Monoclonal Antibodies Against Lgr5 Identify Human Colorectal Cancer Stem Cells§

Authors


  • Author contributions: K.K.: conception and design, collection and/or assembly of data, data analysis and interpretation, and manuscript writing; W.D.L.: conception and design and provision of study material; P.R.P.: provision of study material, collection and/or assembly of data, and data analysis and interpretation.; H.R.: collection and/or assembly of data; H.C.: conception and design and final approval of manuscript; J.P.M.: conception and design, data analysis and interpretation, and final approval of manuscript.

  • Disclosure of potential conflicts of interest is found at the end of this article.

  • §

    First published online in STEM CELLSEXPRESS August 7, 2012.

Abstract

In colorectal cancer (CRC), a subpopulation of tumor cells, called cancer stem cell (CSC) fraction, is suggested to be responsible for tumor initiation, growth, and metastasis. The search for a reliable marker to identify these CSCs is ongoing as current markers, like CD44 and CD133, are more broadly expressed and therefore are not highly selective and currently also lack function in CSC biology. Here, we analyzed whether the Wnt target Lgr5, which has earlier been identified as a marker for murine intestinal stem cells, could potentially serve as a functional marker for CSCs. Fluorescence-activated cell sorting-based detection of Lgr5, using three newly developed antibodies, on primary colorectal tumor cells revealed a clear subpopulation of Epcam+Lgr5+ cells. Similarly, primary CRC-derived spheroid cultures, known to be enriched for CSCs, contain high levels of Lgr5+ cells, which decrease upon in vitro differentiation of these CSCs. Selection of the Lgr5high CRC cells identified the clonogenic fraction in vitro as well as the tumorigenic population in vivo. Finally, we confirm that Lgr5 expression is dependent on the Wnt pathway and show that Lgr5 overexpression induces clonogenic growth. We thus provide evidence that Lgr5 is, next to a functional intestinal stem cell marker, a selective marker for human colorectal CSCs. STEM CELLS2012;30:2378–2386

INTRODUCTION

The driving force behind tumor growth is thought to reside in the cancer stem cells (CSCs), also referred to as tumor-initiating cells. CSCs are thought to be responsible for metastasis and relapse after therapy. These cells form a minority of the tumor cells in colorectal cancer (CRC) and can be identified using markers like CD133 [1–3], CD24/29 [4], or a combination of CD44 with CD166 [5]. Although these markers can identify CSCs, their function in stem cell biology remains unclear and at this point they only serve as biomarkers [6]. A search for a functional CSC marker is therefore warranted. Recently, we found that colon CSCs contain elevated Wnt signal transduction as compared to their more differentiated progeny and can be isolated using a Wnt-reporter construct [7]. Enhanced Wnt signaling in CSCs is a result of elevated nuclear β-catenin localization, which is mirrored in primary tumor samples that display heterogeneous nuclear β-catenin localization as well [8, 9]. Even though this shows that Wnt signal intensity can be used to identify CSCs, isolation of viable tumor cells from a primary tumor based on the localization of β-catenin is not feasible.

A potentially more successful approach to isolate CSCs is to use cell surface-expressed Wnt target genes. Previously, we described the seven-transmembrane protein Lgr5, also referred to as GPR49, as a Wnt target that could reliably identify normal intestinal stem cells [10]. Using lineage tracing, Lgr5-expressing cells can be shown to differentiate into all cell types, such as goblet cells, entero-endocrine cells, and Paneth cells, present in the intestinal tract [10] and thereby qualify as intestinal stem cells. Although we initially identified Lgr5 in a screen for Wnt targets [11], Lgr5 also enhances the canonical Wnt pathway as it serves as the receptor for R-spondin1-4, which, by binding to Lgr5, enhances LRP6 phosphorylation [12, 13]. More recent evidence indicates that R-spondin not only binds Lgr4 and Lgr5 but also at the same time interacts with the cell surface E3-ligases ZNRF3 and RNF43, which regulate the cell surface expression of the Wnt receptor complex [14]. Binding of R-spondin to Lgr4/5 and the E3-ligase induces their internalization and thereby prevents the inhibitory ubiquitilation of LRPs [15]. The E3 ligases therefore serve as inhibitors of the Wnt pathway and in agreement we observed that deletion of both enzymes in the intestinal stem cell leads to intestinal hyperplasia and adenoma formation [15]. The intestinal stem cell marker Lgr5 thus serves as a marker for intestinal stem cells but also has a functional role in facilitating a pivotal stem cell signal transduction pathway.

