Many solid cancers display cellular hierarchies with self-renewing, tumorigenic stemlike cells, or cancer-initiating cells (CICs) at the apex. Whereas CICs often exhibit relative resistance to conventional cancer therapies, they also receive critical maintenance cues from supportive stromal elements that also respond to cytotoxic therapies. To interrogate the interplay between chemotherapy and CICs, we investigated cellular heterogeneity in human colorectal cancers. Colorectal CICs were resistant to conventional chemotherapy in cell-autonomous assays, but CIC chemoresistance was also increased by cancer-associated fibroblasts (CAFs). Comparative analysis of matched colorectal cancer specimens from patients before and after cytotoxic treatment revealed a significant increase in CAFs. Chemotherapy-treated human CAFs promoted CIC self-renewal and in vivo tumor growth associated with increased secretion of specific cytokines and chemokines, including interleukin-17A (IL-17A). Exogenous IL-17A increased CIC self-renewal and invasion, and targeting IL-17A signaling impaired CIC growth. Notably, IL-17A was overexpressed by colorectal CAFs in response to chemotherapy with expression validated directly in patient-derived specimens without culture. These data suggest that chemotherapy induces remodeling of the tumor microenvironment to support the tumor cellular hierarchy through secreted factors. Incorporating simultaneous disruption of CIC mechanisms and interplay with the tumor microenvironment could optimize therapeutic targeting of cancer.
Colorectal cancer is the third leading cause of cancer-related death in the United States, with ∼141,210 new cases and 49,380 deaths in 2011 (American Cancer Society, 2011). Despite clinical advances, 50% of stage III and 95% of stage IV colorectal cancer patients will die from their disease (American Cancer Society, 2011). Improving survival for patients afflicted with colorectal cancer will require more effective and durable responses to adjuvant chemotherapy. Advances in the genetics of colorectal cancers have defined molecular targets altered during the development and progression of colorectal cancers, but have translated into targeted therapeutics with only modest efficacy. Tumor suppressor pathways account for most common genetic lesions, but these have proven difficult to target pharmacologically. Molecularly targeted therapies, like the anti–epidermal growth factor receptor (EGFR) agents cetuximab and panitumumab augment the activity of conventional chemotherapy but are not curative (Arnold and Seufferlein, 2010). Resistance to chemotherapy may be associated with the outgrowth of clones harboring advantageous genetic lesions, but cellular diversity derived from nongenetic sources also contributes to recurrent tumor growth (Weaver et al., 2002; Matsunaga et al., 2003; Bissell and Labarge, 2005). Cancers exist as complex systems composed of multiple cell types that collectively support and maintain tumor growth. Nontransformed elements may display relatively few genomic lesions and be more likely to display sustained responses to therapy, suggesting advantages to their use as therapeutic targets (Shaked et al., 2006, 2008; Yamauchi et al., 2008; Gilbert and Hemann., 2010; Hao et al., 2011; Shree et al., 2011; Straussman et al., 2012; Gilbert and Hemann., 2011; Acharyya et al., 2012; Nakasone et al., 2012; Hölzel et al., 2013; Bruchard et al., 2013). Indeed, the microenvironment has become a major focus in modeling the growth of cancer and therapeutic response. Inhibition of tumor vasculature through blockade of endothelial proliferation signals has clinical benefit, leading to the development of bevacizumab, a humanized anti–vascular endothelial growth factor (VEGF) antibody (Winder and Lenz, 2010). Another important compartment of tumor stroma is cancer-associated fibroblasts (CAFs). CAFs originate from heterogeneous cell types, including bone marrow–derived progenitor cells, smooth muscle cells, preadipocytes, fibroblasts, and myofibroblasts (Orimo and Weinberg, 2007; Worthley et al., 2010; Gonda et al., 2010). CAFs support tumorigenesis by stimulating angiogenesis, cancer cell proliferation, and invasion (Gonda et al., 2010; Worthley et al., 2010). They are also an important player in therapeutic resistance (Crawford et al., 2009; Porter et al., 2012), and fibroblasts can serve as a source for cytokines released in the cancer-initiating cell (CIC) microenvironment (Vermeulen et al., 2010). Furthermore, irradiated CAFs have been previously reported to promote tumor growth in breast (Barcellos-Hoff and Ravani, 2000) and lung cancers (Hellevik et al., 2013). It is thus logical that disruption of CAFs in the tumor microenvironment would influence clinical tumor behavior.
Cancers are maintained over the long term by a subpopulation of cancer cells, the CICs (Barker et al., 2009; Ricci-Vitiani et al., 2009; Blanpain, 2013). Like tissue-specific stem cells, the identification and characterization of CICs is evolving: the current definition is based on functional assays focused on recapitulation of the parental tumor upon xenotransplantation. The features of self-renewal, differentiation, and sustained proliferation are inherent within the regeneration of the tumor organ system (Magee et al., 2012). Interpatient variation in the genetics and epigenetics of colorectal cancers is so divergent that no identical mutational signatures have been reported for patients (Sanchez et al., 2009; Ogino et al., 2012; Sadanandam et al., 2013). It is therefore not surprising that markers to distinguish CICs from more differentiated progeny have not been absolutely informative across all tumors. Further, most CIC enrichment markers mediate interactions between a cell and its microenvironment, suggesting that the information associated with that marker may be lost after removal from the tumor microenvironment. Whereas CD133 (Prominin-1) had been reported by some groups to selectively identify colorectal CICs (O’Brien et al., 2007; Ricci-Vitiani et al., 2007; Elsaba et al., 2010; Fang et al., 2010), Shmelkov et al. (2008) reported that CD133 failed to inform identification of the CICs. Other groups have reported that CD44 (Dalerba et al., 2007; Du et al., 2008; Yeung et al., 2010; Ohata et al., 2012), CD166 (Dalerba et al., 2007; Vermeulen et al., 2008), CD66c (Gemei et al., 2013), Lgr5 (Barker et al., 2007; Vermeulen et al., 2008; Takahashi et al., 2011), or aldehyde dehydrogenase (ALDH; Huang et al., 2009; Deng et al., 2010) inform CIC characteristics. Regardless of the marker used, CICs are enriched for tumorigenic potential, indicating that these subgroups of tumor cells drive colorectal cancer maintenance and must be targeted to inhibit tumor growth.
