Hedgehog Signaling and the Gli Code in Stem Cells, Cancer, and Metastases

See allHide authors and affiliations

Sci. Signal.  22 Nov 2011:
Vol. 4, Issue 200, pp. pt9
DOI: 10.1126/scisignal.2002540
A Presentation from the 1st International HEALING Meeting: Hh-Gli Signaling in Development, Regeneration, and Disease, Kolymbari, Crete, 23 to 25 June 2011.


The Hedgehog (Hh)–Gli signaling pathway is an essential pathway involved in development and cancer. It controls the Gli code—the sum of all activator and repressor functions of the Gli transcription factors. Through the Gli code, and Gli1 in particular, it modulates the fate and behavior of stem and cancer stem cells, as well as tumor growth and survival in many human cancer types. It also affects recurrence and metastasis and is enhanced in advanced tumors, where it promotes an embryonic stem (ES) cell–like gene expression signature. A central component of this signature, Nanog, is critical for glioblastoma and cancer stem cell survival and expansion. Gli1 activity is also enhanced by several oncogenic proteins, including Ras, Myc, and Akt, and by loss of tumor suppressors, such as p53 and PTEN. The oncogenic load boosts Gli1 levels, which supports tumor progression, and promotes a critical threshold of Gli1 activity that allows cells to enter the metastatic transition. In colon cancers, this transition is defined by enhanced Hh-Gli and, surprisingly, by repressed Wnt-Tcf signaling. Together our data support a model in which the Gli code, and Gli1 in particular, acts as a key sensor that responds to both Hh signals and the oncogenic load. We hypothesize that, in turn, the Gli-regulated ES-like factors induce a reprogramming event in cancer stem cells that promotes high invasion, growth and/or metastasis. Targeting the Gli code, the autoregulatory Gli1-Nanog module and interacting partners and pathways thus offers new therapeutic possibilities.

Presentation Notes

Slide 1: Science Signaling logo

The slideshow and notes for this Presentation are provided by Science Signaling (

Slide 2: Hedgehog signaling and the Gli code in stem cells, cancer, and metastases

Hedgehog (Hh)-Gli signaling is conserved and is involved in many aspects of animal development and patterning. It is also notably involved in the control of cancer stem cells and tumor growth. It is this duality that makes this pathway exceptionally interesting, because defects in Hh-Gli signaling can lead to loss of cells, as in degenerative diseases, and inappropriate increases in signaling can lead to cancer. For example, Hh-Gli signaling controls cerebellar development and granule neuron precursor proliferation in the external germinal layer of the cerebellum, as revealed with my former postdoctoral fellow Nadia Dahmane, now at the Wistar Institute in Philadelphia, Pennsylvania (13), and I, and affects the formation of brain tumors, such as gliomas, that she and others in my lab first showed express the Hh ligand Sonic hedgehog (Shh) and require Shh signaling (4, 5). My lab is interested in understanding how cell fate and function are controlled by the Gli code that we define as the balance of all activator and repressor functions of the Gli transcription factors (68).

Slide 3: The skeleton of the classical Hedgehog-Gli pathway

Genetic and molecular analyses have defined a conserved pathway. Whereas structures and exact biochemical functions remain to be elucidated for most components, the logic of the skeleton of the conserved pathway is shown here. Signaling starts with the production and reception of Hh ligands. In general terms, Hh signaling starts with the production and secretion of Hh ligands, such as Shh. This is followed their reception by Patched (Ptc) and subsequent activation of the G protein–coupled receptor–like transmembrane protein Smoothened (Smo). Genetically, Ptc is a repressor of Smo activity, and Hh binding to Ptc inhibits its activity so that Smo is then free to signal intracellularly. Smo localizes to primary cilia and signals intracellularly through several components, including kinases, to regulate the three Gli zinc finger transcription factors. The single Smo gene is essential, and Smo protein activity leads to the enhancement of activator forms of Gli and the diminution of the abundance of Gli repressor forms, which alters the Gli code. Of the three Gli proteins in mice and humans, Gli1 is the last step of the pathway and the key final output of Hh. In the absence of ligand, Smo is inactive, and generally, Gli1 is not transcribed and Gli3 is cleaved to generate repressor isoforms (Gli3Rs). Upon ligand reception and Smo function, Gli3 is repressed, Gli3 repressor function inhibited, and Gli1 transcribed so that the final output is, generally, the transcription of Gli1 target genes. In addition to Suppressor of fused (Sufu), which can sequester the Gli1 protein in the cytoplasm and thus render it unable to transactivate targets, there are other components that fine-tune Hh-Gli signaling that are not shown here. Notably, Smo activity also represses Sufu function. The pathway also has several feedforward and feedback loops, such as Gli1-postitive autoregulation to enhance the signal, and Gli1-induced expression of Ptc to dampen and shorten signaling strength and duration. The diagram also shows several of the tools used to manipulate signaling in the experiments presented in this talk, such as short hairpin (shRNAs), the Smo inhibitor cyclopamine, small interfering RNAs (siRNAs), and cDNAs with their site of action.

Slide 4: Key questions

The key questions addressed in this talk will follow from the background I will present in the next few slides and focus on the interaction of Hh-Gli signaling and the Gli code with other signaling pathways known to be important for stem cells, development, and cancer. In particular, we will investigate the interaction of Hh and Wnt signaling. A second question that I will address is the nature and critical role of an embryonic stem cell–like “stemness” gene signature—which we first described in gliomas (5) and which has since been described in several other tumor types [e.g., (9)]—in the control of cancer stem cell behavior and which we show is regulated by Hh-Gli1 signaling.

Slide 5: Autocrine Hh-Gli in sporadic human cancers with and without Hh pathway mutations

I would like to present now several central results from my lab that highlight the essential function of the Gli code and Hh signaling in cancer and stem cells. First, we have provided ample and direct evidence for autocrine, juxtacrine, or both forms of Hh signaling within the epithelial tumor cells of primary sporadic human carcinomas that have known mutations in the Hh-Gli pathway, such as basal cell carcinomas (BCCs) (10) but also for those without known Hh pathway mutations, such as gliomas (4, 5) or metastatic melanomas (11). This is in contrast with the largely unsubstantiated suggestion of exclusive paracrine signaling in human tumors that has been proposed by others and is based solely on indirect evidence in mouse-human chimeras (1214).