Several papers have suggested that Lgr5 can also serve as a CSC marker [4, 7, 16, 17], but evidence to directly support this claim is lacking. Lgr5 is one of the Wnt targets that is differentially expressed on mRNA between CSCs and nontumorigenic differentiated tumor cells [7]. In addition, we confirmed that spheroid cultures derived from primary CRC express Lgr5 protein when analyzed by immunofluorescent staining [4]. However, to determine the validity of Lgr5 as a CSC marker, functional antibodies that recognize the extracellular part of Lgr5 are needed. So far, several commercial antibodies are available, but none of them seemed suitable for fluorescence-activated cell sorting (FACS) analysis, which is necessary to sort viable cells and to provide evidence for the presence of CSCs in in vitro and in vivo assays. Here, we show that with three monoclonal antibodies developed against human Lgr5, we can identify Lgr5 on primary tumors as well as spheroids CSC cultures. Our data show that the Lgr5high cells represent the clonogenic fraction as evidenced in in vitro limiting dilution assays (LDA) and are more tumorigenic in vivo. Moreover, overexpression or knockdown of Lgr5 enhances or decreases clonogenicity, respectively. Altogether, we show that Lgr5 is a functional CSC marker in human CRC.

MATERIALS AND METHODS

Antibody Development

Rat monoclonal antibodies were raised against full-length human LGR5 protein. Mapping and testing of the antibodies is described in [13]. Shortly, affinity for human Lgr5 was tested with LGR5-exo-Fc. In addition, epitope mapping was performed using clones with C-terminal deletions and hybrids of human–mouse Lgr5 exodomain. Three antibodies passed all tests: 1D9, 4D11, and 9G5, from now on referred to as Ab1, Ab4, and Ab9. These were biotinylated to facilitate detection with streptavidin-based approaches.

Isolation of Primary Colorectal Tumor Cells

Tumor samples were obtained under standard medical ethical procedures of the Academic Medical Center and were cut in small pieces and enzymatically dissociated by incubation with 1.5 mg/ml of collagenase II (C6885, Sigma-Aldrich, St Louis, MO, http://www.sigmaaldrich.com) and 20 μg/ml of hyaluronidase (H4272, Sigma-Aldrich, St Louis, MO, http://www.sigmaaldrich.com) at 37°C. Removal of nondissociated parts was achieved by straining the cells through a 70-μm strainer.

Tissue Culture

Colorectal cell line LS174T was cultured in RPMI-1640 (Gibco-Invitrogen, Carlsbad, CA, http://www.invitrogen.com) supplemented with 8% fetal calf serum (FCS), 2 mM glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin. As control, these LS174T were transfected with pcDNA3.1-Lgr5 and a stable line was created by selecting for the expression of the transgene with 1 mg/ml G418. CSC spheroid cultures were maintained as described previously [4]. CSCs were differentiated in vitro as described [18]. The previously described T-cell factor/lymphoid enhancer factor (TCF/LEF) reporter driving expression of green fluorescent protein (GFP) (TOP–GFP) was a gift from Laurie Ailles [19] and was transduced into the CSC cultures as described previously [7]. For glycogen synthase kinase 3 (GSK3) inhibition treatment, CSC cultures were freshly plated with the addition of 4 μM CHIR99021 (Axon Medchem, Groningen, The Netherlands, http://www.axonmedchem.com) or dimethyl sulfoxide (DMSO) and analyzed after 48 hours.

