Differential expression profile of CXCR3 splicing variants is associated with thyroid neoplasia. Potential role in papillary thyroid carcinoma oncogenesis?

Papillary thyroid cancer (PTC) is the most prevalent endocrine neoplasia. The increased incidence of PTC in patients with thyroiditis and the frequent immune infiltrate found in PTC suggest that inflammation might be a risk factor for PTC development. The CXCR3-ligand system is involved in thyroid inflammation and CXCR3 has been found upregulated in many tumors, suggesting its pro-tumorigenic role under the inflammatory microenvironment. CXCR3 ligands (CXCL4, CXCL9, CXCL10 and CXCL11) trigger antagonistic responses partly due to the presence of two splice variants, CXCR3A and CXCR3B. Whereas CXCR3A promotes cell proliferation, CXCR3B induces apoptosis. However, the relation between CXCR3 variant expression with chronic inflammation and PTC development remains unknown. Here, we characterized the expression pattern of CXCR3 variants and their ligands in benign tumors and PTC. We found that CXCR3A and CXCL10 mRNA levels were increased in non-metastatic PTC when compared to non-neoplastic tissue. This increment was also observed in a PTC epithelial cell line (TPC-1). Although elevated protein levels of both isoforms were detected in benign and malignant tumors, the CXCR3A expression remained greater than CXCR3B and promoted proliferation in Nthy-ori-3-1 cells. In non-metastatic PTC, inflammation was conditioning for the CXCR3 ligands increased availability. Consistently, CXCL10 was strongly induced by interferon gamma in normal and tumor thyrocytes. Our results suggest that persistent inflammation upregulates CXCL10 expression favoring tumor development via enhanced CXCR3A-CXCL10 signaling. These findings may help to further understand the contribution of inflammation as a risk factor in PTC development and set the basis for potential therapeutic studies.