Slide 6: Gli1 is ubiquitously expressed in sporadic human basal cell carcinomas

This image shows the discovery by my lab that sporadic human BCCs harbor an active Shh-Gli pathway, as seen by the expression of Gli1 in the tumor cells themselves, seen singly or in nodules. In this in situ hybridization performed by Nadia Dahmane in my lab, an antisense RNA probe for detecting Gli1 transcripts in a human tumor section was used, black indicates the highest expression, and yellow indicates expression above background (10). Because Gli1 transcription is a reliable marker of pathway activation (15), the data indicate pathway activity in tumor cells. It also shows that single Gli1-positive (Gli1+) tumor cells that are not routinely seen in Mohs surgery (wherein the surgeon resects normal-looking tissue bordering the tumor until no histologically identifiable tumor masses are detected in the resected tissue) can be seen through visualization of Gli1 expression, which makes Gli1 an excellent tumor biomarker.

Slide 7: Evidence for autocrine Shh-Gli signaling in epithelial tumor cells

Another example of activated Hh signaling in a tumor context is illustrated by these RNA in situ hybridization analyses of metastatic melanoma lesions from human skin (top panel), which were performed by Barbara Stecca, a former postdoctoral fellow in the lab who is now a principal investigator at the Istituto Toscano Tumori, in Florence, Italy. These tissue sections show expression of endogenous transcripts of Shh, the receptor Ptch1, and the three transcriptional mediators Gli1–3. All are expressed in the melanoma cells themselves, which form grapelike bunches, as shown with hematoxylin-and-eosin (HE) staining, and express the canonical melanoma markers microphthalmia-associated transcription factor (MITF) and the melanocyte antigen MELAN-A (11). In the lower left panel, an example of our analyses of human sporadic prostate cancer in which Shh, Ptch1, and Gli1 are all prominently localized to the tumor mass (t) and not the adjacent stroma (s). This work was performed by Pilar Sánchez, a former postdoctoral fellow in the lab who is now a principal investigator at the Instituto Carlos III in Madrid, Spain (16). On the lower right is shown a further example wherein we documented expression of pathway components in the tumor nodules of local sporadic human colon cancers. Here, we show Gli1 and Ptch1 expression in tumor cells that form a characteristic lumen and also stain positively for the presence of the Shh protein; control (sense) RNA probes do not produce any signal (17).

Slide 8: Gli1 is expressed in local and metastatic colon cancer epithelial cells

Detecting Gli1 protein with antibodies has been exceedingly difficult, and most commercial antibodies recognize overexpressed protein—for instance, through cell transfection with cDNA-containing vectors, but not endogenous Gli1. However, Frédéric Varnat, now a chief scientific officer at a biotech company in France, and others in the lab were able to detect endogenous Gli1 in the epithelial tumor cells (t) of colon cancers both in the primary tumor site (t, top) and in liver metastases (t, bottom) with both polyclonal and monoclonal antibodies directed against the same Gli1 epitope. Note that the stroma (s) does not show any clear labeling (17).

Slide 9: Glioma stem cells produce invasive gliomas that recapitulate the original human tumors

Yet another example is provided by an analysis of human glioma cells implanted into the brains of immunocompromised mice, which express Hh-Gli pathway components (5). The injected cells were transduced with LacZ-expressing lentivectors prior to injection and are highlighted in brain sections after tumor growth and parenchymal invasion by the blue color obtained through the X-Gal reaction (top). The bottom panel shows the results of RNA in situ hybridizations of tissue slices with antisense riboprobes (top row) and sense control riboprobes (bottom row). Hh-Gli signaling components are strongly expressed by the implanted human cells. Together, the many cases tested in my lab for the expression of Hh-Gli pathway components in single-cell resolution assays demonstrate that the Hh-Gli pathway is active in the tumor cells themselves. Whether cell subpopulations within the human tumor stroma have a very low amount of pathway activity remains to be determined.

Slide 10: Hh-Gli signaling is required by many types of sporadic human cancers

A second important finding from my lab is the elucidation of the requirement for Hh-Gli signaling by many kinds of human sporadic tumors. To this end, we have taken the time and put forth a lot of effort to obtain patient samples directly from hospitals through our excellent medical collaborators (at New York University Medical Center, Hôpital Universitaire de Genève, and Hôpital Pitié-Salpêtriére de Paris) and to develop lentivector-based tools for manipulating the human Hh-Gli pathway in a cell-autonomous manner in vivo and in vitro.

Slide 11: Human Gli1 induces skin and CNS tumors

The first link between the Hh-Gli pathway and the development of sporadic tumors came not from work with human cells but from experiments in Xenopus laevis (frog) embryos. At the time, I was injecting Gli1 mRNA into early cleavage stage frog embryos to test for the ability of this single transcription factor to respecify fates of dorsolateral cells within the neural tube to a floor plate phenotype, where Gli1 is expressed early in development (18). In the experiments shown on this slide, I injected in vitro transcribed human Gli1 mRNA into the dorsal or ventral animal blastomeres along with LacZ mRNA to trace the injected cells and their descendants. When injected ventrally, Gli1 induced epidermal tumors, and when injected dorsally Gli1 induced tumors in the central nervous system (CNS). Both were blue after reacting with 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-gal), which indicated that they were LacZ-positive (LacZ+) and thus derived from the injected cells. The tumors contained an increased number of 5-bromo-2′-deoxyuridine (BrdU)–positive proliferating cells (4, 10).

Slide 12: Exogenous human Gli1 induces CNS tumors that require endo­genous frog Gli1

Using this assay, we could show that Gli1 was sufficient to induce tumor formation by altering somatic cells. These results parallel the findings of the Scott lab with ectopic Shh expression or loss of Ptch1 function in mice and were later extended by those of the Toftgård lab with ectopic expression of Gli1 in the mouse skin (1921). However, the frog tumor assay allowed us for the first time to test for the requirement of endogenous frog Gli1 for tumor development. The injected human Gli1 mRNA and the protein transcribed from that message are degraded quickly after injection, yet we observed tumors appear after gastrulation. These tumors expressed endogenous frog Gli1, so I reasoned that perhaps it could be involved in perpetuating the tumor-inducing ability of the injected human Gli1. Indeed, coinjection of a morpholino antisense oligonucleotide specific for frog Gli1, but not a control morpholino, along with the human Gli1 mRNA resulted in the absence of tumor growth and the correct normal fate specification of the injected blastomeres, as seen by the normal distribution of blue cells (4). We extended these experiments to human cells and tested for the first time whether human tumor cells could be sensitive to Hh-Gli pathway inhibition. Indeed, treatment of medulloblastoma and glioma cells (4) with cyclopamine, a plant alkaloid that specifically inhibits the key Hh transducer Smo (22), resulted in inhibition of proliferation. A year later, the Beachy lab extended our findings with primary medulloblastomas. Additional work from our lab and others has also shown that Shh-Gli signaling is required by many cancer types, including primary and cell-line prostate cancers (16, 23, 24), melanomas (11), meduloblastomas (4, 25), gliomas (5, 26), and hematological malignancies (27).