Staining, FACS analysis, Sorting, and LDA

Cells were washed in phosphate buffered saline (PBS) containing 1% bovine serum albumin and 0.02% sodium azide, after which they were stained simultaneously with the three different biotin-labeled Lgr5 antibodies (Ab1, Ab4, and Ab9) with a concentration of 1 μg/ml in PBA for 20 minutes at 4°C. As isotype control, a biotin-anti-rat IgG2b (553898l, BD Biosciences, San Diego, CA, http://www.bdbiosciences.com) was used at the same concentration. After washing with PBA, cells were stained with streptavidin-allophycocyanin (APC) (17431782, eBioscience, San Diego, CA, http://www.ebiosciences.com), diluted 1:500, for 20 minutes at 4°C. Double staining was done with EpCAM-fluorescein isothiocyanate (FITC) (F0860/Ber-EP4, Dako), AC133 (130090826, Miltenyi Biotec), or CD24 (555428, BD Biosciences, San Diego, CA, http://www.bdbiosciences.com). Dead cells were excluded by propidium iodide (PI) (1 μg/ml). FACS analysis was performed on the FACSCanto (BD Biosciences), while data analysis was performed in Flowjo. FACS sorting for the in vitro and in vivo LDA was performed on the FACSAria (Becton Dickinson, Franklin Lakes, NJ, http://www.bd.com). For the in vitro LDA, cells were sorted either based on their Wnt-reporter (TOP-GFP) activity or on their Lgr5-positivity. Cells were deposited in a 96-well plate as 1, 2, 4, 8, 10, 12, 16, 20, and 24 cells per well in at least eight wells per condition. After 2 weeks, the presence of spheroids in each well was scored. Clonogenic fraction was calculated by extreme limiting dilution analysis (ELDA) using http://bioinf.wehi.edu.au/software/elda/index.html.

Immunofluorescent Staining

LS174T and LS174T-Lgr5 were plated overnight in a 48-well plate. Next day, cells were washed in PBS and fixed in 4% paraformaldehyde (PFA) for 20 minutes at 37°C. After washing with PBS, aspecific binding was blocked by 10% FCS in PBS for 1 hour at RT. Lgr5 or isotype control antibodies (1 μg/ml) were incubated for 1 hour at RT in blocking buffer, followed by an incubation of 1 hour at RT with Alexa546-streptavidin (1:1,000, S12225, Molecular probes, Eugene, OR, http://probes.invitrogen.com) as a secondary antibody. DAPI was used to counterstain for the nuclei. Pictures were taken on an Axiovert 200-M microscope (Carl Zeiss, Cambridge, UK, www.carl-zeiss.co.uk). CSCs maintained in Matrigel were prepared for immunofluorescent staining by first fixing them for 10 minutes in 4% PFA followed by permeablization with 0.5% Triton X-100 in PBS at 4°C for 10 minutes. After multiple wash with PBS, Matrigel was incubated with primary antibody overnight at 4°C. The following antibodies were used: cytokeratin 20 [CK20 (Genetex, Irvine, CA, http://www.genetex.com)], Lgr5-epitope C terminus (Genetex, Irvine, CA, http://www.genetex.com), and Mucin2 (Santa Cruz Biotechnology, Santa Cruz, CA, http://www.scbt.com). Matrigel was subsequently washed three times and incubated for 1 hour RT with Alexa 488-anti IgG (Invitrogen, Carlsbad, CA, http://www.invitrogen.com). 4′,6-diamidino-2-phenylindole (DAPI( was used for nuclei counter staining and pictures were taken on Leica TCS-SP2 confocal microscopy.

