Activated Thyroid Hormone Promotes Differentiation and Chemotherapeutic Sensitization of Colorectal Cancer Stem Cells by Regulating Wnt and BMP4 Signaling Free

T3 is a pleiotropic hormone, which controls several cellular processes, including growth, development, and homeostasis (1). The main thyroid product thyroxine (T4) is inactive until converted into the active hormone T3 via type 1 or type 2 deiodinase (D1 and D2), while type 3 deiodinase (D3) converts T4 and T3 into inactive metabolites (2, 3).

T3 primarily acts as a transcription factor upon binding to its nuclear thyroid hormone receptors. Thyroid hormone receptors heterodimerize with many other nuclear receptors and bind the chromatin to thyroid hormone receptor response elements (TRE) for the positive or negative regulation of target genes transcription (4).

Many in vitro and in vivo studies have indicated that the thyroid status affects tumorigenesis. Type 3 deiodinase (D3) is an oncofetal protein frequently expressed in proliferating and neoplastic cells, where it controls aspects of diseases, injury responsiveness, and tumorigenesis (5). Congruently, the actions of the deiodinases provide tissue-specific regulation of thyroid hormone action at an intracellular level (2). In different tumoral contexts, deiodinases are under the control of relevant pathways in cancer, such as Wnt and Shh (5).

Many in vitro and in vivo studies have indicated that thyroid status affects tumor formation, growth, and metastasis in experimental laboratory animals and humans (5). However, the relationship between the cell-specific mechanisms, which control thyroid hormone ligand availability and properties of cancer cells, is still unknown.

Colorectal cancer stem cells (CR-CSC) represent a small subset of cells within the tumor mass with self-renewing potential and the ability to engraft and generate tumors in immunodeficient mice (6, 7). According to the CSC model, these cells are difficult to kill and their relative insensitivity to chemotherapeutic drugs may explain the frequent failure of conventional treatments used against advanced tumors (8).

The intestine is a highly dynamic tissue, characterized by rapid and continuous regeneration and supported by crypt intestinal stem cells (9). The Wnt signal, which rigorously controls the sequential events that leads to the transition from normal colon mucosa to adenocarcinoma, is one of the major forces that maintain the stem cells' fate and capacity to self-renew as well as their ability to escape conventional chemotherapy-induced apoptosis (10, 11). We recently demonstrated that the Wnt–β-catenin pathway drives an inverse, coordinated regulation of D2 and D3 in colon cancer cells (12).

Here, we demonstrate that CR-CSCs are highly sensitive to intracellular T3. After T3 treatment or D3 depletion, CR-CSCs undergo differentiation, a process that under normal serum conditions requires intracellular T4 to T3 conversion. This is achieved through increases in the BMP-4 levels and its downstream targets and significant attenuation of the Wnt pathway. Strikingly, increasing intracellular T3 results in reduced clonogenic and tumorigenic potential and establishes a higher sensitivity of CR-CSCs to conventional chemotherapeutics.

Sphere purification and propagation from patients with colorectal cancer were assessed as described previously (8, 13, 14). Cells were monthly tested for mycoplasma contamination as described previously (14). To evaluate the asymmetric division, single CR-CSCs were also labeled with PKH26 dye (2 × 10⁻⁶ mol/L, Sigma), cultured for up to 14 days, and subjected to flow cytometry analysis to yield the PKH positivity. BMP4, at a concentration of 100 ng/mL, in combination with rT3, was added to CR-CSCs and cultured up to 48 hours. Human colorectal adenocarcinoma cells (CaCo-2, obtained in 2005 from ATCC and authenticated by RT-PCR analyses), cultured in adherent conditions in presence of DMEM medium and supplemented with 2 mmol/L l-glutamine and 10% FBS (ATCC).

The BMP4 reporter plasmid was generated by PCR on genomic DNA with two sets of oligonucleotides (5mBMP4pU and 5mBMP4pL, and 3mBMP4pU and 3mBMP4pL). Two regions were amplified: 859 bp (containing the first exon) and 552 bp (containing the 3′ UTR of the mouse bmp4 gene). The PCR products were digested with SacI/XhoI and cloned in pGL3basic (5′-UTR BMP4) or TKpGL3 (3′-UTR BMP4).

