Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Cell Rep. Author manuscript; available in PMC 2017 September 27.
Published in final edited form as:
PMCID: PMC5616185

The Ulk3 kinase is critical for convergent control of cancer associated fibroblast activation by CSL and Gli


The connection between signaling pathways activating cancer associated fibroblasts (CAFs) remains to be determined. Metabolic alterations linked to autophagy have also been implicated in CAF activation. CSL/RBPJ, a transcriptional repressor that mediates Notch signaling, suppresses gene expression program(s) leading to stromal senescence and CAF activation. Deregulated Gli signaling can also contribute to CAF conversion. Here we report that compromised CSL function depends on Gli activation for conversion of human dermal fibroblasts into CAFs, separately from cellular senescence. Decreased CSL up-regulates the expression of the Ulk3 kinase, which binds and activates Gli2. Increased Ulk3 also induces autophagy, which is unlinked from Gli and CAF activation. Ulk3 up-regulation occurs in CAFs of several tumor types and Ulk3 silencing suppresses the tumor enhancing properties of these cells. Thus, Ulk3 links two key signaling pathways involved in CAF conversion and is an attractive target for stroma-focused anti-cancer intervention

Keywords: Tumor microenvironment, Cancer stroma, CAF (Cancer associated fibroblast), CSL/RBPJ, Autophagy, Gli


Stepwise acquisition of genetic alterations is implicated in epithelial tumor development, yet many of these changes can be found in apparently normal tissues, pointing to the importance of concomitant stromal changes (Bissell and Hines, 2011; Hanahan and Weinberg, 2011; Martincorena et al., 2015). Notch signaling controls cell fate commitment, differentiation and tumorigenesis (Allenspach et al., 2002; Artavanis-Tsakonas et al., 1999; Dotto, 2009). While, Notch tumor suppressive function in stratified epithelia is well established (Dotto, 2008), recent evidence indicates an opposite role in stromal fibroblasts (Hu et al., 2012; Junttila and de Sauvage, 2013; Procopio et al., 2015). Specifically, the CSL/RBP-Jκ protein (CSL), a transcriptional repressor mediating Notch signaling, is key for the negative control of cancer associated fibroblast (CAF) activation. Deletion of CSL gene in the mesenchymal skin compartment of mice results in multifocal keratinocyte tumor development preceded by dermal atrophy, matrix alterations and inflammation (Hu et al., 2012). CSL loss in primary human fibroblasts results in a similar CAF phenotype, that is associated with a p53-dependent program of cellular senescence (Procopio et al., 2015). The concomitant down modulation of CSL and p53 activities leads to the expansion of altered stromal fibroblasts and cancer cells as seen in field cancerization, a condition of major clinical significance consistent of multiple and recurrent tumors (Dotto, 2014).

The Hedgehog (Hh) pathway is linked to the development of several types of epithelial cancers (Rubin and de Sauvage, 2006). A paracrine mechanism has been implicated in CAFs activation, with Hh ligands secreted by cancer cells activating the glioma-associated transcription factors (Gli) in surrounding stromal cells (Junttila and de Sauvage, 2013; Theunissen and de Sauvage, 2009). Interestingly, while this chain of events should enhance tumor formation (Theunissen and de Sauvage, 2009), an opposite restraining function of stromal Hh-activation in pancreatic and bladder cancer has also been reported (Ozdemir et al., 2014; Rhim et al., 2014; Shin et al., 2014). Mammalian cells express three Gli proteins (Aberger and Ruiz, 2014). Under basal conditions, the Gli2 transcription factor is cleaved into a transcriptional repressor, with Hh stimulation blocking this process and converting it into activator. Expression of the Gli1 transcription factor is then induced, as signal amplifier. In Drosophila, the activation of Gli homologue Cubitus Interruptus (Ci) is well established (Ruel et al., 2003; Wang et al., 2000), with a kinase (Fused) releasing it from an inhibitory complex (Costal2/SuFu) and promoting its nuclear translocation. In mammalian cells, the mechanisms responsible for Gli activation are less understood. Recently, the unc-51-like-kinase 3 (Ulk3), an inducer in human fibroblasts of autophagy (Young et al., 2009), has been proposed to fulfill the role of Fused, through association and phosphorylation of Gli2 (Maloverjan et al., 2010; Rubin and de Sauvage, 2006). Importantly, in cancer cells Gli proteins can also be activated by Hh-receptor independent mechanisms involving PI3K/AKT (Metcalfe and de Sauvage, 2011), MAPK/ERK (Seto et al., 2009), S6K (Wang et al., 2012) and KRAS activation (Nolan-Stevaux et al., 2009; Stecca et al., 2007).

Recent evidence suggests that altered metabolic properties of CAFs, resulting from increased autophagy/mitophagy and associated shift to aerobic glycolysis, can contribute to their tumor enhancing properties (Kalluri, 2016; Martinez-Outschoorn et al., 2017). How these processes relate to CAFs activation remains to be determined. We report here a so far unexpected link between the CSL and Gli signaling pathways, with the Ulk3 kinase as an attractive target for stroma-focused anti-cancer intervention, controlling the CAF effectors gene expression program separately from the autophagy/mitophagy processes.