RNA and Polymerase Chain Reaction

RNA was extracted from LS174T and the CSC cultures with Trizol reagent (Invitrogen) according to the manufacturer's protocol. After checking quality and quantity of the RNA by NanoDrop ND-1000, cDNA was prepared with reverse transcriptase III (Invitrogen) according to manufacturer's protocol. Quantitative polymerase chain reaction (qPCR) was performed on the LC480 II (Roche, Basel, Switzerland, http://www.roche-applied-science.com) with the following primers: Lgr5-F: 5′-AATCCCCTGCCCAGTCTC-3′, Lgr5-R: 5′-CCCTTGGGAATGTATGTCAGA-3′, GAPDH-F: 5′-AAGGTGAAGGTCGGAGTCAAC-3′, GAPDH-R: 5′-TGGAAGATGGTGATGGGATT-3′.

In Vivo Mouse Experiments

Mice were maintained and experimented on in accordance with the guidelines of and after approval from the Animal Ethical Committee (DEC) of the Academic Medical Institute. After sorting, cells were mixed 1:1 with growth factor-reduced matrigel (BD Biosciences) and injected subcutaneously into male athymic BALB/cOlaHsd-Foxn1nu mice (Harlan Laboratories, AN Venray, the Netherlands. http://www.harlan.com) in a LDA. Per group, eight mice were included which were injected with either 1,000, 100, or 10 Lgr5high or Lgr5low cells.

Establishment of LGR5 Knockdown Cells

To generate stable knockdown of Lgr5, the lentiviral vector pLKO.1-TRC containing the following oligos were used: TRCN0000011586/shB4 5′-CCGGCCATCC AATTTGTTGG GAGATCTCGA GATCTCCCAA CAAATTGGAT GGTTTTT-3; TRCN0000011585/shB5 5′-CCGGCCATAG CAGTTCTGGC ACTTACTCGA GTAAGTGCCA GAACTGCTAT GGTTTTT. Lentiviral particle was produced by transiently transfecting 293T cells with pVSV, pMDL, and pREV. After 48 hours of transfection, the supernatant was filtrated and then used to transduce CSC spheroid cultures in the presence of 8 mg/ml polybrene. Following 16 hours of incubation with virus, the medium was changed, and the cells were allowed to grow in CSC medium for 2 days. On day 3, the selection was started with 1 μg/ml puromycin for 7 days. Knockdown efficiency was analyzed by qPCR for Lgr5.

RESULTS

Testing Monoclonal Lgr5 Antibodies

Rat monoclonal antibodies directed against human Lgr5 were developed and their respective epitopes were mapped as described in [13]. These antibodies (Ab1, Ab4, and Ab9) can detect three different epitopes on the extracellular domain [13]. To test if these antibodies could recognize Lgr5 on CRC cells, we made use of the cell line LS174T, as this line endogenously express Lgr5 mRNA. Lgr5-transfected LS174T served as a positive control. Overexpression of Lgr5 in the stable LS174T-Lgr5 line was confirmed by qPCR (Fig. 1C). To test the reactivity of the Lgr5 antibodies, the antibodies were first biotinylated and used to stain LS174T-Lgr5 cells. Staining was performed either with each of the Lgr5 antibodies individually or by all three antibodies simultaneously, to determine whether the combined use of these antibodies, which bind to different epitopes [13], could enhance reactivity. FACS analysis of Lgr5 on LS174T-Lgr5 with the separate clones resulted in a clear detection as compared to isotype control. Ab1 displayed the lowest activity, whereas Ab4 showed the highest staining intensity (Fig. 1A). Staining combining all three antibodies gave the highest intensity, confirming that these antibodies recognize different parts of the protein and give the best detection of Lgr5 when used together (Fig. 1A). To analyze if these antibodies could also detect endogenously expressed Lgr5, LS174T was used. FACS analysis gave similar results as obtained with LS174T-Lgr5: staining with Ab1 resulted in the lowest detection, Ab4 in the highest, but the triple staining gave a better result than any of the single stains (Fig. 1B). Although the intensity of staining was significantly lower in LS174T cells as compared to the Lgr5-overexpressing line, it is consistent with the level of mRNA present in these cells (Fig. 1C) and thus indicates that endogenously expressed Lgr5 can be detected with these antibodies. In addition, triple staining ensures maximum recognition of Lgr5.