INTRODUCTION
Thyroid cancer has one of the fastest growths in incidence when compared to other malignant and endocrine neoplasia [1,2], where papillary thyroid cancer (PTC) represents approximately 80% of thyroid malignancies [3,4]. Based on recent data showing a sharp increase in its rate of incidence, PTC is predicted to become the fourth most common malignancy by 2030 in United States [5]. Usually, 20% of patients develop lymph node metastasis (LNM) which is associated with increased patient morbidity and mortality [6,7].
A remarkable progress has been made in the understanding of the oncogenic mechanisms that promote thyroid cancer development and progression. The inflammatory state of premalignant and malignant lesions play decisive roles in tumor initiation and progression [8]. Specific chronic inflammatory conditions that increase the risk of cancer development by promoting the infiltration of inflammatory cells to premalignant lesions and the oncogene signaling that initiate cancer by inducing the inflammation-related genes expression are known mechanisms that link inflammation with cancer [9]. Chronic thyroiditis has been reported in 20 to 50% of PTC cases [10,11] and is associated to an increased risk of developing PTC [12][13][14]. Also, the higher incidence of PTC in patients with chronic lymphocytic thyroiditis or Hashimoto autoimmune thyroiditis (HT) [15][16][17] has suggested a role of inflammation as a predisposing factor in the development of thyroid cancer [18][19][20]. Even in the absence of any signs of HT, remarkable peritumoral lymphocytic infiltrate or focal lymphocytic thyroiditis is frequently found in PTC [18] which is significantly higher and more frequent than in benign thyroid lesions [21]. As lymphocytic infiltrate is generally found in the peritumoral environment of PTC, it has been suggested that it favors the development of PTC [22]. Thyroid cancers arising in the context of HT exhibit a better prognosis than PTCs without chronic lymphocytic thyroiditis [23,24]. In the absence of typical signs of HT, poorly differentiated and anaplastic thyroid carcinomas, which are associated with a worse prognosis, generally exhibit a reduced lymphocytic infiltrate compared to PTCs [11]. However, there is still no clear evidence whether thyroid inflammation promotes tumor development and/or plays a protective role against thyroid cancer progression.
Phenomena related with inflammation, such as elevated cytokine and chemokine levels, have been suggested to predispose to RET proto-oncogene rearrangements (RET/PTC), a classic oncogene that induced malignant transformation in thyrocytes [18]. In turn, RET/PTC promoted PTC oncogenesis through a transcriptional program that included upregulation of chemokines and their receptors creating an inflammatory environment that triggers a pro-tumorigenic response [18,[25][26][27][28][29]. Furthermore, increased levels of proteins from this "inflammatory program" have been detected in primary thyroid tumors of patients with LNM [25,30]. Therefore, accumulating evidence suggests that there is a strong association between inflammation and increased risk to neoplastic transformation and progression in thyroid cancer [18][19][20][31][32][33].
Inflammation is a key component of the tumor microenvironment and chemokines are part of the inflammatory network of mediators associated to neoplasia [34]. Frequently, chemokines and chemokine receptors are found in tumors and often their expression and signaling are deregulated [35,36]. In fact, chemokine receptor signaling in malignant cells promotes tumor growth, invasion and metastasis [37][38][39][40][41]. Spread of tumor cells to chemokine gradients is restricted to specific patterns of chemokine receptors and to chemokine availability in the tumor microenvironment [42,43]. Inflammatory cytokines produced by tumor cells and/or by tumor associated leukocytes may contribute to malignant progression [44]. Growing amount of evidence indicate that IFN-ɣ inducible chemokines (CXCL9, CXCL10 and CXCL11) and CXCR3 (their main receptor) [45,46] play an important role in the initial stage of autoimmune thyroiditis [47,48].
In thyroid tissues, secreted levels of CXCL10 are associated with T helper 1 (Th-1) cell infiltration which is commonly found in autoimmune thyroiditis and has been closely related to thyroid tumors [49]. Th-1 lymphocytes recruited to a tumor site may be responsible for enhanced production of IFN-γ and tumor necrosis factor-α (TNF-α) which in turn stimulates CXCL10 secretion from a variety of cells helping to maintain and potentiate cytokine production in the tumor microenvironment [50]. Since IFN-γ induces the secretion of CXCR3 ligands and CXCL10 in normal [51,52] and PTC thyrocytes [53], it has been proposed that CXCR3 chemokines may promote thyroid malignant transformation and/or thyroid cancer progression [54][55][56][57]. In fact, the CXCR3 system plays a pivotal role in the malignancy process. The CXCR3 angiostatic activity is complemented with its ability to potentiate the anti-tumor immunity in several tumor www.impactjournals.com/oncotarget systems [58][59][60]. However, upregulation of the receptor has been reported in many human tumors and is frequently associated with LNM and poor prognosis [61]. IFN-γ inducible chemokines have shown to promote malignancy by inducing cell migration of solid tumors to metastatic sites [62][63][64]. In lymph nodes, CXCL9 and CXCL10 upregulation is induced by activated Th-1 lymphocytes that secretes IFN-γ [62,65,66]. Moreover, CXCR3 expression levels in primary tumor were causatively involved in LNM [62]. These observations suggest that host and tumor cells can exploit IFN-γ inducible chemokines to promote their neoplastic transformation and cancer progression through CXCR3 signaling.
CXCR3 chemokines may have divergent effects on tumor cells in part dictated by the array of two functionally CXCR3 splicing variants; CXCR3A and CXCR3B [46,67]. While CXCR3A signaling (by CXCL10 and CXCL11) promotes cell viability, proliferation, angiogenesis and inhibits apoptosis, CXCR3B promotes an angiostatic effect and apoptosis [46]. The relative expression of CXCR3 variants is important for regulating proliferation and survival in several cancer cells [67][68][69] and correlates with tumor dissemination and metastasis [70][71][72][73]. In human renal, ovarian, prostate and breast cancer tissues, CXCR3A is upregulated while CXCR3B is markedly decreased [67,68,73,74]. Although mRNA levels for CXCR3, CXCL10 and CXCL11 were increased in thyroid cells carrying mutations in the RET/PTC-RAS-BRAF axis [27], no studies have addressed the contribution for the CXCR3-ligand system in PTC development and LNM.
In this study, we determined the expression pattern of CXCR3 variants and their ligands in PTC and in benign thyroid tumors and compared them with non-malignant thyroid tissue. In addition, to investigate if the isoform expression profile is associated with cancer progression or inflammation, we analyzed this pattern in patients with and without LNM or chronic thyroiditis. Here we demonstrate, to our knowledge, for the first time that protein levels of both CXCR3 isoforms are progressively upregulated in benign and malignant tumors where CXCR3A expression remains as the predominant isoform and promotes cell proliferation in Nthy-ori-3-1 by CXCL10 and CXCL11 stimulation. CXCL10 is upregulated only in PTC patients with thyroiditis and is significantly induced by IFN-γ in normal and cancer epithelial cells. Overall, our findings suggest a potential role of inflammation in PTC tumorigenesis through CXCR3A-CXCL10 enhanced proliferative signaling. CXCR3 expression was analyzed by immunohistochemistry in 30 PTC tumors (14 nonmetastatic PTC (nMPTC) and 16 metastatic PTC (MPTC) and in their contralateral tissue. Histological features for PTC specimens and distribution of PTC tumors analyzed by immunohistochemistry are shown in Table 1. CXCR3 was mainly detected in follicular cells ( Figure 1A, section d and inset). The quantitative analysis of immunoreactivity showed a 4-fold increase of average staining intensity for CXCR3 expression in PTC (409.5±179.8) vs. PTC-CLT (121.2±89.3, p<0.0001) ( Figure 1B). CXCR3 immunoreactivity showed reduced risk to develop extrathyroidal extension (OR 0.09, 95% IC 0.01-0.74) ( Table 2). No associations were found between CXCR3 staining with the others histopathological features evaluated (tumor size, lymph vascular invasion, LNM and thyroiditis). Additionally, the percentage of positive cells for CXCR3 immunostaining was higher in PTC (85%) than in PTC-CLT (51%, p<0.0001) ( Figure 1C). The frequency graph of CXCR3 positive cells showed that most of PTC   Significative results are showed in bold font.

Increased CXCR3 expression in human PTC tissues
specimens presented high frequencies (% of CXCR3 immune reactive cells between 67-100%), whereas PTC-CLT cases were mainly distributed in moderate (34-66%) and low frequencies (0-33%) ( Figure 1D). A representative image of low, moderate and high percentage of positive CXCR3 reactive cells is shown in Supplementary Figure 1.
When the CXCR3 frequency of positive cells was analyzed between MPTC and nMPTC specimens, no differences were found ( Figure 1E). Thyroiditis was found in 30% of PTC specimens and all cases with chronic inflammation showed a high frequency for CXCR3 reactive cells. No differences were found in CXCR3 staining intensity between cases with or without thyroiditis. Moreover, PTC-CLT with or without thyroiditis, showed similar CXCR3 intensity levels (Supplementary Figure 2).

