Loss of One or Two PATZ1 Alleles Has a Critical Role in the Progression of Thyroid Carcinomas Induced by the RET/PTC1 Oncogene

POZ/BTB and AT-hook-containing zinc finger protein 1 (PATZ1) is an emerging cancer-related gene that is downregulated in different human malignancies, including thyroid cancer, where its levels gradually decrease going from papillary thyroid carcinomas (PTC) to poorly differentiated and undifferentiated highly aggressive anaplastic carcinomas (ATC). The restoration of PATZ1 expression in thyroid cancer cells reverted their malignant phenotype by inducing mesenchymal-to-epithelial transition, thus validating a tumor suppressor role for PATZ1 and suggesting its involvement in thyroid cancer progression. Here, we investigated the consequences of the homozygous and heterozygous loss of PATZ1 in the context of a mouse modeling of PTC, represented by mice carrying the RET/PTC1 oncogene under the thyroid specific control of the thyroglobulin promoter RET/PTC1 (RET/PTC1TG). The phenotypic analysis of RET/PTC1TG mice intercrossed with Patz1-knockout mice revealed that deficiency of both Patz1 alleles enhanced thyroid cancer incidence in RET/PTC1TG mice, but not the heterozygous knockout of the Patz1 gene. However, both RET/PTC1TG;Patz1+/− and RET/PTC1TG;Patz1−/− mice developed a more aggressive thyroid cancer phenotype—characterized by higher Ki-67 expression, presence of ATCs, and increased incidence of solid variants of PTC—than that shown by RET/PTC1TG; Patz1+/+ compound mice. These results confirm that PATZ1 downregulation has a critical role in thyroid carcinogenesis, showing that it cooperates with RET/PTC1 in thyroid cancer progression.


Introduction
Thyroid cancer is the most common type of endocrine malignancy, and one of the few tumor types for which incidence has been increasing over the past 20 years and is predicted to be the Pten [19], or thyroid-specific combined mutations of BRAF V600E and PIK3CA H1047R [20], confirming the multi-step carcinogenesis model.
In this study, we generated mice characterized by thyrocyte-specific expression of the RET/PTC1 oncogene in conjunction with one or two null alleles of Patz1. The absence or even the reduction of Patz1 expression enhanced RET/PTC1-induced thyroid carcinogenesis, promoting the rapid development of aggressive carcinomas, including ATC and solid variant of PTC. Therefore, this genetically engineered mouse model recapitulates key features of the human ATC and aggressive PTC and may represent a suitable model for the development of innovative therapeutic approaches.

Loss of One or Two Patz1 Alleles Enhances Thyroid Tumor Aggressiveness in RET/PTC1 Mice
In the time-window of 10-17 months of age, thyroid carcinomas were present in 100% of RET/PTC1 TG ;Patz1 −/− , whereas they were diagnosed in 54% and 58% of RET/PTC1 TG ;Patz1 +/+ and RET/PTC1;Patz1 +/− mice, respectively, that also show the presence of hyperplasia/goiter (Table 1). The thyroid carcinomas developed by these mice closely phenocopy the human pathology. As shown in Figure 2 Notably, 3 out of 14 tumors (21%) in RET/PTC1 TG ;Patz1 +/− mice showed features of either ATC or PDTC, whereas only one PDTC-but no ATC-was found in RET/PTC1 TG ;Patz1 +/+ . The pathological features of the PDTCs were those of the solid type, but without the nuclear features of PTC, and with the presence of necrosis and high mitotic activity (data not shown). Spindle cell morphology, with frequent giant cells, was observed in ATCs ( Figure 2e). Interestingly, in both cases of ATC developed by RET/PTC1 TG ;Patz1 +/− mice, the anaplastic phenotype co-existed with a solid variant of PTC (Figure 2f), suggesting that it has arisen from pre-existing aggressive papillary carcinomas. In summary, the spectrum of thyroid malignant neoplasms developed by RET/PTC1 TG mice carrying one, two, or no Patz1-null alleles included classical and solid variants of PTC, PDTCs, and ATCs ( Figure 3). Importantly, at the observed time-window of age, the more aggressive solid variant of PTC was present in 3 out of 14 carcinomas (21%) in RET/PTC1;Patz1 +/− mice, compared with 1 out of 12 carcinomas (8%) in RET/PTC1 TG ;Patz1 +/+ mice. Both classical and solid variants of PTC were present in equal percentage in RET/PTC1 TG ;Patz1 −/− , while the classical variant was predominant on the solid one in either RET/PTC1 TG ;Patz1 +/− (~3:1 ratio) or RET/PTC1 TG ;Patz1 +/+ (10:1 ratio) ( Figure 3). Therefore, even though heterozygous loss of Patz1 does not change the time-dependent incidence of thyroid tumors in RET/PTC1 TG mice, histopathological analyses revealed a significant difference in thyroid tumor phenotype distribution developed in RET/PTC1 TG ;Patz1 +/− mice compared to compound RET/PTC1 TG ;Patz1 +/+ mice (p < 0.0001, chi-square test) and RET/PTC1 TG ;Patz1 −/− (p < 0.0001, chi-square test) (Figure 3), indicating a progressive cooperation between Patz1 allelic loss and RET activation toward an increasingly malignant phenotype resulting in the development of aggressive, metastatic tumors closely resembling human thyroid carcinomas.

