Differential expression of cell cycle regulators in CDK5-dependent medullary thyroid carcinoma tumorigenesis.

Medullary thyroid carcinoma (MTC) is a neuroendocrine cancer of thyroid C-cells, for which few treatment options are available. We have recently reported a role for cyclin-dependent kinase 5 (CDK5) in MTC pathogenesis. We have generated a mouse model, in which MTC proliferation is induced upon conditional overexpression of the CDK5 activator, p25, in C-cells, and arrested by interrupting p25 overexpression. Here, we identify genes and proteins that are differentially expressed in proliferating versus arrested benign mouse MTC. We find that downstream target genes of the tumor suppressor, retinoblastoma protein, including genes encoding cell cycle regulators such as CDKs, cyclins and CDK inhibitors, are significantly upregulated in malignant mouse tumors in a CDK5-dependent manner. Reducing CDK5 activity in human MTC cells down-regulated these cell cycle regulators suggesting that CDK5 activity is critical for cell cycle progression and MTC proliferation. Finally, the same set of cell cycle proteins was consistently overexpressed in human sporadic MTC but not in hereditary MTC. Together these findings suggest that aberrant CDK5 activity precedes cell cycle initiation and thus may function as a tumor-promoting factor facilitating cell cycle protein expression in MTC. Targeting aberrant CDK5 or its downstream effectors may be a strategy to halt MTC tumorigenesis.

diagnosed at late metastatic stages due to the absence of specific symptoms and are therefore often fatal. MTC arises from the thyroid parafollicular cells (C-cells), which secrete calcitonin. Most MTC cases (75%) are sporadic and about 40% are caused by somatic mutations in the RET proto-oncogene, 15% by mutation in the RAS gene, 10% by mutations in other genes and 35% by unknown causes [3][4][5]. Overall the etiology of sporadic MTC is poorly understood. Hereditary forms of MTC represent about 25% of cases and result from germline mutation in the RET proto-oncogene [6]. These genetic forms of MTC are often associated with other types of NE cancers and they are referred to as Multiple Endocrine Neoplasia of Type 2 (MEN 2). Surgical resection of the thyroid is the best treatment currently available for early stage disease but recurrence is common, particularly in sporadic MTC. The prognosis for advanced forms of MTC is poor with a five-year survival rate of 30%. FDA-approved drugs include the tyrosine kinase inhibitors, Vandetanib [7] and Cabozantinib [8], however their efficacy is limited [8,9]. Therefore a better understanding of the drivers of MTC progression, especially in the absence of RET or RAS mutations, is needed to develop more effective treatment strategies. Toward this goal, it is paramount to elucidate additional molecular mechanisms underlying MTC and identify new targets for therapy development.
We recently reported that cyclin-dependent kinase 5 (CDK5) was involved in MTC pathogenesis [10,11]. CDK5 is a serine/threonine kinase that is highly expressed in the brain and regulates neuronal function [12] but its role in cell cycle and cancer has not been well explored. CDK5 is activated by interaction with its cofactor, p35 [13], which can be cleaved by the calcium-dependent protein kinase, calpain, to produce p25. The resulting p25-CDK5 complex engenders aberrant activity with a different range of substrates. CDK5, p35 and p25 are expressed in other tissues besides brain and have been implicated in various forms of neoplasms, including thyroid [10,11], pancreatic [14,15], pituitary [16], breast [17], prostate [18,19], and lung [20] cancers. In particular, CDK5 contributes to MTC by inactivating the tumor suppressor retinoblastoma protein (Rb), which is a 'gatekeeper' of the cell cycle [10], thereby suggesting a crucial role for CDK5 in the regulation of the cell cycle.
We have generated a novel conditional MTC mouse model in which overexpression of p25 (p25OE) in mouse thyroid C-cells invokes aberrant CDK5 activity and MTC tumorigenesis [10,21]. Importantly, in these mice, arrest of p25OE completely halts MTC growth, thereby transforming tumors from a malignant to benign state. Mice harboring arrested tumors exhibit normal survival rates, whereas mice with proliferating MTC die within 30 weeks of transgene induction. A comparison of genes and proteins that are differentially expressed between malignant and benign tumors can help unravel the molecular basis for MTC tumorigenesis. Therefore in this study we investigate further the role of CDK5 in MTC pathogenesis by using an integrated approach including the novel MTC mouse model, human MTC cell lines and patient samples.