CXCR3A and CXCL10 mRNA levels increase in non-metastatic PTC tissues
For qPCR and western blot analyses the histological features for benign and PTC specimens and their corresponding distribution is shown in Table 3. Primer sequences for CXCR3 variants and their ligands are listed in Table 4. Next we analyzed the mRNA levels for CXCR3A and CXCR3B in benign nodules and in their contralateral tissue (B-CLT) (control tissue). CXCR3A was slightly but not significantly increased and CXCR3B was reduced by 7-fold compared to B-CLT (-6.801±15.41, p=0.0073) ( Figure 2A). Then, we investigated if changes in CXCR3 mRNA profile could be associated to tumor development by comparing the expression of CXCR3A and CXCR3B in nMPTC with B-CLT. Expression of CXCR3A was significantly increased (1.651±3.539, p=0.0091) whereas CXCR3B was decreased (-2.903±4.368, p=0.0011) ( Figure 2B). This indicates that in PTC the expression profile of CXCR3 variants is represented by increased levels of the pro-tumoral CXCR3A together with a reduction of the anti-tumoral CXCR3B isoform. To determine if the profile exhibited by the CXCR3 variants is associated to tumor metastatic progression, mRNA levels of CXCR3 variants were compared between MPTC and B-CLT ( Figure 2C) and between MPTC and nMPTC ( Figure 2D). A significant decrease of CXCR3B expression was observed in MPTC (-3.84±8.66, p=0.0139) when compared to B-CLT, while CXCR3A levels remained constant ( Figure 2C). In contrast, when mRNA levels were compared using nMPTC as control tissue, a significant decrease of CXCR3A expression was observed in MPTC (-4.876±5.279, p<0.0001) while CXCR3B did not significantly change ( Figure 2D). To investigate the potential role of CXCR3 ligands in the process of neoplastic transformation and tumor progression, the mRNA levels of CXCR3 ligands were assessed in benign and PTC tissues. We analyzed ligands expression in benign nodules (B), nMPTC and MPTC by comparing to B-CLT. In benign nodules, CXCL11 was increased by two folds (2.401±17.29), although without statistical significance whereas CXCL4 (-13.13±18.74, p=0.0121) and CXCL9 (-4.132±6.374, p=0.0044) were decreased ( Figure 2E). In addition, the analysis of CXCR3 ligands revealed that the expression of CXCL10 and CXCL11 were increased in nMPTC (1.883±2.950, p=0.0023; 3.307±9.782) while CXCL4 mRNA levels were decreased (-4.126±10.33) ( Figure 2F). When comparing CXCR3 ligands expression in nMPTC with or without thyroid inflammation respect to B-CLT, a significant increase of all ligands only in nMPTC with thyroiditis (CXCL4: 3.236±3.855, p=0.469; CXCL9: 5.021±4.166, p=0.0039; CXCL10: 2.713±1.574, p=0.0039; CXCL11: 8.121±10.21, p=0.0098) was observed ( Figure 2G). In contrast, all ligands were downregulated in nMPTC without thyroiditis. Comparison of CXCR3 ligands expression in MPTC respect to B-CLT showed that CXCL4, CXCL9 and CXCL10 were significantly reduced ( Figure 2H). Similar results were obtained when nMPTC was used as a control tissue. All ligands were decreased with statistical significance in MPTC, with major decreases for CXCL9 (-12.86±21.07, p=0.0002), CXCL10 (-12.58±23.08, p=0.0015) and CXCL11 (-10.35±23.08, p=0.0171), all high affinity CXCR3A ligands [46]. Summarizing, the results indicate that CXCR3A is upregulated in nMPTC, while both CXCR3 variants and all their ligands are significantly decreased in MPTC. In nMPTC, the presence of thyroiditis conditioned the increased expression of all CXCR3 ligands, but CXCL10 levels increased significantly when compared to B-CLT.