Patz1-Null Mutation Enhances Proliferation of Thyroid Cancer Cells in RET/PTC1 TG Mice
To deeper investigate phenotypic differences in RET/PTC1 TG mice carrying one, two, or no Patz1-null alleles, we analyzed expression of Ki-67 (a typical marker of cell proliferation) by immunohistochemistry on tumor samples from each genotypic group. The results have shown a significant increase in the percentage of Ki-67 positive cells (p < 0.01; Tukey's multiple comparisons test) in thyroid tumors from either RET/PTC1 TG ;Patz1 +/− or RET/PTC1 TG ;Patz1 −/− mice, compared to those from RET/PTC1 TG ;Patz1 +/+ mice. Conversely, no significant differences were observed between RET/PTC1 TG ;Patz1 −/− and RET/PTC1 TG ;Patz1 +/− mice ( Figure 4). These results indicate a cooperative role of Patz1-null mutation with RET/PTC1 in enhancing thyroid cancer cell proliferation, accounting for the higher aggressive phenotype of tumors developed by RET/PTC1 TG mice, heterozygous or homozygous for the Patz1-knockout mutation, compared to single mutant RET/PTC1 TG mice.
Interestingly, PATZ1 expression was either reduced or lost in thyroid carcinomas derived by either RET/PTC1 TG ;Patz1 +/+ or RET/PTC1 TG ;Patz1 +/− mice, respectively, compared to normal thyroids or hyperplastic lesions ( Figure 5). In particular, PATZ1 positivity was associated with more differentiated tumor areas in a same papillary tumor with solid aspects (data not shown). This is consistent with previous results in human thyroid carcinomas [11,12], and may explain why we did not observe significant differences in the proliferation index of RET/PTC1 TG ;Patz1 +/− and RET/PTC1 TG ;Patz1 −/− thyroid cancers. According to our previous data showing that the maintenance of PATZ1 expression in thyroid cancer cells was associated with expression of E-cadherin [11], we found that the rare cases of PATZ1 retention in thyroid cancer cells were associated with immuno-histochemical detection of E-cadherin, which was not or barely detectable in all the other thyroid cancer samples analyzed, where also PATZ1 expression was lost ( Figure 5).

