Synergistic repression of thyroid hyperplasia by cyclin C and Pten

Differentiated thyroid carcinomas, such as follicular or papillary tumors, are the most common endocrine malignancies (Aschebrook-Kilfoy et al., 2013). Given the requirement of large quantities of H₂O₂ for hormone production in the thyroid, it is not surprising that oxidative stress is a contributing factor in thyroid cancer initiation and progression (reviewed in Kim, 2015). Although prevalent, these diseases are relatively easily treated by established approaches. Conversely, the undifferentiated anaplastic disease still represents a challenge to effectively treat. Many studies support a model that thyroid cancer represents the progression from the well-differentiated disease type to a more aggressive tumor type due to acquisition of additional mutations (Pacini and DeGroot, 2000).

Thyroid carcinogenesis is characterized by both oncogene activation as well as loss of tumor suppressor activity. Activation of the oncoprotein B-Raf is associated with a high percentage of thyroid tumors (Kimura et al., 2003; Knauf et al., 2005). More recently, tumor suppressors have been reported that restrict hyperplasia and carcinogenesis. For example, mutations in the tumor suppressor phosphatase and tensin homolog (Pten) (Li et al., 1997; Steck et al., 1997) have been identified in many tumor types (Bonneau and Longy, 2000; Di Cristofano et al., 2001). In a murine model, Pten loss resulted in thyroid neoplasia closely resembling the human disease (Antico Arciuch et al., 2011; Yeager et al., 2008). In addition, combining Pten mutation with loss of p53 (also known as TP53) function greatly accelerated carcinogenesis in this mouse model system (Antico Arciuch et al., 2011) suggesting that multiple pathways can influence thyroid tumor development. For example, the Jak-Stat3 signaling pathway was initially implicated in the acceleration of many tumor types (O'Shea et al., 2015). However, recent studies have revealed a role for the Jak-Stat pathway, and Stat3 in particular, in retarding thyroid tumor progression (Couto et al., 2012). These findings indicate that Jak-Stat can exert very different activity dependent on cell type and cell context controls.

Cyclin C–Cdk8 is a highly conserved protein kinase that associates with the RNA polymerase II holoenzyme that both positively and negatively regulates gene expression in mice (Li et al., 2014; Stieg et al., 2019) and other systems (reviewed in Bourbon, 2008; Nemet et al., 2014). This dual activity is illustrated by its role in both tumor suppression and promotion (Xu and Ji, 2011). For example, cyclin C–Cdk8 co-stimulates the Wnt/β-catenin pathway in colon cancer (Firestein et al., 2008). Conversely, cyclin C–Cdk8 suppresses tumor progression in T-cell acute lymphoblastic leukemia (T-ALL) (Li et al., 2014) by negatively regulating Notch signaling (Fryer et al., 2004). Cyclin C–Cdk8 has also been implicated in the Jak-Stat pathway by phosphorylating Stat3 on Ser727 (Bancerek et al., 2013), a phosphorylation mark that is also mediated by other kinases including the Erk MAPKs (Chung et al., 1997). Given the diversity of signaling kinases, it is perhaps not surprising that this modification can have a positive or negative impact on Stat3-dependent transcriptional activation (Chung et al., 1997; Wen et al., 1995).

In addition to its role in transcription, cyclin C also has a Cdk8-independent role in mediating stress-induced mitochondrial fragmentation and mitochondrial outer membrane permeability (MOMP) (Ganesan et al., 2019; Wang et al., 2015), the commitment step to intrinsic mitochondrial-dependent regulated cell death type 1 (RCD-1) (Galluzzi et al., 2015). This function is highly conserved, as cyclin C is required for mitochondrial fission and stress-induced cell death in budding yeast (Cooper et al., 2014) and mammalian cells (Jezek et al., 2019). Finally, deleting cyclin C protects mouse embryonic fibroblast (MEF) cells from RCD-1 induced by oxidative stress or the anti-cancer drug cisplatin (Wang et al., 2015). These results suggest that, in addition to its transcriptional function, the mitochondrial function of cyclin C may also play a role in suppressing tumor formation. The cyclin C locus (Ccnc) maps to 6q21 (Demetrick et al., 1995), a chromosomal location that is frequently deleted in several types of cancers including high-grade non-Hodgkin's lymphomas (Offit et al., 1993) and osteosarcomas (Ohata et al., 2006). In addition, 6q21 is lost in 33% of poorly differentiated thyroid tumors and 27% of anaplastic malignancies (Wreesmann et al., 2002). Interestingly, Ccnc was not mutated in the more easily treated well-differentiated thyroid cancers. These results suggested that Ccnc loss supports tumor progression, but not initiation. In this report, [Pten; Ccnc]^thyr−/− double-mutant mice were generated and thyroid tumor progression monitored. These studies revealed a synergistic increase in thyroid hyperplasia in double-mutant animals. Further analysis supported a combined role for these two factors in maintaining p21 (also known as CDKN1A) and p53 tumor suppressor factors in thyrocytes.