CAF conversion by loss of CSL function depends on Gli activation

The conversion of normal stromal fibroblasts into CAFs is likely to be a multistep process, with loss of CSL transcriptional repression inducing a large battery of CAF effector genes associated with stromal cell senescence, as a fail-safe mechanism limiting cancer/stromal cell expansion (Procopio et al., 2015). An unanswered question is the possible integration of different signaling(s) leading to CAF conversion. Like Notch/CSL, Hh-Gli signaling is deregulated in tumor stroma and can contribute to CAF activation (Junttila and de Sauvage, 2013; Theunissen and de Sauvage, 2009). Under basal conditions, the glioma-associated transcription factor 2 (Gli2) is cleaved into a transcriptional repressor, while block of this process coincides with Gli2 activation and induction of gli2 and gli1 expression as signal amplifiers (Aberger and Ruiz, 2014). We found that, in human dermal fibroblasts (HDF), the short repressive form of Gli2 was dramatically reduced, while the full-length transcription-activating protein was induced by 24–48 hours of siRNA-mediated csl silencing (Figure 1A), at a time coinciding with induction of CAF effector genes but preceding that of cellular senescence, which occurs only after 4–5 days (Procopio et al., 2015). In parallel, gli1 and gli2 expression was up regulated in HDFs and to a similar extent also in human gingival fibroblasts (HGF) (Figures 1B, 1C and S1A).

Figure 1
CAF activation by CSL silencing depends on Gli activation

The above findings are of clinical and functional significance. Patient-derived CAFs have higher Gli2 levels than HDFs, which are inversely related to CSL expression (Figure 1D). Global analysis of gene expression of HDFs vs. CAFs present notable similarities when compared to the expression profile after down modulation of CSL function in HDFs (Procopio et al., 2015), and thus in our studies we focused on a limited-number of CAF genes to characterize the mechanism of HDF conversion to CAF. Functionally, induction of CAF effector genes by csl knockdown was blocked in HDFs with concomitant gli2 silencing (Figure 1E), or in mouse embryo fibroblasts (MEF) with disruption of the gli1 and gli2 genes (gli 1/2 −/−) (Lipinski et al., 2008) (Figures 1F and 1G). Notably, up-regulation of senescent effectors by csl loss was unaffected by gli2 silencing, thus implicating Gli activation selectively in induction of CAF effector genes (Figure S1B and S1C).

An essential property of fibroblasts with deleted or silenced CSL is that they enhance the growth of adjacent cancer cells (Procopio et al., 2015). To test whether these properties are Gli-dependent, we employed a cancer/stromal cell expansion model we developed, based on mouse ear injections of weakly tumorigenic squamous cell carcinoma (SCC) cells admixed with fibroblasts with various genetic manipulations (Procopio et al., 2015). We determined every three days for three weeks the growth of EGFP-expressing SCC13 cells admixed with gli +/+ or gli 1/2 −/− MEFs, both plus/minus csl silencing. As previously reported for HDFs with csl silencing (Procopio et al., 2015), SCC13 cells admixed with gli +/+ MEFs with csl knockdown formed significantly larger lesions than controls (Figures 1H1J), while the tumor enhancing effects of csl silencing were lost with MEFs with gli1/2 gene deletion (Figures 1H–1J and S1E). Hence our data indicate that CAF conversion of stromal fibroblasts by csl loss requires Gli activity.

Ulk3 is a direct CSL target gene, which links csl silencing to Gli and CAF activation

In cancer cells, Gli proteins can be activated by Hh-receptor dependent and independent mechanisms (Aberger and Ruiz, 2014). The Ulk3 kinase was recently proposed to fulfill the same role in mammalian cells as the Drosophila Fused in Gli2 activation, through possible association and phosphorylation (Maloverjan et al., 2010; Rubin and de Sauvage, 2006). In agreement with these studies, we confirmed that Ulk3 and Gli2 can be recovered by co-immunoprecipitation in Ulk3-overexpressing cells, in which activated phosphorylated Gli2 (Xing et al., 2014) was detected (Figures 2A and 2B). We also determined the association of endogenous Ulk3 and Gli2 by proximity ligation assays (PLA) in HDFs in which Ulk3 expression was increased by serum starvation (Figures 2C and 2D).

Figure 2
Ulk3 is a direct CSL target gene which links csl silencing to Gli activation

Upon csl silencing, levels of ulk3 expression were significantly induced in several HDFs strains, as well as in HGFs, while expression of two other family members, ulk1 and ulk2, was unaffected (Figures 2E–2G and S2A–S2C). Notably, ulk3 was also induced after csl silencing in MEFs with gli1/2 gene deletion (Figures 2H and 2I), in which induction of CAF effector genes was blocked (Figures 1F and 1G). Bioinformatic analysis of the 5000bp promoter region of the Ulk3 gene upstream of the initiating ATG revealed the presence of three putative CSL binding sites. Chromatin immunoprecipitation assays (ChIP) of HDFs showed little or no binding of CSL to the first upstream site (Site 1), while the other two were strongly bound, one upstream of the transcription start site (Site2) and the other in the first intron (Site3) (Figure 3J). Histone marks of active chromatin configuration (H3K27ac and H3K4me3) were highly enriched at these sites upon csl knockdown (Figures 3K and S2D). Up-regulation of ulk3 by csl silencing is of functional significance, as Gli activation and induction of CAF effector genes were suppressed in cells with concomitant csl and ulk3 knockdown (Figure 3A). Consistent with previous reports (Maloverjan et al., 2010), we found that the activation of Gli reporter activation required Ulk3 kinase activity in HDFs (Figure S2E). In contrast, determinants of cellular senescence (cdkn1a, cdkna2 and cdkn2b) were induced by CSL loss also in concomitance with ulk3 silencing (Figures S2F and S2G). Conversely, increased ulk3 expression by lentiviral infection was sufficient to induce gli and CAF marker genes (Figures 3B and 3C), with induction of the latter being suppressed in cells with gli2 silencing (Figure 3D).