CICs do not exist in isolation, but rather reside in an interactive niche with multiple cell types, including fibroblasts (Vermeulen et al., 2010; Medema and Vermeulen, 2011), endothelial cells (Lu et al., 2013), and immune cells (Hölzel et al., 2013). Each component contributes to the overall function and maintenance of the tumor and has potential roles in CIC resistance and recurrence. Mechanisms driving CIC maintenance and resistance are still being defined, but cell–cell interactions mediated through numerous molecular mechanisms, including cytokines and chemokines, are critical (Todaro et al., 2007; Vermeulen et al., 2010; Li et al., 2012). Cytokines and chemokines have the capacity to function as both paracrine and autocrine factors, supporting these secreted molecules as ideal mediators of interactions between the cellular hierarchy and other tumor cellular components. Indeed, we have described IL-6 as a key cue derived from more differentiated tumor cells to maintain glioblastoma CICs, which express IL-6 receptors (Wang et al., 2009). Mesenchymal stem cells and tumor-associated macrophages secrete IL-6 and CXCL7 in breast cancer to stimulate CIC growth and dispersal (Liu et al., 2011). These interactions are reciprocal, as CICs create supportive niches for stroma through the recruitment of mesenchymal stem cells via IL-1 secretion. In return, mesenchymal stem cells secrete IL-6 and IL-8 to promote CIC maintenance (Li et al., 2012).
Here, we first confirm that chemotherapy preferentially targets non-CICs due to cell autonomous resistance of CICs, but furthermore uncover a novel negative impact of chemotherapy in the stimulation of CAFs to create a chemoresistant niche by releasing cytokines, including IL-17A, as a CIC maintenance factor. These results have important clinical implications as most chemosensitizing approaches focus on disrupting cell autonomous molecular mechanisms without consideration of the interplay with the microenvironment that may display differential molecular dependence and temporal course, suggesting more complex therapeutic paradigms may be required to improve patient outcomes.
Interrogating the cellular hierarchy in human colorectal cancers
Heterogeneity in tumors is derived from a combination of genetic and epigenetic variance. A subset of tumor cells within colorectal cancers may display stem-like features. These CICs are defined functionally through self-renewal and phenocopy the complex parental tumor on secondary growth. The use of cell surface markers to enrich or deplete CICs is essential to segregate tumorigenic and nontumorigenic cells to demonstrate a cellular hierarchy. As several markers have been reported to selectively identify colorectal CICs, we first interrogated the ability to enrich for CICs in colorectal cancer patient specimens using several of the previously reported candidates (CD166, CD133, ALDH activity, and CD44; Dalerba et al., 2007; O’Brien et al., 2007; Ricci-Vitiani et al., 2007; Du et al., 2008; Vermeulen et al., 2008; Shmelkov et al., 2008; Huang et al., 2009; Fang et al., 2010; Elsaba et al., 2010; Yeung et al., 2010; Deng et al., 2010; Ohata et al., 2012). To avoid selection effects of cell culture, we interrogated markers in acutely dissociated patient specimens derived from either primary tumors or metastases (Fig. 1 A). Some markers, such as CD166, were not consistently expressed across tumors (not depicted) in concordance with other studies, suggesting specificity for colorectal cancer subtypes (Horst et al., 2009; Park et al., 2012). CD133 and ALDH activity measured by the Aldefluor assay were detectable but did not specifically segregate the capacity to generate tumor spheres (a surrogate marker of self-renewal associated with increased tumor initiation; unpublished data) or tumor initiation. Notably, CD44 consistently enriched for self-renewing, tumorigenic neoplastic cells (i.e., CICs). CD44 identified a minority of tumor cells (Fig. 1 A) that overlapped with cells that expressed the epithelial marker, EpCAM (Fig. 1 B), validating an epithelial origin as well as Aldefluor activity (Fig. 1 C) and CD133 antigen expression (Fig. 1 D). CD44high/+ cells also had elevated levels of putative CIC genes (Blache et al., 2004; Jay et al., 2005; Takahashi and Yamanaka, 2006; Huang et al., 2009; Deng et al., 2010; Guo et al., 2012), including c-Myc, ALDH1, Sox9, Oct4, Musashi1 (Msi1), and BMI1, when compared with CD44low/− cells (Fig. 1 E). These data indicate that CD44high/+ colorectal cancer cells coexpress other CIC markers. The coexpression of CD44, EpCAM, Aldefluor, and CD133 was validated in four independent samples (Fig. 1 F). CD44 is a receptor for components of the extracellular matrix. Like most receptors, its expression is down-regulated in response to ligand engagement. To assess the difference of in vivo versus in vitro environmental effects on the fraction of CD44high/+ cells, single cell suspensions from freshly dissociated patients’ colorectal tumors and their cultured counterparts were analyzed by FACS for the expression of CD44. As expected, the fractions differed based on growth conditions, but a general concordance in the relative degree of CD44high/+ fractions within each condition was observed (Fig. 1 G). The clinical characteristics of these samples are summarized in Table S1.
To further evaluate the ability of CD44 to enrich for CICs, we next examined key functional assays (tumorsphere formation and tumor propagation). Using several human colorectal cancer specimens from different sites (primary and metastatic), stages, and genetic backgrounds, we compared the phenotype of CD44high/+ and CD44low/− cells. CD44high/+ cells formed tumorspheres within 14 d at high efficiency compared with CD44low/− cells (Fig. 1 H). In a limiting dilution assay, CD44high/+ showed a higher estimated stem cell forming frequency compared with CD44low/− (Fig. 1 I). When implanted into immunocompromised hosts, CD44high/+ cells formed tumors that phenocopied the parental tumor (Fig. 1 J). 2 × 104 CD44high/+ cells initiated tumors in at least 80% of immunocompromised mice injected from four different colorectal cancer patient specimens, whereas 2 × 105 CD44low/− cells were not sufficient to form tumors (Fig. 1 K). Collectively, these data suggest that CICs can be prospectively identified and characterized in the CD44high/+ cell population, whereas non-CICs are CD44low/-.
CICs display cell autonomous resistance to chemotherapy
The ability to prospectively enrich and deplete CICs permits comparative analysis of selected tumor cell phenotypes. Although therapeutic resistance is not a defining feature of CICs, relative chemoresistance has been reported for CICs (Lakshman et al., 2004, 2005; Todaro et al., 2007; Dylla et al., 2008; Dallas et al., 2009). We interrogated the responses of matched CICs and non-CICs derived from human colorectal carcinomas to a clinically relevant chemotherapy combination (FOLFOX: fluorouracil [5-FU], oxaliplatin, and leucovorin; Fig. 2). In two human tumor systems, the basal viability of CICs was similar to matched non-CICs (Fig. 2 A). Upon chemotherapy treatment, both CICs and non-CICs showed a decrease in relative viability, but CICs were significantly less sensitive to chemotherapy. To determine the cellular phenotype responsible for this change, we quantified apoptosis and found that basal apoptosis was relatively similar in CICs and non-CICs, but chemotherapy-induced apoptosis was significantly reduced in the CICs relative to non-CICs (Fig. 2 B). The relative resistance of CICs to chemotherapy suggests that CICs may be enriched after chemotherapy treatment. Indeed, the CIC immunophenotype (CD44high/+) was increased 1.5–3-fold after chemotherapy treatment in 4 human tumor models (Fig. 2 C and not depicted). Collectively, these results support a cell-autonomous relative chemoresistant phenotype of colorectal CICs.