Differential gene expression analysis of tumors from an inducible medullary thyroid carcinoma mouse model
We have previously described a novel mouse model for MTC in which tumor progression and arrest are induced by overexpressing, and interrupting, green fluorescent protein-tagged p25 (p25-GFP) in thyroid C-cells [10]. Proliferating tumors display abnormally elevated CDK5 activity and are malignant. In contrast, arrested tumors are benign and exhibit much lower levels of CDK5 activity. Consistent with elevated cell proliferation, PET/CT imaging revealed 2.7-fold elevation in metabolic activity for proliferating malignant thyroid tumors compared to arrested benign tumors ( Figure 1A). To gain more understanding of the molecular mechanisms underlying p25-CDK5-induced MTC proliferation, we conducted a microarray study of the differential mRNA expression in malignant versus benign tumors. Unsupervised clustering analyses identified 116 genes that were up-regulated, while 7 genes were down-regulated in malignant MTC compared to benign tumors ( Figure 1B, Tables S1 and S2). Gene ontology analyses revealed a significant up-regulation of genes involved in cell cycle, cellular assembly and organization, DNA replication, DNA recombination, DNA repair, cellular movement, cellular death and cell survival ( Figure 1C).
We previously showed that p25-CDK5induced tumorigenesis was associated with elevated Rb phosphorylation, which inactivates Rb and leads to increased transcription of E2F target genes [10]. Consistent with these observations, differential gene analysis showed that the expression of E2F target genes such as those encoding Aurora kinase 1, Polo-like kinase 1, Survivin (BIRC5) and the cell cycle regulators CDK1, cyclin-A1 and cyclin-E2 were 3-to 4-fold up-regulated in malignant tumors ( Figures 1D -1E, Table S1).
To validate the results of the differential gene expression analyses, eGFP was used as an internal control, as eGFP mRNA levels increase in proliferating tumors due to tetOp-tTA-mediated p25-GFP transgene overexpression. Consistent with its targeted regulation, the eGFP gene exhibited the highest differential expression level (log2 = 6.88, Table S1). Reverse transcription real time PCR (RT-qPCR) analysis confirmed this increased eGFP mRNA expression ( Figure S1A). Heat map shows segregation of arrested and proliferating tumors using unsupervised clustering analysis of gene expression data. Green color indicates down-regulated genes and red color, up-regulated genes. The X-axis on volcano plot represents expression of all genes in the microarray, with cutoffs displayed at +/-log2 = 2.00. The Y-axis plots a t-test transform, representing the significance of the gene expression difference for each gene. The line represents the boundary for p-values above -log10 = 3.00. Red points represent genes of interest, which exceeded both indicated cutoffs. C) Pathway analysis of up-regulated genes in proliferating tumors. The X-axis represents the p-value associated with each gene category identified from Ingenuity Pathway Analysis (IPA). The higher the value, the less likely that the gene group appears in the proliferating tumors by chance. The red line represents a standard cutoff value of p = 0.01. D) IPA analysis showing the relationship between retinoblastoma protein (RB1) and up-regulated downstream molecules that were identified in the gene expression array analysis. Yellow color indicates moderate levels of expression and red color indicates high levels of expression. Relations between Rb and downstream molecules are as follows: orange lines indicate molecules that are inhibited when Rb is in an active, hypophosphorylated state; grey lines indicate that the effect of Rb on the molecule is not predicted. E) Examples of cell cycle regulators with quantitation from proliferating vs. arrested tumors. Data are represented as mean +/-SEM. www.impactjournals.com/oncotarget