Figure 3
Ulk3 activity is required for induction of Gli1/2 and CAF effectors by csl silencing

An important property of CAFs is the ability to enhance proliferation of neighboring cancer cells. In co-culture assays, proliferation of skin-derived SCC cells (SCC13) was enhanced to a much greater extent by HDFs with silencing of csl individually than in combination with ulk3 (Figures 3E and 3F). Thus, ulk3 is a direct CSL target gene, which is required for Gli and CAF activation by csl down modulation.

CSL silencing and increased Ulk3 induce cellular autophagy and mitophagy separately from Gli and CAF activation

Similarly to other Ulk kinase family members, Ulk3 can plays a key role in initiation of autophagy (Russell et al., 2013; Young et al., 2009). Consistent with the observed up-regulation of ulk3, csl silencing in HDFs induced autophagy, as assessed by a number of biochemical events, including processing of microtubule-associated protein-1 light chain 3 (LC3) (Klionsky et al., 2016), Beclin-1 phosphorylation (Russell et al., 2013), and down-modulation of the p62 cargo and TOM20 mitochondrial proteins (Bingol et al., 2014; Klionsky et al., 2016) (Figure 4A). Induction of these events was associated with an increase of autophagosome formation, which occurred also in HDFs with csl silencing treated with the lysosomal inhibitor Bafilomycin, indicative of an enhanced “on rate” (Figures S3A-S3D). Cellular mitophagy is a selective form of autophagy with a decreased number of mitochondria leading to a metabolic shift towards aerobic glycolysis (Martinez-Outschoorn et al., 2017). After csl silencing, both the mitochondrial Tom20 immunofluorescence signal and mitochondrial labelling with MitoTracker were decreased (Figure 4B). Consistent with enhanced mitophagy and glycolytic transition, HDFs with silenced csl had enhanced expression of pyruvate kinase isozyme 2 (PKM2) and lactate dehydrogenase A (LDHA), two key enzymes responsible for lactate production (Figure 4C), and higher lactate levels (Figure 4D). A similar induction of autophagy, decreased mitochondria network and increased glycolytic switch were observed in HDFs with increased ulk3 expression (Figures 4E–4I).

Figure 4
Csl silencing and increased Ulk3 induce cellular autophagy and mitophagy separately from Gli and CAF activation

An important question is whether the above processes are linked to induction of the CAF effector gene program or proceed as parallel independent events. Indicative of the second possibility, we found that the silencing of csl induced CAF-effector genes to a similar extent in HDFs with or without silencing of the autophagy essential genes atg5, atg7 or becn1 (Figure 4J) and in MEFs with atg5 gene knockout compared to wild type MEFs (Figure S3E). Similarly, increased ulk3 expression by lentiviral infection induced CAF effector genes in HDFs with silenced atg5 (Figure 4K). Thus, both csl silencing and ulk3 up-regulation enhance autophagy, which is however dispensable for CAF effectors genes induction.

Ulk3 as a target to suppress CAF activation and cancer/stromal cell expansion

An important question was whether ulk3 expression is increased also in clinically occurring CAFs. Analysis of published gene expression profiles of CAFs derived from head and neck SCCs (HNSCC) showed that expression of ulk3 and gli is consistently higher than in normal fibroblasts derived from the same body site, with similar up-regulation in CAFs from prostate (Ashida et al., 2012) and breast (Finak et al., 2008) cancers (Figures 5A and S4A).

Figure 5
Ulk3 levels are elevated in CAFs

In skin, laser capture micro-dissection (LCM) followed by RT-qPCR analysis showed consistently increased ulk3 expression in several SCC-adjacent stromal cells (in areas we tested as devoid of macrophage and leukocyte contamination (Procopio et al., 2015)), relative to stroma areas of distant normal skin (Figure 5B). Results were confirmed by immunofluorescence analysis, showing significantly higher number of Ulk3-expressing fibroblasts (Vimentin positive cells) in SCC surrounding stroma versus unaffected areas of same or different individuals (Figures 5C and S4B). Similarly up-regulated ulk3 and gli levels - with concomitantly decreased csl - were also found in CAFs isolated from skin SCCs relative to normal dermal fibroblasts from same or different individuals (Figures 5D–5F and S4C). Results were further validated in a mouse model of AK based on mesenchymal deletion of the csl gene (Hu et al., 2012). LCM and RT-qPCR analysis showed significant ulk3 up regulation in stromal fibroblasts of skin lesions that developed in these mice at 3 months of age, relative to distant unaffected skin (Figure 5G).