Discussion
It has been previously reported that PATZ1 is downregulated in thyroid cancer cell lines and tissues compared to normal thyroid cell lines and tissue, and its expression is inversely correlated with the degree of malignancy of thyroid carcinomas being lower in PDTC and ATC compared with PTC [10,11,23], then suggesting a tumor suppressor role of PATZ1 in the progression from PTC to ATC. Such a hypothesis has been supported by in vitro studies showing that restoration of PATZ1 in rat and human malignant thyroid cells, including PTC and ATC cell lines, inhibits cell proliferation, migration, and invasion, that are, conversely [10,11,23], enhanced by PATZ1 silencing in both normal and malignant thyroid cells [11].
Then, in order to validate the tumor suppressor role of PATZ1 in thyroid carcinogenesis in vivo, we crossed a mouse model of PTC, carrying the RET/PTC1 oncogene under the thyroid-specific control of the bovine thyroglobulin promoter [13] with mice knockout for the Patz1 gene [21].
In humans, PTC can be subdivided in several histologic variants, showing distinct patterns of growth and clinical behavior, which include: (i) classical, with papillary architecture and Psammona bodies (scarred and calcified remnants of infarcted papillae), the most common; (ii) follicular, with cells organized in follicles, accounting for approximately 10% of all PTCs; (iii) oncocytic or Hurthle-cell, characterized by cells with abundant eosinophilic granular cytoplasm as a result of accumulation of altered mitochondria, accounting for about 3-10% of all differentiated thyroid cancers, and present also as a variant of FTC; (iv) tall-cell, with cells two to three times as tall as they are wide, showing abundant eosinophilic cytoplasm, occurring in about 10% of PTCs; (v) cribriform morular, associated with familial adenomatous polyposis, with interspersed balls of squamoid cells or morules; (vi) solid, characterized by solid sheets, more common in children and associated with the Chernobyl nuclear accident and sometimes defined as poorly differentiated carcinoma with insular patterns; (vii) columnar, with elongated nuclei in tall cells, very rare [24].
Our results show that homozygous deletion of the Patz1 gene worsens outcome in RET/PTC1 mice, since RET/PTC1 TG ;Patz1 −/− mice develop thyroid carcinomas four months earlier than RET/PTC1 TG ;Patz1 +/+ controls, and induces a thyroid cancer phenotype characterized by the presence of a higher number of proliferating cells and an increased incidence of the solid variant with respect to controls. Interestingly, RET/PTC1 TG mice heterozygous for the Patz1-knockout mutation do not significantly differ from RET/PTC1 TG ;Patz1 +/+ control mice as far as thyroid tumor incidence is concerned, but their thyroid cancer phenotype is significantly more aggressive than that of controls. Indeed, it is undistinguishable from that of RET/PTC1 TG ;Patz1 −/− in terms of Ki-67 expression. Moreover, these mice have developed ATCs and an increased number of solid variants of PTC compared to RET/PTC1 TG ;Patz1 +/+ compounds. Therefore, RET/PTC1 TG ;Patz1 +/− mice show a thyroid cancer phenotype intermediate between that of RET/PTC1 TG ;Patz1 +/+ and RET/PTC1 TG ;Patz1 −/− mice.
A local lymph node metastasis was occasionally observed in a RET/PTC1 TG ;Patz1 +/− mouse carrying a solid variant of PTC. However, we did not systematically analyze lymph nodes or distant organs of all mice carrying thyroid tumors. Therefore, we cannot exclude that the number of metastases could be even higher than that one observed in our study. Future studies specifically focused on metastatization will be helpful to clarify this issue.
It is worth noting that the solid variant of PTC is associated with a less favorable prognosis than classical PTC [25]. Indeed, recent studies have shown that the presence of the solid component in PTC, regardless of the proportion, is associated with adverse clinical parameters and a shorter disease-free survival [26,27]. At a molecular level, the solid component is enriched in the expression of cancer stem cell markers ATP-binding cassette G2 (ABCG2) and multidrug resistance associated protein 1 (MRP1) that were absent or significantly lower expressed in the papillary component of the same tumor, whereas they are frequently overexpressed in ATC and are related to adverse clinical outcomes [28,29]. This is consistent with the idea that the solid component is less differentiated and may be a progression toward a poorly differentiated or anaplastic phenotype.
The ability of PATZ1 to negatively regulate the EMT process [11,12] likely accounts for the higher malignant phenotype of the RET/PTC1 TG mice carrying a complete or a partial impairment of PATZ1 function. Consistently, it has been reported that EMT plays a key role in the development of ATC [30][31][32][33], and there are evidences in animal models that ATC can occur from preexisting PTC, passing through a PDTC [16,34]. Therefore, accordingly, we report that both E-cadherin and PATZ1 expression are absent in all PDTCs and ATCs developed by these mice, similar to that already described in human thyroid cancer, where PATZ1 expression is downregulated or lost in ATCs [11,12]. On these bases, we could speculate that PATZ1 loss is required to convert the classical variant of PTC into the solid one, accounting for the acquisition of a less differentiated and more aggressive phenotype. However, further experiments that more deeply investigate both upstream and downstream pathways involving PATZ1 in thyroid carcinogenesis will be necessary to fully address this hypothesis.
It is worth noting that Patz1-knockout mice, previously characterized in a mixed c57BL/6J × 129SvJ genetic background [22], did not spontaneously develop thyroid carcinomas during their lifespan [10], while they do with the new mixed FVB/N × c57BL/6J × 129SvJ genetic background. Indeed, we showed here that RET/PTC1 WT ;Patz1 +/− mice develop thyroid carcinomas even with a minor incidence and a longer latency than RET/PTC1 TG ;Patz1 +/− mice. This is consistent with previous reports in another mouse model of PTC, the transgenic mice expressing the BRAF V600E oncogene under the control of the bovine thyroglobulin promoter, in which the genetic background of the mouse strain has a crucial role on phenotype determination [35,36]. A role of RET/PTC-RAS-BRAF signaling pathway as an initial event in thyroid carcinogenesis has been described [37]. The observed PATZ1 reduction during thyroid cancer development could be a consequence of an activation of the RET/PTC-RAS-BRAF signaling. Indeed, we have previously demonstrated that oncogenic RAS downregulates PATZ1 during thyroid carcinogenesis and that PATZ1 overexpression inhibits the malignant phenotype of thyroid cells transformed by the oncogenic Ras [24]. However, our data on Patz1 +/− mice that do not express RET/PTC1 in thyroid cells (Figure 1b) suggest that PATZ1 downregulation could be itself an initial event of thyroid carcinogenesis, independently from RET/PTC1.
In conclusion, even if with the limits of a small number of mice carrying the homozygous null mutation of Patz1 (due to the embryonic lethality) and the lack of a systematic study on the metastatic behavior, the results presented here provide compelling evidence that impairment of PATZ1 expression promote the occurrence and aggressiveness of thyroid tumors in RET/PTC1 TG mice, and per se may also be an initial event in thyroid carcinogenesis. Moreover, this genetically engineered mouse model, by recapitulating key features of the human ATC and aggressive PTC, may represent a helpful model for future research either in vivo or in vitro (by generating tumor-derived cell lines) in the development/evaluation of new therapeutic approaches.
All mice were maintained under standardized nonbarrier conditions in the Laboratory Animal Facility of Istituto dei Tumori di Napoli (Naples, Italy). All studies were conducted in accordance with the 3Rs principle and Italian regulations for experimentations on animals (prot. no. 997/2014 approved by the Italian Ministry of Health on 3 March 2014).