Tumor response to treatment is associated with increased frequency of CAFs
Prior studies have suggested that cytotoxic therapy alters the stromal compartment in colorectal cancer (Washington et al., 2009; Edge et al., 2009), suggesting that this remodeling may be an extrinsic regulator of CICs. We therefore interrogated the stromal response in a cohort of colorectal patients treated with cytotoxic therapy. We investigated CAFs based on their proposed roles in supporting tumorigenesis (Olumi et al., 1999; Allinen et al., 2004; Orimo et al., 2005; Erez et al., 2010; Vermeulen et al., 2010; Calon et al., 2012). CAFs are not homogeneous, but rather display distinctive characteristics based on tumor type and stage (Micke and Ostman, 2004; Desmoulière et al., 2004; Sugimoto et al., 2006; Orimo and Weinberg, 2007; Xing et al., 2010). To model CAF evolution in colorectal cancer after cytotoxic treatment, we quantified CAFs during the course of therapy in human patients. In matched specimens from identical patients before and after cytotoxic treatment, we stained for CAF markers α-smooth muscle actin (αSMA) and vimentin (Sappino et al., 1988; Rønnov-Jessen et al., 1996; Serini and Gabbiani, 1999; De Wever and Mareel, 2002; Nakayama et al., 2002; Orimo and Weinberg, 2007; Worthley et al., 2010; Gonda et al., 2010). Immunohistochemistry demonstrated that fibroblast markers, including vimentin and αSMA, increased after treatment, whereas the epithelial marker EpCAM decreased (Fig. 3 A). Quantitative measures of the relative proportion of fibroblasts to the epithelial component (αSMA versus EpCAM) further demonstrated that CAFs were significantly increased after treatment (Fig. 3 B). We validated these findings using transcriptional profiling on a larger cohort of samples derived from patients who were either untreated or treated with cytotoxic regimens (17 samples each group; Fig. 3 C). These results confirm that CAFs are enriched during post-therapy tumor growth. Clinical characteristics and treatment data for the cohort of patients used above are summarized in Tables S2 and S3.
As shown in our findings and in previous studies (Dallas et al., 2009; Dylla et al., 2008; Lakshman et al., 2004; Lakshman et al., 2005; Todaro et al., 2007), CICs display cell autonomous therapeutic resistance, but the significant increase in the percentage of CAFs after cytotoxic treatment suggests that treatment may also induce the remodeling of the tumor microenvironment and CAFs may provide microenvironmental cues instructing the cellular hierarchy. A previous study demonstrated that an immortalized colon fibroblast cell line derived from normal colon increases CIC proliferation (Vermeulen et al., 2010). To extend these findings to human CAFs, we prospectively isolated viable CAFs from freshly resected patient colorectal cancer specimens using the cell surface protein, platelet-derived growth factor receptor α (PDGFRα; Erez et al., 2010; Micke and Ostman, 2004; Fig. 3 D). We then confirmed the cell immunophenotype via immunofluorescence of multiple activated fibroblast markers, such as PDGFRα, vimentin, αSMA, FAP (Park et al., 1999; Calon et al., 2012), and FSP1 (Grum-Schwensen et al., 2005; Sugimoto et al., 2006); we also confirmed the absence of the epithelial marker EpCAM (Fig. 3 E), the smooth muscle marker smoothelin (Gonda et al., 2010), and the pericyte marker high molecular weight caldesmon (Nakayama et al., 2002; not depicted). These patient-derived CAF cultures expressed activated fibroblast lineage markers without significant contamination by epithelial cells.
Chemotherapy-stimulated CAFs promote CIC growth
To determine the impact of CAFs on CICs, we performed a series of co-culture studies with the two cell types in proximity but not in direct contact. Both CICs and non-CICs had modest increases in cell viability when co-cultured with CAFs, demonstrating a general effect of CAFs in promoting tumor growth (unpublished data). However, co-culture of chemotherapy-treated CAFs increased the viability of CICs compared with vehicle-treated control CAFs (Fig. 4 A). As CAFs are phenotypically distinct from normal tissue fibroblasts and systemically administered chemotherapy induces genotoxic stress tissue-wide to create selection pressure in microenvironmental elements, prior studies may underestimate the role of CAFs in therapeutic resistance and recurrent tumor growth. In support of this hypothesis, we found that even normal fibroblasts (i.e., not CAFs) treated with chemotherapy increased CIC viability relative to untreated fibroblasts (Fig. 4 A). As cell–cell interactions such as those between CICs and CAFs can be mediated through paracrine factors that regulate the tumor hierarchy, we hypothesized that chemotherapy-induced CAFs secrete factors to maintain CICs. Indeed, we found that conditioned media from CAFs mimicked the effects of CAF co-culture on CIC growth (Fig. 4 B).
To determine the effects of CAFs on in vivo tumor growth, we implanted 3,000 CICs alone or in combination with CAFs pretreated with chemotherapy or vehicle control. In concordance with cell culture results, chemotherapy activation of CAFs augmented the ability of CICs to initiate tumors in immunocompromised mice (Fig. 4 C). The presence of CAFs increased tumor incidence and reduced the latency of tumor formation by CICs with increased tumor size in mice co-implanted with CICs and chemotherapy-pretreated CAFs (Fig. 4 C). Moreover, xenografts generated by CICs co-implanted with chemotherapy-treated CAFs displayed increased proliferative indices (Fig. 4 D) and the CIC immunophenotype (percentage of CD44high/+ cells by FACS analysis; Fig. 4 E). To establish the effects on CICs in functional assays, we performed secondary in vitro tumorsphere formation assays (Fig. 4 F) and measurement of resistance to cytotoxic treatment (ATP activity; Fig. 4 G) finding induction of increased cell survival and growth with chemotherapy-treated CAFs compared with the effects of untreated CAFs, suggesting specific effects on CIC maintenance. These data collectively support a model in which chemotherapy activates CAFs in the microenvironment to stimulate CIC growth and stimulate tumor growth.