Cells were transiently transfected (FuGENE 6; Roche) with a mixture of inducible TCF/LEF–responsive firefly luciferase and constitutively expressed Renilla luciferase (40:1), or with a negative control containing a mixture of noninducible firefly luciferase and constitutively expressed Renilla luciferase (40:1). The relative quantification of gene expression was calculated on triplicate reactions using the comparative C_t method (ΔΔC_t).

Cell migration was measured using growth factor–depleted Matrigel-coated (BD Biosciences) transwell inserts. Dissociated CR-CSCs (5 × 10³), transduced with a D3-specific miRNA (iD3) or control scramble miRNA (iCTR), were placed on 8-μm pore size Matrigel-coated transwell (Corning). NIH-3T3-stem cell–conditioned medium was used as a chemoattractant and plated in the bottom compartment of transwell. After plating, migrated cells were counted up to 72 hours. Cell viability was evaluated on spheres transfected with a D3-specific RNAi pool.

The mRNAs were extracted with TRIzol reagent (Life Technologies, Ltd). The cDNAs were prepared with Superscript III (Life Technologies, Ltd) as indicated by the manufacturer. The cDNAs were amplified by PCR in an iQ5 Multicolor Real Time Detector System (Bio-Rad) with the fluorescent double-stranded DNA-binding dye SYBR Green (Bio-Rad). The relative amounts of gene expression were calculated with cyclophillin A expression as an internal standard (calibrator). The results, expressed as N-fold differences in target gene expression, were determined as follows: N *target = 2(C_t sample C_t calibrator).

Cells were resuspended in ice-cold NP40 lysis buffer, fractioned on SDS-polyacrylamide gels, and blotted on nitrocellulose membranes. Membranes were exposed to specific antibodies for Notch full length (sc-6014-R, Santa Cruz Biotechnology), cleaved Notch-ICD (Cell Signaling Technology, 2421), E-cadherin (rabbit polyclonal; Cell Signaling Technology), pAkt (9271, Ser 473, rabbit IgG; Cell Signaling Technology), Akt (9272, rabbit IgG; Cell Signaling Technology), pGSK3 (Ser 9, rabbit polyclonal; Cell Signaling Technology), GSK3 (rabbit polyclonal; Cell Signaling Technology), PTEN (138G6, rabbit IgG; Cell Signaling Technology), or tubulin (T9026, Sigma), and detected using HRP-conjugated anti-mouse or anti-rabbit antibodies (Amersham Biosciences). p21 (sc-397) and p27 (sc-1641) antibodies were purchased from Santa Cruz Biotechnology. Densitometry analyses were performed by Scion Image. Results were expressed as protein/tubulin optical density ratio.

For the immunofluorescence, CR-CSC cytospins, untreated and treated with T3 and rT3, were fixed with 2% paraformaldehyde, permeabilized with 0.1% Triton X-100, and washed in PBS. Thereafter, the slides were exposed to antibodies against cytokeratin 20 (Ks20.8, mouse IgG2a; Dako Cytomation) and BMP4 (3H2, mouse IgG1; Novocastra), both diluted in PBS + 3% BSA and 0.05% Tween 20 (PBS-T). Cells were treated with fluorochrome-conjugated anti-mouse antibodies (Invitrogen) plus RNaseA (200 ng/mL, Sigma) and counterstained with Toto-3 iodide (Invitrogen). Samples were analyzed on a Nikon C1-Si confocal microscope equipped with EZ-C1 software.

Paraffin-embedded sections of xenografts were subjected to specific antibodies for CK20 (Ks20.8, mouse IgG2a; Dako Cytomation), Ki67 (MIB-1, mouse IgG1; Dako Cytomation), CD133 (AC133, mouse IgG2b; Miltenyi Biotec), or isotype-matched controls at appropriate dilutions. Apoptotic events were determined by terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick-end labeling assay using the In Situ Cell Death Detection, AP Kit (Boehringer Mannheim). DNA strand breaks were detected by 5-bromo-4-chloro-3-indolyl-phosphate (BCIP, Dako Cytomation) substrate.

Cell viability was assessed by using a CellTiter-Glo Luminescent Cell Viability Assay Kit (Promega) on sphere cells, treated with T3 or rT3, for up to 6 days, and then exposed to oxaliplatin (100 nmol/L) and 5-fluorouracil (50 μg/mL), alone or in combination. Cell death was evaluated by orange acridine (50 μg/mL)/ethidium bromide (1 μg/mL) staining.

To evaluate the proportion of differentiated and undifferentiated cells, CR-CSCs were cultured in stem cell medium in adherent conditions with 100 ng/mL BMP4 for up to 72 hours.