Previous studies described the generation of floxed alleles of Ccnc (Ccnc^fl/fl) (Wang et al., 2015) and Pten (Pten^fl/fl) (Yeager et al., 2007). To determine whether cyclin C plays a role in thyroid cancer progression, Ccnc^fl/fl mice were crossed with Pten^fl/fl; thyroid peroxidase (Tpo)-Cre mice to generate mouse lines with thyroid-specific deletion(s) of Ccnc and/or Pten (denoted Ccnc^thyr−/− and Pten^thyr−/−). Tpo is a thyroid-specific gene that is induced between 14.5 and 16.5 days post coitus (dpc) corresponding to the final differentiation stages of this gland (De Felice et al., 2004). As previously reported (Antico Arciuch et al., 2011), Pten^thyr−/− mice displayed no survival defect for up to 60 weeks (Fig. 1A). Similar survival results were obtained for Ccnc^thyr−/− mice. However, [Pten; Ccnc]^thyr−/− mice exhibited early lethality starting at 6 weeks with no animal surviving past 26 weeks. Animal death in this approximate time frame was observed previously when Pten thyroid-specific deletions were combined with either Kras activating (Miller et al., 2009) or Tp53 deletion (Antico Arciuch et al., 2011) alleles. In both reports, pathological analysis identified hyperplasia and oncogenesis in the thyroid. Therefore, the thyroid glands from 20-week-old single- and double-mutant mice were examined (Fig. 1B). Ccnc^−/− thyroids (including trachea tissue) were similar in size to wild-type controls (Fig. S1A). Consistent with previous reports (Yeager et al., 2007), the Pten deletion resulted in a 2.5-fold size increase over wild-type or Ccnc^thyr−/− thyroids (Fig. 1B). However, the [Pten; Ccnc]^thyr−/− double-mutant mice exhibited very enlarged thyroids, nearly 6-fold more massive than Pten mutants alone. These results indicate that loss of both cyclin C and Pten activity exhibits a synergistic increase in thyroid size.

To further quantify the impact of Ccnc and Pten deletion on thyroid size, the thyroids of single- and double-mutant 12-week-old male mice were examined. We chose a 12-week endpoint as the majority of the [Pten; Ccnc]^thyr−/− mice did not display any overt pathologies at this time. In addition, males were chosen for this study as sexual differences were previously observed in phenotypic expression of thyroid disease in Pten^thyr−/− mutant animals (Yeager et al., 2007). As observed with the 20-week-old animals, Pten^−/− thyroids were 2-fold larger than Ccnc^−/− single-mutant organs (Fig. 1C). Again, the double-mutant organs displayed an 8-fold increase in thyroid size compared to either single-mutant indicating that cyclin C and Pten function in a synergistic manner to suppress hyperplasia in thyroids. Furthermore, analysis of [Pten; Ccnc]^thyr−/− females revealed a similar increase in thyroid size (Fig. S1B) indicating that sex is not a factor with this phenotype.

To further investigate the enlarged thyroid phenotype, hematoxylin and eosin (H&E) staining was used to examine thyroid morphology from mice with the different genotypes described above. As previously observed (Yeager et al., 2007), deleting Pten resulted in increased follicle size (compare Fig. 2A with 2B) with multiple layers of the surrounding follicular cells (blue arrows, Fig. 2B). Thyroid-specific deletion of Ccnc did not qualitatively change the thyroid size or morphology (compare Fig. 2A with 2C) although additional layers of follicular cells were frequently observed (blue arrows, Fig. 2D). Follicular enlargement was dramatically increased in the Pten^−/−; Ccnc^−/− double knockout thyroids compared to wild-type or either single mutant (Fig. 2E,G). In addition, the double knockout mice exhibited thick fibrotic cell layers (yellow arrow, Fig. 2G) and potential invasion of thyrocytes into a blood vessel (green arrow, Fig. 2F). However, no clear tumors were observed in the other four mice examined. These results suggest that, although thyroid-specific knockout of Ccnc and Pten together promote extensive hyperplasia, these mice did not die as result of tumor formation per se. One possibility for death is that the enormous size of the thyroid might lead to collapse of the trachea. This possibility is supported by the lack of metastasis following gross examination (data not shown) and the absence of well-formed tumors in the thyroid. The synergistic effects of deleting both genes raise the possibility that Ccnc and Pten play redundant roles in restricting follicle enlargement and accumulation.