A progressive increase in protein balance of CXCR3A/CXCR3B from benign tumors to PTC tissues
Immunohistochemistry analysis revealed that total CXCR3 levels were increased in PTC (Figure 1), however, the antibody used in this analysis did not account changes in the CXCR3 variants profile. To compare the proteins levels of CXCR3A and CXCR3B in benign and malignant thyroid tissue, western blot analysis was performed by using a CXCR3 mAb (clone 1-C6) that recognizes CXCR3A (41 kDa) and CXCR3B (44 kDa) [75] as independent bands (Supplementary Figure 5). A representative western blot analysis is shown in Figure  3A. In B-CLT, benign and PTC, CXCR3A bands were frequently stronger than CXCR3B whereas in B-CLT CXCR3B expression was weakly detected. In benign tumors, nMPTC and MPTC tissues, both CXCR3 variants showed a significant increase when compared to B-CLT (CXCR3A: p=0.0050, p=0.0014 and p=0.0353; CXCR3B: p=0.001, p<0.0001 and p<0.0001) ( Figure  3B). To investigate the profile of CXCR3 variants in tumor progression, CXCR3A and CXCR3B levels were compared between nMPTC and MPTC. Data showed that both variants were significantly downregulated www.impactjournals.com/oncotarget In all graphs, data is shown as mean ± s.e.m. Statistical significance for Pfaffl's analyses was performed using the Wilcoxon test ( * p<0.05, ** p<0.01, *** p<0.001). www.impactjournals.com/oncotarget in MPTC (CXCR3A: 1.1190±0.4086, p=0.0184; CXCR3B: 0.8669±0.3755, p=0.0205) ( Figure 3B). In fact, the association of CXCR3A and CXCR3B protein levels with histopathological features of PTC, tested by UVA (univariate analysis), showed that CXCR3A and CXCR3B upregulation was associated with the absence of lymphnode metastasis in PTC (OR 0.19, 95% IC 0.05-0.78 for both) ( Table 5).
To analyze the relative fold changes of CXCR3 variants over B-CLT, CXCR3A and CXCR3B/α-tubulin  CXCR3A and CXCR3B levels were determined as in (B) and the CXCR3A/CXCR3B ratio (protein balance) was plotted. In all graphs, data is shown as mean ± s.e.m. Ratios were analyzed by Mann-Whitney test ( * p<0.05, ** p<0.01, *** p<0.001). ratio of each patient was divided by a B-CLT average ratio. Although in nMPTC, CXCR3B presented a higher fold change in protein levels (2.71) compared to CXCR3A (2.04), the latter remained as the predominant expressed isoform ( Figure 3C). We also determined whether significant changes in CXCR3A and CXCR3B levels found in PTC could be related to the presence of thyroidinfiltrating lymphocytes. Interestingly, the increase in CXCR3A and CXCR3B protein levels in nMPTC was higher in patients without thyroiditis (CXCR3A: 1.7680±0.7505, p=0.0002; CXCR3B: 1.3480±0.5230, p<0.0001) than those with thyroiditis (CXCR3A: 1.2210±0.3714, p=0.0459; CXCR3B: 1.0460±0.5350 p=0.0024) ( Figure 3D), suggesting that most of the increased protein levels of CXCR3A and CXCR3B might come from epithelial thyroid cells rather than from the inflammatory infiltrate in PTC. It has been previously reported that the IFNγ-inducible chemokines and their receptor CXCR3 potentiate the initiation of HT [13]. Our results showed an increased level of CXCR3 ligands in nMPTC tissues with thyroiditis, observation that strongly supports the notion that inflammation plays a key role in neoplastic transformation and cancer development through the CXCR3/ligand axis. To determine if changes in the CXCR3 expression profile are indeed associated to thyroid chronic inflammation, we analyzed the protein profile of CXCR3A and CXCR3B in tissues with thyroiditis and compared them to B-CLT, benign and nMPTC tumors. In this case, the contralateral tissue of benign tumors without thyroiditis (B-CLT-T) was used as the control tissue. Results showed a progressive increase in protein levels for CXCR3A and CXCR3B. Control tissue exhibited the lowest amount of protein followed by thyroiditis, benign and nMPTC tumors, the latter presenting the highest protein levels ( Figure 3E). A significant increase of CXCR3B expression was found in thyroiditis compared to B-CLT (0.5747±0.433, p=0.0199), while both CXCR3 variants showed a statistical increment in nMPTC tumors compared to thyroiditis (CXCR3A: 1.593±0.7143, p=0.0171; CXCR3B: 1.240±0.5445 p=0.0004) ( Figure  3E).
The analyses of protein balance (CXCR3A/ CXCR3B ratio) in benign tumors and cancer thyroid tissues showed a predominant CXCR3A expression over the CXCR3B isoform ( Figure 3F). A significant and progressive decrease in the protein ratio was observed from B-CLT to MPTC ( Figure 3F) due to a greater fold change increase of the CXCR3B variant. A major decrease in the protein balance was observed when comparing B-CLT and benign nodules (p=0.0292), and between nMPTC (p=0.0005) or MPTC (p<0.0001) with B-CLT. No major difference in the protein ratio was found between nMPTC and MPTC. Thus, results suggest that changes in   In all graphs, data is shown as mean ± s.e.m. Statistical analysis for chemokine secretion levels was performed using Mann-Whitney test and for CXCL10 immunohistochemistry was performed using a two-tailed paired t-test ( * p<0.05, ** p<0.01, *** p<0.001). the ratio of the CXCR3 variants is associated to neoplastic transformation and tumor growth rather than tumor metastasis.

CXCL11 is the predominant secreted ligand in thyroid epithelia but only CXCL10 is significantly induced by IFN-ɣ in PTC cell line
Next we sought to evaluate whether IFN-ɣ could promote a differential response in secreted levels of CXCR3 ligands in Nthy-ori-3-1 and TPC-1 cells. Under serum free media conditions, we found that CXCL10 levels in Nthy-ori-3-1 and TPC-1 cells (8.276±19.94 and 5.642±15.64 pg/ml, respectively) were lower than that of CXCL11 (47.08±3.585 and 46.72±2.799 pg/ml respectively) ( Figure 5A). No statistical differences were Figure 6: Schematic modeling of the CXCR3 variant expression profile that is associated to thyroid benign neoplasia and PTC. A progressive increase in protein balance for CXCR3A (red) and CXCR3B (blue) from control tissue to benign and malignant (PTC) tumors is shown, with consistent predominance of CXCR3A. The increase of CXCR3A and CXCR3B in benign tumor suggests a neoplastic transformation role. Higher levels of CXCR3A in nMPTC suggest a role in malignant transformation or tumor growth. The decrease of CXCR3A and CXCR3B in MPTC compared to nMPTC suggests that CXCR3A may not be associated to thyroid tumor dissemination but with tumor growth. Inflammation was conditioning for CXCR3 ligand increased availability in PTC. CXCR3A and CXCL10 are highly expressed in nMPTC consistent with a condition that may favor neoplastic transformation and tumor growth. found for these ligands between both cells lines in starving media. In contrast, CXCL10 was robustly induced by IFN-ɣ and this increase was significantly higher in TPC-1 compared to Nthy-ori-3-1 (189.8±39.47 pg/ml, p=0.0286 vs. 129.6±12.41 pg/ml, p=0.0357). Therefore, results suggest that CXCL10 expression is highly induced by an inflammatory stimulus, such as IFN-ɣ, in normal and cancer thyroid epithelial cells.