Slide 13: Smo blockade decreases endogenous tumor size in vivo, which improves the animal’s health

Studies with mouse models of cancer provide further support for the involvement of Hh-Gli signaling in tumorigenesis. We first utilized a mouse model initially developed by the Scott lab in which mice heterozygous for a loss-of-function mutation in Ptch1 develop medulloblastomas. A cohort of Ptch+/−;p53−/− littermates was split, and one-half were treated systemically with intraperitoneally injected cyclopamine. Those that received the drug had no tumors or very small ones. In contrast, the other half that received only the carrier (cyclodextrin) had large tumors. The latter showed signs of illness, whereas the former were perfectly healthy (28). This work was performed by Pilar Sánchez and me in the lab and recapitulated contemporaneous findings from another group using the same mouse model with a proprietary Hh signaling inhibitor (29).

Slide 14: NRasQ61K;Ink4a−/−mouse melanomas express Gli1 and are eliminated by Smo inhibitors

Barbara Stecca and I showed that sporadic mouse melanomas that appear in mice carrying an oncogenic NRasQ61K transgene and lacking the tumor suppressor Ink4a express endogenous Gli1 and require Smo function for growth in vitro and in vivo and that tumor growth is abolished by peritumoral injection of cyclopamine (11). Together, the frog, mouse, and human data make a strong case for the requirement in cancer of Hh-Gli signaling in an autocrine or juxtacrine mode or both.

Slide 15: Hh-Gli in recurrence and metastases

A third critical issue is that Hh signaling and the Gli code, the sum of all positive and negative Gli activities, regulates not only tumor growth, but also recurrence and metastases, two key medical problems of cancer.

Slide 16: Colon cancer primary epithelial cells require Hh-Gli function

To evaluate the requirements for Hh-Gli signaling in colon cancers, we used patient-derived cancer cells in which we manipulated Hh-Gli activity through lentivector-mediated expression of cDNAs encoding Hh signaling components or shRNAs or siRNAs that target messages encoding specific components for degradation. Colon cancer cells were obtained from fresh tumors and shown to homogeneously express cytokeratin antigens that confirmed their epithelial identity (top). In vitro, modulation of Hh-Gli signaling with siRNAs (B) or shRNAs (C) specific for Gli1, Gli2, or Gli3, or with cyclopamine (D) resulted in consistent changes in cell behavior. Stimulating Hh-Gli signaling with Ptch1-specific shRNA (shPtch1) or Gli1 cDNA, for instance, resulted in increased proliferation and decreased apoptosis. In contrast, when the pathway was inhibited with cyclopamine, Gli1-specific siRNA (siGli1), Smoothened homolog–specific shRNA (shSmoh), or cDNA encoding Gli3R, for example, proliferation decreased and apoptosis was enhanced. This work, performed by Frédéric Varnat and me, also showed that the survival of colon cancer liver metastases is particularly sensitive to pathway inhibition (17). Our work also clarified the role of Hh-Gli in colon cancer, which had been quite controversial and based on limited in vitro data from very few cell lines (30, 31).

Slide 17: Interference with Hh-Gli signaling blocks tumor growth in vivo, and enhanced signaling promotes tumor growth

In vivo, xenografts with a widely used diploid colon cancer cell line (Ls174T) showed that tumor size paralleled the amount of pathway activity. Increasing signaling by expressing either Gli1 or shPtch1 led to enhanced tumor growth, whereas decreasing signaling by expressing shSmoh or Gli3R abolished tumor growth (top panel). Note that the effect of shSmoh was rescued by concomitant expression of Gli1, which acts downstream of Smoh. Similar results were obtained with primary patient-derived high-grade colon adenocarcinoma cells (CC14), and both cell types displayed similar growth curves under different conditions (lower graphs) (17).

Slide 18: Limited blockade of Hh-Gli signaling prevents tumor recurrence

To investigate the role of Hh-Gli signaling in tumor recurrence, we used xenografts of cells from a colon cancer cell line to test whether cyclopamine could prevent all tumor growth in a sustained fashion. After the formation of palpable tumors, cyclopamine was administered, and the tumors disappeared, as expected. However, the tumors recurred when cyclopamine treatment was interrupted. We then tested for tumor recurrence after 5, 10, or 20 additional days of treatment. The tumors recurred aggressively after a 5- or 10-day additional treatment, but not after 20 days’ additional treatment. In the latter cohort, the mice remained healthy and tumor-free for over a year (17). These results are similar to those we obtained with melanoma recurrence in a mouse model (11). Together, they suggest that a sustained, but time-limited, blockade of Hh-Gli signaling is required to prevent tumor growth and that any tumor-initiating cells driving recurrence also require pathway activity. This is an important issue that I will return to later in this talk.

Slide 19: Interference with Hh-Gli signaling blocks metastatic growth in vivo

To explore the role of Hh-Gli signaling in metastases, we adopted a well-known metastatic assay that uses injection of cancer cells into the circulation through the tail vein and subsequent surveillance for tumor formation in the lungs. This assay does not recapitulate the entire metastatic process, but it does provide a test for the critical step of target organ seeding and metastatic growth. To visualize even small metastases, tumor cells were transduced with a LacZ lentivector so that the cells appeared blue after reacting with X-gal. In the sample shown on the slide, we used metastatic melanoma cells that readily colonize the lungs of immunocompromised mice after tail vein injection. Two weeks after injection, mice were split into two groups. One received systemic cyclopamine treatment via intraperitoneal injection, and the other received carrier (cyclodextrin) only. Animals treated with cyclopamine, but not carrier only, displayed few or no metastases (11), which indicated that metastatic growth requires Hh-Gli activity. The animals treated systemically with cyclopamine remained healthy. These results provide the proof of concept for a time-limited anticancer treatment through Hh-Gli pathway interference. Moreover, they support the use of cyclopamine or its derivatives as antitumor and antimetastatic agents.

Slide 20: Signaling integration by the Gli code

A fourth key issue to appreciate is that the Gli code is regulated by Hh and non-Hh inputs, which include peptide growth factors, oncogenes, and tumor suppressors.