Gene expression analysis of cell cycle regulators in malignant and benign mouse Mtc
While CDK5 is well characterized for its role in the central nervous system, less is known about its role in the cell cycle. We and others have found that CDK5 could regulate the activation state of the tumor suppressor Rb, thereby implicating CDK5 in the regulation of cell cycle progression [10,22,23]. Here, the gene expression analysis revealed a significant up-regulation of genes encoding cell cycle proteins, including CDKs, cyclins and endogenous cyclin-dependent kinase inhibitors (CKI), in malignant mouse MTC, suggesting that p25-CDK5-dependent MTC tumorigenesis is associated with alterations in cell cycle regulation.
To validate the expression array data and investigate further the role of CDK5 in the cell cycle, we measured the relative expression of genes encoding CDK, cyclins and CKI by RT-qPCR analyses. Elevated Cdk1, Cdk2 and Cdk4 mRNA levels were detected in proliferating malignant tumors. Cdk5 mRNA levels remained unchanged despite p25 overexpression. Somewhat unexpectedly, Cdk6 mRNA expression was increased in arrested benign tumors ( Figure 2A). Similarly, cyclin-D1 and p35 gene products, Ccnd1 and Cdk5r1, were up-regulated in benign tumors ( Figure 2B). In contrast, the genes encoding cyclin-A1, -B1, -E1 and -E2, i.e. Ccna1, Ccnb1, Ccne1 and Ccne2, were up-regulated in malignant tumors. Finally, CKI gene expression analysis revealed elevated mRNA expression in malignant tumors for p16 INK4a , p18 INK4c , p19 INK4d and p21 CIP/WAF1 encoding genes, i.e. Cdkn2a, Cdkn2c, Cdkn2d and Cdkn1a, while p15 INK4b and p27 KIP1 encoding genes, i.e. Cdkn2b and Cdkn1b mRNA levels were unchanged between conditions ( Figure 2C). Thus there is an overall elevation in the expression of several mRNAs encoding CDK, cyclins and CKI in proliferating malignant tumors compared to arrested benign MTC, thereby suggesting a role for CDK5 activity in the cell cycle. To determine the role of CDK5 activity in the expression of cell cycle regulators, we assessed the changes in protein level of cyclins, CDKs, and CKI in mouse tumors and in MTC-SK cells following blockade of CDK5 activity. We have indeed previously demonstrated that CDK5 activity is reduced in arrested tumors by stopping p25-GFP transgene expression and in MTC-SK cells by overexpressing a kinase-dead CDK5 construct [10]. The changes in expression of cell cycle regulators in human MTC tumors compared to normal thyroid tissues are also reported.

Protein expression analysis of cell cycle regulators in malignant and benign mouse Mtc
To further confirm the effects of p25-CDK5 upregulation on expression of cell cycle regulators, we evaluated protein expression levels in proliferating malignant versus arrested benign mouse MTC ( Figure 3). Quantification of immunoblots revealed CDK2 protein levels were elevated in proliferating tumors as previously observed [10]. However, CDK4 expression was unchanged between conditions ( Figure 3A). CDK6 levels were more variable from sample to sample and thus no significant change in CDK6 was detected between malignant and benign MTC ( Figure 3A). Consistent with mRNA findings, protein levels of cyclin-A1 [10], cyclin-B1 and cyclin-E2 were all up-regulated in malignant MTC ( Figure  3B). Cyclin-D1 protein expression was also variable from sample to sample and overall its expression was not significantly altered ( Figure 3B). As observed at the mRNA levels, CKI protein expression was predominantly up-regulated in malignant mouse tumors. In particular, the expression of p15 INK4b , p18 INK4c , p19 INK4d and p21 CIP/WAF1 was elevated in proliferating tumors compared to those in the arrested state ( Figure 3C). In contrast, the expression of p16 INK4a and p27 KIP1 was not significantly changed between conditions. Overall, these results are consistent with the observations from the differential gene expression analysis.