Figure 1.

Testing monoclonal Lgr5 antibodies. Three newly developed antibodies were tested for staining of Lgr5. (A): Colorectal cancer cell line LS174T overexpressing Lgr5 (LS174T-Lgr5) was stained with the three Lgr5 either separately or together and detection was analyzed by fluorescence-activated cell sorting. The graph represents the MFI. Histograms represent isotype control staining profile (black line) versus specific staining profile as indicated (red shade). Staining with all three antibodies simultaneously gives the best detection efficacy. (B): Endogenously Lgr5 expressing cell line LS174T is stained with the three Lgr5 antibodies separately or together. MFI is shown in the accompanying graph. (C): Expression of Lgr5 mRNA, comparing LS174T and LS174T-Lgr5. (D): Immunofluorescent staining of LS174T and LS174T-Lgr5. Lgr5 staining is shown in red, nuclei are stained in blue. Bar indicates 20 μM. Abbreviations: APC, allophycocyanin; PI, propidium iodide; MPI, mean fluorescence intensity.

Next, we determined whether these antibodies could be used to detect Lgr5 on CRC in Western blotting or immunohistochemistry. Unfortunately, all three antibodies failed to detect Lgr5 by immunoblotting (data not shown). Immunofluorescence staining on both LS174T and Lgr5-transfected LS174T revealed that Lgr5 can be readily detected on Lgr5-overexpressing LS174T cells by Ab4 and Ab9, while Ab1 is not able to do this (Fig. 1D). The detection of Lgr5 on endogenously Lgr5-expressing LS174T cells is in general much lower, which is related to the low expression of the protein in these cells. However, it is still detectable, confirming that these antibodies can also be used for immunofluorescence staining.

Detection of Lgr5 on Human CRC Cells

The use of Lgr5 as a CSC marker for human CRC requires detection of the endogenous molecule on the surface of primary tumor cells. Pieces of primary colorectal tumors were retrieved from surgery specimens, enzymatically dissociated, and subsequently stained for the epithelial marker Epcam and for Lgr5. The Epcam+ cells in CRC specimens (Fig. 2A, upper panel) contained only a small subpopulation of Lgr5+ cells, varying between 1.9% and 11.1% (Fig. 2A, lower panels). This low percentage is consistent with previous observations which suggest that CSCs are shown to constitute only a small fraction of the total tumor [1, 3–5, 20, 21].

Figure 2.

Lgr5 expression on human colon cancer stem cells. (A): Primary tumor samples were dissociated and stained for Epcam and Lgr5 (all three antibodies together). Epithelial tumor cells were selected for further analysis by Epcam-positivity (upper panels), wherein the Lgr5+ fraction was determined (lower panels). Histograms represent isotype control staining profile (black line) versus specific staining profile as indicated (red shade). (B): Primary tumor samples were dissociated and cultured under spheroid conditions to enrich for CSCs. Detection of Lgr5 by the antibodies on these spheroid cultures showed that all cells expressed Lgr5. Abbreviations: APC, allophycocyanin; CSC, cancer stem cell; EpCAM, Epithelial cell adhesion molecule; FITC, fluorescein isothiocyanate.

Next, we determine whether the Lgr5+ tumor cells were indeed CSCs. Initial enrichment of CSCs can be performed by spheroid culture of primary CRC material. In short, primary tumor cells were dissociated and cells were cultured on nonadherent plates in special media supplemented with epithelial growth factor (EGF) and basic fibroblast growth factor (bFGF) [22]. We have previously shown that this method results in growth of tumor cells in spheroids and enriches for the CSCs as these spheroid cells express CSC markers, like CD133 and CD44 [4]. Although we cannot formally prove that the cells growing in the spheroid cultures are the cells that originally were identified as CSC marker-positive in the primary specimens, we have shown that such cultures only work when they are initiated with CSCs present [4], suggesting that we have selective expansion of preexisting CSCs. FACS staining with the monoclonal antibodies revealed that these spheroid cultures also express high levels of Lgr5+ cells (Fig. 2B), indicating that Lgr5 expression is also enriched in these CSC-containing cultures.

