Autocrine Secretion of Progastrin Promotes the Survival and Self-Renewal of Colon Cancer Stem–like Cells

Cancer stem cells (CSC) have been implicated in the initiation, maintenance, and propagation of hematologic and solid tumors (1, 2). They represent a source of phenotypic heterogeneity in several cancer types (3), where they play a role in metastasis development (4) and resistance to current therapies (5). The recent literature suggests that the CSC phenotype may be tightly regulated by extrinsic cues coming from the surrounding microenvironment (6), including other tumor cells (7), suggesting that the regulation of differentiation and cell plasticity by microenvironment-driven signals controls phenotypic heterogeneity within tumor cell populations. Thus, the demonstration that non–stem tumor cells are able to convert into self-renewing tumor-initiating cells (8) challenges the hierarchical concept that constitutes a pillar of the CSC paradigm, pointing towards the existence of a CSC state rather than fixed CSC entities (9). Yet, these findings also suggest that homeostatic proportions of identifiable tumor-initiating and nontumorigenic cell subpopulations exist in individual tumors and that drugs targeting stem-like populations only are likely to trigger plasticity until this homeostatic equilibrium is reattained, thus allowing posttreatment relapse. A better understanding of signals that maintain homeostatic proportions of CSCs and/or regulate phenotypic plasticity between non-stem and stem-like cells is therefore essential to inform future combination therapies aiming to minimize relapse.

Extrinsic cues controlling the phenotype of colorectal CSCs include soluble factors (secreted by stromal or tumor cells) that directly regulate key signaling pathways for the CSC phenotype, such as Wnt or Notch (6, 7). We hypothesized that progastrin, secreted by human colorectal tumors (10, 11) and exhibiting strong tumor-promoting functions in colorectal cancer (12, 13), may play a role in CSC regulation. We previously demonstrated that progastrin regulates the Wnt and Notch pathways in colorectal cancer (12, 14), suggesting that it could affect the phenotype of colon CSCs, which rely on these pathways for their survival (15). Progastrin was shown to promote the proliferation of progenitor cells in the mouse colonic epithelium (16) and a link has been suggested between progastrin expression and populations of cells expressing CD133 (17), DCAMKL1 (18), or CD44 (19) in mouse colonic crypts and human cancer cell lines. Yet, expression of these markers is not restricted to CSCs (20–23), and the functional role of progastrin on CSC self-renewal and tumor-initiating potential has not been documented.

Here we demonstrate that colorectal CSCs require this peptide to maintain their tumor-initiating and self-renewal ability and demonstrate that this regulation acts through the promotion by progastrin of a highly glycolytic phenotype in colon CSCs. We further establish that downregulation of progastrin expression or neutralization of its activity alters the long-term tumor-initiating capacity of these cells in vitro and in vivo.

Human colorectal cancer cell lines T84, SW620, HCT116, HT29, and RKO (obtained from the ATCC between 2005 and 2010) were cultured in a humidified atmosphere at 37°C and 5% CO₂ in complete DMEM. Cell line authenticity was last tested in December 2015 using STR profiling (LGC standards). Patient-derived colorectal cancer cells (CPP1-19-24-25) were prepared from fresh biopsies (Human ethics agreement #2011-A01141-40) as described under Supplementary Methods.

One hundred cells per well were seeded in ultra-low adherent 96-well plates (Corning) in 100 μL of M11 medium. After 7 days, photos of each sphere (>40 μm) were taken and sphere size was measured using the ImageJ software.

Cells were transiently transfected once, seeded, and resulting spheres were passaged every 7 days. For antibody treatment, cells were treated once a day with 3 μg/mL polyclonal anti-human IgG (Jackson ImmunoResearch, 309-005-008) or progastrin-selective antibodies. Counting was performed 7 days later. CSC frequency was determined using the Extreme Limiting Dilution Analysis (ELDA; ref. 24) 7 days after seeding in M11 medium (10 wells/condition).