Genotyping
Three polymerase chain reactions (PCR) were performed on genomic DNA from tail clippings. Primers used to amplify a 203-bp DNA fragment from RET/PTC1 transgenic mice as previously reported (KD2: 5 -AGTTCTTCCGAGGGAATTCC-3 and TPC4: 5 -GTCGGGGGGCATTGTCATCT-3 ) [14]. A set of three primers was used to detect both normal and mutant Patz1 alleles. To detect the Patz1-knockout allele, a 450-bp DNA fragment was amplified between a sequence inside the neomycin cassette and a sequence downstream of the replaced region of the Patz1 gene, using the primers Pak5b: 5 -GCCTTCTTGACGAGTTCTTC-3 and Pa3: 5 -CCACACCATCAAAGTTGG-3 . To detect the Patz1 wild-type allele, a 385-bp DNA fragment was amplified between a sequence overlapping the replaced region in the knockout mutant and a sequence downstream of it and common to both wild-type and knockout alleles, using the primers PaN5: 5 -AAGCAAGTGGCTTGTGAG-3 and Pa3. For all PCR cycling conditions were as follows: 1 95 • C followed by 35 cycles of 15" 95 • C; 15" 55 • C; 15" 72 • C.

Histopathology and Immunohistochemistry
Histological evaluation of the thyroid gland was performed on all mice. Representative tissues were fixed overnight in 10% neutral buffered formalin, processed by routine methods, and embedded in paraffin. Sections (5 µm) were stained with ematoxylin and eosin. Blinded (i.e., without knowledge of genotype) histological evaluation included classification of tumor morphology (papillary, classical or solid, poor differentiated or ATC). Immunohistochemical staining was performed on 5 µm paraffin sections. Endogenous peroxidase was inhibited by 0.3% hydrogen peroxide in methanol for 30 min. For antigen retrieval, slides were microwaved in a DAKO autostainer in 0.01 M citric acid for 10 min and then quenched in 1% H 2 O 2 .
Polyclonal rabbit primary antibodies were against Ki-67 (ab 15580, 1:200, Abcam, Cambridge, UK), PATZ1 (custom ab R1P1, Primm [11]), and anti-E-cadherin (610181, BD Transduction Laboratories, BD Italia, Milan, Italy). The secondary antibody for all primary antibodies was biotinylated goat anti-rabbit antibody (Vector Laboratories, Burlingame, CA, USA). Specific binding was amplified using the streptavidin-biotin immunoperoxidase technique (DAKO). Chromogen reaction was developed with 3-3 diaminobenzidine (DAB) solution (DAKO), and nuclei were counterstained with Mayer's hematoxylin. Negative controls were performed by omitting the primary antibody. Ki-67 expression was quantified by counting the percentage of immunopositive cells with respect to total cells, evaluating the mean of at least three fields for each tumor. At least four mice per group were included in this analysis. Conversely, immunohistochemical analysis of both PATZ1 and E-cadherin expression was only qualitative.

Statistical Analysis
Log-rank (Mantel-Cox) test was applied to analyze differences in Kaplan-Meier survival curves. Tumor distribution of the three Patz1 genotypes of RET/PTC1 mice was analyzed to determine whether differences were statistically significant at 10 to 17 weeks of age. Pearson's x 2 test was performed on data expressed as percentage. Ordinary one-way ANOVA followed by Tukey's multiple comparisons test was applied for Ki-67 analysis. All tests were assessed using GraphPad Prism 6 software, La Jolla (CA), USA. Statistical significance was indicated by p < 0.05.