CXCL10 protein levels are upregulated in PTC
To determine whether CXCL10 expression partially originates from epithelial cells in PTC tumors, immunohistochemistry assays were performed. A major increase in CXCL10 immunoreactivity was detected in PTC epithelial cells compared to PTC-CLT ( Figure 5B, section c and d). A higher magnification image of section d is shown in Figure 5B. Quantitative analysis showed a 16-fold increase in cell intensity per cell for CXCL10 expression in PTC (82.24±52.54 vs. 5.08±3.77, p<0.0001) ( Figure 5C). No association was found between CXCL10 staining and other histopathological features ( Table 6). The percentage of cells expressing CXCL10 as determined by immunostaining was higher in PTC (56.46%) than in PTC-CLT (6.74%, p<0.0001) ( Figure 5D). The frequency graph of CXCL10 positive cells showed that most of PTC specimens presented moderate or high frequencies (% of CXCL10 immune reactive cells between 34-66% and 67-100%, respectively), whereas PTC-CLT cases were only distributed in low frequencies (% of CXCL10 immune reactive cells lower than 34%) ( Figure 5E). When the CXCL10 frequency of positive cells was analyzed between MPTC and nMPTC specimens, cases with moderate frequency were very similar in both groups. In contrast, specimens with higher frequency of CXCL10 were more abundant in nMPTC than in MPTC ( Figure 5F). CXCL10 expression was found in 100% of PTC specimens and all of them showed high intensity (data not shown).