Total RNA was extracted using the RNeasy Mini/Micro Kit (Qiagen) and treated with DNAse-1. cDNA was synthetized using M-MLV or SuperScript II (when RNA < 250 ng; Invitrogen). Relative gene expression was measured by real-time PCR as described previously (12). GAPDH expression was used as an internal calibrator. Primer sequences are provided in Supplementary Methods.

RNA expression profiling of cell lines was performed using Affymetrix Human gene ST 2.0 DNA microarrays (see Supplementary Methods). Raw transcriptome data were deposited in the ArrayExpress repository (http://www.ebi.ac.uk/arrayexpress/, accession number: E-MTAB-3215).

siRNA transfection was performed in suspension, using Lipofectamine RNAiMAX and siRNA duplexes (siβgal and siPG, Invitrogen), according to manufacturer's recommendation. Six hours after transfection, cells (1–6 cells/μL) were seeded in ultra-low adherent flasks or 96-well plates. siRNA sequences were described previously (12).

Polyclonal anti-progastrin antibodies were produced and purified by Eurogentec as detailed under Supplementary Methods. We used ELISA to confirm their lack of binding to other peptides in the gastrin family, and their ability to reverse the effects of progastrin (Supplementary Fig. S2D) was validated on p38 MAPK phosphorylation (see Supplementary Methods).

In vivo experiments followed French guidelines for experimental animal studies (DSV agreement C34-172-27).

T84 shCT/shPG cells were pretreated with doxycycline (1 μg/mL) 48 hours before injection. ALDH^high cells were resuspended in DMEM/Matrigel (v/v) and injected subcutaneously into the right flank of female balb/C nude CD1 mice (Charles River Laboratories). For secondary injection, tumors were dissociated as described for human tumors, and a Ficoll gradient was used to remove dead cells/debris. Doxycycline (2 mg/mL) was provided in the drinking water 96 hours before injection and throughout the experiment. Tumors were measured twice a week using a caliper [volume = (length × width × thickness)/2]. Tumor samples were fixed in 4% paraformaldehyde and paraffin-embedded.

SW480 and CPP19 cells transfected with shPG or full-length human progastrin (as in ref. 12) were grown as monolayers in FBS-containing medium. ALDH^high cells were injected subcutaneously into the right flank of female balb/C nude CD1 mice, and tumors were measured as above.

Staining with ALDEFLUOR (STEMCELL Technologies), with CD44 (Miltenyi Biotec, 130-098-110), EpCAM (Miltenyi Biotec, 130-080-301), CD133 (Miltenyi Biotec, 130-098-829), LGR5 (BD, #562913), and Annexin V (Fluoroprobe) antibodies was performed according to manufacturer's instructions. Negative cells were respectively defined using the ALDH inhibitor diethylaminobenzaldehyde (DEAB) or isotype controls. Dead cells were excluded on the basis of 7-aminoactinomycin D (7-AAD, Invitrogen) exclusion.

Cellular bioenergetics [including oxygen consumption (OCR) and extracellular acidification rates (ECAR)] were determined on an XF24 Extracellular Flux Bioanalyzer (Seahorse Bioscience). Cells per well (n = 50,000) were seeded into 24-well XF24 cell culture microplates and grown overnight. Three wells containing medium alone were used as background. One hour before the assay, growth medium was replaced with unbuffered assay medium (pH = 7.4, Seahorse Bioscience) containing 25 mmol/L glucose (Sigma-Aldrich) and placed in a non-CO₂ incubator for CO₂ and temperature stabilization. ECAR and OCR were analyzed during a mitochondrial stress test using consecutive administration of oligomycin (2 μmol/L), FCCP (1.5 μmol/L), and rotenone/antimycin A (2 μmol/L). To determine the impact of glycolysis on progastrin-regulated self-renewal, ALDH^high/neg T84 cells were transfected with GAST- or βgal-selective siRNA, and their stem cell frequency was quantified using ELDA in the presence of DMSO (0.01%) or 0.05 μmol/L antimycin A.