CR-CSCs were transduced with lentiviral-TOP-dGFP-reporter as reported previously (14). TOP-dGFP high and low (Wnt^high and Wnt^Low) populations were enriched by FACSaria (BD Biosciences). The relative quantification of gene expression was calculated in triplicate reactions using the comparative C_t method (ΔΔC_t). BLOCK-iT Lentiviral RNAi Expression System (Life Technologies, Ltd) was used to the expression of miRNA silencing D3 (iD3) or negative control miR (iCTRL) into dissociated sphere cells, as previously reported (12).

Dissociated sphere cells were cultured in presence or absence of FBS and exposed to 30 nmol/L rT3 or T3, on ultra-low adhesion 96-well plates at a single cell concentration. rT3 and T3 were added every 48 hours at the end of experiments. Wells containing either none or more than one cell were excluded for analysis. Sphere formation in culture was monitored and counted under light microscope for up to 21 days. The CSC frequency was calculated using the ELDA algorithm (http://bioinf.wehi.edu.au/software/elda/).

The RT² Profiler PCR array was performed to simultaneously evaluate the mRNA levels of 89 genes responsive to and related to WNT signal transduction, following the manufacturer's protocol (PAHS-243Z, PAHS-043Z; SuperArray Bioscience). Cycle threshold values were normalized using the average of 5 housekeeping genes in 96-well plates (B2M, HPRT1, RPL13A, GAPDH, and ACTB). The comparative cycle threshold method was used to calculate the relative quantification of gene expression (cells treated with T3 and rT3 in presence of 10% FBS medium versus untreated or cultured with 10% FBS). RT² Profiler PCR Array Data Analysis was represented by clustergrams based on Pearson correlation of 2^ΔC_t.

Cell-cycle analysis was evaluated on dissociated CR-CSCs as previously described (13).

Dissociated CR-CSCs (5 × 10⁵) and those transfected with a D3-specific RNAi pool were injected subcutaneously with Matrigel Growth Factor Reduced (BD Biosciences) at a 1:2 ratio in a total volume of 100 μL. Tumor size was calculated once a week up to 10 weeks according to the following formula: (π/6) × larger diameter × (smaller diameter)².

Colorectal cancer specimens were obtained from 14 patients (age range 50–57 years) undergoing cancer resection (Table 1), in accordance with the ethical standards of the Institutional Committee on human experimentation (Convention on Human Rights and Biomedicine, Oviedo, 4.IV.1997). None of the patients was under thyroid hormone treatment before surgery. Informed consent for all patients was obtained prior to their participation in this study. All research was conducted within the Ethical principles for Medical Research involving human subjects expressed by the Declaration of Helsinki (revised version, Seul 2008, nn.14–15).

Male NOD.CB17-Prkdcscid/J (NOD/SCID) mice, 7 weeks old, were acquired from Charles River Laboratories. Mice were maintained, in accordance to the institutional guidelines of the University of Palermo Animal Care committee, in an animal house authorized by the Italian Ministry of Health (DGSAF #0020301-P-03.10.2014).

Differences between samples were assessed by a Student two-tailed t test for independent samples. P values less than 0.05 were considered significant. Relative mRNA levels (in which the first sample was arbitrarily set as 1) are reported as results of real-time PCR, in which expression of cyclophilin A was used as a housekeeping gene. All experiments were repeated and analyzed three to five times. Extreme limiting dilution assay analysis was performed to determine the statistical differences in stem cell–like frequency among the treatment groups.