H&E staining revealed enlarged thyroids with the accumulation of multiple layers of follicular cells in the [Ccnc; Pten]^thyr−/− double knockout mice (Fig. 2). These results suggested that the double-mutant cells were undergoing extensive proliferation. To address this possibility, thyroid sections prepared from 12-week-old male mice of the different genotypes were reacted with an antibody recognizing the proliferation-associated antigen Ki-67. Wild-type and Ccnc^−/− thyroids displayed similar frequencies of Ki-67-positive cells (0.3% and 0.8%, respectively, Fig. 3A), consistent with the nearly normal size of the Ccnc^−/− organ. As previously reported (Antico Arciuch et al., 2011), the Pten^−/− tissue exhibited an elevated frequency of Ki-67-positive cells (2.8%). However, the double-mutant thyroid displayed a 3.9-fold increase in the number of Ki-67-positive cells over the Pten^−/− thyroids. These results suggest that the cause of the elevated follicular cell number is enhanced proliferation. To further evaluate this possibility, the replicative index of the thyroids were analyzed in living animals by analyzing 5-bromo-2′-deoxyuridine (BrdU) incorporation. Following a 2-hour incorporation time, animals were killed and thyroids dissected. Consistent with the Ki-67 results, BrdU staining of these tissue samples also revealed significantly elevated proliferation in [Pten; Ccnc]^thyr−/− double knockout thyrocytes compared to what is found in wild-type or either single mutant (Fig. 3B). Taken together, these findings indicate that the increased size of the double-mutant thyroids is due to aberrant elevated proliferation.

Oxidative stress is a contributing factor in many cancers, including in the thyroid (Poncin et al., 2008; Valko et al., 2006). We found that cyclin C translocates from the nucleus to the mitochondria in response to several stressors including H₂O₂ or the anti-cancer drug cisplatin (Wang et al., 2015). Both compounds share the characteristic of inducing reactive oxygen species (ROS) as part of their cytotoxic activity (Martins et al., 2008). In addition, thyroid disease is associated with ROS generation by Duox enzymes, which are necessary for hormone production (Lambeth, 2007). Therefore, inactivation of the cyclin C re-localization pathway may lead to reduced cell death and/or enhanced proliferation. To test whether the cyclin C re-localization system was functioning in thyroid cancer cell lines in vitro, we monitored the subcellular localization of cyclin C and mitochondrial morphology in four mouse thyroid cancer cell lines. The anaplastic thyroid cancer (ATC) lines T1903 and T1860 are deleted for both Pten and Tp53 tumor suppressors (Antico Arciuch et al., 2011). The poorly differentiated thyroid carcinoma (PDTC) cell lines D316 and D445 contain an activated Kras allele (Kras^G12D) in addition to homozygous Tp53 deletion (Champa et al., 2016). Actively dividing cells were treated with H₂O₂ (0.4 mM, 4 h), fixed and examined for the subcellular localization of cyclin C, the mitochondria and nuclei. These studies found that cyclin C was predominantly nuclear prior to stress application as expected for a transcription factor (Fig. 4). Following H₂O₂ exposure, partial cyclin C nuclear release was observed in all four cell lines. In addition, a statistically significant increase in mitochondrial fragmentation was observed in three of the four cell lines, consistent with the ability of cyclin C to induce mitochondrial fission with efficiencies similar to those observed previously with immortalized MEF and HeLa cultures (Wang et al., 2015). Although mitochondrial fragmentation increased in D445 cells, the magnitude did not reach statistical significance, perhaps due to the elevated fission rate observed in unstressed cells. Merging the cyclin C and mitochondrial signals revealed overlap in the T1860 and D316 ROS-stressed cultures (arrows in magnified images). However, the overlap between the mitochondrial and cyclin C signals was less prominent in T1903 and D445 cells. Although the reason for this observation is unknown, this result may reflect differences in dwell times for cyclin C at the mitochondria or a more rapid response than could be seen at the time point examined. Taken together, these results indicate that the cyclin C re-localization pathway is still active in these tumor cell lines.