Statistical tests were performed using GraphPad Prism Version 6, using Mann–Whitney, ANOVA, or Student t test for in vitro studies, with appropriate post hoc tests for multiple comparisons. In vivo survival curves were derived from Kaplan–Meier estimates, and the curves were compared by log-rank tests. A 5% cutoff was used to validate result significance.

We first used colorectal cancer cell lines and cells isolated from patients with colorectal cancer to determine whether progastrin expression was detected in conditions that promote CSC function in vitro. While expression of the progastrin-encoding gene (GAST; not shown) as well as progastrin secretion (Fig. 1B) were detectable when cells were grown as monolayers in serum-supplemented medium (2D), both were strongly elevated in cells grown as colonospheres (3D; Fig. 1A and B).

As an alternative, we enriched CSCs based on their elevated aldehyde dehydrogenase (ALDH) activity. ALDH is a detoxifying enzyme involved in the resistance to alkylating enzymes. Similar to other cancers (25), selection based on its activity leads to enrichment of functional CSCs in colorectal cancer, as ALDH^high colorectal cancer cells repeatedly generate colonospheres in vitro and tumors in immunocompromised animals (26, 27). ALDH^high colorectal cancer cell lines and patient-derived colorectal cancer cells displayed enriched ALDH1A1 mRNA levels (Supplementary Fig. S1A and S1B) and higher levels of GAST mRNA (Fig. 1C) than ALDH^neg cells, although the latter clearly expressed detectable levels of GAST mRNA compared with human fibroblasts and glioblastoma cells, where no specific GAST signal was detected using RT-qPCR (Supplementary Fig. S1C).

To further confirm that CSCs expressed elevated levels of progastrin, we quantified GAST mRNA in colorectal cancer cells sorted using other proposed CSC markers. Cells enriched using CD44/EpCAM (28) or LGR5 (29) expressed significantly higher GAST mRNA compared with marker-negative populations (Supplementary Fig. S1D–S1G). GAST expression was not consistently higher in cells enriched using CD133/1 (Supplementary Fig. S1H–S1K), a more controversial CSC marker in cultured colorectal cancer cells (21, 22). The percentage of CD133/1–positive cells was higher in colorectal cancer cell lines than in most patient-derived cells (Supplementary Fig. S1I), and GAST mRNA was elevated in CD133^high cells only in samples where CD133 expression enriched a small cell subpopulation (CPP1, Supplementary Fig. S1I–S1K). These results strongly suggest that progastrin expression is increased in colorectal CSCs, since growing cells as colonospheres or selecting ALDH^high (Supplementary Fig. S1L and S1M; refs. 27, 30), CD44^high/EpCAM^high (28) or LGR5^high (29) tumor cells enriches cells with functional CSC characteristics. Accordingly, progastrin was highly expressed in ALDH^high cells and in areas surrounding them within colorectal cancer samples (Fig. 1D). These findings suggest that progastrin secretion is not an exclusive characteristic of colon CSCs but is nevertheless higher in these cells than in their non-CSC counterparts.

To determine whether CSCs could respond to progastrin as part of an autocrine and/or paracrine loop, we determined the impact of progastrin on the colonosphere formation ability of T84 and CPP1 cells. Colonosphere formation was strongly impaired following transfection with a GAST gene–specific siRNA (Fig. 2A and Supplementary Fig. S2A and S2B), with a 3- to 4-fold reduction of CSC frequency quantified using ELDA (Fig. 2B)), and this effect was reversed when GAST siRNA-treated cells were incubated with recombinant human progastrin (Fig. 2C and D and Supplementary Fig. S2C and S2D). The size of remaining colonospheres was smaller in progastrin-depleted cells than in controls (Supplementary Fig. S2B), corroborating other studies demonstrating proproliferative and antiapoptotic effects of progastrin (19, 31). Similar results were obtained upon neutralization of progastrin using selective polyclonal antibodies (Fig. 2E and F and Supplementary Fig. S2D and S2E) and upon inducible shRNA-mediated progastrin downregulation (Supplementary Fig. S2F and S2G), the former indicating that this effect is linked to an autocrine/paracrine function of the peptide.