Slide 21: Hh-Gli activity integrates peptide growth factor signaling

The first evidence for this idea was found in our lab through an exploration of the role of Gli2 in early Xenopus development with Rachel Brewster, a former postdoctoral fellow in our lab and now a principal investigator at the University of Maryland, Baltimore County, Maryland (32). We found that Gli2 was expressed mainly in the ventrolateral mesoderm of early gastrulae and that increasing its abundance in the ectodermal animal cap region by injecting Gli2 mRNA induced the ectopic expression of mesodermal markers. Moreover, when injected into the lateral mesoderm, enhanced Gli2 abundance altered the specification of these cells toward more posterior and ventral fates, resulting in ectopic tails. Tadpoles that developed from injected embryos thus had supernumerary tail structures formed by the injected cells, which had been labeled by coinjection of LacZ (arrows, B to D). Gli2 and Gli3, which are coexpressed in ventral posterior mesoderm, induced expression of the transcription factor Brachyury (Xbra), and both Xbra and Gli2 were induced by fibroblast growth factor (FGF). This effect was not seen with Gli1 or with Shh (4). Gli2 and Gli3 thus function in the well-characterized FGF-Xbra autoregulatory loop, and the data placed Gli factors downstream of peptide growth factor–receptor tyrosine kinase (RTK) function.

Slide 22: Hh-Gli activity integrates EGF-Ras-MEK-Akt signaling

Additional and independent support for Gli transcription factors acting downstream of RTKs derived from an analysis of the interaction of epidermal growth factor (EGF) and Shh in neurosphere formation, which was carried out in the lab by Verónica Palma, a former postdoctoral fellow in our lab and now a principal investigator at Universidad de Chile, Santiago, Chile) (33, 34). Neurospheres are floating clones grown in vitro that are derived from a single self-renewing stem cell that also gives rise to non–stem cell precursors and differentiated cells. Thus, a clonal neurosphere will contain some stem cells but will be primarily composed of other precursors and differentiated cells. The number of such stem cells in a neurosphere can then be tested in secondary neurosphere–forming assays. Whereas other labs had failed to see any interactions between EGF or FGF and Shh in neurospheres, we decided to lower the amount of exogenous EGF from that used in standard neurosphere cultures, and this allowed us to detect a positive interaction. Low constant doses of EGF or Shh allowed increasing doses of Shh or EGF, respectively, to boost proliferation. Other labs subsequently provided additional evidence and further explored the mechanisms for such synergism (35). Because EGF signaling can be oncogenic and acts upstream of Ras, we decided to test whether oncogenic Ras could indeed boost Shh signaling generally, and Gli1 function specifically. Indeed, we found that oncogenic Ras-dependent signaling through mitogen-activated or extracellular signal-regulated protein kinase kinase (MEK) or the serine-threonine kinase Akt boosted Gli1 activity, as seen here in a Gli-dependent luciferase reporter assay. The boosting of Gli1 by these oncogenes remains sensitive to inhibition by the Gli1 blockers Suppressor of fused homolog (Sufuh) and Gli3R (11). We and others have now documented such interactions in many human cancer types, and here I show results with human U87 glioma cells.

Slide 23: Exogenous and endogenous Hh-Gli signaling require endogenous Ras-MEK-Akt function

As in gliomas, human melanoma cells, which often harbor activating mutations in N-Ras or B-Raf, also show Ras-Gli1 synergism, and the mechanism of action is, at least in part, related to the subcellular localization of Gli1. As a transcription factor, it acts in the nucleus, but it shuttles in and out of the nucleus and can be retained in the cytoplasm by various factors, including Sufuh. This results in a fine balance of the abundance of nuclear versus cytoplasmic Gli1, and thus, transcriptional activity reflects the strength of signaling. An oncogenic mutant form of Ras (Q61→K; designated as Ras* on the slide) drives enhanced nuclear localization of Gli1, which is reversed by Sufuh. This is seen both with exogenous (above) and endogenous (below) Gli1 when nuclear localization of endogenous Gli1 is inhibited by pharmacological blockers of Akt1 or MEK1 function (11, 36). Work by other labs has provided additional evidence for interactions between the Hh-Gli and the Akt and MEK pathways in normal development (37).

Slide 24: Additive and synergistic effects of Smoh inhibitors with Akt and MEK inhibitors

We were also able to show that inhibition of Hh-Gli signaling and blockade of Akt or MEK1 activities could have synergistic effects in primary and cell line gliomas (5). This sets the stage for a combinatorial therapy wherein potential side effects of each drug are minimized because lower doses of each can be used, and the effect in targets cells where the inhibitors synergize is enhanced.

Slide 25: p53 regulates Gli1 activity

A major step forward in the idea of the oncogenic load—the loss of tumor suppressors and the activation of oncogenes—regulating the Gli code in human cancer (8) came from studies performed by Barbara Stecca and me in the lab. Using a mouse model in which Gli1 could be conditionally activated, we found that enhanced Gli1 or decreased p53 activity caused similar increases in neural stem cell numbers and that these two showed synergy (upper panel) (38). A synergistic effect of reduced p53 and enhanced Gli1 was also observed on tumor growth in vivo as shown in the lower panel, by using orthotopic grafts of U87 glioma cells (38). Tumors are outlined in the photographs.

Slide 26: Gli1 and p53 establish a negative autoregulatory loop that controls stem cell and tumor cell numbers

Additional work with other human tumor types allowed us to further expand our findings to show that such synergism can take place in stem cells and is seen in CD133-positive (CD133+) colon cancer stem cells (top) (17). In essence, all of these findings, plus the ability of Gli1 to inhibit p53 function by elevating the abundance of the ubiquitin ligase Mdm2 (38, 39), allowed us to discover the existence of a negative regulatory loop between Gli1 and p53.

Slide 27: p53 regulates the stability and structure of human Gli1

How p53 inhibits Gli1 function is just beginning to be understood. Structural studies allowed us to detect a previously unknown isoform of Gli1 that lacks the N terminus of the full-length protein, as well as C-terminally truncated forms that mimic the repressor forms of Gli2 and Gli3. This is consistent with activity of a C-terminally truncated Gli1 cDNA (40) and with the context-dependent positive and negative functions of Gli1 as shown through the work of Vân Nguyen, a former postdoctoral fellow in the lab, in frog embryos (41). Using antisera raised against Gli1, we could show that p53 alters the abundance and phosphoryl­ation status of the N-terminally truncated Gli1130 isoform. Decreased p53 or enhanced Shh signaling driven by decreased Ptch1 or Sufuh abundance through RNA interference resulted in increased abundance of the unphosphorylated form of Gli1130, which suggests that p53 promotes the phosphorylation of Gli1130 to an inactive state and suggests the existence of a heretofore unidentified phosphatase that may be regulated by Shh signaling (38).

Slide 28: Regulation of the Gli code by the oncogenic load

Overall, the data summarized up to now supports a model in which the amount of active Gli1 is a critical determinant of cell behavior and cell fate. It is normally regulated in stem cells by Hh signaling and other inputs, such as the described repression by oncogenes and activation by peptide growth factor–RTK inputs. In contrast, in cancer, the oncogenic load leads to the superinduction of Gli1 that results in enhanced stem cell self-renewal and cancer cell survival and proliferation, which promote tumor growth, recurrence, and metastasis (17, 36, 42).