Evaluation of the relationship between cDK5 activity and cell cycle protein expression levels
The gene and cell cycle protein expression analyses described above indicate a possible relationship between CDK5 activity and the regulation of cell cycle protein expression in proliferating versus arrested mouse MTC tumors. To directly assess the effect of CDK5 activity on the expression of cell cycle regulators in human MTC, we compared cell cycle protein expression in a human sporadic, non-RET mutated MTC cell line, MTC-SK [24], which was transfected with either a construct encoding kinase-dead CDK5 (Kd-CDK5) or with a control plasmid.
We previously demonstrated that overexpressing kinasedead CDK5 in MTC-SK cells abrogates CDK5 activity and stops cell proliferation [10].
As a prerequisite observation, most of the CDKs, cyclins and CKI that were detected in the mouse tumors were also expressed in MTC-SK cells transfected with a control plasmid ( Figure 4). The only exception was cyclin-D1 which was not detected in MTC-SK cells and was unaffected by CDK5 activity in growing versus arrested mouse tumors. Interestingly, the expression of CDK2, CDK4 and CDK6 was diminished following Kd-CDK5 overexpression ( Figure 4A). Moreover, the protein levels of cyclin-A1, cyclin-B1 and cyclin-E2 were all reduced in Kd-CDK5-transfected cells ( Figure 4B). Finally, the expression of p15 INK4b , p19 INK4d and p21 CIP/WAF1 was decreased in transfected MTC-SK cells, whereas the protein levels of p16 INK4a , p18 INK4c and p27 KIP1 were not significantly changed ( Figure 4C). These observations are consistent with the analysis of cell cycle protein

Analysis of cell cycle protein expression in human Mtc tissue
While hereditary MTC is mainly caused by RET proto-oncogene mutations and consequent deregulation of the RET signaling pathway, the molecular basis for sporadic MTC is not well understood. Some sporadic MTC cases harbor mutations in the RET or Ras genes, but others do not. We previously reported that CDK5 was involved in MTC tumorigenesis, and found that high levels of CDK5 and its activators, p35 and p25, occur predominantly in the sporadic compared to the hereditary form of the disease [10]. Having established that CDK5 activity correlates with elevated cell cycle protein expression levels in proliferating mouse MTC and in a human sporadic MTC cell line, we compared protein levels of CDKs, cyclins and CKI in sporadic and hereditary MTC patient tissues ( Figure 5). Sporadic specimen did not exhibit RET mutations (see method section). We found that CDK2 levels were elevated in sporadic cases, but decreased in hereditary forms of MTC. In contrast,  CDK4 and CDK6 showed no significant changes in any of the groups analyzed ( Figure 5A). Cyclin-D1 levels were increased in sporadic MTC, while cyclin-A1 and cyclin-E2 were unchanged in MTC compared to control thyroid samples ( Figure 5B). Cyclin-B1 could not be detected in control thyroid or MTC specimens. p15 INK4b and p16 INK4a expression was increased in both, sporadic and hereditary MTC. The levels of p18 INK4c , p19 INK4d and p21 CIP/WAF1 were elevated in sporadic but not hereditary MTC. Finally p27 KIP1 expression was unchanged in MTC samples compared to control samples. Proliferating Cell Nuclear Antigen was expressed at the same level in sporadic and hereditary tumors, thereby confirming that the observed changes were not just a measure of growth fraction ( Figure S1B). The changes in cell cycle regulator expression that were observed in sporadic MTC tissues are consistent with those observed in the human MTC cell line and in mouse tumors ( Table 1). The results suggest that the MTC mouse model may more accurately model the molecular mechanisms underlying sporadic MTC than hereditary forms. Thus CDK5 may play a more important role in sporadic than familial MTC tumorigenesis.
To characterize further CDK5 activity in human tumors, we compared the phosphorylation state of the known CDK5 substrates, inhibitor-1 (Ser-6) [25], Tau (Thr205) [26] and STAT-3 (Ser727) [27] between sporadic and hereditary tumors ( Figure 6). We found these substrates were phosphorylated at their respective CDK5 sites, equally in both forms of MTC. Thus CDK5 activity is not uniformly elevated in response to p25 overexpression, and as p35 expression remains normal in both forms of MTC [10], the phosphorylation state of physiological CDK5 substrates may be unaffected.
Together the analysis of human tumors, cell line and mouse MTC suggests that up-regulation of cell cycle protein expression via a CDK5-mediated mechanism may contribute to sporadic MTC pathogenesis. The role of CDK5 in hereditary MTC tumorigenesis such as those arising from familial mutations in RET is less clear. These findings also suggest that cell cycle proteins may serve as useful biomarkers for sporadic forms of neuroendocrine thyroid cancer.