To assess the sensitivity of CSCs to T3 signaling, we initially demonstrated the expression of thyroid hormone receptors TRα and TRβ and transporters in freshly purified CD133⁺ cells from two different colorectal cancer samples (Supplementary Figs. S1–S3). Thereafter, we analyzed deiodinases expression in CR-CSCs during FBS-induced differentiation. Both D2 and D3 were expressed and oppositely regulated during spontaneous differentiation (Fig. 1A). D3 mRNA was at its highest level in proliferating CR-CSCs and diminished during differentiation. Contrarily, D2 expression, barely detectable in proliferating colorectal cancers, drastically increased in differentiated cells. We have previously demonstrated that subpopulations of CR-CSCs with high expression of Wnt (Wnt^High) differ from those with a low expression of Wnt (Wnt^Low), with respect to their stemness and ability to differentiate (15). In FACS sorted cells, we observed that Wnt^High cells are characterized by a strong expression of type 3 deiodinase, and a corresponding reduction in putative thyroid hormone–target genes (Supplementary Fig. S2), thus confirming that D3 expression marks the highly proliferating subset of CR-CSCs and is regulated by the Wnt pathway (Fig. 1B). To assess the effect of exogenous T3-treatment, we cultured CR-CSCs under different T3 conditions and evaluated their morphology after 3 and 6 days. T3 treatment alone is sufficient to induce morphologic cell changes similar to what was observed when using FBS treatment, that is, large, flatted, polygonal-shape cells. In contrast, rT3 treatment (by abolishing D2-mediated T4 to T3 conversion and the consequent reduction in intracellular T3) impairs FBS-induced differentiation by markedly reducing the expression of CK20 (Fig. 1C). Overall, these data indicated that while exogenous T3 treatment is sufficient to promote cell differentiation, FBS-induced differentiation occurs via endogenous T3, which derives from T4-to-T3 conversion.

It has been proven that BMP4 is a major promoter of differentiation in normal colonic stem cells (13). To get insights into the mechanisms by which T3 induces cellular differentiation, we investigated the existence of a putative interplay between T3 and BMP4 signals. Notably, we observed that T3 treatment sustains BMP4 synthesis while rT3 inhibits FBS-mediated BMP4 induction (Fig. 2A). We hypothesized that BMP4 could be a novel T3-target gene as suggested by a ChIP-seq and in silico analyses (unpublished data and Fig. 2B, top). To prove this, we cloned a 859-bp (part of 5′-UTR) and a 552-bp DNA region (part of 3′-UTR) containing two putative thyroid hormone receptor–binding sites upstream of the luciferase reporter gene (5′-UTR BMP4 and 3′-UTR BMP4, Fig. 2B, top). Functional assays demonstrated that T3 strongly induces the activation of both the 5′ and 3′ regions of the BMP4 gene (Supplementary Fig. S4). Finally, chromatin immunoprecipitation (ChIP) assay confirmed that the 5′-UTR and 3′-UTR region of the BMP4 gene are physically associated with TRα in colorectal cancer cells and this interaction is potentiated by T3 (Fig. 2B, bottom). Altogether, these results indicate that BMP4 is a novel T3-responsive gene in colorectal cancer cells.

Considering that BMP4 induces a noncanonical pathway that involves PI3K/AKT activation and modulates the Wnt signaling, we tested the effects of T3 depletion on PI3K/AKT. Western blot analysis revealed that T3 attenuation by rT3 potently induces phosphorylation of AKT and GSK-3β in FBS- and BMP4-treated cells. Accordingly, E-cadherin was induced by T3 and downmodulated by T3 signal reduction (Fig. 2C).

Interestingly, BMP4 was able to induce the expression of the T3-producing enzyme D2, and reduce the T3-inactivating enzyme, D3 (Fig. 2D). Such a double regulation generates an auto-sustaining loop driven by BMP4, which is a T3 target and simultaneously aims to increase intracellular T3 by modulating deiodinases (Fig. 2E). Overall, these data indicate that T3 induces differentiation of colorectal cancer cells by reducing BMP/PI3K/AKT signaling, a process that is auto-sustained by BMP.

We evaluated the clonogenic potential of colorectal cancer spheres, which were characterized for the expression of the putative and functional stem cell markers CD133 and CD44v6 (Supplementary Fig. S3) cultured at different T3 concentrations. As shown in Fig. 3A, the clonogenic capacity of colorectal cancer cells was significantly reduced in the presence of serum (by 54%) or T3 (by 58%), compared with those cultured in stem medium. In contrast, rT3 treatment significantly preserved the clonogenic capacity, even in the presence of serum. CSC frequency confirmed the ability of T3 to induce differentiation (1/13.43 vs. 1/6.75) and thereby decrease the clonogenic activity of CR-CSCs (Fig. 3A, bottom).

Of note, we observed that the exogenous exposure to FBS or T3 significantly decreased the quiescent cell compartment in the G₀–G₁ phase of CR-CSCs (Supplementary Fig. S5). This is in accordance with the notion that T3 awakens CR-CSCs from their dormant state by enhancing the p21 and p27 expression levels (Supplementary Fig. S6A and S6B).