Chemotherapy stimulates CAF cytokine secretion
These data demonstrated the presence of biologically important components in CAF secretome, so we performed a comparative analysis of cytokine expression in conditioned media of CAFs at baseline or treated with chemotherapy in three different patient specimens, as well as in normal fibroblasts. Although there were several molecules unique to each CAF, there were only three common molecules to every chemotherapy-induced CAF culture: the secreted DTK (TYRO3) receptor, IL-17A, and TGFβ (Fig. 5 A). A prior study suggested that a single normal, immortalized fibroblast line stimulated a CIC model through secretion of hepatocyte growth factor (HGF; Vermeulen et al., 2010). In our analysis, HGF was not significantly associated with chemotherapy response, suggesting that HGF may contribute to initial tumor growth but is less likely to be associated with chemotherapy response (Fig. 5 B). In contrast, sDTK, IL-17A, and TGFβ were expressed at low levels in quiescent tumor-derived and normal colon fibroblasts, but the expression of each molecule dramatically increased after chemotherapy in CAFs but not normal colon fibroblasts (Fig. 5 C). This regulation appeared to occur at the level of transcription, as similar increases were detectable for each target at the mRNA level (Fig. 5 D). Among the three top up-regulated cytokines (IL-17A, TGFβ, and GAS6 [DTK ligand]), TGFβ and GAS6 did not directly stimulate CIC growth (unpublished data) so we focused on IL-17A. CAF expression of IL-17A protein was further confirmed by immunofluorescence double staining of αSMA and IL-17A of matched human patient samples before and after cytotoxic treatment (Fig. 5 E).
To extend these observations to a clinical paradigm, we quantified IL-17A mRNA in cohorts of colorectal cancer tumor specimens from patients either untreated or treated (17 patients in each group; Fig. 6 A; clinical data are summarized in Tables S2 and S3). We also studied IL-17A protein expression before and after therapy in an additional four matched human tumors by immunohistochemistry (Fig. 6, B and C; and data not depicted). In each case evaluated, there was an increase in stromal expression of IL-17A protein, which was additionally confirmed in a cohort of 40 tumor specimens from either untreated or treated patients (20 patients in each group; Fig. 6 D). Collectively, these results strongly suggest that CAFs respond to chemotherapy by the expression of defined factors including IL-17A that could instruct the cellular hierarchy.
IL-17A can contribute to CICs maintenance through IL-17A receptor
The vast majority of colorectal cancers are spontaneous but a subset of colorectal cancers occurs within the context of inflammatory bowel disease, particularly ulcerative colitis. This minority has been extensively modeled using genetically engineered mouse models of colitis to induce colorectal cancer. In these models, IL-17A is secreted by T-helper IL-17 (Th17) cells to regulate other immune elements and promote tumorigenesis (Wu et al., 2009). Our studies focus on models that represent spontaneous colorectal tumors in which inflammation is important but may be less central to disease pathology and, thus, other stromal elements may serve greater roles. We found that chemotherapy-treated CAFs directly augmented CIC growth in the absence of inflammatory cells, suggesting that IL-17A might directly regulate CICs. As prior studies have not determined involvement of IL-17A in CICs, we next sought to evaluate the direct contribution of IL-17A signals to CIC maintenance. First we analyzed IL-17A expression in CICs treated with vehicle or chemotherapy in four different patients by cytokine array analysis and we showed the presence of moderate autocrine production of IL-17A that could contribute to their own hierarchy (Fig. 7 A). To clarify the potential role for autocrine effects of IL-17A, we examined the effects of IL-17A blockade on CIC’s viability in the absence of CAFs using a neutralizing antibody (Fig. 7 B). In these studies, autocrine IL-17A appeared to function as a minor component in CIC maintenance relative to stromal IL-17A. These data suggest that the dominant role of IL-17A is a paracrine signal.
As the biological effects of IL-17A require target cell expression of the cognate receptor, we assessed IL-17A receptor (IL-17RA) expression in isolated CICs cultured short term as tumorspheres. The majority of CICs expressed IL-17RA, as measured by immunofluorescent staining and FACS analysis, whereas non-CICs less commonly expressed IL-17RA (Fig. 7, C and D), supporting a possible functional paracrine loop of CAF-derived IL-17A to IL-17RA on CICs.
IL-17A promotes protumorigenic CIC behavior
To interrogate the functional significance of IL-17A on CICs and non-CICs, we added exogenous IL-17A ligand to mimic the effects of IL-17A produced by CAFs. IL-17A enhanced CIC tumorsphere formation measured by in vitro limiting dilution (Fig. 8 A), with more modest effects on non-CICs. In addition, IL-17A modestly increased the size of the spheres formed compared with the untreated group (Fig. 8 B). Nuclear localization of β-catenin has been reported as a potential functional marker of CICs (Vermeulen et al., 2008). We therefore examined the effects of IL-17A signaling on the percentage of cells expressing nuclear β-catenin. IL-17A induced nuclear β-catenin localization in CD44high/+ cells but not CD44low/− cells, confirming in limiting dilution studies (Fig. 8, C and D). IL-17A promoted increased CIC viability in ATP-dependent assays, cell survival, and DNA synthesis (Fig. 8, E–G). These effects were not detected in non-CICs (Fig. 8, H–J). Another phenotype associated with CICs, label retention, was increased in response to IL-17A treatment (Fig. 8 K). As CICs may contribute to tumor invasion and metastasis, we also demonstrated that exogenous IL-17A nearly doubled the CIC migration rate as measured by a scratch assay (Fig. 8 L).
To further evaluate the effects of IL-17A signal transduction on CICs, we generated genetically modified cells with IL-17A overexpression to mimic the effects of CAFs or IL-17RA knockdown to block the reception of IL-17A signals. We validated our viral transduction protocol using a GFP-containing vector. Efficacy of viral transduction was measured by FACS. 91.5% of transfected CICs were positive for GFP (unpublished data). IL-17A cDNA or IL-17RA shRNA resulted in an ∼80% up-regulation or reduction of IL-17A or IL-17RA mRNA, respectively (Fig. 9, A and B). When these genetically modified CICs were studied in an in vitro limiting dilution assay, IL-17A overexpression significantly increased tumorsphere formation, whereas IL-17RA knockdown decreased sphere formation (Fig. 9 C). Exogenous IL-17A also stimulated tumorsphere formation but this effect was more modest than ligand overexpression (Fig. 9 C), likely because of the sustained effects of secreted ligand. These data suggest that IL-17A broadly augments protumorigenic CIC behaviors.
To address the contribution of IL-17A in the in vivo growth of CICs using a clinically relevant method, we used an IL-17A neutralizing antibody in combination with chemotherapy (5-FU and Oxaliplatin). Tumors were established by injecting CICs in combination with CAFs subcutaneously in immunocompromised mice. The addition of the IL-17A neutralizing antibody augmented the efficacy of chemotherapy on tumor growth (Fig. 9, D and E), confirming CIC growth dependence upon chemotherapy treatment on IL-17A. These results provide a direct relevance for IL-17A targeting, even in the absence of a functional immune system.