Lgr5 Is Downregulated During Differentiation

One of the prerequisites for a reliable CSC marker is its downregulation upon cellular differentiation. In vitro differentiation of CSCs can be achieved by plating the tumor cells, retrieved from the primary spheroid cultures, on adherent plates, removing EGF and bFGF and adding FCS. After a week, adhered tumor cells express differentiation markers like intestinal alkaline phosphatase, mucin2 (Muc2) and CK20, while they had lost expression of CSC markers, like CD133 [18]. We hypothesized that Lgr5, as a CSC marker, should also be downregulated during differentiation.comparing Lgr5 staining of spheroid cultures to differentiated cells confirmed this hypothesis: Lgr5 is highly expressed in the spheroid cultures, but its expression is decreased upon in vitro differentiation (Fig. 3A). This decrease in Lgr5 expression was confirmed by qPCR and immunohistochemistry (Fig. 3B, 3D). Similarly, CD24 and CD133, two other frequently used CSC markers in CRC, were also reduced upon in vitro differentiation (Fig. 3C), while conversely expression of the differentiation markers Muc2 and CK20 was induced (Fig. 3D).

Figure 3.

Lgr5 expression correlates to differentiations status. (A): Spheroid cultures were differentiated by plating the cells adherent, removing epidermal growth factor and basic fibroblast growth factor and adding fetal calf serum. Staining showed that Lgr5 was downregulated upon differentiation, analyzed here for two different CSC lines. The graph represents the ratio of the MFI of the specific Lgr5 staining and the isotype control. (B): Lgr5 mRNA expression of spheroid and differentiated cultures were analyzed by quantitative polymerase chain reaction (qPCR). As input control, GAPDH was used. (C): Cells from colon cancer spheroid culture showed decrease of cancer stem cell marker CD133 and CD24 upon differentiation in serum containing medium. Histograms represent isotype control staining profile (black line) versus specific staining profile as indicated (red shade). (D): Confocal microscopy analysis of colorectal cancer (CRC) spheroids cultured with cancer stem cell medium (upper panels) or differentiation medium (lower panels). Spheroids cultured in differentiation medium demonstrate the loss of LGR5 expression and increased expression of secretory goblet lineage marker Muc2 and mature colonocytes marker CK20. Bar indicates 50 μM. (E): CRC spheroid culture transduced with a TCF/LEF-GFP (TOP-GFP) construct was previously used to demonstrate that the TOP-GFPhigh fraction identifies the CSC and thus undifferentiated fraction, whereas the TOP-GFPlow fraction expressed differentiation markers [7]. Here, this TOP-GFP spheroid culture was stained for Lgr5. Lgr5 expression was found to correlate to Wnt-activity. Graph represents MFI of LGR5 staining. (F): qPCR for Lgr5 on sorted TOP-GFPhigh versus TOP-GFPlow cells. GAPDH was used as housekeeping gene. (G): Lgr5 expression correlated with CSC marker CD133 and CD24. The graphs represent the MFI of both markers in LGR5high and LGR5low fraction. Abbreviations: APC, allophycocyanin; CSC, cancer stem cell; CK20, cytokeratin 20; GFP, green fluorescent protein; MFI, mean fluorescence intensity; muc2, mucin2.