To assess whether the observed increase in ulk3 expression and Gli activation in CAFs is functionally linked to decreased csl, we infected these cells with a csl-inducible lentivirus, we previously showed to suppress CAF marker expression (Procopio et al., 2015). As shown in Figure 5H increased csl expression caused a significant down-regulation of ulk3 and gli levels.

A much-needed development for cancer therapy is the identification of strategies affecting cancer stroma (Goruppi and Dotto, 2013). We found that siRNA-mediated silencing of ulk3 in two different CAFs strains down-modulated key CAF effector genes and decreased gli1, gli2 expression (Figure S5A), suggesting that ulk3 could be an attractive target to counteract CAF activation. We tested this possibility by the same cancer/stromal cell expansion assays employed above, by parallel injections into mouse ears of SCC13 cells admixed with two different strains of patient-derived CAFs plus/minus shRNA-mediated ulk3 silencing. As shown in Figures 6A and S5B, ear lesions formed in the presence of CAFs with ulk3 knockdown were consistently smaller than with corresponding CAFs controls. This was paralleled by a lower number of cancer cells positive for the p63 differentiation marker and reduced ki67 proliferative index (Figure 6B). Importantly, key CAF markers, like α-Smooth Muscle Actin (SMA), Periostin (POSTN) and Tenascin C (TNC) were expressed to a much lower extent in lesions with ulk3-silenced CAFs than corresponding controls, with increased macrophage infiltration and lesser angiogenesis, as assessed by antibodies against CD68 and CD31 markers, respectively (Figures 6C and S5C). These data indicate that ulk3, whose expression is elevated in clinically derived CAFs, is required for the tumor growth-enhancing activity of these cells.

Figure 6
Loss of Ulk3 suppresses CAFs tumor-enhancing properties


Genes with critical cell and tissue regulatory functions are tightly controlled by the convergence of multiple positive and negative signals. While a large number of CAF effector genes have been identified, an unanswered question is the integration of different signaling pathways involved in their control. In stromal fibroblasts, CSL represses senescence- and CAF-effector genes, with induction of the first class of genes occurring as the result of loss of CSL repression and increased p53 activity (Procopio et al., 2015). We show here that relief of CSL repression of CAF effector genes is not sufficient for induction of their expression, for which activation of Gli1/2 transcription factors is also required. Increased expression of the Ulk3 kinase, which occurs in CAFs from various cancer types, provides a link between compromised CSL function and Gli activation, which can occur in concert with other previously reported Hh-receptor dependent and independent mechanisms (Metcalfe and de Sauvage, 2011; Seto et al., 2009); (Wang et al., 2012) (Nolan-Stevaux et al., 2009; Stecca et al., 2007). Autophagy can create a pro-tumorigenic microenvironment rich of metabolic precursors directed from CAFs to tumors (Lisanti et al., 2010; Zhao et al., 2013). As Ulk3 is also an inducer of this process, it provides an attractive target for stroma-focused anti-cancer intervention, to suppress combined aspects of CAF activation.

CSL is endowed with an intrinsic transcription repressive function and CAF effector genes are induced as a consequence of CSL down-modulation (Hu et al., 2012; Procopio et al., 2015), as it can occur upon exposure to pro-carcinogenic stimuli such as UVA or smoke-derived compounds (Menietti et al., 2016). We show that loss of CSL transcription repressive function is by itself not sufficient for induction of these genes and for the resulting tumor growth enhancing activity of fibroblasts, for which Gli activation is needed. Aberrant activation of Hh/Gli signaling has been implicated in up to one-third of all human cancers and is considered to be both a consequence and driver of tumor-stromal interactions(Gupta et al., 2010; Ruiz i Altaba et al., 2002). Hh ligands have been implicated as paracrine mediators of cancer development(Rubin and de Sauvage, 2006). Earlier studies indicated that production of these molecules by colon, pancreatic and ovarian cancers cells can promote tumor growth indirectly by activating Hh signaling in surrounding stroma(Theunissen and de Sauvage, 2009). However, recent evidence points to a restraining role of Hh stimulation in pancreatic and bladder cancer stroma (Ozdemir et al., 2014; Rhim et al., 2014; Shin et al., 2014). In our own analysis of multiple HDFs strains plus/minus csl silencing and skin-derived CAFs versus matched and unmatched controls, expression of Hh ligands was variously modulated. Similar variations were found in published profiles of CAFs from other cancer types. An attractive possibility is that, irrespectively of Hh levels, Gli activation in stromal fibroblasts can occur through increased activity of the Ulk3 kinase, encoded by a direct CSL target gene up-regulated by csl down modulation.

Ulk3 was recently proposed to fulfill the same central role in mammalian cells as the Drosophila Fused kinase in Gli2 activation(Maloverjan et al., 2010). Besides confirming these results, we found a detectable association of endogenous Ulk3-Gli2 in HDFs and showed that increased ulk3 expression was both required and sufficient for Gli and CAF activation. Separately from this, we uncovered an interplay between CSL and Ulk3 in control of autophagy that may apply, with context-dependent adjustments, to other cellular systems in which Notch signaling has been pharmacologically implicated (Barth and Kohler, 2014; Chen et al., 2016; Song et al., 2015; Yao et al., 2015). Importantly, while autophagy/mitophagy have been previously implicated in metabolic reprogramming of CAFs (Kim et al., 2011; Russell et al., 2013), our genetic evidence indicates that they are not significantly involved in the gene expression program leading to CAF activation.