Cyclin C exhibits stress-induced nuclear release in budding yeast (Cooper et al., 2014, 2012) similar to that observed in mammalian cells (Wang et al., 2015). Subsequent studies revealed that the yeast cyclin C is retained in the nucleus through interaction with the Med13 protein (Khakhina et al., 2014; Stieg et al., 2018). The binding domain on cyclin C directing this interaction is termed the holoenzyme-associating domain or HAD (Cooper and Strich, 1999). We recently demonstrated that addition of a cell-penetrating stapled HAD mimetic peptide (S-HAD) induced cyclin C nuclear release in the absence of stress (Jezek et al., 2019). The finding that cyclin C is present in thyroid cancer cell lines prompted the question of whether S-HAD had a similar activity in PDTC cell line D316. Treating D316 cells with S-HAD (10 µM) for 2 h resulted in a significant increase (P=0.01, Student's t-test, three independent cultures) in mitochondrial fragmentation (68±9%) compared to control (22±6%) (mean±s.d.; see Fig. 5A for representative images). In addition to inducing mitochondrial fission, S-HAD treatment also increased the sensitivity of HeLa cells to cisplatin by 2-fold (Jezek et al., 2019). Therefore, we tested the impact that inducing cyclin C nuclear release had on cisplatin sensitivity in either immortalized MEF or D316 cells. Both cell lines were treated with S-HAD (10 µM) for 2 h prior to cisplatin addition (15 and 30 µM for MEF and D316 cells for 24 and 48 h, respectively). RCD-1 frequency was measured by determining the proportion of cells that are both positive for annexin V and negative for propidium iodide (PI). This criteria allows the removal of necrotic cells from this analysis. The MEF and D316 cell lines both exhibited a significant, but modest, increase (30% and 50%, respectively) in cisplatin-induced RCD-1 (Fig. 5B,C). Treatment with S-HAD alone did not induce cell death (Fig. 5B). These results indicate that manipulating the cyclin C subcellular localization alters mitochondrial morphology and the sensitivity of thyroid tumor cell lines to cisplatin. An alternative explanation for these results is that nuclear release of cyclin C could result in changes in gene transcription through inactivation of Cdk8. To test this possibility, D316 cells were treated with two different Cdk8 inhibitors, Senexin A (Porter et al., 2012) and compound 32 (Koehler et al., 2016), before cisplatin addition. These experiments found no change in cisplatin toxicity (Fig. S2A) or in overall cyclin C and Cdk8 levels following Cdk8 inhibition (Fig. S2B). Taken together, these results suggest that mitochondrial localization of cyclin C, not altered cyclin C–Cdk8-mediated gene regulation, is responsible for the increased sensitivity of D316 cells to cisplatin.

We previously reported that deleting Ccnc in MEF cells resulted in protection from cisplatin-induced RCD-1 (Wang et al., 2015). To test whether D316 cells exhibited a similar phenotype, cyclin C levels were reduced using Ccnc-specific siRNA treatment (48 h) as determined by western blot analysis (Fig. 5D). To determine whether a reduction in cyclin C levels affected cisplatin sensitivity, the Ccnc knockdown D316 cells were treated with cisplatin (30 µM) for 24 h. RCD-1 efficiency was determined by quantifying the percentage of the population that were annexin V positive and PI negative. Surprisingly, reducing cyclin C levels had no effect on RCD-1 efficiency (Fig. 5E). These results suggest that the RCD-1 control pathways differ in untransformed MEF cells versus thyroid cancer cells.

Our results suggest that there is a mitochondria-independent role for cyclin C in restraining hyperplasia in follicular cells. Previous studies have indicated that cyclin C–Cdk8 phosphorylates Stat3 on Ser727 (Bancerek et al., 2013) and that Stat3 suppresses tumor progression in thyroids (Couto et al., 2012). To investigate a potential role for cyclin C–Cdk8 in Stat3 regulation in the thyroid, protein extracts were prepared from thyroids isolated from 18–20-week-old mice and probed for the presence of Stat3 and the Ser727-phosphorylation species by western blotting. Compared to wild-type thyroids (Fig. 6A, lanes 1,2; Fig. 6B, lanes 8,9), Stat3 levels did not appreciably change in Pten^+/− (Fig. 6A, lanes 6,7, quantified in Fig. 6C) or Pten^−/− (Fig. 6A, lanes 3–5; Fig. 6B, lanes 10–12) thyroids. Homozygous or heterozygous Ccnc status (+/+ or +/−) did not alter this result. Interestingly, the double [Pten; Ccnc]^thyr−/− mutant thyroids exhibited significant reductions in Stat3 levels (Fig. 6B, lanes 13–15, quantified in Fig. 6C) that were not observed in either single mutant.