Serial sphere passaging demonstrated that siRNA-mediated progastrin depletion performed during the initial passage induced a long-term impairment of the sphere-forming potential of colorectal cancer cells when compared with controls (Fig. 2G and H), indicating that self-renewal of sphere-forming colorectal cancer cells is severely altered in the absence of progastrin.

As our results and those of others point to ALDH activity as a functional marker to enrich stem-like cells in colorectal cancer, we assessed whether the effect of progastrin on self-renewal was due to a direct regulation of the ALDH^high cell subpopulation. siRNA-mediated progastrin downregulation significantly decreased ALDH^high cell proportions within colorectal cancer cell lines and patient-derived cells (Fig. 3A and B), suggesting that progastrin is essential to stabilize the proportion of CSC-like cells under homeostatic conditions.

The recent literature suggests that stochastic transitions between non-stem and stem-like cells contribute to the maintenance of phenotypic equilibrium within cancer cell populations (8, 32). To gain further insight concerning the effect of progastrin on ALDH^high CSC maintenance, we analyzed the individual capacity of ALDH^high and ALDH^neg cells to regenerate population heterogeneity when grown under nonchallenging adherent culture conditions. Purified ALDH^high cells generated populations comprising similar homeostatic proportions of ALDH^high and ALDH^neg cells to those found in their parental cells within 7 days of culture (Fig. 3C). The proportion of each subpopulation once equilibrium had been reached was regulated by progastrin, with proportions of ALDH^high cells significantly increased in cells incubated with recombinant progastrin and decreased upon treatment with a progastrin-neutralizing antibody (Fig. 3C). This result suggests that progastrin hampers or slows down the differentiation of self-renewing ALDH^high cells toward ALDH^neg colorectal cancer cell phenotypes.

Purified ALDH^neg cells gave rise to ALDH^high subpopulations, although without reaching the initial equilibrium proportions by 7 days (Fig. 3C). This result suggests that ALDH^high cells can be generated from ALDH^neg colorectal cancer cells grown under conditions that facilitate cell survival, albeit at significantly slower rate than the ALDH^high to ALDH^neg conversion. In any case, the rate of transition from ALDH^neg to ALDH^high within 7 days of FACS-mediated purification was unaffected by recombinant progastrin or by progastrin-neutralizing antibodies (Fig. 3C), suggesting that progastrin did not regulate the ability of ALDH^neg cells to generate ALDH^high cells within this timeframe.

To assess whether this lack of effect on ALDH^neg cells was paralleled by a differential ability of progastrin to engage signaling in these cells, we analyzed the ability of progastrin to regulate the phosphorylation of p38MAPK, which plays a role in regulating CSC characteristics such as drug resistance (33) and is phosphorylated in response to progastrin (34, 35). p38MAPK phosphorylation was promoted by recombinant progastrin and inhibited by progastrin-neutralizing antibodies in ALDH^high T84 and CPP19 cells but remained unaffected in ALDH^neg cells (Supplementary Fig. S3A), suggesting that ALDH^neg colorectal cancer cells do not respond or differentially respond to progastrin compared with ALDH^high cells. Taken together, these results suggest that progastrin strongly promotes the maintenance of the ALDH^high tumor-initiating phenotype without promoting the transition of ALDH^neg tumor cells into stem-like cells.

As progastrin regulated homeostatic proportions of ALDH^high stem-like cells without affecting the ability of ALDH^neg cells to revert to an ALDH^high phenotype, we focused on ALDH^high cells to determine whether progastrin could directly regulate their self-renewal and/or survival.