DISCUSSION
It has been suggested that inflammation promotes an increased risk to neoplastic malignant transformation in thyroid tissues [18] and may be involved in thyroid cancer progression [20]. IFN-γ-mediated CXCL10 secretion is associated to thyroiditis and is closely related to thyroid tumors [49]. However, the role of CXCR3 spliced variants and their relation with thyroid cancer and inflammation has not been addressed. We report a progressive increase in protein balance of CXCR3 variants when comparing control tissue to benign tumors and malignant thyroid neoplasia where CXCR3A was always the predominant isoform. We also report that inflammation was conditioning for CXCR3 ligand increased availability in PTC. Finally, we show that CXCR3A and CXCL10 are highly expressed in non-metastatic PTC and in the PTC epithelial cell line TPC-1, consistent with a condition that may favor neoplastic transformation and tumor growth.
Our initial data showed increased protein levels of CXCR3 and CXCL10 in PTC. CXCR3 and CXCL10 immunoreactivity was mainly localized in epithelial cells within the tumor, but some minor staining was also observed in the inflammatory infiltrate. Thus, we cannot ignore a CXCR3A contribution from immune cells to the PTC-CXCR3A profile. In fact, a higher proportion of the CXCR3+/CCR5+ thyroid-infiltrating lymphocytes gained by fine needle aspiration biopsy has been recently reported in malignant nodules as compared to benign nodules [76] confirming some CXCR3 expression due to an inflammatory reaction surrounding thyroid carcinoma, as previously described [77]. However, our results suggest that the increase of CXCR3 expression in PTC is mainly due to an increased receptor expression in epithelial cells and not an exclusive contribution from the inflammatory infiltrate or stroma. Most of our PTC specimens with high CXCR3 staining were cases without thyroiditis and no major differences in CXCR3 immunoreactivity were found between PTC with or without thyroiditis. In addition, CXCR3A and CXCR3B protein levels in nMPTC were higher in patients without thyroiditis supporting that increased expression of CXCR3 is predominantly from epithelial cancer cells. In fact, we found increased protein levels for total receptor in TPC-1 when compared to Nthyori-3-1 cells. In ovarian cancer, the protumoral variant was observed in the inflammatory infiltrate, but was mainly detected in cancer cells [74]. Taken together, these results suggested that the increment of CXCR3 in PTC might be mainly epithelial with a minor contribution from Th-1 lymphocytes or stromal cells. Our findings prompted us to analyze the expression profile of CXCR3 in a papillary thyroid cancer cell line (TPC-1) as well as in benign and malignant thyroid neoplasms to determine if expression profiles are associated to thyroid malignant phenotypes.
Our results showed that in control tissue, if there is an imbalance of CXCR3 (favoring CXCR3A over CXCR3B levels), which gradually decreased as tissues becomes neoplastic (benign) and then malignant (nMPTC and MPTC). The increase of CXCR3A and CXCR3B protein levels between control and benign tumor tissue suggests a neoplastic transformation role for the CXCR3 variants profile. The balance of protein levels of CXCR3A and CXCR3B found in nMPTC suggest that similar levels for CXCR3 variant might have a role during malignant transformation or tumor growth. In benign tumors and PTC, proteins levels for CXCR3A and CXCR3B were significantly increased when compared to B-CLT, but CXCR3A was the most expressed variant with highest levels in nMPTC. Furthermore, since (i) CXCR3 staining was mainly observed in epithelial cells within the tumor, (ii) the protein levels for CXCR3A were significantly higher in TPC-1 than in Nthy-ori-3-1 and (iii) CXCR3B protein levels in TPC-1 were significantly lower than www.impactjournals.com/oncotarget CXCR3A and (iv) CXCR3B levels in TPC-1 were lower than Nthy-ori-3-1), we propose that increased CXCR3 levels found in PTC by immunohistochemistry analysis correspond to the CXCR3A variant mainly expressed in PTC epithelial cells with a minor CXCR3A expression from inflammatory cells [74,78,79].
Despite that CXCR3A was predominant in PTC we detected an unexpected increase of CXCR3B. In TPC-1, CXCR3B protein levels were not increased when compared to Nthy-ori-3-1 cells. These results support that CXCR3B increment found in benign tumors or PTC may originate from other cell types within the tumor rather than epithelial cells. CXCR3B is primarily expressed in endothelial cells [46], in fibroblasts and T-lymphocytes but it has been found in several endocrine malignant neoplasia [80]. In epithelial breast cancer cells, decreased mRNA levels for CXCR3B were induced by upregulation of CXCL10 [67]. Consistently, we found increased CXCL10 at mRNA and protein levels in PTC tissues while CXCR3B mRNA levels were significantly downregulated. In fact, previous reports have shown that a significant proportion of freshly isolated PBMC (peripheral blood mononuclear cells) expressed CXCR3 at the plasma membrane while transcripts remained undetectable [81,82]. Unexpectedly, CXCR3B proteins were increased in all tissues when compared to B-CLT. This may be due to a reduced receptor destination to proteasomes or lysosomes. Taken together, these results suggest a mechanism for CXCR3B mRNA downregulation induced by CXCL10 in PTC thyrocytes, reinforcing that CXCR3B protein increment found in PTC may not mainly come from tumoral epithelial cells. In ovarian and renal cancer, CXCR3B was preferentially detected in endothelial cells [46,74], supporting the notion that CXCR3B activation may inhibit the PTC tumor associated angiogenesis. Other source for increased CXCR3B in PTC could be thyroid cancer stem cell-like populations, since CXCR3B has been found upregulated in this cells from breast cancer [70]. CXCR3B expression is frequently downregulated in several cancer types [67,73,74,83], consistently with a reduced anti-proliferative and limited angiostatic activity [46]. Over-expression of CXCR3B in cancer cells significantly inhibited cell proliferation and promoted apoptosis [45,74,84] and in renal [57,85] and gastric cancer [86] it was associated with tumor necrosis extension and immune-angiostatic activity respectively, demonstrating an anti-tumoral role for CXCR3B upregulation.
Changes in CXCR3 profile expression may be related to initial stages of cancer development or its progression to metastatic disease. CXCR3A has been associated to tumor progression in prostate, ovarian, breast and renal cancer [73,74,83]. When compared to nMPTC, a significant decrease of CXCR3A and CXCR3B was observed in MPTC but CXCR3A continued being the predominant variant. These results suggest that CXCR3A expression may not be associated to tumor thyroid dissemination. However, it has been reported that cells from metastatic primary tumors that have undergone epithelial-mesenchymal transition (EMT) display a reduced proliferation rate that is linked to dissemination and survival of metastatic cells [87] which would be consistent with a reduced CXCR3A expression. Further studies are necessary to understand how a weakened CXCR3A expression in MPTC could favor cell migration and tumor invasion.
Opposite effects for CXCR3 ligands depend on the array of spliced-variants and their localization (stroma and/or tumor cells) [80]. In nMPTC, CXCL10 and CXCL11 upregulation was associated to an increased CXCR3A expression, supporting a potential role for CXCR3A signaling during malignant neoplastic transformation. In fact, CXCR3A overexpressing cells promoted cell proliferation. Moreover, CXCL10 and CXCL11 induced significant increase in cell growth compared to cells transfected with empty vector. CXCL9, CXCL10 and CXCL11 has shown to induce cell proliferation on HMC, which selectively express CXCR3A [88,89]. Likewise, CXCR3A overexpression in HEK293 cells promoted proliferation in basal and CXCL10 treated cells and PTX strongly reduced the proliferation in both conditions [46].
All CXCR3 chemokines were significantly reduced in MPTC when compared to nMPTC. The declining expression of CXCR3A and CXCR3B together with the low levels of CXCR3 ligands detected in MPTC, suggest that in cancer thyrocytes, CXCR3 signaling may not be involved in tumor dissemination and LNM.
Chronic inflammation in the tumor microenvironment plays a crucial role in tumorigenesis. Inflammatory cytokines and chemokines in tumor microenvironments attract immunosuppressive cells, and enhance EMT, tumor angiogenesis and cancer-stem cell formation. TNF-α is a master switch from chronic inflammation to cancer [90,91]. Recent data has shown that TNF-α significantly increased the expression of CXCR2 and CXCR3 and their related ligands in renal cancer cells enhancing the migration, invasion and EMT [92]. Considering that in thyroid tissues, recruited Th-1 lymphocytes are responsible for enhanced IFN-γ and TNF-α production, which in turn stimulates CXCL10 secretion in thyroid cells perpetuating the autoimmune process [93], thus TNF-α stimulation of thyrocytes would clarify whether inflammatory microenvironment upregulates CXCR3 expression promoting thyroid tumorigenesis. Interestingly, CXCR3 ligands were statistically increased only in nMPTC patients with thyroiditis, indicating that the inflammation microenvironment is a prerequisite for increased CXCR3 ligand availability. In fact, CXCL10 was significantly induced by IFN-γ in Nthy-ori-3-1 and TPC-1, as has been previously reported in normal and PTC primary thyrocytes [53]. Inflammation promoted a switch on the CXCR3 ligand profile in normal and cancer epithelial cells, where CXCL10 became the predominant chemokine. Main cellular sources of CXCL10 in PTC tissues are Th-1 lymphocytes and mast cell infiltrate [94,95], the later which has been associated with a pro-tumorigenic role in PTC [95,96].
The role of chemokines and their receptors in malignant transformation in the tumor microenvironment is still debatable. Chronic inflammation causes DNA damage in proliferating cells [32,97,98] leading to neoplastic transformation [32]. CXCL10 is member of the interferon-related DNA damage signature, linking CXCL10 expression with pro-tumorigenic functions. Inflammation may promote CXCR3 upregulation in PTC by inducing its promoter demethylation [99] or by demethylation of intron CpG sites resulting in increased CXCR3A levels in cancer cells [61,100]. In damaged thyroid tissue with cytokines surrounding the premalignant lesion, early recruitment and activation of Th-1 leukocytes and mast cells, may lead to constitutive secretion of CXCL10 that in turn promotes, by paracrine or autocrine manner, thyrocytes (and other cells) survival through CXCR3A signaling. We propose that persistent inflammation and sustained CXCL10 secretion, may upregulate CXCR3A in epithelial cells ( Figure 6) promoting the malignant neoplastic transformation through constitutive CXCR3A-CXCL10 signaling. Then, in malignant tissue, tumor growth and cell survival may be supported by CXCR3A-CXCL10 axis, contributing to thyroid cancer progression but not to tumor dissemination. The function of increased CXCR3B expression could be related to a decreased tumor associated angiogenesis and to tumor dependent immunity. Further studies are pending to understand the mechanism underlying the increased CXCR3A and CXCR3B expression within PTC tissues and their role in cancer development and progression.
In conclusion, to our knowledge, this is the first report showing a switch in the CXCR3A/CXCR3B balance in PTC and its association with inflammation and thyroid cancer progression. The current study shows that CXCR3A and CXCL10 are highly expressed in PTC and in a PTC derived epithelial cell line (TPC-1), consistent with a condition that may favor tumor development and progression. In addition, CXCL10 is strongly induced by IFN-ɣ in CXCR3A over-expressing cells (TPC-1) suggesting that CXCR3A-CXCL10 enhanced signaling might induce tumor progression promoting signals in PTC.