Next, the phospho-Ser727 signal was examined in these same extracts. A 50% reduction in the single Pten^−/− (Fig. 6A, lanes 3–5; Fig. 6B, lanes 10–12) mutant thyroids (quantified in Fig. 6D) was observed. The level of the phospho-Ser727 species was reduced even further in Ccnc^−/− single mutants to 25% of wild-type levels (Fig. 6A, lanes 6–7, quantified in Fig. 6D). Analysis of the double-mutant thyroids (Fig. 6B, lanes 13–15) revealed an additive reduction in Stat3 phosphorylation compared to either single mutant. These findings suggest that these two factors support Stat3 phosphorylation through separate pathways. Taken together, these results indicate that Stat3 levels and Ser727 phosphorylation is supported by cyclin C and Pten. To confirm these findings, immunohistochemistry (IHC) was performed probing for phospho-Ser727 in fixed tissue slices. As observed by western blot analysis, positive signals were observed in Ccnc^+/+ tissues regardless of Pten status (Fig. 6E). However, this signal was below the limits of detection in Ccnc^−/− thyroid sections. Although Pten is required for normal Stat3 phosphorylation, these results indicate that cyclin C–Cdk8 plays the predominant role in maintaining this modification.

The results just described indicate that cyclin C–Cdk8 is required for Stat3 Ser727 phosphorylation in the thyroid. However, as Ser727 phosphorylation was reduced in Ccnc^−/− mutant thyroids, this result may not fully explain the dramatic hyperplastic phenotype observed only with the double mutant. Cyclin C–Cdk8 inhibits Notch signaling via destruction of the intracellular part of the Notch receptor [ICN; also known as the Notch intracellular domain (NICD)] (Fryer et al., 2004). Notch signaling has been reported to impede thyroid tumor progression. For example, Notch activation promotes differentiation and is downregulated in some thyroid tumors (Ferretti et al., 2008). In addition, pharmacologically stimulating Notch1 in ATC cell lines produced a dose- and time-dependent decrease in proliferation (Patel et al., 2014). However, Notch also represses Pten transcription (Palomero et al., 2007; reviewed in Hales et al., 2014) leaving open the question of the role of this pathway in thyroid cell expansion. Therefore, we measured Notch signaling in Ccnc^−/− and Ccnc^+/− thyroids. Quantitative PCR (qPCR) was used to monitor mRNA levels of two Notch targets, Hes5 (Kobayashi and Kageyama, 2014) and Sox9 (Capaccione et al., 2014). Thyroid RNA was prepared from three 20-week-old mice of each genotype then subjected to qPCR analysis for Hes5 and Sox9 mRNA levels, using Gapdh transcription as the control. These experiments found no significant difference between Hes5 and Sox9 mRNA levels in Ccnc^−/− and Ccnc^+/− tissues (Fig. S3). Therefore, our results argue against upregulation of the Notch pathway contributing to hyperplasia in the cyclin C-depleted thyroids.

Next, we examined the levels of p53 and p21, tumor suppressors that act early in the transformation process. As above, we isolated thyroids with different Pten and Ccnc genetic compositions and probed total protein extracts for the presence of p53 and p21 by western blotting. In the absence of either Pten or Ccnc, no changes in p53 or p21 levels were observed (Fig. 7A). Surprisingly, the double-null thyroids displayed p21 and p53 levels at or below the limits of detection (Fig. 7B). These results indicate either Pten or cyclin C is sufficient to maintain the levels of these tumor suppressors. To gain insight into the mechanism behind this reduction in protein levels, qPCR was again employed to determine whether the mRNA levels of Tp53 or Cdkn1A were altered in these mutant tissues. In these experiments, two or three thyroids of each possible genotype were examined and the results averaged. These experiments revealed no significant differences between wild-type and double-mutant thyroids with respect to Tp53 or Cdkn1A mRNA levels (Fig. 7C,D). These results suggest that cyclin C and Pten maintain p53 and p21 levels through a post-transcriptional mechanism. Taken together, these findings provide a potential mechanism for the hyper-proliferation observed only in the double-mutant thyroids (see Discussion).

This report details the first analysis of a tumor suppressor function for cyclin C in a solid tumor mouse model. This study revealed that deleting Ccnc in the thyroid caused only a modest increase in hyperplastic growth. However, in combination with Pten ablation, a dramatic increase in organ size was observed to the point that the animal succumbed to this aberrant growth in ∼20 weeks. Two roles have been previously described that are consistent with a tumor suppressor role for cyclin C. First, cyclin C–Cdk8 suppresses T-ALL progression in mice through downregulation of Notch signaling (Li et al., 2014). The second function is Cdk8 independent, and occurs in the cytoplasm when cyclin C is released from the nucleus following oxidative stress to induce extensive mitochondrial fragmentation and RCD-1 execution (Ganesan et al., 2019; Jezek et al., 2019; Wang et al., 2015). However, neither of these mechanisms appears to be responsible for the extensive neoplastic growth observed in this study. Rather, we find alterations in the Jak-Stat signaling pathway and a dramatic reduction in both p53 and p21 levels only in the double-mutant thyroids (Fig. 7). These results suggest a new mechanism by which cyclin C–Cdk8 and Pten suppress early events in thyroid tumor development.