A much-needed development for cancer therapy is the identification of strategies affecting cancer stroma (Goruppi and Dotto, 2013). As in HDFs with csl silencing and CAFs from skin SCCs, analysis of publically available gene expression profiles indicated that ulk3 expression is elevated in CAFs derived from several different cancer types, including HNSCC, prostate and breast cancer. Functionally, genetic suppression of ulk3 in patient-derived CAFs was sufficient to inhibit CAF-effector gene expression as well as Gli2 activation, and suppressed the growth-enhancing and pro-tumorigenic properties of these cells on neighboring cancer cells. While similar effects were observed with a compound reported to inhibit Ulk3 kinase activity, further translational studies await the identification of compounds with greater specificity. Thus Ulk3 kinase by linking two key signaling pathways involved in CAF conversion represents an interesting target for stroma-focused anti-cancer interventions.


Cell Manipulations

HDFs were prepared from discarded skin samples of abdominoplasty patients at Massachusetts General Hospital (Boston, Massachusetts, USA) with institutional approval (2000P002418) or previously obtained (Procopio et al., 2015). Conditions for culturing cells, viral infection, siRNA-mediated gene silencing, RT–qPCR and ChIP were previously reported (Brooks et al., 2014; Hu et al., 2012). For derivation of CAFs and NFs, surgically excised discarded skin SCC samples and non-affected skin samples, were dissociated with Liberase TL (Roche). For co-culture and in vivo approaches, skin SCC13 cells (Restivo et al., 2011) were infected with an EGFP-expressing lentivirus (Tiscornia et al., 2003). Each strain derived from a patient was identified with a number, which is indicated in the different panels. Spontaneously immortalized MEFs plus/minus combined deletion of the gli1 and gli 2 genes (Lipinski et al., 2008) were infected with ulk3 silencing or empty vector control lentiviruses and selected with puromycin. After 2 weeks cells were analyzed by immunofluorescence and for gene expression, then used for the in vivo studies.

Immunofluorescence, autophagy studies and cell assays

Immunoblots and immunofluorescences were performed as in (Kong et al., 2010; Procopio et al., 2015). For immunofluorescence cells were seeded on coverslips, fixed in 3% PFA and processed as in (Kong et al., 2010). Immunohistochemistry of tumor and tissue sections was performed as in (Brooks et al., 2014; Hu et al., 2012; Procopio et al., 2015) and quantification of ki67 and p63 protein levels was made using the watershed algorithm ( and ImageJ, NIH. Quantification of all other tissue immunofluorescence staining was performed using ImageJ.

Autophagosome evaluation studies were performed as in (Kong et al., 2010). Two different strains of HDFs were infected with pHcRed1-LC3 encoding for the far-red fluorescent protein (Clontech) fused to LC3.

Colorimetric determination of lactate was performed using an EnzyChrom L-lactate assay kit (BioAssay Systems). The cell culture supernatants of HDFs with csl silencing or overexpressing Ulk3-V5 were collected after three days of culture. Each sample reaction was carried out in triplicate and a standard curve of lactate was used to determine the lactate concentration.

For mitochondrial labeling HDFs plus/minus csl silencing for three days were pulse labeled with 100nM MitoTracker Red FM (Thermofisher), fixed with 3% PFA in PBS and analyzed using confocal microscopy (Zeiss Observer Z1) using a Zen Pro 2.3 software. DAPI stained the nuclei.

The siRNA and shRNA sequences used are in TABLES1 and TABLES2. The oligonucleotides used in RT-qPCR are provided in TABLES3. The oligonucleotides used in ChIP experiments are provided in TABLES4. A detailed list of all the antibodies and the condition(s) used is in TABLE S5. Unprocessed original scans of immunoblots are shown in Supplementary Figures S6–S8.

Human samples and LCM experiments

Normal human skin samples and samples of squamous cell carcinoma (SCC) were obtained at the Department of Dermatology, Massachusetts General Hospital, as discarded parts not needed for diagnosis. All samples were processed as approved by the Institutional Review Board. Matched normal skin and SCC paraffin samples used for LCM and RT-PCR were provided by the Department of Dermatology, University of Tubingen with institutional review board approvals and informed consent, previously repoted in (Procopio et al., 2015). LCM was made using an Arcturus XT micro-dissection system (Applied Biosystems) as in (Hu et al., 2012; Procopio et al., 2015). Gene expression was normalized to 36β4, and the absence of contaminating leucocytes in the selected areas was previously confirmed with CD45 oligos in RT-qPCR assays(Procopio et al., 2015).

Animal studies

Experiment with gli1 and 2 −/− MEFs: Mouse ear injections with the MEFs were carried out in 8 to 10-week-old female SKID (CB17sc-m) mice (Taconic), as in (Procopio et al., 2015). EGFP expressing SCC13 cells (1 × 105) were admixed with equal numbers of MEFs (either gli +/+ or gli 1/2 −/−) with shRNA-mediated silencing of csl, or a control. Cells were injected 5μl per site using a 33-gauge micro syringe (Hamilton). Starting the day after injection, the mice ears were imaged using a fluorescent stereomicroscope (Leica MZ-FLIII), every three days for 21 days. The mice were sacrificed after 24 days and images of the ears taken using bright field and fluorescence stereomicroscopy. Experiment with patient-derived CAFs: Mouse ear injections of cells were carried out in SKID mice admixing EGFP expressing SCC13 cells (1 × 105) with equal numbers of CAFs (strain#1 or strain#3) infected with an ulk3 silencing lentivirus or empty vector control. Mice were sacrificed after 3 weeks. All animal studies were approved by MGH institutional animal care and use committee (2004N000170).