IL-17A contributes to CIC therapeutic resistance
In concordance with IL-17A as an important regulator of the CIC phenotype, we found that IL-17A promoted basal CIC growth and protected CICs from chemotherapy-induced growth inhibition (Fig. 10 A). A neutralizing IL-17A antibody blocked the effects of chemotherapy treated CAFs on CIC growth in comparison to an IgG control (Fig. 10 B), confirming the importance of IL-17A for CAF effects on CICs. To define potential mechanisms through which IL-17A regulates CIC biology, we interrogated gene expression profiles of CICs isolated from three different patients treated with and without IL-17A. Heat maps illustrate the common effects of IL-17A (Fig. 10 C) on the repression and activation of subsets of genes within CICs. When these genes were analyzed to determine the pathways altered by IL-17A treatment, NF-κB signaling was found to be highly elevated (IPA Ingenuity Systems; Fig. 10 D). We confirmed activation of the NF-κB pathway and its downstream target ERK1/2 at the protein level through the analysis of the levels of the phosphorylation status of NFκBp65 (RelA) and p42-44-ERK1/2 in CICs treated with vehicle or with IL-17A for 12 h (representative immunoblot, Fig. 10 E). The specificity of this effect was confirmed by the addition of an IL-17A–blocking antibody to restore the basal phosphorylation level confirming the direct connection between IL-17A and the NF-κB activation (Fig. 10, D and E). These data suggest that IL-17A up-regulation of NF-κB is a possible contributing mechanism through which CIC maintenance is promoted and that targeting of NF-κB or its target genes may inhibit the effects of CAF produced IL-17A after cytotoxic treatment. As NF-κB has been associated with CIC maintenance (Schwitalla et al., 2013), our results add an additional layer of microenvironmental signals that may enforce tumor growth.
Stem cells are essential contributors to maintenance of tissue homeostasis and recovery from injury. However, the sustained proliferative potential of stem cells must be tightly constrained to prevent cancer. The functional location or niche in which a stem cell resides provides maintenance cues and restrains uncontrolled growth. When a stem cell exits its niche, it undergoes lineage specification and differentiation, providing an essential role for the microenvironment in normal cellular hierarchies (Schuijers and Clevers, 2012). As cancers mimic development and wound responses (i.e., “the wound that does not heal”; Dvorak, 1986; Schäfer and Werner, 2008), two scenarios in which stem cells function, it is not surprising that cancers display cellular hierarchies and coopt stem cell regulatory mechanisms. This has been shown first in leukemia (Bonnet and Dick, 1997) and, more recently, in several solid cancers, including colorectal cancers (Dalerba et al., 2007). Single hematopoietic stem cells repopulate the entire systemic blood system, and these cells rapidly home to the bone marrow niche for instructive cues to sustain this process. Organ regeneration by single stem cells is less evident in solid tissues, and recent work in the colon has shown that intestinal epithelial stem cells display a synergistic relationship with supportive CD24+ (or cKit+) accessory cells, similar to Paneth cells in the small intestine (Sato et al., 2011). Stem cell–Paneth cell doublets potently proliferate relative to single stem cells (Sato et al., 2011). This relationship demonstrates a functional complementation between specialized client cells and accessory cells that provides signals to promote client cell adaptation and defend the homeostatic state under noxious conditions. This hypothesis has been advanced by Medzhitov and Horng (2009) within the context of macrophage function in inflammatory responses. Based on this background, our results suggest that colorectal cancers respond to noxious conditions (cytotoxic treatment) by stimulating accessory cells (CAF) to provide signals (IL-17A) to support a homeostatic state (maintenance of the cellular hierarchy). Disruption of this circuit through targeted therapies may sensitize the chemotherapy-resistance commonly found in colorectal cancers.
Although several studies have examined the cellular hierarchy within colorectal cancers, many have used cell lines that have been forced to acquire self-renewal mechanisms in the absence of the tumor microenvironment. Our studies were devoted to dissecting a single cell–cell interaction in a more complex system. An elegant genetically engineered mouse model of colon cancer described a role of a long term passaged, immortalized normal colon fibroblast cell line in the self-renewal of early transformed colon cells associated with expression of HGF (Vermeulen et al., 2010). As most lethal colorectal cancers are diagnosed at more advanced stages, we used a clinical relevant strategy based on human CAFs (rather than a modified normal fibroblast) and colorectal cancer cells directly derived from patients. Using these resources, we recognized increased CAF frequency in patient tumors after cytotoxic treatment, suggesting that CAFs may function in regenerating the aberrant organ system in colorectal cancers. Chemotherapy-stimulated CAFs produced a secretome that enhanced CIC proliferation and tumorigenesis. As cytokines offer important instructive cues from supportive accessory cells, we focused our attention on these molecules within the secretome. We found that chemotherapy potently induced IL-17A in CAFs that directly augmented the growth of CICs. Of note, we also found that chemotherapy induces expression of CAF TGFβ, a chemokine with a well-established role in the cancer microenvironment. Alterations in TGFβ pathway components affect 40–50% of all colorectal cancers (Markowitz et al., 1995; Markowitz and Bertagnolli, 2009). In the tumor microenvironment, TGFβ induces transdifferentiation of normal fibroblasts to CAFs, i.e., morphological changes and up-regulation of α-SMA. (Lieubeau et al., 1994; Tuxhorn et al., 2002). In addition, TGFβ activates colorectal cancer CAFs to secrete various cytokines that exert a prosurvival signal on cancer cells (Calon et al., 2012) and stromal TGFβ signaling is required for the initiation of colorectal cancer metastasis (Calon et al., 2012). Although TGFβ has been linked to CIC growth in glioblastoma (Peñuelas et al., 2009), we did not find that TGFβ directly enhanced colorectal CIC growth (unpublished data). However, our results support potential interactions of multiple cellular components within tumors as TGFβ induces IL-17A in CD8+ splenocytes in tumor-bearing mice (Schäfer and Werner, 2008), suggesting that chemotherapy treatment may stimulate IL-17A expression both directly by CAFs and indirectly through TGFβ effects on the immune system. This model would explain why targeting IL-17A largely inhibited the direct effects of CAFs on CICs, thus augmenting the importance of this pathway in colorectal cancer.
Prior delineation of the role of IL-17A in colorectal cancer has come from either models of colitis-associated cancers (Hyun et al., 2012), induction of tumor formation from colonic bacteria (Wu et al., 2009; Grivennikov et al., 2012), and promoting tumor resistance to antiangiogenic therapy (Chung et al., 2013). These studies have largely focused on the interplay between IL-17A and the immune system, including Th17 cells. As we were able to isolate the paracrine relationship between CAFs and tumor cells, future studies will combine these observations with interactions between the immune system and both CAFs and tumor cells.