Loss of cyclin C function results in resistance to oxidative stress in both yeast (Krasley et al., 2006) and mammalian cells (Wang et al., 2015). This role in RCD-1 in mammalian cells appears to be direct, as cyclin C is required for efficient mitochondrial recruitment of the pro-RCD-1 protein Bax (Jezek et al., 2019). However, we found that reducing cyclin C levels had no impact on cisplatin sensitivity (Fig. 5E). A recent report revealed that activation of the death receptor cell death pathway had little impact on ATC cell lines (Gunda et al., 2014). However, inhibiting both B-Raf and Akt1 signaling, in combination with death receptor activation, induced substantial apoptosis. Therefore, RCD-1 efficiencies might already be reduced even with normal cyclin C function. This possibility would explain the relatively modest increase in RCD-1 when cyclin C is ‘activated’ by the S-HAD peptide. Further studies are required to test the effect of S-HAD treatment in other conditions to identify enhancers of S-HAD-induced cell death.

Thyroids produce high levels of H₂O₂ to generate the hormone thyroxin. The H₂O₂ is produced through the Nox4, Duox1 and Duox2 family of cellular oxidases (Carvalho and Dupuy, 2013). There is increased H₂O₂-induced oxidative damage in the thyroid, as compared to what is seen in other endocrine organs, which manifests itself as elevated DNA damage and increased cancer incidence (Song et al., 2007). This results in the classic ‘mitogen, mutagen, carcinogen’ tumorigenic pathway in which low level ROS serves as a second messenger for growth promotion. As ROS levels increase, DNA damage and resulting oncogenic mutations can occur (Fig. 8). The tumor suppressors p53 and p21 prevent reentry into the mitotic cell cycle. We found that only deleting both Pten and Ccnc dramatically reduces p53 and p21 levels, consistent with the observed thyroid hyperplasia. This synergistic effect formally suggests that these factors have redundant functions to control p53 and p21 levels. Turnover of p53 is mediated by the ubiquitin ligase Mdm2 (Haupt et al., 1997; Honda et al., 1997). Previous studies have reported that activation of Akt1 promotes Mdm2 activity, thus stimulating p53 degradation (Zhou et al., 2001b). Thus, Akt1 activation through Pten deletion may stimulate Mdm2, promoting p53 degradation (top line, Fig. 8). However, Pten deletion alone was insufficient to reduce p53 or p21 levels in thyroid tissues, suggesting an additional cyclin C-dependent pathway is at work stabilizing these tumor suppressors.

The Jak-Stat signal transduction pathway is activated by the IL-6 cytokine family and has a complex role in thyroid cancer development (O'Shea et al., 2015). Stat3 has been correlated with tumor suppression (Couto et al., 2012; Kim et al., 2012) and chemoresistance (Francipane et al., 2009) of thyroid cancer. Stat3 is reported to stimulate Arf (also known as Cdkn2a) expression in murine prostate cancer models (Pencik et al., 2015). This result may be instructive to the present study, as deleting Pten and Stat3 individually resulted in a small increase in prostate size while the double mutant exhibited a >10-fold increase in gland mass (Pencik et al., 2015). Arf inhibits Mdm2 function, thus increasing p53 stability (Zhang and Xiong, 2001). This study identified Stat3 as an inducer of Arf function. Our finding that cyclin C–Cdk8 is required for Ser727 phosphorylation places it upstream of Stat3 function (Fig. 8, bottom line). Therefore, constitutive activation of Akt1, combined with loss of Stat3 activity, could enhance Mdm2-dependent p53 degradation. These results suggest that parallel regulatory systems could prevent hyperplastic growth in thyroid and prostate. However, differences in the two systems were observed upon examination of p53 levels in mutant prostate epithelium. Pten^−/− prostates display elevated p53, consistent with the Pten inactivation cellular senescence (PICS) response (Chen et al., 2005). We did not observe this response in Pten^−/− thyroids. Taken together, these findings suggest that prostate and thyroid share some regulatory strategies with respect to suppressing early neoplastic growth in these tissues.