Another method to study differentiation is to make use of CRC spheroid cultures that are transduced with a GFP reporter linked to a TCF/LEF responsive element (TOP-GFP). Previously, we showed that CSCs in CRC can be identified by high Wnt reporter activity, as the TOP-GFPhigh cells in the spheroid cultures express CSC markers and are clonogenic as well as tumorigenic, whereas the TOP-GFPlow cells express differentiation markers and loose clonogenicity and tumorigenicity [7]. So, already within the spheroid cultures, a hierarchy exists between CSCs and more differentiated tumor cells. Lgr5 expression was detected with the monoclonal antibodies in these TOP-GFP-transduced spheroid cultures and correlated with TOP-GFP intensity. We found that TOP-GFPhigh cells indeed display high Lgr5 expression, whereas TOP-GFPlow cells are Lgr5low (Fig. 3E). In addition, Lgr5 mRNA expression was indeed higher in the TOP-GFPhigh fraction as compared to the TOP-GFPlow, confirming the antibody staining (Fig. 3F). Finally, expression of Lgr5 correlated with the CSC markers CD133 and CD24 (Fig. 3G).

Taken together our data therefore indicate that Lgr5 expression correlates with other CSC markers and that downregulation of Lgr5 is observed upon in vitro differentiation of CSCs, suggesting that it indeed marks only the undifferentiated CSC fraction in tumors.

Lgr5 Identifies the Clonogenic Fraction In Vitro

Another key feature of CSC markers is that tumor cells with high expression of this marker represent the more clonogenic fraction compared to marker-low/negative cells. Analyzing the clonogenic potential in vitro can be performed by LDA, in which cells based on marker expression are plated at different cell numbers to analyze their potential to form new spheres. As positive control in this assay, the TOP-GFP-transduced spheroid cultures were used, since we showed previously that TOP-GFPhigh cells are more clonogenic than TOP-GFPlow cells [7]. In agreement with these observations, we found that the clonogenicity of TOP-GFPhigh cells was significantly higher than of TOP-GFPlow cells (Fig. 4A). When spheroid cells were sorted based on their Lgr5 expression, we found that Lgr5high cells had also a significant higher clonogenic potential compared to Lgr5low cells (Fig. 4A). This difference in clonogenicity was not due to a difference in cycling cells between the two populations as this was identical as determined by Ki67 staining of the sorted Lgr5high and Lgr5low cells (Supporting Information Fig. 1). This is therefore indicating that Lgr5 expression can indeed be used to purify the more clonogenic fraction in vitro. However, the Lgr5low cells were not completely devoid of clonogenic growth, which was not due to mis-sorting of Lgr5high cells as this remained well below 1% (Supporting Information Fig. 2). Importantly, the spheroids that emanate from the Lgr5low cells contained Lgr5high cells as well when analyzed a week after sorting (Fig. 4B). This suggests that the expression of Lgr5 is flexible and can be re-acquired by the tumor cells.

Figure 4.

Lgr5 expression identifies the clonogenic and tumorigenic fraction. (A): TOP-GFP spheroid cultures were sorted for the 10% highest or 10% lowest fraction of either the GFP expression, or the Lgr5 expression or both and deposited into a limiting dilution assay (LDA). Outgrowth of spheroids was analyzed after 2 weeks, and clonogenic capacity and significance was calculated by extreme limiting dilution analysis (ELDA). ****, p ≤ .0001. (B): The 10% highest and 10% lowest fraction of the Lgr5 expression was sorted from the spheroid cultures. After 7 days of culturing, the Lgr5 expression from both populations was measured. (C): The 10% highest and 10% lowest fraction of the Lgr5 expression was sorted from the spheroid cultures and injected subcutaneously into immunocompromized mice. For Lgr5, 10, 100, or 1,000 cells were injected (n = 8 per group). The table shows which mice developed tumors. (D): By ELDA, the in vivo clonogenic potential and its significance was calculated. ****, p ≤ .0001. Abbreviation: GFP, green fluorescent protein.