Endogenous progastrin depletion in ALDH^high cells or treatment with recombinant progastrin, respectively, decreased or increased their sphere-forming ability (Fig. 3D and Supplementary Fig. S3B), suggesting that progastrin promotes the self-renewal of these cells. In contrast, incubation with recombinant human progastrin or with medium conditioned by ALDH^high cells was insufficient to allow colonosphere formation by ALDH^neg cells, again suggesting that progastrin alone is insufficient to drive their phenotype toward functional CSCs or that ALDH^neg cells are altogether not sensitive to progastrin.

We then investigated whether progastrin also promoted the survival of ALDH^high colorectal CSCs. Since the proportion of ALDH^high cells was already strongly decreased 4 days after transfection of colorectal cancer cells with progastrin-selective siRNA (see Fig. 3A and B), we quantified the proportion of ALDH^high cells undergoing apoptosis 72 hours after transfection with control or progastrin-selective siRNA. The proportion of Annexin V–positive cells was significantly increased following transfection with progastrin siRNA compared with controls (35.7 ± 1.2% vs. 18.3 ± 1.5% respectively, n = 4; Fig. 3E), suggesting that progastrin is instrumental to maintain ALDH^high stem-like cell survival.

To determine whether, beyond decreasing the proportion of ALDH^high cells in colorectal cancer cell populations, progastrin downregulation also inhibited their self-renewal in vivo, immunocompromised mice were xenografted with ALDH^high T84 cells expressing doxycycline-inducible control (shCT) or progastrin-targeting (shPG) shRNA constructs. Efficiency of progastrin-selective shRNAs to downregulate GAST mRNA expression (Supplementary Fig. S2D), TIC frequency (Supplementary Fig. S2E), and ALDH^high cell proportions (Supplementary Fig. S4A and S4B) was confirmed in vitro. Live cells that remained ALDH^high following a 48-hour doxycycline induction of progastrin-selective shRNA were injected (250 or 25 cells/100 μL) into the flank of balb/C nude CD1 mice (Fig. 4A). Tumors appeared later in mice injected with 250 ALDH^high shPG cells compared with controls (median tumor-free survival: shPG, 31 days; shCT, 23 days) but tumor incidence was similar (6 of 6) in both groups 50 days postinjection (Fig. 4B). In contrast, tumor incidence was lower in mice injected with 25 ALDH^high shPG cells (1 of 6 mice) than in control ALDH^high cells (4 of 6 mice; Fig. 4B), and tumor growth was reduced by progastrin depletion (Fig. 4D). GAST mRNA levels were lower in tumors collected from the shPG group at day 50, compared with controls (shCT; Supplementary Fig. S4C). To assess whether the diminished tumor initiation potential of ALDH^high cells following progastrin downregulation leads to long-term decreased self-renewal, xenografted tumors from mice injected with 250 shCT or shPG ALDH^high cells were collected at day 50 and enzymatically dissociated. Following Ficoll gradient–based live cell isolation, 25 or 250 unsorted cells were reinjected in new balb/C nude CD1 mice (Fig. 4A). Second-generation xenografts appeared later than their first-generation counterparts that were generated from ALDH^high cells (median tumor-free survival times: 60 and 39 days in 25 shCT and 250 shCT second-generation groups - Fig. 4C, compared with 31 and 23 in first-generation controls; Fig. 4B). Zero of 6 and 2 of 6 tumors were detected up to 98 days after reinjection of 25 and 250 shPG cells, respectively, in comparison with 5 of 6 and 6 of 6 tumors in animals injected with 25 and 250 shCT cells (Fig. 4C). The tumor-initiating potential remained severely impaired in shPG cells by 100 days after secondary injection (i.e., before mice from the shCT group had to be culled on ethical grounds), with a greater than 47-fold reduction of estimated TIC frequency in the shPG group (1 in every 693) compared with controls (1 in every 14.5; P = 8.46E⁻⁰⁷). These results indicate that progastrin downregulation leads to a robust and long-term reduction of tumor-initiating frequency in human colorectal cancer cells and suggest that an impairment of ALDH^high cell CSC potential is instrumental to this process. To confirm that progastrin indeed promotes tumor-initiating properties of colorectal cancer cells, we performed additional experiments using the SW480 colorectal cancer cell line and CPP19 patient-derived cells made to overexpress full-length human progastrin. Although these cell lines were less tumorigenic than T84 cells at low concentrations, tumor incidence was increased for both cell types in progastrin-overexpressing cells compared with controls (Supplementary Fig. S4).