Statistical analysis

Data are presented as mean± SEM, mean ± SD, or ratios among treated and controls, with two to three separate HDFs strains in independent experiments as indicated in the Figure legends. For gene expression and functional testing assays, statistical significance of differences between experimental groups and controls was assessed by two-tailed unpaired or paired t-test, and one sample t-test. P values < 0.05 were considered as statistically significant.

Supplementary Material



We are grateful to Robert Lipinski for the MEFs gli1/2 −/− and gli +/+ and to Noburo Mitzushima for the MEFs atg5 −/− and atg5 +/+. We thank Liuqing Yang for the P-Gli2 antibodies, Alessia Di Nardo for the shatg5 vector, Shanell Moijta for technical support with histology and Wolfram Hoetzenecker, for providing SCC samples used in LCM. A. C. is supported by a Marie Curie fellowship from the Italian Association for Cancer Research and the European Union FP7 Marie Curie Program. This work was supported by grants from the National Institutes of Health (R01AR039190; R01AR064786; the content not necessarily representing the official views of NIH), the Swiss National Science Foundation (310030_156191/1) and European Research Council (26075083) to GPD. The Authors declare that there are no competing financial interests.



Supplemental information includes five figures, extended experimental procedures and four tables.


S.G., M.P., S.J. and A.C. performed the experiments and contributed to analysis of the results. V.N. provided clinical samples. S.G. and G.P.D. designed the study and wrote the manuscript.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.