This study found that the p21 and p53 reductions in the double-mutant thyroids were due to a post-transcriptional mechanism. The control of p21 turnover is complex, and involves both ubiquitin-dependent and independent pathways (Abbas and Dutta, 2009). Akt1-mediated phosphorylation of p21 prevents its nuclear relocation and cell cycle arrest activity (Zhou and Hung, 2002). However, stability differences were not observed between the nuclear and cytoplasmic species (Zhou et al., 2001a) arguing that the reduction in p21 levels is not simply due to mislocalization. On the other hand, Akt-mediated phosphorylation on p21 Ser146 increases p21 stability (Li et al., 2002) as part of its pro-growth function. However, another study reported that Mdm2 and Mdmx triggers p21 degradation through a ubiquitin-independent, but proteasome-dependent, mechanism (Jin et al., 2003, 2008). This mechanism allows Mdm2 (and Mdmx) to direct the destruction of both p53 and p21. One potential model is that Mdm2 helps recruit p21 to the 20S proteasome particle, in particular the C8α subunit (Touitou et al., 2001). Currently, a connection between the cyclin C–Cdk8 and Pten regulatory axis and Mdm2/MDMX function has yet to be elucidated.

In conclusion, this study describes the first suppressor role for cyclin C–Cdk8 in a solid tumor. In addition, our results uncovered a new regulatory connection between cyclin C–Cdk8 and Pten to maintain thyroid follicular cell quiescence through maintenance of the p21 and p53 tumor suppressors.

All animal experiments were conducted with the institutional animal care and use committee (IACUC) review and conducted in the 129Sv background. All strains were backcrossed at least eight times prior to analysis. Littermates with the indicated genotypes were used as controls. Sex and ages of the animals are given in the text for each experiment where appropriate.

The anaplastic thyroid cancer (ATC, T1903 and T1860) and poorly differentiated thyroid cancer (PDTC, D445 and D316) mouse thyroid cancer cell lines were as previously described (Antico Arciuch et al., 2011; Champa et al., 2016). Wild-type and Ccnc^−/− immortalized MEF cells were as previously described (Wang et al., 2015). The presence of activated Kras (Kras^G12D) allele and deleted Pten, Tp53 and Ccnc were confirmed via PCR analysis of genomic DNA. These cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin (Invitrogen, Grand Island, NY, USA). To transiently silence cyclin C, D316 cells were simultaneously reverse-transfected with three different siRNA sequences (s391, s392 and a custom siRNA, sense strand, 5′-GUUAUUGCCACUGCUACGGtt-3′) (Ambion, Austin, TX; 60 pmol each), at 50,000 cells/well in a 12-well plate using Lipofectamine RNAiMAX (Thermo Fisher Scientific) according to the manufacturer's protocol. Cells were used for experiments 2 days after the transfection.

The floxed Ccnc (Ccnc^fl) (Wang et al., 2015), Pten (Pten^fl) and the thyroid-specific Cre (Tpo-Cre) (Yeager et al., 2007) alleles have been previously described. Genotyping was accomplished using genomic tail DNA purified using the Phire genotyping kit (Thermo Fisher Scientific). The genotyping primers for Ccnc alleles were: Ccnc2 (5′-TAATCGACCAGACAGTACGGGAGTC-3′), SDL2 (5′-GGTAGTTTATCTGAACTGATGAAAACACATC-3′) and Lox1 (5′-GGAAGCAGAAGCAACAGGAATCTG-3′). The Pten alleles were identified using Pten-lox forward (5′-TGTTTTTGACCAATTAAAGTAGGCTGTG-3′), Pten-lox reverse (5′-AAAAGTTCCCCTGCTGATGATTTGT-3′), TPO-Cre forward (5′-TGTTTCTGACCAGTCAGGAC-3′) and Tpo-Cre reverse (5′-CTCGTTGCATCGACCGGTAATG-3′) as described previously (Yeager et al., 2007).

Cells were cultured on poly-L-lysine-coated cover slips for 2 days then treated as follows. H₂O₂ (0.4 mM) was added to serum-free medium 4 h prior to treatment. Cells were stained with 100 nM MitoTracker Red CMXRos (Thermo Fisher Scientific) for 30 min then fixed with 4% paraformaldehyde for 20 min, permeabilized with 0.2% Triton X-100 for 15 min, blocked with 2% BSA, and incubated with 4 mg/ml anti-cyclin C antibody (Thermo Fisher Scientific, PA5-16227) at 4°C overnight and 1 mg/ml Alexa Fluor 488-conjugated secondary antibody (Thermo Fisher Scientific, A11008) for 1 h at 23°C. The cells were mounted with 4′,6-diamidino-2-phenylindole (DAPI)-containing medium (Vector Labs, Burlingame, CA) and the images were acquired and processed with a Nikon Eclipse 90i microscope (Melville, NY) equipped with a Retiga Exi CCD camera and NIS software.