Slide 29: Hh signaling and the Gli code in stem cells

A fifth central issue is the role of Hh signaling and the Gli code in the control of stem cells and cancer stem cells.

Slide 30: Hh-Gli signaling modulates mouse neural and human glioma stem cell size and number

Evidence for a role of Shh in the control of stem cell self-renewal was first derived from work on neural stem cells. Verónica Palma and I showed that Shh signaling was a critical regulator of the number and size of clonogenic neurospheres formed in vitro by culturing of neural stem cells (left) (33, 34, 43). This suggested Shh control of two important parameters in normal neurospheres: sphere number, which indicates control of stem cell self-renewal, and clone size, which suggests a role in controlling the proliferation of stem cell–derived precursors, because most of the cells in a neurosphere are not stem cells. The latter perfectly fits with the results on brain precursors in the ventricular zones of the CNS and external germinal layer of the cerebellum that we had obtained previously (1, 4). Analyses of human cancer stem cells revealed that Shh signaling controls the same two key parameters: Enhanced signaling induced by application of exogenous ligand augmented clone size and number, whereas blocking signaling with cyclopamine or shSmoh resulted in decreased clone size and number (5). These results, however, were obtained in vitro.

Slide 31: The abundance of Gli1 dictates cerebellar stem cell number

Use of our inducible Gli1 mouse model also allowed us to show that the number of neurospheres derived from brain tissue with enhanced Gli1 in vivo correlated with the actual amount of Gli1 expression (38), which indicated a close relation between Gli1 and stem cell self-renewal. These results expand our initial findings using Shh ligand and the Smoh blocker cyclopamine (1, 4, 33, 34) and further highlights the importance of Gli1 levels in the control of stem cell behavior.

Slide 32: A red-green in vivo competition assay for tumor growth and stem cell tracking

To test the idea that Hh-Gli signaling regulates stem cell behavior in vivo, I designed a new assay, called the red-green competition assay, in which human tumor cells obtained from patients, for instance, are transduced with either red fluorescent protein (RFP)– or green fluorescent protein (GFP)–expressing lentivectors, and then the cells are mixed and coinjected into mice. This can be done with whole-tumor populations or with stem cells, such as CD133+ glioblastoma or colon cancer populations. The resulting tumors are harvested, and an aliquot is used to quantify the RFP-positive (RFP+) and GFP-positive (GFP+) populations by fluorescence-activated cell sorting (FACS), and then the mixture is reinjected into a second host directly or after further stem cell enrichment by, for example, magnetic activated cell sorting (MACS) for CD133 or other antigens. Thus, tumor and stem cell behavior can be tracked in vivo with internal controls and for as long as necessary. Moreover, in this system, there is always a tumor formed by RFP+ cells (unless non–cell autonomous signals from GFP+ cells negatively affect RFP+ cells), and thus all the important signals from the tumor stroma are present. Finally, this is a competition assay and allows for the evolution and selection of the fittest tumor cells, much as is likely to occur in endogenous tumors in the patient.

Slide 33: A novel in vivo red-green competition assay

Use of the red-green competition assay with colon cancer stem cells showed that although RFP+ and GFP+ populations were maintained through two serial passages in vivo (top), blocking Shh signaling with shSmoh in GFP+ cells prevented tumor contribution by GFP+ cells (middle panel), and enhancement of pathway activation in GFP+ cells by shPtch1 resulted in expansion of the GFP+ pool (bottom) (17). Similar results have been obtained with gliomas (44). Thus, the evidence points to an absolute requirement for Shh signaling in cancer stem cells.

Slide 34: How does Hh-Gli signaling interact with the Wnt-Tcf pathway in colon cancer?

These results and the conclusions presented here led us to ask about the interaction between Shh-Gli function and another key oncogenic and developmental pathway, the Wnt-Tcf pathway, which has been proposed to be of central importance in the initiation and maintenance of human colon cancers. In these tumors, Wnt activation through loss of the Wnt mediator adenomatous polyposis coli (APC) or activation of β-catenin precedes other oncogenic changes, such as Ras mutation and loss of p53. Because we had recently shown that Shh signaling, and especially Gli1 activity, are essential for colon cancers of all grades, including liver metastases (17), and José Mullor, a former postdoctoral fellow in our lab and now a principal investigator at Instituto de Investigación Sanitaria La Fe in València, Spain, co-workers in the lab and I had shown that Wnt genes are targets and mediators of Gli function in other systems (45), the question of Hh-Gli–Wnt-Tcf interactions arose anew.

Slide 35: Intestinal tumorigenesis in mice through activation of Wnt signaling by loss of Apc

To begin to address a possible interaction between these pathways in colon cancer, we used the mouse model of colon adenoma in which animals lacking the tumor suppressor and Wnt pathway inhibitor APC develop fatal massive tumorigenic lesions in the small intestine. Previous elegant work had demonstrated that loss of Apc in mice mimics loss of Apc in humans and drives adenoma formation (46, 47).

Slide 36: Smo inhibitors prolong disease-free survival

Apc-floxed mice carrying tamoxifen-driven Cre activity were defloxed, and tumors soon appeared and survival plummeted. In contrast, those treated systemically with cyclopamine survived longer, which suggested that Shh-Gli activity is required downstream of or in parallel to activation of Wnt-Tcf downstream of APC (48).

Slide 37: Genetic deletion of Smo rescues the lethality of loss of Apc

Similar results were obtained by introducing a floxed allele of Smo. Whereas defloxed Apc mice developed tumors and died, simultaneous defloxing of Smo rescued the Apc loss-of-function phenotype (48). Complementary results were obtained with a hypomorphic allele of Smo (49).

Slide 38: Crypt analyses show that loss of Smo rescues overall defects triggered by loss of Apc but not the nuclear localization of β-catenin

Analyses of intestinal crypt phenotypes showed that concomitant loss of Smo rescued the enhanced proliferation, loss of goblet cells (Alcian blue stain), and increase in lysozyme-expressing Paneth cells that are seen in Apc mutant intestines but did not rescue the nuclear localization of β-catenin (48). These results indicate that Hh-Gli activity is required downstream of β-catenin for Wnt pathway–induced tumorigenesis.