Patients and tissue samples
Surgical tissue specimens included malignant (PTC, with or without LNM (metastatic PTC, (MPTC) and nonmetastatic (nMPTC) respectively)) and benign (B) thyroid tumors. Fresh benign and PTC tumors and their respective contralateral tissues (B-CLT and PTC-CLT) were obtained from patients undergoing total thyroidectomy. Patients signed an informed consent approved by the ethics committee of the Pontificia Universidad Católica de Chile according to the principles expressed in the declaration of Helsinki. Tissue samples were divided and placed in 4% paraformaldehyde or protein lysis solution, or preserved in a RNA stabilization solution (Ambion, Austin TX). Inclusion criteria were patients at least 18 years old and with a confirmed benign or malignant diagnosis by histopathological examination (biopsy report). Exclusion criteria were patients with follicular, medullary and anaplastic thyroid cancer. Thyroid biopsy chart showing benign and PTC cases distribution for IHQ, qPCR and WB analysis is shown in Supplementary  Figure  3). To establish our gold standard control tissue, we compared the mRNA levels of CXCR3A and CXCR3B between B-CLT and PTC-CLT tissues. Further, sub analysis comparing the expression for CXCR3 variants was performed between nMPTC-CLT and MPTC-CLT. Expression of CXCR3A was significantly increased in nMPTC-CLT when compared to B-CLT, whereas CXCR3B levels were only slightly increased (Supplementary Figure 4), indicating that PTC-CLT may not represent a qualified control. Therefore, we considered B-CLT as the gold standard for non-malignant tissue (control) in the mRNA and protein profiles analyses.

Real-time PCR (qPCR) reaction and qPCR analyses
Expression of CXCR3 variants and ligands was analyzed by real time RT-PCR (qPCR) using Brilliant II SYBR® Green QPCR Master Mix (Stratagene) following the manufacturer instructions. All reactions were performed in triplicate using 10 ηg of cDNA. ACTB and RNA18S were used as control genes. Primer sequences for CXCR3 variants and their ligands were designed using the NCBI's Primer BLAST (http://www.ncbi.nlm.nih.gov/ tools/primer-blast) and are listed in Table 4. Ct values were analyzed by Pfaffl's relative quantification model [103] presented as relative fold changes of mRNA amount in target tissue respect to reference tissue (Supplementary Materials).