The cells were seeded in 12-well plates at a density of 0.5×10⁵ cells/well 2 days before H₂O₂ treatment. H₂O₂ (0.4 mM) was added to cells immediately following a switch to serum-free medium. Senexain A and compound 32 Cdk8 inhibitors were added to give a 10 µM final concentration for 24 h prior to stress treatment. For cisplatin treatment, the drug was added to normal culture medium at a concentration of 15 µM and 30 µM to MEF and HeLa cultures, respectively. Annexin V (BD Biosciences) assays were conducted as described by the manufacturer and quantified using a fluorescence activated cell counter (Accuri C6, BD). S-HAD treatment (10 µM, 4 h) was conducted prior to addition of cisplatin. Error bars indicate standard deviation, and statistical analysis was performed using the Student's t-test with P<0.05 considered significant.

Hematoxylin and eosin (H&E) staining was performed as previously described (Antico Arciuch et al., 2011). Ki-67 staining was conducted essentially as previously described (Saad et al., 2006). Immediately after euthanizing the mice, thyroids were collected and all connective tissue was removed. The tissues were then washed twice in formalin, weighed and fixed in formalin overnight. Then the tissues were stored in 70% ethanol for future examination. Approximately 200 cells were counted for each genotype from at least two animals. For BrdU incorporation, mice were injected with BrdU stock (5 mg/ml) at a final concentration of 10 mg/kg of body weight. After 2 hours, the animals were euthanized and the thyroids were harvested and fixed in formalin for further examination. All the tissues were embedded in paraffin and sectioned at 6 µm. Sections were subjected to antigen retrieval in 0.1 mM sodium citrate and counterstained with hematoxylin. Stat3 phosphorylation status was detected using Ser727 phospho-specific antibody (1:50; Thermo Fisher Scientific, 44-384G) in conjunction with an ImmPRESS Excel amplified HRP polymer staining kit (Vector Labs, MP-7601) using Methyl Green as a counterstain. All stained sections were photographed at 40–200× magnification and analyzed using ImageJ software.

Mouse thyroid tissues were homogenized in RIPA buffer [150 mM NaCl, 50 mM Tris-HCl pH 8, 1% Nonidet P-40 substitute, 0.5% sodium deoxycholate and 0.1% SDS] containing 1% protease inhibitor cocktail (Sigma P8340, St Louis, MO), 1% EZBlock^® phosphatase inhibitor cocktail IV (BioVision, Milpitas, CA), 10 mM NaF, 10 mM β-glycerol phosphate and 2 mM Na₃VO₄. Homogenates were incubated for 2 h at 4°C and centrifuged at 14,000 g for 20 min at 4°C to separate soluble proteins from aggregates and cell debris. Protein concentration was determined by performing a Bradford assay (Bio-Rad, Hercules, CA USA). Samples were dissolved in sample buffer (100 mM Tris-HCl pH 6.8, 4% SDS, 20% glycerol and 2 mg/ml Bromophenol Blue) supplemented with 100 mM dithiothreitol, boiled for 5 min, and probed for phospho-Ser727 Stat3 (Thermo Fisher Scientific, 44-384G, Waltham MA USA), Stat3 (Thermo Fisher Scientific, 13-7000), p53 (BD Biosciences, 554157), or p21 (Abcam ab188224, Cambridge, MA USA) at 1:5000 dilution. Western blots were run on 12% SDS-PAGE gels at 10 µg per lane, transferred to a PVDF membrane, and visualized by film exposure using alkaline phosphatase-conjugated rabbit (Abcam, ab97061) or mouse (Abcam, ab97027) secondary antibody and CDP-Star (Thermo Fisher Scientific) as a substrate. Blots were stripped and reprobed between each primary antibody application. Gapdh (Abcam, ab8245) was used as a loading control. Quantification of western blot signals was accomplished using phosphorimaging (Fuji Inc.) using Gapdh levels as an internal standard. Specific signals were obtained by subtracting background on a lane-by-lane basis. Results from two or three isolated organs were averaged and compared to each genotype indicated.

10–20 mg of frozen thyroids dissected from euthanized 18–20-week-old animals were disrupted using a rotating pestle system and total RNA prepared with an RNeasy kit (Qiagen, Germantown, MD). Poly(A)+ mRNA was enriched from these samples, and converted into cDNA using oligo dT primers. qPCR reactions and primers for Hes5, Sox9 and Gapdh analysis were obtained from GeneCopoeia (Rockville, MD) and used as per the manufacturer's instructions. Analysis was conducted using SYBR Green assays using an Applied Biosystems StepOne system. ΔCT values were calculated from target CTs subtracted from the value for the internal Gapdh control. All studies were conducted with three biological samples in duplicate.

We thank Genetech Inc. for supplying the compound 32 Cdk8 inhibitor. We also thank the Histology Core Resource at Albert Einstein Medical Center for help in preparing samples for analysis.