Slide 39: Hh-Gli signaling is dominant over Wnt-Tcf signaling in human colon cancer cells in vitro

Direct analyses of human colon cancer cells started with the use of the same cell line used by the Clevers lab to demonstrate the essential role of Wnt-Tcf signaling in this cancer type (50). Ls174T colon cancer cells were stably transfected with a dominant-negative isoform of Tcf4 (dnTCF4) (50) that was rendered conditional by fusion with the hormone-binding domain of a mutant estrogen receptor (dnTCF4ERT2 ; kindly made available to us by E. Batlle, Institut de Recerca en Biomedicina, Barcelona). dnTcf4 activity will appear only when the protein translocates into the nucleus in the presence of tamoxifen (TAM). Addition of tamoxifen abolished tumor cell growth in vitro on standard plastic dishes as expected. However, the Hh-Gli pathway was dominant, because enhancing its activation by expressing shPtch1 or Gli1 stimulated growth both with and without Tcf activity, and blocking Hh-Gli signaling with shSmoh or Gli3R abolished growth regardless of Tcf activity (51). As expected for their relative positions in the Hh-Gli pathway, the effects of shSmoh were rescued by Gli1.

Slide 40: A major switch in pathway activation at the metastatic transition of human colon cancers

A major surprise came from the analyses of gene expression in patient-derived colon cancers of different World Health Organization–defined TNM (tumor, node, metastases) stages carried out in the lab by Frederic Varnat. Quantitative reverse transcription–polymerase chain reaction (QRT-PCR) of CD133+ and CD133 colon cancer cell populations showed that the Shh-Gli pathway signature [expression of Shh, Ptch1, Gli1, Gli2, and Hedgehog-interacting protein (Hip)] is enhanced preferentially in CD133+ cancer stem cells and in cells from advanced stage tumors in which the patients have node or liver metastases, or both, that are in the TNM3 or TNM4 stage and in the liver metastases themselves (top bar graphs) (17). In contrast, a Wnt-Tcf signature formed by key transcriptional targets such as Lgr5, Sox4, c-Myc, and Axin2 were elevated in nonmetastatic tumors (TNM1 and TNM2 stages), as expected from the well-established role of Wnt signaling in adenomas (50). It was surprising that the Wnt signature was greatly repressed in local colon tumors taken from patients that already had node or liver metastases (TNM3 and TNM4) (bottom bar graphs) (51). This was totally unexpected because the prevailing dogma indicated that Wnt-Tcf signaling was required in colon cancers from the birth of the tumor until the death of the patient. This striking result was consistent with the essential role of Hh-Gli signaling in colon cancer (17) but suggested that advanced cancers shed their addiction to Wnt-Tcf activity.

Slide 41: The Wnt-Tcf pathway is not universally required for colon carcinoma growth in vivo

To test whether loss of Wnt-Tcf was required for human cancer colon growth in mouse xenografts, we used two independent systems to block Tcf activity solely in the grafted human cells. First, we used the tamoxifen-inducible dnTcf system described above and then also the doxycycline-inducible system of the Clevers lab (50). The first system is based on retention of dnTcf in the cytoplasm until nuclear translocation is induced by tamoxifen treatment. The second system is based on the transcription of dnTcf only in the presence of doxycycline. Both gave the same results, and here I show results using the tamoxifen system. The archetypal Ls174T colon cancer cells, which show an absolute requirement for Tcf activity in vitro, were completely unaffected by Tcf blockade in vivo. More important, patient-derived primary colon cancer CC14 cells and patient-derived liver metastasis mCC11 cells were also totally unaffected by normal or suppressed Tcf activity. Furthermore, blocking Tcf, and thus Wnt-Tcf activity, even enhanced the growth of a third patient-derived TNM3 colon cancer (CC36) in mice (51). Only tumors derived from the colorectal adenocarcinoma cell line DLD-1 regressed upon activation of dnTcf (51). Therefore, although additional in vivo data with numerous primary tumors should resolve the penetrance of these phenotypes, our results are striking and call into question the dogma of a universal requirement for Wnt-Tcf function in medically relevant colon cancers. Indeed, it is not the early, gut-localized tumors, but the distant metastases that remain incurable.

Slide 42: Blockade of Wnt-Tcf enhances metastatic growth of human colon carcinomas

Because a massive decrease in the Wnt-Tcf signature is seen in tumors from patients who have metastases, the essential question to address was whether altered Wnt-Tcf function could affect metastatic behavior. To this end, we used the system we had employed previously in which LacZ-transduced human cancer cells are injected into the tail vein of immunocompromised (nude) mice. Injection of Ls174T cells led to the development of very few and very small metastatic lesions in the lung (top left), each of which consisted of fewer than 10 cells. In contrast, Ls174T and mCC11cells expressing the introduced dnTcf4 construct showed enhanced metastatic seeding and growth or, in the case of cells that had previously shown no metastatic activity in this assay (CC36 and CC14), displayed new metastatic activity (51). We conclude that human colon cancers need to down-regulate Wnt-Tcf signaling in order to become metastatic. Wnt-Tcf inhibitors may thus not only be inefficient but may worsen the situation of patients with advanced cancers or with metastatic risk!

Slide 43: The in vitro milieu imposes an early nonmetastatic state to metastatic cells

But why then do the same cells respond differently to Tcf blockade in vitro versus in vivo? Note that we have obtained the same in vitro results as previous studies, but that the in vivo Tcf blockade had never been tested. A possible answer to this conundrum came from the detection of differential gene expression in vitro versus in xenografts. Cells cultured on plastic and with serum (that is, under normal in vitro conditions) more strongly expressed the Wnt-Tcf signature genes (top) as compared with the corresponding tumors in vivo. And this was true for both CD133+ (red) and CD133 (blue) populations. In contrast, the Hh-Gli signature (bottom) was greatly enhanced in vivo as compared with in vitro, especially in CD133+ stem cells. These changes were seen in all three samples tested. Indeed, we found that the signatures of patient tumors were recapitulated by xenografts and not by culture conditions in either two- or three-dimensional cultures (51). Thus, we conclude that advanced metastatic colon cancer cells acquire a nonmetastatic early tumor phenotype that is imposed by the in vitro conditions and that previous in vitro results by other groups (50) should not be extrapolated to apply to in vivo and patient conditions.

Slide 44: Model for the regulation of tumor progression and the metastatic transition by the Gli code

We therefore propose a model in which it is the degree of positive function of the Gli code or of Gli1 as the final positive output of the Hh pathway that dictates the behavior of cancer stem cells and tumors in general. The small amounts of Hh-Gli signaling observed in adenomas and nonmetastatic carcinomas are required for tumor development in cooperation with Wnt-Tcf. However, the increasing oncogenic load that cells acquire during tumor progression drives the activity of Gli1 above a threshold that is critical for the conversion of nonmetastatic to metastatic carcinomas at what we call the metastatic transition (from TNM2 to TNM3). Enhanced Gli1 then drives repression of Wnt-Tcf activity and induces epithelial-to-mesenchymal transition (EMT). Indeed, elevated Hh-Gli or decreased Wnt-Tcf signaling drive an inverse change in the other pathway, and enhanced Hh-Gli signaling drives EMT in epithelial colon cancer cells (17, 51).