DIscUssION
MTC is a devastating disease for which new treatments are urgently needed. Having generated a novel, inducible animal model for MTC, we set to elucidate the molecular mechanisms underlying this cancer. Array-based transcriptome analysis comparing proliferating malignant and arrested benign mouse MTC revealed deregulation of cell cycle regulator expression (Figure 1). A focused examination uncovered a CDK5dependent increase in expression of genes encoding cell cycle regulators, including CDK2, cyclin-D1, p15 INK4b , p16 INK4a , p18 INK4c , p19 INK4d and p21 CIP/WAF1 . RT-qPCR and protein expression analyses confirmed their up-regulation at mRNA and protein level in p25-overexpressing malignant mouse MTC. Furthermore, the same set of cell cycle proteins are down-regulated in human MTC cells lacking CDK5 activity, suggesting aberrant CDK5 activity is necessary and sufficient to drive expression of these markers. Importantly, there was no difference in levels of CDK5 itself, thereby underlining that it is the aberrant activity of the kinase that is critical to these malignancy-associated effects rather than its expression level. Analysis of human MTC tissues from sporadic and hereditary cases showed that the cell cycle proteins, CDK2, cyclin-D1, p18 INK4c , p19 INK4d and p21 CIP/WAF1 were consistently overexpressed in human sporadic, but not hereditary, forms of MTC (Table 1). Although we could not demonstrate that CDK5 activity is higher in sporadic than in hereditary tumors, the integrated analysis of mouse tumors, the human cell line and patient samples suggest that CDK5 activity may be responsible for the increase in cell cycle protein expression that is detected in sporadic, non-RET MTC tissues. Our observations do not allow excluding a role for CDK5 in RET-dependent hereditary MTC. Cdk5 and its activators are indeed expressed in some hereditary tumors [10] and CDK5 has also been suggested to contribute to the proliferation of TT cells, which are derived from a hereditary form of MTC (RET mutation on codon 634) [28]. It will be important in the future to determine whether CDK5 is involved in RETmediated MTC pathogenesis and how.
Here we find that all the members of the INK4 family of CKI, which are four ankyrin-repeat domains proteins inhibiting only cyclin-D-CDK4/6 function [29], are up-regulated in MTC patients. In particular p15 INK4b and p16 INK4a expression is increased in non-RET sporadic and RET-mutated, hereditary MTCs while p18 INK4c and p19 INK4d levels are elevated in non-RET sporadic MTC only. Interestingly p18 INK4c and p19 INK4d have potential CDK5 phosphorylation sites whereas p15 INK4b and p16 INK4a do not [30], thereby supporting our hypothesis that CDK5 may be more important for sporadic than hereditary MTC. The INK4 family members have previously been related to MTC. Base pair change mutations in exon 2 of the p15 INK4b -encoding gene were found in some MTC cases [31]. Somatic mutations in exon 3 of the p18 INK4c -encoding gene disrupting p18 INK4c interaction with CDK4 or CDK6 were also detected in MTC and pheochromocytoma patients [32]. In addition, mutations in other exons of the p18 INK4c -encoding gene or within p19 INK4d -encoding gene have been reported in MTC patients [33]. Mutant mice carrying deletions of p18 INK4c -encoding gene as well as the p27 KIP1 -encoding gene develop multiple endocrine tumors including MTC and pheochromocytoma [34,35]. Finally, deletion of the p18 INK4c -encoding gene and/or the p27 KIP1encoding gene amplifies the tumorigenic effect of RET mutations causing hereditary forms of MTC as studied in mouse models and cell culture [35,36]. Together with our findings, these observations lead us to propose that inactivation or loss of p18 INK4c and p19 INK4d might be a feature of RET-dependent hereditary MTC, while overexpression of these proteins might characterize non-RET sporadic MTC. Analysis of large cohorts of RET and non-RET MTC specimen will be necessary to validate this hypothesis.
In this study, we present compelling evidence that CDK5 functions in the regulatory mechanisms that govern cell cycle progression and contribute to MTC tumorigenesis. The fact that a given set of cell cycle proteins display the same expression profile in three different MTC models, namely mouse model tissues, human sporadic MTC cell lines, and MTC patient samples, serves to underscore the relevance of our observations to mechanisms underlying this disease.
The cell cycle is initiated after the restriction point, or step of no-return, at which cells irreversibly commit to DNA replication. Bypassing the restriction point requires phosphorylation of Rb by CDKs, namely cyclin-D1-CDK4/6. Thereafter, cell cycle progression is reliant on coordinated action of cyclin-CDK complexes at defined phases throughout the cell cycle. Based on our current and previous studies, we propose that the aberrant activity of CDK5 may facilitate bypass of the restriction point through initial deactivation of Rb, via phosphorylation, which in turn promotes E2F target gene expression. In support of this notion, we show here that E2F target genes, including CDK2, p15 INK4b , p19 INK4d and p21 CIP/WAF1 [37,38] are up-regulated in p25-overexpressing malignant mouse MTCs, while the expression of these proteins is reduced following inhibition of CDK5 activity.
In summary, this study suggests that aberrant CDK5 activity promotes sporadic forms of MTC by modulating the expression of cell cycle proteins and points to these as possible biomarkers or targets for the development of new therapies.