To gain mechanistic insight into the regulation of ALDH^high cells by progastrin, we compared the mRNA expression profile of ALDH^high cell–derived T84 colonospheres collected 72 hours after transfection with GAST or β-galactosidase–selective siRNA. The expression of 140 genes was significantly altered by the experimental downregulation of progastrin (92 upregulated and 48 downregulated, Supplementary Table S1). Clustering of all samples strongly separated GAST-siRNA from control siRNA samples and delineated an expression gradient along the three time points (6, 24, and 72 hours) for each type of siRNA, indirectly validating the robustness and biologic significance of this gene list (Supplementary Fig. S5A). Ontology analysis of this gene list using the DAVID functional annotation tool indicated that progastrin regulates many genes involved in metabolic processes (including reduction/oxidation and fatty acid metabolism) and genes, notably histones, involved in chromatin organization (Supplementary Table S2).

In view of the demonstrated role of metabolic processes in the regulation of stem cell and CSC self-renewal (36), we further explored consequences of GAST downregulation on the metabolic activity of ALDH^high CSCs. Using transmission electron microscopy or a selective antibody directed against a human mitochondrial protein, we found that the number of mitochondria was increased in colonospheres treated with GAST siRNA (Fig. 5A), as well as in residual tumor xenografts generated from shPG cells (Fig. 5B) compared with controls. We used a Seahorse XF bioanalyzer to establish whether this resulted in modifications of cellular bioenergetics within enriched ALDH^high or ALDH^neg cells. ALDH^high cells displayed significantly higher basal and maximal glycolysis (as measured via ECAR) than ALDH^neg cells (Fig. 5C and Supplementary Fig. S5B), whereas their basal mitochondrial OCR were similar. Taken together these results demonstrate a cellular metabolic switch toward glycolysis in ALDH^high cells. Interestingly, while there was no difference in basal mitochondrial oxygen consumption, ALDH^neg cells exhibited a significantly greater oxygen reserve, as demonstrated via a two-fold increase in maximal oxygen consumption (Fig. 5C and Supplementary Fig. S5B).

siRNA-mediated progastrin downregulation strongly affected the glycolytic profile of ALDH^high cells, decreasing both basal and maximal glycolysis to levels equivalent to that observed in ALDH^neg cells. In contrast to glycolysis, siPG treatment of ALDH^high cells had no effect on overall mitochondrial activity (Fig. 5D and Supplementary Fig. S5B). Incubation with progastrin-neutralizing antibodies also significantly inhibited ALDH^high cell glycolysis in the HT-29 cell line, and a similar trend was observed in HCT116 cells (Supplementary Fig. S5C and S5D). The expression of mRNA encoding lactate dehydrogenase B (LDH-B), hexokinase 2 (HK2) and to a lesser degree the ubiquitous glucose transporter GLUT1 were significantly decreased in ALDH^high cells. In contrast, the tricarboxylic acid (TCA) cycle regulator isocitrate dehydrogenase 1 (IDH1) was significantly increased (Fig. 5D). Finally we used the oxidative phosphorylation inhibitor antimycin A to prevent the decrease in glycolysis and the increase in oxidative phosphorylation induced by the PG siRNA in ALDH^high cells. Antimycin A (0.05 μmol/L) prevented the inhibition of sphere-forming ability induced by progastrin downregulation in these cells (Supplementary Fig. S5E), suggesting that regulating mitochondrial metabolism was instrumental to the promoting role of progastrin on ALDH^high CSC self-renewal. Taken together these results indicate that progastrin promotes the ability of ALDH^high cells to maintain a pattern of mitochondrial metabolism that is reminiscent of that characterizing CSCs (36).