IL-17A has been the subject of intense research not only in basic laboratory studies, but has already translated into the development of pharmacologic inhibitors, primarily for autoimmune conditions. Although targeting IL-17A has been effective in treating psoriasis (Leonardi et al., 2012; Papp et al., 2012), anti–IL-17A agents have proven less effective against Crohn’s disease, suggesting that combination strategies may be required for some diseases in which IL-17A is functionally important (Hueber et al., 2010). Our studies suggest that IL-17A may also function indirectly in chemotherapy resistance through remodeling of the tumor microenvironment, suggesting a potentially more nuanced role and application within therapeutic paradigms. Indeed, we are currently developing preclinical targeting strategies to compare both genetic and pharmacologic inhibition in chemotherapy-induced tumor remodeling. These results suggest that chemotherapy may also serve as a “double-edged sword” due to effects on enriching CICs that are relatively less chemosensitive and altering the tumor microenvironment through effects on CAFs. We are actively investigating these elements as well as the interactions between tumor genotype and these tissue stress responses. To date, we have found no differences in the responses of CAFs to chemotherapy in IL-17A production or response of CICs to IL-17A treatment based on their genetics (e.g., KRAS mutations, microsatellite stable versus unstable). Collectively, these results and those of other groups studying a role of IL-17A in the early induction of colorectal tumorigenesis strongly suggest that cytokines produced in response to cytotoxic treatment may function to support the development and maintenance of colorectal cancers and offer novel therapeutic targeting strategies.
MATERIALS AND METHODS
Isolation and culture of CICs and CAFs.
All human tissues were acquired from primary human colorectal tumor patient specimens according to human experimental guidelines. Human protocols were approved by the Institutional Review Board of the Cleveland Clinic Foundation (protocol number: IRB 4134). Tumor specimens were maintained through subcutaneous xenografts in the flanks of NOD-SCID/IL2Rγnull (NSG) mice. Tumors were dissociated using a papain dissociation system (Worthington Biochemical). CD44high/+ cells (enriched in CICs) were enriched by flow cytometry (FACS Aria II) and grown as tumorspheres at 37°C in an atmosphere of 5% CO2. CICs were cultured in serum-free media with basic fibroblast growth factor (bFGF, 10 ng/ml; R&D Systems) and epidermal growth factor (EGF, 10 ng/ml; R&D Systems). For isolation of CAFs, surgical specimens were similarly dissociated into single-cell suspension, and PDGFRα-expressing cells were sorted using flow cytometry (FACS Aria II). Cells were then cultured in DMEM with 10% BSA. For cell counting before each experiment, a single-cell suspension was achieved using TrypLE (Invitrogen) dissociation.
Flow cytometric analysis.
Flow cytometry was performed using a FACSAria II Cell Sorter (BD). To enrich for CICs. Single cells were labeled with a phycoerythrin-conjugated monoclonal antibody against CD44 (Miltenyi Biotec), and then analyzed for the expression of phycoerythrin-conjugated monoclonal antibody against CD133 (Miltenyi Biotec), FITC monoclonal antibody against EpCAM (R&D Systems), and ALDH with ALDEFLUOR assay (ALDAGEN). To enrich for PDGFRα-positive CAFs, single cells were labeled with phycoerythrin-conjugated monoclonal antibody against PDGFRα (R&D System). To analyze the expression of IL-17RA within the CIC population, single cells were double labeled with FITC-CD44 (Miltenyi Biotec) and APC-IL17RA (R&D Systems). Dead cells were eliminated by using the viability dye DAPI. Side scatter and forward scatter profiles were used to eliminate cell doublets. Isotype controls were used to establish proper gates. The “high positive” population was gated in a range of 4–10% (depending on the percentage of the total positive cell population of each sample) of the tail of the positive cells.
CICs from 3 different patients were treated with vehicle, 100 ng/ml IL-17A ± 15 ng/ml IL-17A blocking antibody (R&D Systems). Lysates were harvested 12 h after treatment and immunoblotted. The membrane was simultaneously probed for anti–NF-κB p65 (Cell Signaling Technology), p42/44ERK1/2 (Cell Signaling Technology), and anti–β-actin (Sigma-Aldrich). Secondary detection was accomplished using IRDye 800 and IRDye 680 detected on the Odyssey imaging system (Li-Cor).
Total cellular RNA was isolated with the RNeasy kit (QIAGEN) and reverse-transcribed into cDNA using the SuperScript III Reverse Transcription kit (Invitrogen). Real-time PCR was performed on an Applied Biosystems 7900HT cycler using SYBR Green Master mix (SA Biosciences) and gene-specific primers, as follows: β-actin forward (5′-AGAAAATCTGGCACCACACC-3′) and reverse (5′-AGAGGCGTACAGGGATAGCA-3′), CD44 forward (5′-GCCCTTCCATAGCCTAATCC-3′) and reverse (5′-CTTTGGTGTCTCCCAGAAGC-3′), Sox-9 forward (5′-CCTGCCCTTCTGGTTCCG-3′) and reverse (5′-TCCTCGTCCCTCTCTTTCTTCAG-3′), ALDH1 forward (5′-TGGCTTATCAGCAGGAGTGT-3′) and reverse (5′-GCAATTCACCCACACTGTTC-3′), IL-17A forward (5′-AAGACCTCATTGGTGTCACTGCTAC-3′) and reverse (5′-ATCTCTCAGGGTCCTCATTGCG-3′), IL-17RA forward (5′-AGACACTCCAGAACCAATTCC-3′) and reverse (5′-TCTTAG AGTTGCTCTCCACCA-3′), OCT4 forward (5′-TCTCCCATGCATTCAAACTGAG-3′) and reverse (5′-CCTTTGTGTTCCCAATTCCTTC-3′), BMI1 forward (5′-CCAGGGCTTTTCAAAAATGA-3′) and reverse (5′-GCATCACAGTCATTGCTGCT-3′), c-MYC forward (5′-TCAAGAGGCGAACACACAAC-3′) and reverse (5′-GGCCTTTTCATTGTTTTCCA-3′), MSI1 forward (5′-CCAATGGGTACCACTGAAGC-3′) and reverse (5′-ACTCGTGGTCCTCAGTCAGTCAGC-3′), DTK forward (5′-TGCTGGCAGAGGACATGA-3′) and reverse (5′-ACAGACAGCTGGCCCAAG-3′) TGFβ forward (5′-GCCTTTCCTGCTTCTCATGG-3′) and reverse (5′-TCCTTGCGGAAGTCAATGTAC-3′).
In vitro limiting dilution assay.
CICs were flow sorted into 96-well plates at a final cell number per well of 1, 10, 50. Tumorsphere formation was evaluated 7 and 14 d after sorting, and wells were scored positive or negative for the presence of at least one tumorsphere. The estimated stem cell frequency was calculated using extreme limiting dilution analysis (Hu and Smyth, 2009).
In vivo tumorigenic potential.