Subsequently, we monitored cell division by analyzing the distribution of the membrane dye PKH26 expression via FACS (16). While 81% of cells in stem medium underwent asymmetric division, the vast majority of them, grown in stem medium plus T3 as well as in FBS, arrested cell division (Fig. 3B), the latter completely reversed via rT3 treatment. Altogether, these results suggest that T3 alters the percentage of asymmetric cell division in CSCs, thus affecting their ability to maintain the stem-like properties.

As Notch is a critical regulator of asymmetric cell division and cell fate in colorectal cancer cells (17, 18), we sought to assess whether T3 might induce variations in Notch activity. Hence, we measured the expression levels of the active Notch (Notch-ICD). As shown in Fig. 3C and Supplementary Fig. S7, T3 or D3 depletion greatly induced the intracellular cleavage and activation of Notch. FBS-induced Notch activation was completely revoked by rT3, thus indicating a key role of intracellular T3 in the regulation of Notch pathway.

Given the relevance of the Wnt pathway with regard to the functional state of CR-CSCs (15, 19) and its cross talk with the BMP-4 pathway, we aimed to investigate the possible influence of T3 on the Wnt pathway. For this purpose, we compared the transcriptional profile of Wnt targets and Wnt pathway–relevant genes in T3-treated sphere cells. Clustergrams showed Wnt activation in CSCs when untreated and treated with FBS plus rT3. Vice versa, T3 treatment resulted in a considerable reduction of relevant Wnt targets, including β-catenin, associated with the upregulation of negative regulators of the Wnt receptor signaling (Supplementary Fig. S8). This analysis significantly points to T3 as a potent repressor of Wnt signaling.

To assess the capacity of T3 to inhibit the tumorigenic potential, we measured the effects of T3 treatment on the CR-CSCs engrafting capacity. We injected CR-CSCs, pretreated with 30nmol/L T3 for 6 days, into immunodeficient NOD/SCID mice. Untreated CR-CSCs generated palpable tumors within 3–4 weeks, which were strikingly reduced in T3-treated cells (Fig. 3D). Tumors derived from T3-treated cells show elevated CK20 expression levels and abundant cell death. Furthermore, the CD133 and Ki67 expression was drastically reduced in CR-CSCs exposed to T3 (Supplementary Fig. S9), indicating a substantial effect on proliferating cells endowed with self-renewal capacity.

D3 is the T3-inactivating enzyme often overexpressed in human cancer (5). To assess the role of D3 in the modulation of intracellular T3 and its effect on tumorigenic potential, we infected CR-CSCs with a lentivirus able to efficiently knockdown D3 (12). D3 knockdown in CR-CSCs cells dramatically reduced the in vitro clonogenicity and their invasive capability as well as tumor growth in immunocompromised mice (data not shown). D3-depleted CR-CSCs failed to give rise to detectable tumor outgrowth. In line with the observed T3 effects, D3 depletion potently reduced the xenograft's growth, which is consistent with an increase of the intracellular T3 availability in the absence of D3 (Fig. 3E). These data underline that T3 signaling is a strong determinant in the CSCs' ability to promote tumorigenesis in vivo.

We have previously reported that CSCs are widely resistant to chemotherapeutic drugs (8). To assess whether such a complex regulation of T3 on the BMP4/Wnt pathways in CR-CSCs might influence their ability to respond to chemotherapy, we investigated whether T3 is able to alter the resistance to oxaliplatin (OX), 5-fluorouracil (5-FU), and their combination (FOX), at clinically relevant doses. Time course treatment showed that, while untreated spheres were largely inert to chemotherapeutic drug-induced apoptosis, T3 treatment caused an increased percentage of cell death (up to 75%), when combined with FOX. RT3 treatment did not alter the cell resistance to drugs (Fig. 3F).

Despite improvements in therapeutic strategies, colorectal cancers remain the third leading cause of cancer-related deaths in western countries, due to the failure in curing the metastatic disease. Although canonical cancer treatments have been designed to reduce the rapidly growing tumor cells, CSCs are spared and are still able to mediate cancer relapse after chemotherapy and radiation (20). This suggests that curative therapies can be effective only by targeting the subpopulations of those tumor cells with self-renewing potential.

Drug-induced differentiation represents a promising approach to hamper CSCs' self-renewal ability. Although differentiation therapies do not selectively kill CSCs, they make them more sensitive to the conventional therapies and ultimately eradicate the tumor-driving cell population. Contrary to the haematologic malignancies, the clinical use of differentiation-inducing agents to treat solid tumors is very limited (21, 22).