Immunohistochemistry
CXCR3 and CXCL10 protein levels in tissue samples were analyzed in formalin-fixed, paraffinembedded samples obtained from PTC tumors and their contralateral tissue (PTC-CLT). Clinical information of PTC specimens analyzed by immunohistochemistry is summarized in Table 1. Primary mouse monoclonal (mAb) anti-human CXCR3 antibody (antibody that not account to CXCR3 variant profile due an almost complete overlap between variants aminoacid sequences) [104] (Abcam, ab64714) and polyclonal (pAb) anti-human CXCL10 antibody [67] (R&D Systems, AF-266-NA) were used. Further details of the procedure are described in Supplementary Materials.
The immunoreactivity intensity values were determined using the Image-Pro Plus program version 6.0 (Media Cybernetic Inc.) and the staining intensity of CXCR3 and CXCL10 in tumor tissue was normalized by total number of nuclei counted in each microphotograph and scored. To quantify immunoreactivity, 3 representative digital microphotographs were taken from sections of tumor and matched non-neoplastic tissue. The average normalized values were plotted and referred as intensity of CXCR3 or CXCL10 per cell. Frequency graphs for CXCR3 and CXCL10 were obtained by counting the total immunoreactive positive cells in at least 3 fields of each slide. Percentage of positive cells was obtained by dividing the number of stained cells by total cell number in each field. The percentage of positive cells of CXCR3 or CXCL10 in tumor tissue or in the PTC-CLT was scored based on the following criteria: low: (0-33% of positive cells), moderate (34-66% of positive cells) and high (67-100%). To assess the best cutoff value for the Odds Ratio analysis, a receiver operating characteristic (ROC) curve was constructed and analyzed (data not shown). Optimal sensitivity to specificity ratios were observed using the cutoff values of 90% of positive cells for CXCR3 and 58% of positive cells for CXCL10 and the immunoreactivity values of 493 for CXCR3 and 58.8 for CXCL10.

Quantitative detection of chemokines
Cells plated in 6 cm dishes were grown to 70% confluence. After 24 hours, cells were washed, cultured in serum-deprived media and treated with IFN-ɣ (10000 U/mL) for 24 hours. Control cells were maintained in serum-deprived media. Supernatants were harvested and cleared by centrifugation at 800 g at 4°C. Chemokine concentration was determined in triplicates by quantitative immunoassay ELISA kit (QuantiKine ELISA kit; R&D Systems) following the manufacturer's instructions.

Plasmids
Plasmids containing the complete open reading frame of CXCR3A or CXCR3B genes were obtained by isolating the human sequences from benign patients by RT-PCR reaction. PCR fragments were then cloned into pcDNA 3.1/V5-His © TOPO ® TA (ptopo, Invitrogene) (USA). Variants sequences were under the control of CMV and T7 polymerase and alternatively fused to the V5 epitope, adding 45 extra aminoacids. pmCherry-V5 plasmid was a gift from Dr. R. Fuentealba (Universidad Autonoma, Chile).

MTT assays
Cell proliferation was measured by MTT cell proliferation assay (Trevigen, Gaithersburg, USA). Following transfection cells were plated (3×10 3 cells/well) in 96-well plates with no phenol red RPMI medium mixed with 10% fetal bovine serum and then cultured at 37°C for 24 h. Then CXCL10 and CXCL11 were added at 100 ng/ mL and the cells were cultured for 48 h and 10 μl of MTT reagent was added to each well. When purple crystals of formazan became visible under the microscope, 100 μl of detergent reagent was added, and the cells were incubated for 2 h. Absorbance of cells in each well was observed at 570 nm under an absorption spectrophotometer (Autobio Labtec, China) and corrected against blanks (culture medium). Cells transfected with empty vector (pTopo) were considered as control. All experiments were conducted independently for 3 times. The reading at 570 nm is directly proportional to cell proliferation (number of viable cells).

In vitro translation
In vitro translation reactions were carried out in nuclease-treated rabbit reticulocyte lysate (RRL; #L4960, Promega, Madison, WI) according to the manufacturer instructions using 1 μg of RNA in each reaction at 70% v/v of RRL supplemented with 0.1 mM of an amino acid mixture minus leucine (#L9951, Promega), 0.1 mM of an amino acid mixture minus methionine (#L9961, Promega) and 40 U of RNAsin (#E00382, Thermo Scientific, ThermoFisher Scientific). The final volume in all reactions was 50 μL. As a negative control, a reaction without RNA was used. All the reactions were carried out during 90 minutes at 30°C and after that, the reactions were stopped placing the tubes at -20°C.

Statistical analysis
Statistical significance for experiments in which data were compared to a standard or predetermined value (Pfaffl's graphs), was performed by using a Wilcoxon test (for non-parametrical data) or a two-tailed t-test (for parametrical data). In experiments where data between several groups were compared, Mann-Whitney test (for non-parametrical data) or a paired or unpaired two-tailed t-test (for parametrical data) was performed. Data are presented as means ± standard error of the mean (s.e.m.) with sample size (n) indicating the number of samples (patients) or (n) indicating the number of independent experiments (cell lines). Significance of mean comparison was considered when p values were less than 0.05. All data were tested for normality with D'Agostino and Pearson omnibus test. The association between the protein expression of CXCR3, CXCR3A, CXCR3B and CXCL10 with each histopathological feature of PTC was determined by the calculation of the Odds Ratio (OR) by univariate analysis. The 95% of the confidence interval was also calculated for OR, resulting statistically important when did not cross the unit value. All analyses were performed using GraphPad Prism 5, version 5.01 ® software (GraphPad Prism, Inc). No statistical method was used to predetermine sample size therefore it was estimated according to biopsy availability.