Slide 45: Does high Gli1 activity drive an ES cell–like state?

What remains unclear is the genetic programs that Gli1 may drive in metastatic cells, because this is a key medical problem that needs to be solved, and an understanding of their behavior remains a fantastic challenge in biology. As I will now present, Gli1 drives changes in cancer stem cells that we propose underlie the metastatic transition.

Slide 46: How does Hh-Gli regulate cancer stem cell behavior?

QRT-PCR analyses of brain tumors from patients showed that Gli1 is expressed in all brain tumors tested, including astrocytomas of all stages, oligodendrogliomas, and medulloblastomas. All samples had amounts of Gli1 expression higher than or equal to that of normal brain tissue. Note that astrocytoma grade IV (also known as glioblastoma multiforme) had variable amounts of other Shh-Gli1 pathway components (5). These results confirmed and extended our previous analyses (4). What was surprising was detecting for the first time a “stemness” signature in human cancer. Several control genes showed random amounts of expression, whereas the expression of genes that characterize other types of stem cells, notably embryonic stem (ES) cells, showed consistently high expression in grade III gliomas. Within this signature, we highlighted the core ES cell and induced pluripotent stem cell (iPSC) signature formed by Nanog, Oct4, and Sox2. Glioblastomas also had this signature, but it was diluted by the enormous increase in the non–stem cell–derived tumor cells that actually kill the patient. Neurospheres formed by culturing glioblastomas (also known as gliomaspheres), which are enriched in stem cells, did indeed show the same signature (5). This was the first description of a tumor stemness signature that resembled the core ES cell and iPSC signatures, and we call it the “ES-like” signature. But is this signature related to Hh-Gli signaling?

Slide 47: Hh-Gli pathway blockade represses the expression of the cancer ES-like signature

A first hint that this may be the case came from the analyses of two patient-derived glioblastomas (GBM-7 and GBM-8) grown in vitro as gliomaspheres and treated with cyclopamine or the structurally similar compound tomatidine, which does not affect Hh signaling. The ratio of gene expression with cyclopamine as compared with tomatidine was determined by QRT-PCR and showed that the expression of genes encoding components of the Hh-Gli pathway, notably Gli1, were inhibited by cyclopamine. Notably, the ES-like stemness signature behaved like Gli1 and was coherently down-regulated, which showed that the ES-like signature is Hh-Gli–dependent (5).

Slide 48: The core ES and induced pluripotent stem reprogramming gene set is regulated by Hh-Gli, but not Wnt-Tcf, signaling in human cancer

Recently, we have addressed this question more directly by analyzing the expression of the core ES-like signature genes in patient-derived colon cancer cells, both in whole populations and in CD133+ cancer stem cells. Blocking Hh-Gli signaling in a cell-autonomous manner with lentivector-encoded shSmoh or Gli3R decreased Gli1 and Ptch1 expression, as well as expression of the ES-like genes Nanog, Oct4, Sox2, and Klf4. Conversely, enhanced activation of the Hh-Gli pathway resulted in greatly elevated expression of ES-like signature genes. Numbers refer to ratios of expression between experimental and control cells (a value of 1.0 would indicate equal expression in experimental and control cells). Red in the heat map denotes enhanced expression, and blue indicates repressed expression. No effects were detected when Tcf function was blocked by dnTCF4 (51). Thus, we conclude that the cell-autonomous regulation of the ES-like signature is Hh-Gli–dependent.

Slide 49: A Gli–c-Myc positive-feedback loop

The previous slide showed an increase in c-Myc abundance induced by enhanced Hh-Gli1 activity and a reduction in c-Myc abundance induced by diminished Hh-Gli1 activity. Here, I show that a dominant-negative form of c-Myc (dnMyc) represses Hh-Gli signaling and Gli1 expression, whereas increasing c-Myc activity by overexpressing a c-Myc cDNA leads to enhanced Hh-Gli signaling and Gli1 expression. This effect was observed in all primary colon cancer samples and cell lines tested. As in the previous slide, red denotes enhanced expression, and blue indicates repressed expression.

Slide 50: The cancer ES-like signature follows the differential regulation in vitro versus in vivo of Hh-Gli but not Wnt-Tcf

Consistent with the analyses of Wnt-Tcf and Hh-Gli in vitro versus in vivo, the ES-like signature was enhanced in vivo as compared with in vitro, most notably in CD133+ cancer stem cells. Thus the ES-like signature tracked with Hh-Gli but not Wnt-Tcf signaling (51).

Slide 51: Enhanced Gli1 activity in Nestin+ precursors in vivo leads to enlarged brains

Our focus on this Gli-dependent ES-like signature also developed from another angle. Work in the lab analyzing the role of Gli1 in brain development led us to use the doxycycline-inducible system to generate bigenic mice that overexpress Gli1 in Nestin-positive (Nestin+) precursors and stem cells. These animals with increased Gli1 in Nestin+ precursors and stem cells exhibited larger brains, as shown for the cortex (Ctx), tectum (Tct), and cerebellum (Cb) (top). Sectioning these brains revealed that the Nestin+ cell population, which was also labeled by LacZ expression, was greatly expanded, and the ventricular zones showed increased circumvolutions because of the expansion of the tissue within the limited intracranial space (38). These results demonstrate that in mice, as in frog embryos (4, 18), Gli1 is sufficient to alter the behavior of neural precursors in vivo.

Slide 52: Enhancement of Gli1 activity in vivo increases the number of cells that display glioma-like behavior

Further analyses also revealed a pretumorigenic phenotype resulting from Gli1 overexpression. Because the changes in brain anatomy driven by enhanced Gli1 abundance (our system induces a ~2- to 8-fold increase, on average) are lethal, it is not possible to assess the development of tumors as the animals age. However, cells overexpressing Gli1 and labeled with GFP were observed to stream out of the ventricular zone and invade the brain parenchyma in association with CD34-positive (CD34+) blood vessels seen here in a confocal microscopy z-stack reconstruction (38). Such behavior is typical of human gliomas and suggests that increased Gli1 activity drives a precancerous change in behavior.

Slide 53: Enhancement of Gli1 activity in mice increases the number of clonogenic stem cells in vivo and in vitro and boosts Nanog abundance

To test for changes in stem cell behavior, we challenged primary cortical, thalamic, and cerebellar cells from mice overexpressing Gli1 in neural precursors to make neurospheres in vitro, thus allowing for a retrospective assessment of stem cell frequency in vivo. All brain regions tested showed enhanced clonogenic activity after enhanced Gli1 expression, which suggests that Gli1 increased the numbers of stem cells in multiple brain domains. Analyses of gene expression changes in response to enhanced Gli1 expression in vivo showed that of all the tested genes, Nanog showed the greatest change, particularly in cerebellar neurospheres (38). This was striking because Nanog had not been previously shown to be expressed in normal brain cells and was thought to be mostly restricted to the early embryo and the germ line. Recent work, however, confirms the expression of Nanog in cells of the developing mouse cerebellum (52).