All animal studies described were approved by the Cleveland Clinic Foundation Institutional Animal Care and Use Committee and conducted in accordance with the National Institutes of Health Guide for the Care and Use of Animals. For subcutaneous tumors studies, 3 × 103 CICs from a xenograft originally derived from a primary colorectal cancer patient specimen (064) were implanted alone and along with 10 × 103 fibroblasts treated for 3 d with vehicle or chemotherapy: 5-Flurouracil (50 µg/ml) + Oxaliplatin (10 µM) + Leucovorin (1 µM) into the flanks of 20 (5 mice per condition) 6-wk-old female NSG mice were obtained from The Jackson Laboratory. Tumors were monitored, and after 1 wk from the first tumor growth, all the mice were sacrificed and tumors were removed and weighed to evaluate the tumor development, and then dissociated (see following paragraph).
In vivo evaluation of the dependence of CICs on IL-17A.
104 CICs and 3 × 104 CAFs were implanted into mice. When established tumors reached ∼0.3 cm3 in volume, treatment was initiated. Four arms were included: isotype control (2 mg/kg antibody 5 times a week for 2 wk and then once a week for 2 wk), FOLFOX chemotherapy alone (oxaliplatin:, 0.25 mg/kg once a week for 4 wk; 5-FU, 15 mg/kg 5 times a week for 2 wk), IL-17A blocking antibody alone (IL-17A blocking antibody, 2 mg/kg 5 times a week for 2 wk and then once a week for 2 wk), or a combination of the IL-17 antibody and FOLFOX chemotherapy (4 mice per arm). Mice were weighed and the tumor volume was measured every other day for 30 d. Mice were sacrificed and tumors were collected and analyzed.
Cell growth and survival assays.
3,500 freshly sorted CICs and non-CICs cancer cells per well were plated in triplicate into 96-well plates containing appropriate growth media. The next day, vehicle or chemotherapy was added to the plate. Readings of ATP levels for viability (CellTiter-Glo, Promega) or apoptosis (Caspase-Glo 3/7, Promega) were performed using a luminometer (Perkin-Elmer).
Tumorspheres dissociated at single cells were cytospin (Shandon Cytospin4, Thermo) in PBS with 0.5% BSA. Cells were then post-fixed/permeabilized in 4% PFA, blocked in 5% normal goat serum, 0.1% TritonX-100 in 1XPBS and immunostained for anti-IL-17RA (R&D System) overnight at 4°C. CD44high/+ cells sorted from primary xenograft were treated for 12 h with vehicle or IL-17A and stained for β-catenin (Abcam). Cancer associated fibroblasts cultured on coverslips were stained for PDGFRα, VIMENTIN, αSMA, FAP, EpCAM, SMOOTHELIN, high molecular weight CALDESMON (Abcam) overnight at 4°C. Secondary detection was accomplished using highly cross-adsorbed secondary fluorescently labeled antibodies (Invitrogen) for 1 h at room temperature. Nuclei were counterstained with DAPI (Invitrogen). For immunostaining analyses of tissue sections, 10-µm frozen sections were fixed, permeabilized, and stained for Il-17A (RD Systems), KI67 (Abcam) as describe above. Images were taken on an upright microscope (microscope make and model: Leica DM4000B; magnification/aperture: 10× 0.3NA- 20× 0.5NA- 40× 0.75NA; room temperature; imaging medium: air (glass coverslipped slides); fluorochromes: 488 and 568; camera make and model: QImaging Exi aqua; acquisition software: Qcapture pro6).
Paraffin-embedded sections of human colorectal tumors matched before and after treatment from four patients were stained for IL-17A. The percentage of the IL-17A–positive cells versus the total stromal cells was quantified.
Immunohistochemistry/immunofluorescence imaging quantification.
Images of pre- and post-chemotherapy tissues cross sections were acquired using a DM6000 upright microscope (Leica), a 10× 0.3NA objective, and a Retiga 2000R CCD digital camera (Q-Imaging; room temperature, imaging medium: air (glass coverslipped slides), fluorochrome: 488 and 568). Image acquisition was fully automated for up to eight slides using an eight-slide, linearly encoded X, Y, Z-motorized stage (Prior Scientific) managed by Objective Imaging’s Oasis 4i controller, TurboScan software (Objective Imaging), and Image-Pro Plus 7.0 (Media Cybernetics).
During acquisition, high-magnification (auto-focused) image fields were acquired across the entire tissue section and stitched together to form a single high-resolution, large field-of-view (FOV) image. Each image field was background corrected before stitching to ensure continuity and account for lens-induced vignetting. In brief, for each cross section, the stromal channel was loaded, spectrally filtered to cluster regions, and thresholded (by size and morphology) to generate a binary mask. A Euclidean distance transform was then applied to this mask and multiplied by the stromal “skeleton” (medial axis thinning of stromal regions) to determine pixel by pixel stromal thickness. Similarly, the epithelial channel was also loaded and morphologically processed to generate a binary mask. Stromal thickness values and epithelial/stromal segmented areas were subsequently exported to Excel and these segmented regions and thickness measurements (stromal mask based restricted dilation) were pseudocolored and superimposed upon the original images. Finally, for the calculation of intensity distribution, the stromal and epithelial masks were multiplied by their respective extracted channels and thresholded in bins of 10 gray values to calculate area of staining per intensity bin. These values were also exported to Excel for each channel.
CIC growth in co-culture assay.
2 × 104 CAFs pretreated with vehicle (DMSO) or chemotherapy (5-flurouracil (50 µg/ml) + oxaliplatin (10 µM) + leucovorin (1 µM)) for 3 d were plated in triplicate into the top part (with 3 µm pores) of a 24-well co-culture plate in serum-free media. 104 freshly sorted colorectal CD44 high/+ cancer cells were plated in triplicate into the bottom of the co-culture plates. IgG control and IL-17A–blocking antibodies (15 ng/ml; R&D Systems) were used as indicated. After 10 d, ATP levels (CellTiter-Glo; Promega) were performed using a luminometer (Perkin-Elmer). Data are mean ± SD (n = 3); **, P < 0.01; ***, P < 0.001.
Thymidine incorporation assay.
2 × 104 CICs were plated in triplicate in 12-well plates, incubated for 4 h at 37°C, and washed in cold PBS and then 10% TCA. After 1-h incubation at 4°C in 10% TCA, cells were resuspended in 0.2 N NaOH and incubated at room temperature overnight. Samples were then placed in scintillation tubes with 9 ml of scintillation fluid and vortexed to break up thymidine precipitates. The count per minute of the thymidine-positive cells was taken using LS6500 Scintillation Counter (Beckman Coulter).
The cytokine secretion of CAFs or CICs treated with vehicle control (DMSO) or chemotherapy for 3 d was assessed using a 120 human cytokine antibody array from RayBio. This membrane-based cytokine is designed to simultaneously detect expression levels of multiple cytokines, including the one that are normally involved in disease processes like cancer. This array has been previously used by Vermeulen et al. (2008, 2010).