  • Aberger F, Ruiz IAA. Context-dependent signal integration by the GLI code: the oncogenic load, pathways, modifiers and implications for cancer therapy. Semin Cell Dev Biol. 2014;33:93–104. [PMC free article] [PubMed]
  • Allenspach EJ, Maillard I, Aster JC, Pear WS. Notch signaling in cancer. Cancer Biol Ther. 2002;1:466–476. [PubMed]
  • Artavanis-Tsakonas S, Rand MD, Lake RJ. Notch signaling: cell fate control and signal integration in development. Science. 1999;284:770–776. [PubMed]
  • Ashida S, Orloff MS, Bebek G, Zhang L, Zheng P, Peehl DM, Eng C. Integrated analysis reveals critical genomic regions in prostate tumor microenvironment associated with clinicopathologic phenotypes. Clin Cancer Res. 2012;18:1578–1587. [PubMed]
  • Barth JM, Kohler K. How to take autophagy and endocytosis up a notch. BioMed research international. 2014;2014:960803. [PMC free article] [PubMed]
  • Bingol B, Tea JS, Phu L, Reichelt M, Bakalarski CE, Song Q, Foreman O, Kirkpatrick DS, Sheng M. The mitochondrial deubiquitinase USP30 opposes parkin-mediated mitophagy. Nature. 2014;510:370–375. [PubMed]
  • Bissell MJ, Hines WC. Why don’t we get more cancer? A proposed role of the microenvironment in restraining cancer progression. Nat Med. 2011;17:320–329. [PMC free article] [PubMed]
  • Brooks YS, Ostano P, Jo SH, Dai J, Getsios S, Dziunycz P, Hofbauer GF, Cerveny K, Chiorino G, Lefort K, et al. Multifactorial ERbeta and NOTCH1 control of squamous differentiation and cancer. J Clin Invest. 2014;124:2260–2276. [PMC free article] [PubMed]
  • Chen X, Zhang Y, Shi Y, Lian H, Tu H, Han S, Yin J, Peng B, Zhou B, He X, et al. MiR-129 triggers autophagic flux by regulating a novel Notch-1/E2F7/Beclin-1 axis to impair the viability of human malignant glioma cells. Oncotarget. 2016;7:9222–9235. [PMC free article] [PubMed]
  • Costea DE, Hills A, Osman AH, Thurlow J, Kalna G, Huang X, Pena Murillo C, Parajuli H, Suliman S, Kulasekara KK, et al. Identification of two distinct carcinoma-associated fibroblast subtypes with differential tumor-promoting abilities in oral squamous cell carcinoma. Cancer Res. 2013;73:3888–3901. [PubMed]
  • Dotto GP. Notch tumor suppressor function. Oncogene. 2008;27:5115–5123. [PMC free article] [PubMed]
  • Dotto GP. Crosstalk of Notch with p53 and p63 in cancer growth control. Nat Rev Cancer. 2009;9:587–595. [PubMed]
  • Dotto GP. Multifocal epithelial tumors and field cancerization: stroma as a primary determinant. J Clin Invest. 2014;124:1446–1453. [PMC free article] [PubMed]
  • Finak G, Bertos N, Pepin F, Sadekova S, Souleimanova M, Zhao H, Chen H, Omeroglu G, Meterissian S, Omeroglu A, et al. Stromal gene expression predicts clinical outcome in breast cancer. Nat Med. 2008;14:518–527. [PubMed]
  • Goruppi S, Dotto GP. Mesenchymal stroma: primary determinant and therapeutic target for epithelial cancer. Trends Cell Biol 2013 [PMC free article] [PubMed]
  • Gupta S, Takebe N, LoRusso P. Targeting the Hedgehog pathway in cancer. Therapeutic Advances in Medical Oncology. 2010;2:237–250. [PMC free article] [PubMed]
  • Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell. 2011;144:646–674. [PubMed]
  • Hu B, Castillo E, Harewood L, Ostano P, Reymond A, Dummer R, Raffoul W, Hoetzenecker W, Hofbauer GF, Dotto GP. Multifocal epithelial tumors and field cancerization from loss of mesenchymal CSL signaling. Cell. 2012;149:1207–1220. [PMC free article] [PubMed]
  • Junttila MR, de Sauvage FJ. Influence of tumour micro-environment heterogeneity on therapeutic response. Nature. 2013;501:346–354. [PubMed]
  • Kalluri R. The biology and function of fibroblasts in cancer. Nat Rev Cancer. 2016;16:582–598. [PubMed]
  • Kim J, Kundu M, Viollet B, Guan KL. AMPK and mTOR regulate autophagy through direct phosphorylation of Ulk1. Nat Cell Biol. 2011;13:132–141. [PMC free article] [PubMed]
  • Klionsky DJ, Abdelmohsen K, Abe A, Abedin MJ, Abeliovich H, Acevedo Arozena A, Adachi H, Adams CM, Adams PD, Adeli K, et al. Guidelines for the use and interpretation of assays for monitoring autophagy (3rd edition) Autophagy. 2016;12:1–222. [PMC free article] [PubMed]
  • Kong DK, Georgescu SP, Cano C, Aronovitz MJ, Iovanna JL, Patten RD, Kyriakis JM, Goruppi S. Deficiency of the transcriptional regulator p8 results in increased autophagy and apoptosis, and causes impaired heart function. Mol Biol Cell. 2010;21:1335–1349. [PMC free article] [PubMed]
  • Lipinski RJ, Bijlsma MF, Gipp JJ, Podhaizer DJ, Bushman W. Establishment and characterization of immortalized Gli-null mouse embryonic fibroblast cell lines. BMC Cell Biology. 2008;9:49–49. [PMC free article] [PubMed]
  • Lisanti MP, Martinez-Outschoorn UE, Chiavarina B, Pavlides S, Whitaker-Menezes D, Tsirigos A, Witkiewicz A, Lin Z, Balliet R, Howell A, et al. Understanding the “lethal” drivers of tumor-stroma co-evolution: emerging role(s) for hypoxia, oxidative stress and autophagy/mitophagy in the tumor micro-environment. Cancer Biol Ther. 2010;10:537–542. [PMC free article] [PubMed]
  • Maloverjan A, Piirsoo M, Michelson P, Kogerman P, Osterlund T. Identification of a novel serine/threonine kinase ULK3 as a positive regulator of Hedgehog pathway. Exp Cell Res. 2010;316:627–637. [PubMed]
  • Martincorena I, Roshan A, Gerstung M, Ellis P, Van Loo P, McLaren S, Wedge DC, Fullam A, Alexandrov LB, Tubio JM, et al. Tumor evolution. High burden and pervasive positive selection of somatic mutations in normal human skin. Science. 2015;348:880–886. [PMC free article] [PubMed]