This study demonstrates that progastrin, secreted by human colorectal tumors (10, 11), promotes the survival and self-renewal of cancer stem cells. It also shows that progastrin inhibition in ALDH^high CSCs decreases their glycolytic profile down to levels found in non-CSCs, strongly reducing their tumor-initiating ability in vitro and in vivo.

CSCs are partly responsible for the phenotypic and functional intratumor heterogeneity in several human cancers including colorectal carcinoma (37). The presence of distinct cell subpopulations with different levels of tumorigenicity within colorectal tumors was first reported a few years ago (38, 39), and the involvement of CSCs in colorectal tumor progression and recurrence has been documented since in murine and human models (7, 40, 41). Recent studies indicate that functional characteristics of these cells are not solely driven by intrinsic cues but are also under the regulation of signals coming from the microenvironment (6), including from tumor cells themselves (7). The current study highlights the important role of progastrin in this process.

Progastrin overexpression and its proliferative and tumor-promoting effects in colorectal cancer have been well-documented (12, 42, 43) and previous studies linked progastrin production with the expression of CD133, DCAMKL1, or CD44 (17–19), markers that enrich but are not specific for CSCs (20–23). Importantly, these studies did not provide functional data demonstrating a role for progastrin in regulating CSCs hallmarks such as self-renewal. Our study is therefore the first to provide a functional demonstration that the tumor-initiating potential of self-renewing CSCs in advanced colorectal cancer falls under the regulation of progastrin in vitro and in vivo, including in colorectal cancer cells directly isolated from patient samples.

Our results indicate that colorectal CSCs and, to a lesser degree, other tumor cells can secrete progastrin. In turn, progastrin promotes the self-renewal and tumor-initiating ability of CSCs contained within the ALDH^high cell populations. Progastrin secretion may thus form an integral part of the microenvironmental niche that fosters promoting conditions for tumor initiation and self-renewal at stages of the carcinogenic process when stem-like capabilities provide a selective advantage to tumor cells. This hypothesis is corroborated by recent results demonstrating that progastrin overexpression increases the number of Lgr5⁺ cells in an early carcinogenesis mouse model (13) and promotes the proliferation of colonic progenitors by promoting IGF2 secretion by myofibroblasts (44). Similarly, secretion by tumor cells of factors that promote their own proliferation or CSC self-renewal has been reported in the case of IL4 (7) and TGFβ (45), raising interest in these factors as potential therapeutic targets or prognostic biomarkers. Accordingly, in view of the reported role of CSCs during early stages of tumor initiation, the promotion of self-renewal identified here provides a potential rationale for the proposed role of progastrin overexpression as a risk marker of neoplastic transformation in patients with hyperplastic colonic polyps (46).

Homeostatic equilibrium within heterogeneous tumor cell populations involves proliferation and cell death rates of CSCs and non-CSCs, as well as rates of phenotypic transitions between the CSC and non-CSC states (8, 32, 47). Using ALDH activity as a marker to enrich functional CSCs, our results indicate that progastrin promotes ALDH^high CSC survival and self-renewal. In contrast, although we found that ALDH^neg cells were able to generate ALDH^high cells when grown under non-challenging conditions, progastrin did not control this phenotypic plasticity or confer any CSC-like characteristics to ALDH^neg cells within the timeframe of our experiments (7–10 days). The lack of detectable effect on ALDH^neg cells suggests that these cells are likely to be poorly or not sensitive to progastrin. Alternatively, it is also possible that progastrin induces differential signaling events leading to other biologic functions, unexplored in the current study, in ALDH^neg cells.