The Illumina microarray study comprised three matched pairs of CICs samples. Each pair had a treated (100 ng/ml IL-17A for 7 d) and an untreated sample. The normalized gene signals, p-values, and annotations were imported into MS access a database manager and processed. For a matched pair of samples, the gene signal ratio was calculated. Records were then filtered using these criteria.
A record passed the filter and was deemed an increase in gene expression if the signal ratio of treated/untreated samples was greater than or equal to 1.5 and the p-value of the gene in the treated sample was ≤0.05 (i.e., detected as present). A record passed the filter and was deemed a decrease in gene expression if the signal ratio of treated/untreated samples was ≤0.6666666 and the p-value of the gene in the untreated sample was ≤0.05 (i.e., the gene transcript had to be detected as present in the untreated sample if we are to truly believe it has decreased in the treated sample).
Changes found within the three matched pairs were collated into a master table and sorted into groups of genes that had been observed to have changed in all three matched pairs, in two of three and in only one of the three. If all changes were included, the total number would be 821 genes. If we trimmed the data to include only genes that were seen to have changes in at least 2 of the 3 matched pairs, the list fell to 49.
Signal data were median-normalized to prepare for importation into the software Partek Genomics Suite ver6.4. Principal component analyses were conducted to get a sense of how the whole expression profiles behaved within each group.
Signals were filtered using the 821 gene filter list and the 49 gene filter list and the expression pattern, resulting in these respective filters were subjected to hierarchical clustering. Samples were fixed and genes profiles were clustered. A Pearson dissimilarity metric with complete linkage was used to segregate profiles into various hierarchical groups.
Signal data were filtered to include only genes that were seen to have changes in at least two of the three matched pairs and were subjected to hierarchical clustering. Signal ratio > 1.5 and p-value < 0.05.
CICs were stained for 5 min with 1:250 PKH-26 dye (Sigma-Aldrich), blocked with 1% BSA, washed twice and FACS analyzed with a FACS LSR II (BD) equipped with a 488-nm laser and a band-pass 575/26 nm optical filter. CICs were then divided, plated, and stimulated with vehicle and IL-17A (100 ng/ml) and, after 1 wk, the analysis was repeated.
Migration assay and time-lapse imaging.
CICs stimulated with vehicle or 100 ng/ml exogenous IL-17A were plated in 6-well plates and allowed to attach. Upon full confluence, the plate was scratched with a pipet tip and the ability of the cells to migrate in that area was monitored and imaged for 64 h on a Leica DMIRB Inverted Microscope equipped for time-lapse microscopy with a Leica DFC420 camera, temperature controller (37°C), CO2 (5%) incubation chamber (Leica), PeCon incubator (PeCon GmbH), prior motorized stage with linearly encoded controller with x/y/z drive for time-lapse imaging of multiples fields, Uniblitz shutter (Vincent Associates), and MetaMorph Software (Molecular Devices). Images of multiple fields per well were collected every 3 min for 72 h using a dry 10×, 0.3 NA objective lens, and phase-contrast optics.
Vectors and lentiviral transfection.
Lentiviral clones expressing shRNAs specific to IL-17RA (TCRN0000059153: CCGGGTGGAACGAATCTACCCATTACTCGAGTAATGGGTAGATTCGTTCCACTTTTTG TCRN0000059154: CCGGGCTGAACACCAATGAACGTTTCTCGAGAAACGTTCATTGGTGTTCAGCTTTTTG TCRN0000059155: CCGGCGACTGGTTCGAATGTGAGAACTCGAGTTCTCACATTCGAACCAGTCGTTTTTG TCRN0000059156: CCGGCCACAGTTGCTTTGAGCACATCTCGAGATGTGCTCAAAGCAACTGTGGTTTTTG TCRN0000059157: CCGGGCTAAACTGCACGGTCAAGAACTCGAGTTCTTGACCGTGCAGTTTAGCTTTTTG) and nontargeting control shRNA (SHC002) were purchased from Mission Sigma-Aldrich. Lentiviral clones overexpressing IL-17A (RC218057) and control vector (pCMV6) were purchased from OriGene. Using Lipofectamine 2000 (Invitrogen), viral particles were produced in 293T cells with PAX2 and VSVG helper plasmids (Addgene). Media on the 293T cells was changed 18 h after transfection with stem cell medium, and, 1 d later, viral supernatants were collected, filtered, and used or frozen at −80°C.
Statistical significance was calculated with GraphPad Prism Software using a 1-way or 2-way ANOVA with a Bonferroni’s post-test, Student’s t test, or log-rank (Mantel-Cox) test where appropriate (GraphPad Software Inc.). Data are represented as the mean ± SD.
The study was supported by the International Fellowship Grant of American Society of Colon and Rectal Surgeons (to A.M. Jarrar); Italy-USA Onco-Proteomic Research Program (to F. Lotti); CA154130, CA129958 (to J.N. Rich); CA151522 (to A.B. Hjelmeland); NS063057 (to M. Hitomi and A. Vasanji); National Institutes of Health K99/R00 CA157948, V Scholar Award from the V Foundation for Cancer Research (to J. Lathia); T32 GM007250, F30 CA165857 (to K. Sukhdeo). M.F. Kalady is the Krause-Lieberman Chair in Colorectal Surgery.
The authors have no conflicting financial interests.
Author contributions. Conception and design: F. Lotti, A. Jarrar, A.B. Hjelmeland, M.F. Kalady, J.N. Rich. Development of methodology: F. Lotti, A.M. Jarrar, R.K. Pai, M. Hitomi, J. Lathia, K. Sukhdeo; acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): F. Lotti, A.M. Jarrar, R.K. Pai, M. Hitomi, J. Lathia, A. Mace, G.A. Gantt Jr., K. Sukhdeo, J. DeVecchio, A. Vasanji, P. Leahy, M.F. Kalady, J.N. Rich; analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F. Lotti, A.M. Jarrar, R.K. Pai, M. Hitomi, J. Lathia, G.A. Gantt Jr., J. DeVecchio, A. Vasanji, P. Leahy, M.F. Kalady, J.N. Rich; writing, review, and/or revision of the manuscript: F. Lotti, A. Jarrar, J. Lathia, A.B. Hjelmeland, M.F. Kalady, J.N. Rich; administrative, technical, or material support (i.e., reporting or organizing data, constructing database): F. Lotti, A.M. Jarrar, M. Hitomi, A. Mace, G.A. Gantt Jr., K. Sukhdeo, A. Vasanji, P. Leahy, A.B. Hjelmeland, M.F. Kalady, J.N. Rich; study supervision: M.F. Kalady, J.N. Rich; designed and conducted experiments: F. Lotti, A.M. Jarrar, M. Hitomi, A.B. Hjelmeland, M.F. Kalady, J.N. Rich.
cancer associated fibroblast
cancer initiating cell
hepatocyte growth factor
F. Lotti and A.M. Jarrar contributed equally to this paper.