Slide 54: Lentiviral-mediated interference with Smoh inhibits human glioma growth and invasion

A role for Hh-Gli in gliomas was also indicated by direct inhibition of Smoh in human glioblastomas. Lentiviral-driven expression of shSmoh resulted in a massive decrease in tumor volume in intracranial xenografts, whether this was done with constitutively or conditionally activated lentivectors, the latter acting as a control for insertion site effects (5). Thus, Hh-Gli signaling is both able to induce tumorigenic changes in cell behavior and is required for human patient-derived gliomas in vivo. Together, these results raised the question of the role of the ES-like signature in vivo. Indeed, we show that the signature is regulated by Hh-Gli and that human tumors, including glioblastoma and colon cancers, require Hh-Gli signaling. However, it remained unclear whether components of the ES-like–iPSC signature were functionally essential in cancer stem cells.

Slide 55: Nanog localization in human gliomas

We therefore decided to test directly for the requirement of a member of the ES-like signature in human cancer. Given our previous interest and work on homeodomain proteins as central controllers of cell fate in development and in the brain (53, 54), we focused on the human pluripotency homeodomain protein Nanog. Work by Marie Zbinden, Arnaud Duquet, Christophe Mas, Aiala Lorente-Trigos, and others in the lab first defined its expression in various cell lines and patient-derived glioblastoma cells and demonstrated increased expression of Nanog in cells in vitro after transduction with a Nanog cDNA lentivector and decreased expression after transduction of a lentivector expressing an shRNA specific for Nanog (shNanog). Nanog was seen in most cells but with varying abundance, with a limited number of cells showing high expression. This pattern was largely recapitulated by the use of a Nanog-RFP reporter (bottom). We also found Nanog expression to be enhanced in CD133+ stem cells (44). Given these results, we thus set out to test its function in human glioblastoma.

Slide 56: The red-green competition assay adapted for intracranial xenografts

To do so, we used the red-green competition assay, described previously, in orthotopic xenografts.

Slide 57: Nanog is essential for human glioblastoma growth in the mouse brain

Injection of the RFP control cells mixed with cells carrying only a GFP-encoding vector resulted in the growth of a tumor in the brain of host mice that largely maintained the green/red ratio present in the original injected population. However, work performed by Marie Zbinden, a former graduate student in our lab and now a postdoctoral fellow at Harvard University, Cambridge, Massachusetts, showed that expression of an shRNA that targets both Nanog and Nanogp8 (a human retrogene that also encodes the Nanog protein) and thus all Nanog-producing mRNAs in the GFP+ cells resulted in the eradication of GFP+, but not RFP+, cells in three independent patient-derived glioblastomas (44).

Slide 58: The retrogene NanogP8 is a major contributor to Nanog protein in human glioblastoma

Similar results were obtained by Arnaud Duquet, currently a postdoctoral fellow in the lab, using a different shRNA targeting only Nanogp8, which is expressed at higher abundance in tumors as compared with Nanog. Nanog made from Nanogp8 is thus essential for human glioblastoma multiforme. (44).

Slide 59: Nanog function is epistatic to an active Hh-Gli pathway

Because Nanog expression is regulated by Hh-Gli signaling in vitro, we used this system to test for a possible in vivo interaction between Nanog and the Hh-Gli pathway. We found that in vivo, control GFP+ cells are maintained, shPtch1-GFP+ cells increase in numbers, and shSmoh-GFP+ and shNanog-GFP+ cells are eliminated. We then were able to test whether GFP+ tumor expansion by enhanced Hh-Gli signaling, mediated by depletion (knockdown) of Ptch1, required Nanog function: Simultaneous knockdown of Nanog and Ptch1 abolished the GFP+ tumor population in vivo with the tumor being formed solely of RFP+ cells (44). Nanog is thus an essential mediator of Hh-Gli–induced tumorigenesis.

Slide 60: Evidence for a Nanog-Gli1 regulatory loop

Several assays were carried out to try to further dissect the relation between Nanog and Hh-Gli. First, we found that shNanog and shNanogp8 effectively eliminated Nanog protein as expected, but each also eliminated both the Gli130 and Gli100 isoforms, which indicated the presence of a positive-feedback loop between Gli1 and Nanog. Because we had previously demonstrated a negative-feedback loop between Gli1 and p53 (38), we asked whether knockdown of p53 could negatively affect the abundance of Nanog and Nanogp8. Indeed, an shRNA targeting p53 (shp53) effectively abolished p53 expression and induced a large increase in endogenous Nanog, both at the protein and mRNA (Nanog and Nanogp8) levels. Notably, the up-regulation of Gli1 in glioma cells driven by knockdown of p53 was rescued by concomitant knockdown of both Nanog and Nanogp8 (44).

Slide 61: The Nanog-Gli1 module is a central node in the control of cancer growth and cancer stem cell behavior

Together, our results point to the existence of a central regulatory Gli1-Nanog module that is negatively regulated by p53 and that acts as a key signaling node. Because p53 is commonly lost in human cancers, this would unleash the Gli-Nanog module to drive cancer stem cell and tumor expansion. This module is also regulated by inputs that affect Gli1 and Nanog expression, including Hh inputs and the oncogenic load, which is the accumulation of oncogenic events, such as mutations in Ras, EGFR, and B-Raf, and loss of tumor suppressors, such as Pten and p53. The output of this regulatory network is the control of cancer growth and stemness (44).

Slide 62: Ongoing work

We are now expanding this work by testing our hypothesis that the ES-like signature, which essentially recapitulates an iPSC reprogramming signature and prominently includes Nanog, drives a reprogramming-like event responsible for the acquisition of invasive and metastatic behaviors. Our results also offer the possibility of targeting not only Hh and oncogenic signaling, but also the Gli code and the ES-like signature to eliminate tumor growth and cancer stem cells. Indeed, there are efforts in the lab focused on developing new chemical modulators of Hh-Gli signaling (55) and the Gli1-Nanog module.

Slide 63: Acknowledgments

Here, I presented results from my lab performed by current and former lab members, previously in New York City and now in Geneva. Herein are listed these brilliant scientists with whom I have had the privilege to work and share this exciting and ongoing story. I am grateful to the many institutions that have generously supported our work through the years, and thank you for your attention.

Editor’s Note: This contribution is not intended to be equivalent to an original research paper. Note, in particular, that the text and associated slides have not been peer-reviewed.


View Abstract

Navigate This Article