The receptor mediating these effects of progastrin on CSCs remains to be characterized. Previous studies suggested that the ubiquitous protein Annexin II may represent a receptor for progastrin (48). Although our results do not rule out this possibility, we found that expression of Annexin II was decreased or stable under culture conditions where self-renewal is enhanced, such as in colonosphere assays or in ALDH^high subpopulations (Supplementary Fig. S1M and S1N) when compared with controls. In addition, convincing data from Jin and colleagues suggested that progastrin can act at least partly in vivo via activation of CCK2 receptors (16). Yet, we were not able to detect CCK2 receptor mRNA in the colorectal cancer cells used to characterize the effects of progastrin in this study (data not shown), suggesting that progastrin may display both direct and indirect effects and/or that the progastrin signaling networks in mice and human are somewhat different.

Finally, we found that progastrin inhibition using siRNA or neutralizing antibodies induces a metabolic shift in ALDH^high cells, driving them away from their highly glycolytic profile and reinstating mitochondrial oxidative phosphorylation. This shift was not only concomitant with but also essential for the decrease in their self-renewal ability driven by progastrin inhibition. Several studies suggested that glycolytic pathways are enhanced within CSCs (49, 50), leading to the recently coined concept of “metabostemness” (51). Targeting metabolic pathways to restore the predominance of oxidative phosphorylation may thus be a useful therapeutic strategy against CSCs. Indeed, a recent study (52) convincingly demonstrated that metabolic reprogramming (defined as a shift toward glycolysis and a decrease in mitochondrial numbers) is essential to the self-renewal of nasopharyngeal carcinoma CSCs, and our results demonstrate that similar reprogramming may be controlled in colon CSCs by autocrine/paracrine progastrin secretion.

In summary, this study suggests that inhibiting progastrin decreases the proportion of CSCs within colorectal cancer cell populations and reduces the tumor-initiating potential of residual CSCs. Considering the enrichment of these cells upon exposure of colorectal cancer cells to chemotherapy and their role in posttreatment recurrence (40, 53), these result suggest that targeting progastrin could provide significant therapeutic benefit to colorectal cancer patients undergoing antiproliferative therapy.

F. Hollande reports receiving commercial research grant and was a consultant/advisory board member of Servier Pharma and had ownership interest (including patents) in Biorealites sas. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J. Giraud, L. Houhou, D. Joubert, J. Pannequin, F. Hollande

Development of methodology: J. Ollier, C. Ya, J. Pannequin, F. Hollande

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Giraud, L.M. Failla, J.-M. Pascussi, E.L. Lagerqvist, J. Ollier, F. Bertucci, I. Gasmi, J.-F. Bourgaux, M. Prudhomme, T. Mazard, I. Ait-Arsa, D. Birnbaum, A. Pelegrin, C. Vincent, J.G. Ryall

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Giraud, L.M. Failla, J.-M. Pascussi, P. Finetti, F. Bertucci, J.G. Ryall, J. Pannequin, F. Hollande

Writing, review, and/or revision of the manuscript: J. Giraud, J.-M. Pascussi, F. Bertucci, J.G. Ryall, D. Joubert, J. Pannequin, F. Hollande

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Ollier, C. Ya, C. Vincent

Study supervision: J. Pannequin, F. Hollande

Other (provided reagents and analytic tools): A. Pelegrin

The authors thank P. Crespy and C. Planque for their support with in vivo experiments, F. Charrier-Savournin for her help with p38MAPK assays, C. Duperray for flow cytometry, and C. Cazevieille for electron microscopy experiments, F. Gerbe for helpful discussions and L. Boudier, C. Pfeiffer, B. Vire, and D. Greuet for excellent technical assistance. The authors also thank The Centre Regional d'Imagerie Cellulaire (CRIC) for supporting this work.

This work was supported by the French National Research Agency (ANR, PACCT grant), the Association pour la Recherche contre le Cancer (ARC), and by the Department of Pathology, University of Melbourne. J. Giraud was supported by scholarships from the MRT and from FRM. F. Hollande is supported by the National Health and Medical Research Council (NH&MRC) of Australia (#1049561).

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