  • Martinez-Outschoorn UE, Peiris-Pages M, Pestell RG, Sotgia F, Lisanti MP. Cancer metabolism: a therapeutic perspective. Nat Rev Clin Oncol. 2017;14:11–31. [PubMed]
  • Menietti E, Xu X, Ostano P, Joseph JM, Lefort K, Dotto GP. Negative control of CSL gene transcription by stress/DNA damage response and p53. Cell Cycle. 2016;15:1767–1778. [PMC free article] [PubMed]
  • Metcalfe C, de Sauvage FJ. Hedgehog fights back: mechanisms of acquired resistance against Smoothened antagonists. Cancer Res. 2011;71:5057–5061. [PubMed]
  • Nolan-Stevaux O, Lau J, Truitt ML, Chu GC, Hebrok M, Fernandez-Zapico ME, Hanahan D. GLI1 is regulated through Smoothened-independent mechanisms in neoplastic pancreatic ducts and mediates PDAC cell survival and transformation. Genes Dev. 2009;23:24–36. [PubMed]
  • Ozdemir BC, Pentcheva-Hoang T, Carstens JL, Zheng X, Wu CC, Simpson TR, Laklai H, Sugimoto H, Kahlert C, Novitskiy SV, et al. Depletion of carcinoma-associated fibroblasts and fibrosis induces immunosuppression and accelerates pancreas cancer with reduced survival. Cancer Cell. 2014;25:719–734. [PMC free article] [PubMed]
  • Procopio MG, Laszlo C, Al Labban D, Kim DE, Bordignon P, Jo SH, Goruppi S, Menietti E, Ostano P, Ala U, et al. Combined CSL and p53 downregulation promotes cancer-associated fibroblast activation. Nat Cell Biol. 2015;17:1193–1204. [PMC free article] [PubMed]
  • Restivo G, Nguyen BC, Dziunycz P, Ristorcelli E, Ryan RJ, Ozuysal OY, Di Piazza M, Radtke F, Dixon MJ, Hofbauer GF, et al. IRF6 is a mediator of Notch pro-differentiation and tumour suppressive function in keratinocytes. EMBO J. 2011;30:4571–4585. [PubMed]
  • Rhim AD, Oberstein PE, Thomas DH, Mirek ET, Palermo CF, Sastra SA, Dekleva EN, Saunders T, Becerra CP, Tattersall IW, et al. Stromal elements act to restrain, rather than support, pancreatic ductal adenocarcinoma. Cancer Cell. 2014;25:735–747. [PMC free article] [PubMed]
  • Rubin LL, de Sauvage FJ. Targeting the Hedgehog pathway in cancer. Nat Rev Drug Discov. 2006;5:1026–1033. [PubMed]
  • Ruel L, Rodriguez R, Gallet A, Lavenant-Staccini L, Therond PP. Stability and association of Smoothened, Costal2 and Fused with Cubitus interruptus are regulated by Hedgehog. Nat Cell Biol. 2003;5:907–913. [PubMed]
  • Ruiz i Altaba A, Sanchez P, Dahmane N. Gli and hedgehog in cancer: tumours, embryos and stem cells. Nat Rev Cancer. 2002;2:361–372. [PubMed]
  • Russell RC, Tian Y, Yuan H, Park HW, Chang YY, Kim J, Kim H, Neufeld TP, Dillin A, Guan KL. ULK1 induces autophagy by phosphorylating Beclin-1 and activating VPS34 lipid kinase. Nat Cell Biol. 2013;15:741–750. [PMC free article] [PubMed]
  • Seto M, Ohta M, Asaoka Y, Ikenoue T, Tada M, Miyabayashi K, Mohri D, Tanaka Y, Ijichi H, Tateishi K, et al. Regulation of the hedgehog signaling by the mitogen-activated protein kinase cascade in gastric cancer. Mol Carcinog. 2009;48:703–712. [PubMed]
  • Shin K, Lim A, Zhao C, Sahoo D, Pan Y, Spiekerkoetter E, Liao JC, Beachy PA. Hedgehog signaling restrains bladder cancer progression by eliciting stromal production of urothelial differentiation factors. Cancer Cell. 2014;26:521–533. [PMC free article] [PubMed]
  • Song BQ, Chi Y, Li X, Du WJ, Han ZB, Tian JJ, Li JJ, Chen F, Wu HH, Han LX, et al. Inhibition of Notch Signaling Promotes the Adipogenic Differentiation of Mesenchymal Stem Cells Through Autophagy Activation and PTEN-PI3K/AKT/mTOR Pathway. Cellular physiology and biochemistry: international journal of experimental cellular physiology, biochemistry, and pharmacology. 2015;36:1991–2002. [PubMed]
  • Stecca B, Mas C, Clement V, Zbinden M, Correa R, Piguet V, Beermann F, Ruiz IAA. Melanomas require HEDGEHOG-GLI signaling regulated by interactions between GLI1 and the RAS-MEK/AKT pathways. Proc Natl Acad Sci U S A. 2007;104:5895–5900. [PubMed]
  • Theunissen JW, de Sauvage FJ. Paracrine Hedgehog signaling in cancer. Cancer Res. 2009;69:6007–6010. [PubMed]
  • Tiscornia G, Singer O, Ikawa M, Verma IM. A general method for gene knockdown in mice by using lentiviral vectors expressing small interfering RNA. Proc Natl Acad Sci U S A. 2003;100:1844–1848. [PubMed]
  • Wang G, Amanai K, Wang B, Jiang J. Interactions with Costal2 and suppressor of fused regulate nuclear translocation and activity of cubitus interruptus. Genes Dev. 2000;14:2893–2905. [PubMed]
  • Wang Y, Ding Q, Yen CJ, Xia W, Izzo JG, Lang JY, Li CW, Hsu JL, Miller SA, Wang X, et al. The crosstalk of mTOR/S6K1 and Hedgehog pathways. Cancer Cell. 2012;21:374–387. [PMC free article] [PubMed]
  • Xing Z, Lin A, Li C, Liang K, Wang S, Liu Y, Park PK, Qin L, Wei Y, Hawke DH, et al. lncRNA directs cooperative epigenetic regulation downstream of chemokine signals. Cell. 2014;159:1110–1125. [PMC free article] [PubMed]
  • Yao J, Zheng K, Li C, Liu H, Shan X. Interference of Notch1 inhibits the growth of glioma cancer cells by inducing cell autophagy and down-regulation of Notch1-Hes-1 signaling pathway. Medical oncology (Northwood, London, England) 2015;32:610. [PubMed]
  • Young AR, Narita M, Ferreira M, Kirschner K, Sadaie M, Darot JF, Tavare S, Arakawa S, Shimizu S, Watt FM, et al. Autophagy mediates the mitotic senescence transition. Genes Dev. 2009;23:798–803. [PubMed]
  • Zhao X, He Y, Chen H. Autophagic tumor stroma: mechanisms and roles in tumor growth and progression. Int J Cancer. 2013;132:1–8. [PubMed]