA) according to the manufactures' instructions. RNA was quantified at spectrophotometry, and its quality was checked by the analysis of 260/280 nm and 260/230 nm ratios. 1000 ng of each RNA sample was reverse transcribed using a Superscript reverse transcriptase II Kit (ThermoFisher Scientific), with random hexamers mixture as primers. Appropriated endogenous controls for qRT-PCR were selected with Applied Biosystems ® TaqMan ® Express Endogenous Control Plates (ThermoFisher Scientific) among 32 constitutively expressed different genes tested on healthy and pathological tissues. Among the 32 candidate genes, two genes (ACTB and B2M) were indicated as the most reliable (appropriate and similar cycle threshold -Ct-values between tumor and normal thyroid tissue samples), and were thus selected as endogenous controls for all experiments, as optimal balance between analysis accuracy and samples mRNA availability. RNA was analysed for FAM83B, ACTB, B2M expression using an Applied Biosystems QuantStudio 12K Flex Real-Time PCR System (Thermofisher Scientific) and the following assay Hs00289694_m1, Hs99999907_m1 and Hs99999903_m1 (Thermofisher). Relative FAM83B expression levels were calculated using the 2-ΔΔCt method. The geometric mean of Ct values was used as averaging method of the control genes, as previously reported 37,38 . For each patient the retrotranscription and Real-Time PCR were repeated at least three times, with each sample run in triplicate. A pool of 20 healthy thyroid tissues (10 from women and 10 from men) was run in each experiment together with the samples, and used as reference control.
For RNA of migrating cells, confluent cells seeded in 6-well plates were scratched with a p1000 tip multiple times to obtain alternating confluent cells and free space areas. After 24 h, when most of the cells were moving through the free spaces, samples were lysed together with unscratched samples as control. www.nature.com/scientificreports/ transcriptase II Kit (ThermoFisher Scientific), with random hexamers mixture as primers. PAX8 and NIS mRNA were amplified in the log-linear phase of PCR by 25-34 cycles using 100 ng of each cDNA and primers previously reported 39 . In all PCR analyses, the amount of cDNA in each sample was normalized using ACTB as an internal control. The PCR products were separated by electrophoresis on 2% agarose gels stained with Midori Green Advance DNA Stain (Nippon Genetics Europe GmbH, Düren, Germany). The signal intensities of gene bands were analyzed with Image J software (version 2.1.0/1.53c; National Institutes of Health, Bethesda, MD). Full-length agarose gel images are shown in Supplementary Fig. 5.

Immunohistochemistry (IHC).
Immunohistochemical studies were carried out on one out of four consecutive 4 µm thick sections obtained from paraffin blocks. According to the manufacturer's instructions, tissue sections were incubated overnight with 1:200 dilution of anti-FAM83B antibody (rabbit polyclonal IgG antibody, Atlas Antibodies, Voltavägen, Bromma, Sweden). Slides were evaluated independently by two experts (G.G., L.E.). Image acquisition was performed by LEICA ICC 50 HD.
Wound healing and migration assays. For wound healing assay, samples were processed as previously described 40 . Confluent cells seeded in 6-wells plates were scratched once with a p200 tip with a single vertical movement, washed with PBS and returned to the appropriated medium. At the indicated times post-wound, cells were washed with PBS, fixed with 2% PFA in PBS for 20 min RT and stained with 0.01% Crystal Violet PBS solution for 30 min RT. The samples were washed 3 times in H2O and then let air-dry overnight. Images were acquired basally and at 16 h post-wound with a Samsung J510FN camera on Olympus CK2 microscope with A10PL 10× objective. Wound dimensions were quantified with FIJI 41 . For western blot of migrating cells, confluent cells seeded in 6-well plates were scratched with a p1000 tip multiple times to obtain alternating confluent cells and free space areas. After 24 h, when most of the cells were moving through the free spaces, samples were lysed together with control unscratched samples, as described in "Western blotting" section. For immunofluorescence of migrating cells, confluent cells seeded onto poly-l-lysine (Sigma-Aldrich) coated microscope slides in 6-well plates were gently scratched with a p200 tip. After 6 h, scratched samples together with control unscratched samples, were processed as described in "Immunofluorescence" section.
Western blotting. Cells were lysed in SDS sample buffer (62.5 mM Tris-HCl pH 6.8, 2% sodium dodecyl sulfate) supplemented with protease, phosphatase and proteasome inhibitors, pre-heated at 95 °C 40 . Thereafter, samples were sonicated and protein amount dosed with Pierce BCA protein Assay Kit (Thermo Fisher Scientific). Sixty µg of protein extracts were then separated on NuPage 4-12% Bis-Tris Gels (Thermo Fisher Scientific) and transferred with iBlot System (Thermo Fisher Scientific). Membranes were incubated with blocking buffer (5% nonfat dry milk in TBS-T solution) for 1 h at room temperature and probed overnight at 4 °C with appropriate primary antibodies diluted 1:1000 in blocking buffer: antiFAN83B (Atlas Antiboides), anti-vimentin and anti-SDHA (Abcam), anti-p-ERK1/2, anti-ERK1/2, anti-P-AKT and anti-AKT (Cell Signalling), as previously described 40     In order to validate whether FAM83B expression could predict prognosis in thyroid cancer patients, Kaplan-Meier analysis and Log-rank test were used to assess the association between FAM83B expression and disease-specific survival rate.
For immunofluorescence experiments, co-localization was measured by Pearson's R coefficient in 4 independent experiments with FIJI Coloc2 plugin 39 . For FAM83B intensity quantification, Raw Integrated Density was measured with FIJI after determination of Regions Of Interest (ROIs) based on WGA for total FAM83B, DAPI for nuclear FAM83B, and WGA minus DAPI for cytoplasmic FAM83B.
After normal distribution and variance similarity evaluation, Kruskal-Wallis test followed by Dunn's Multiple Comparison Test, Wilcoxon signed rank test, and Mann Whitney test were used to determine statistical significance with Graphpad Prism 5 (RRID:SCR_002798), as indicated in the corresponding Figure legends. Significance expressed as p values (*p < 0.05, **p < 0.01, ***p < 0.001).