Human samples
Medullary thyroid cancer specimens and control thyroid samples (goiters) were obtained through a human subject institutional-review-board-approved protocol UT Southwestern IRB 052004-044, ''Molecular Analysis of Endocrine Tumors''.

sequencing of human Mtc tumors and of the Mtc-sK cell line
MTC tumors and MTC-SK cells were examined for 'hot spot' somatic RET mutations (exons 10, 11, 13, 14, 15 and 16) by direct sequencing of DNA extracted from tumor specimens using a QIAamp DNA FFPE Tissue Kit (QIAgen) or extracted from cultured MTC-SK cells using DNAeasy Blood and Tissue Kit (QIAgen) according to the manufacturer's instructions. Sequencing was performed as previously described with some modification of primer sequences and conditions [39]. See supplemental methods for detailed information regarding primer sequences and reaction conditions. Sporadic MTC tumors (n = 4) had no mutations in RET exons. Hereditary MTC samples exhibited either a RET C634R mutation (n = 2) or a RET C618S mutation (n = 1), or RET L790F (n = 1). Because of the low number of available samples, tumors were classified as either sporadic (non-RET) or hereditary (RET-mutated). No RET mutations were detected in MTC-SK cells.

Animals
The NSE/p25-GFP line is a bitransgenic mouse model of MTC based on a tetracycline transactivator system and has been described in detail [10,21]. Briefly the neuron specific enolase (NSE) promotor controls expression of a tetracycline transactivator that is inhibited by dietary doxycycline (i.e., Dox off). Thus removing doxycycline (100 mg/L) from drinking water induces p25-GFP expression and MTC development. Subsequent re-addition of doxycycline stops p25-GFP overexpression and arrests tumor growth.

tumor collection and lysate preparation
Proliferating tumors were obtained by inducing p25-GFP expression in NSE/p25-GFP mice for 16 weeks. For arrested tumors, NSE/p25-GFP mice were deprived from doxycycline for 16 weeks and exposed to doxycycline for 4 weeks. Tumors were collected and lysates were prepared as described in [40].

Positron Emission tomography (PEt)/computed tomography (ct) imaging
Mouse PET/CT imaging was performed using a Siemens Inveon PET/CT Multi Modality system (Siemens Medical Solutions, Knoxville, TN) with effective spatial resolution of 1.4 mm at the center of field of view (FOV). All animals were fasted for 12 h prior to PET imaging. Each mouse received 140 µCi of 2-deoxy-2-(18F)fluoro-D-glucose (FDG) in 150 µL in saline intravenously via tail vain injection. The mice were placed on a heat pad before and during image acquisition. PET images were acquired 1h post-injection (p.i.), for 15 min, with animals under 2.5% isoflurane. PET images were reconstructed into a single frame using the 3D Ordered Subsets Expectation Maximization (OSEM3D/MAP) algorithm. CT images were acquired immediately after PET with the FOV centered at the shoulder of the mouse. CT projections (360 steps/rotation) were acquired with a power of 80 kVp, current of 500 µA, exposure time of 145 ms, binning of 4, and effective pixel size of 102 µm. The CT reconstruction protocol used a downsample factor of 2, was set to interpolate bilinearly, and used a Shepp-Logan filter. PET and CT images were co-registered in Inveon Acquisition Workplace (Siemens Medical Solutions) for analysis. Regions of interest (ROI) were drawn manually, encompassing the thyroid in all planes containing the organ. The target activity was calculated as percentage injected dose per gram.