We demonstrate that CR-CSCs with β-catenin activation display high levels of D3, which correlate to an enhanced self-renewal capacity. Drug-induced differentiation could represent a promising approach to hamper the self-renewal ability of CSCs. Our data provide evidence that, while D3 contributes to maintain the undifferentiated status, T3 induces differentiation, affects the Wnt target–related genes, and sensitizes CR-CSCs to the standard chemotherapeutic drugs by downmodulating the AKT/PI3K pathway. Furthermore, D3-induced inactivation of thyroid hormone stabilizes the quiescence status of CR-CSCs, altered by T3 exposure. Targeting D3 abrogated the tumorigenic activity in vivo of CR-CSCs.

It was reported that the Notch1 hinders the β-catenin activation by reducing the levels of the available unphosphorylated (active) form (23). Triggering the Notch signal pathway induces the cytoplasmic cleavage of Notch-ICD, which activates nuclear Notch target genes that mainly promote differentiation (24). It was also demonstrated that the overexpression of Notch has a negative effect on colorectal cancer progression, thus suggesting Notch as a favorable prognostic marker (25). In line with these data, we show that T3 controls the self-renewal pathway of CR-CSCs through the activation of canonical Notch pathway. While T3 induces the cleavage of Notch-ICD, D3 restores the levels of activated β-catenin drastically decreasing the expression levels of Notch that are sustained by T3.

In CR-CSCs, Notch signaling increases asymmetric division, which generates one daughter cell that retains stem cell properties (17, 18). Meanwhile, the second daughter cell undergoes a differentiation process via multiple division rounds. In this scenario, the stem cell pool can be expanded by a series of symmetric cell divisions that is essentially controlled by the Notch inhibitors. Importantly, D3 action, by reducing intracellular T3, increases the frequency of symmetric self-renewing divisions of CR-CSCs, which could explain the enhanced tumor growth.

We have previously demonstrated that BMP4 displays high antitumor activity in colorectal cancer by inducing CSC differentiation, targeting survival, proliferation, and chemoresistance (13). Interestingly T3, a potent BMP4 inducer, impairs clonogenic activity and xenograft tumor outgrowth, suggesting this molecule as a key effector in the activation of the colorectal cancer differentiation program. Being that T3 reduces nuclear β-catenin accumulation, the PI3K/AKT pathway may represent a connection point between BMP4 and Wnt pathways, implying an active crosstalk that balances stemness and differentiation in CR-CSCs. The ability of this hormone to regulate the BMP4 gene activation is confirmed by the ChIP-seq and in silico analysis that provide a feedback control of the differentiation pathway. Likewise, a D3-induced increase of BMP4 inhibitors restores their self-renewal capacity, tumorigenic capacity, and refractoriness to conventional anticancer therapies.

In summary, our findings suggest that intracellular T3 exerts a prodifferentiative effect that prevents CSC expansion and triggers a differentiation program. The latter may result in CSCs depletion, thereby hampering tumor development. While D3 enhances tumor growth, T3 signaling appears particularly effective in inducing differentiation, growth reduction, and chemosensitization of CR-CSCs. The therapeutic effects observed by the combined action on intracellular T3 and chemotherapy further substantiate the necessity to target the stem-like population of cancer cells to improve colorectal cancer treatment and open new avenues for the use of locally manipulated deiodinases for treating proliferative disorders or hormone-sensitive tumors such as colorectal cancers.

No potential conflicts of interest were disclosed.

Conception and design: V. Catalano, G. Stassi, D. Salvatore

Development of methodology: V. Catalano, M. Dentice, C. Luongo, A. Benfante, M. Todaro

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): V. Catalano, M. Dentice, R. Ambrosio, R. Carollo, A. Benfante

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): V. Catalano, M. Dentice, R. Ambrosio, C. Luongo, M. Todaro, D. Salvatore

Writing, review, and/or revision of the manuscript: V. Catalano, R. Ambrosio, G. Stassi, D. Salvatore

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Dentice

Study supervision: G. Stassi

The authors thank Tatiana Terranova for her precious editorial assistance.

This project was supported by the Associazione Italiana per la Ricerca sul Cancro (AIRC; IG 12819 and 5 × 1000, 9979; G. Stassi), IG 13065 (M. Dentice), and IG 11362 (D. Salvatore), by the financial support of the Italian Ministry of University and Research (project PRIN 201223E28B; D. Salvatore).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.