Microarray and pathway analyses
For microarray analysis, total RNA from arrested and proliferating tumors was isolated using RNeasy kit (QIAGEN, Hilden, Germany). Gene expression profiling on each sample was performed using Illumina Mouse WG-6 V3 BeadArrays (San Diego, CA, USA). Beadlevel data were obtained and pre-processed using the R package "mbcb" for background correction and probe summarization. Pre-processed data were then quartilenormalized and log-transformed. Class comparison and unsupervised hierarchical clustering was performed using in-house MATRIX 1.48 (Girard, L. Manuscript in preparation). The volcano plot was developed by transforming p-values obtained in the t-test and plotting the transformed p-value versus the log2 expression value obtained following microarray data processing in MATRIX 1.48. The networks and functional analyses were generated through the use of QIAGEN's Ingenuity Pathway Analysis (IPA® QI Redwood City, www.qiagen. com/ingenuity).

reverse transcription real time Pcr (rt-qPcr)
Tissues of proliferating malignant (n = 7) and arrested benign tumors (n = 6) were collected as described above and homogenized in Trizol reagent (Life Technologies). RNA was extracted using QIAGEN RNeasy Kit according to the supplier's protocol. Purified RNA was treated with RNase-free recombinant DNase I (Roche Diagnostics GmbH). Reverse transcription was performed using iScript select cDNA Synthesis Kit with provided random primers according to manufacturer's instructions (Biorad Laboratories, Hercules, CA, USA). The SYBR-green based DyNAmo Flash SYBR Green qPCR Kit (Thermo scientific, Waltham, MA, USA) containing ROX as an internal reference dye was used for amplification. Real-time PCR reactions were run on an AB 7500 Real-time PCR system (Applied Biosystems, Foster City, CA, USA) and the specificity of each reaction was controlled by melt curve analysis. Primers were purchased from Integrated DNA Technologies (Coralville, IA, USA) and designed as intron-spanning pairs when possible (Table S4). Relative expression levels were calculated according to the 2-ΔΔCt method [41]. Expression levels were normalized to Beta-2 microglobulin as endogenous control and calibrated to average expression level of proliferating tumors for each gene. Unpaired t-tests were used to compare changes in expression for each gene. cell culture, transfections and cell lysate preparation MTC-SK cells were used for cell culture experiments. Those cells were derived from sporadic (non MEN) tumors and do not harbor RET mutations. MTC-SK cells were maintained in culture media containing Ham's F12 (Lonza Group Ltd, Basel, Switzerland), Medium 199 (Sigma, St. Louis, MO, USA) 1:1 v/v and 10% FBS as described in [24]. For transfections, cells were plated at a density of 2.5x10 5 cells/ml and transfected with either 1 μg pCMV-Kd-CDK5 [34] or 1 μg pCMV-EGFP (Clontech Laboratories, Inc., Mountain View, CA, USA) using X-tremeGene HP DNA transfection reagent (Roche Diagnostics GmbH, Mannheim, Germany). Cells were harvested 24 h post-transfection and lysate was prepared as previously described in [42].

Immunoblotting
Immunoblotting was conducted as previously described [43]. Membranes were probed with antibodies to . Antibodies to Inhibitor-1 and pS6-inhibitor-1 were generated and characterized inhouse [25]. The same membrane was probed for several proteins of different molecular weights after stripping in buffer containing 61 mM Tris-Base, 2%SDS and 7% β-mercaptoethanol at 50°C for 30 min. Immunoblots were quantified using Quantity One (BioRad). Samples were normalized to GAPDH (Figures 3 and 4) or to total protein levels as determined by Coomassie blue (CB) staining ( Figure 5).

statistical analysis
Results are presented as mean values and error bars represent ± SEM. Statistical analyses were conducted using two-tailed Student's t-test in GraphPad Prism 6.0 (GraphPad Software, Inc., La Jolla, CA, USA) and p-values < 0.05 were considered as statistically significant.