Lgr5 Selects for the Tumorigenic Cells In Vivo

CSCs are originally defined by their capacity to initiate a phenocopy of the original malignancy upon xenotransplantation into mice [3, 4]. A reliable marker thus is capable of selecting cancer cells that contain the capacity to form tumors in vivo. We therefore isolated the 10% highest and lowest Lgr5-expressing cells from our spheroid cultures and injected these subcutaneously in immunodeficient mice. This analysis was performed in a limiting dilution fashion, meaning that the mice received either 10, 100, or 1,000 Lgr5high or Lgr5low cells (n = 8 per group). Similar experiments performed with tumor cells sorted for TOP-GFPhigh or TOP-GFPlow showed that high expression of the Wnt reporter could select for tumorigenicity: almost 100% of the mice injected with 100 TOP-GFPhigh cells formed tumors whereas only one mouse injected with 100 TOP-GFPlow cells displayed tumor growth [7]. Selection for high Lgr5 expression gave similar results. Subcutaneous injection of for instance 100 Lgr5high cells gave rise to tumors in almost 100% of the mice, whereas less than half of the mice formed a tumor when injected with 100 Lgr5low cells. A similar difference was found when lower or higher cell amounts were injected (Fig. 4C). By ELDA, the tumorigenicity of these cells in vivo was calculated. The Lgr5high cells were 13-fold more tumorigenic as the Lgr5low cells (Fig. 4C, 4D), confirming that the Lgr5high CRC tumor cells represent the tumorigenic fraction in vivo.

Lgr5 Is a CSC Marker with a Function

The above indicated that Lgr5 is a good marker for CSCs but does not indicate whether it has functional relevance for the CSCs. In normal intestinal stem cells, Lgr5 serves as a receptor that binds R-spondins and is crucial for their Wnt-enhancing effects, while at the same time it is a target of the Wnt pathway [13]. We have previously shown that blocking of Wnt cascade in colon carcinoma cell lines diminished the expression of various Wnt target genes [11]. We confirmed this finding with the addition of GSK3 inhibition (which activates Wnt cascade) and showed that Lgr5 expression is increased along with TOP-GFP (Fig. 5A).

Figure 5.

Lgr5 is a functional cancer stem cell (CSC) marker. (A): CSC spheroid cultures were treated with 4 μM of CHIR for 48 hours, and the expression of Lgr5 was analyzed. Graphs represent the normalized MFI of LGR5 (left) and TOP-GFP expression (right). Error bars show SD of triplicate experiment. (B): LS174T-LGR5 cells were sorted for the 10% highest and 10% lowest fraction of Lgr5 expression for limiting dilution assay (LDA). ****, p ≤ .0001. (C): LDA showed complete loss of stem cells frequency in LGR5-KD spheroid cultures (shB4 and shB5). ****, p ≤ .0001. Abbreviations: DMSO, dimethyl sulfoxide; GFP, green fluorescent protein; MFI, mean fluorescence intensity.

To analyze whether Lgr5 expression was required for clonogenicity, we first analyzed the LS174T cells exogenously overexpressing Lgr5. Even though LS174T cells expressing low to almost no Lgr5 were clonogenic, the high expressors were about threefold better in forming new colonies (Fig. 5B), Thus, overexpression of Lgr5 by itself enhanced the clonogenicity of LS174T cells. Vice versa, knocking down Lgr5 in CSCs using short hairpin RNA (shRNA) (Supporting Information Fig. 3) resulted in a complete loss of clonogenic capacity (Fig. 5C). This indicates that Lgr5 is indeed a functional marker for CSC, stimulating Wnt pathway activity and thereby clonogenicity.

DISCUSSION

Here, we show that with newly developed monoclonal antibodies against human Lgr5, we can identify the CSC fraction in CRC. FACS analysis to detect Lgr5 on primary CRC tumor material reveals that the epithelial tumor fraction contains a subpopulation of Lgr5+ cells. After enrichment of CSCs by in vitro spheroid culture, we found that these CSCs express Lgr5, which is downregulated upon in vitro differentiation. Additionally, Lgr5high CRC cells were demonstrated to have higher clonogenic potential in vitro as well as higher tumorigenicity in vivo when compared to the Lgr5low cancer cells. So, the monoclonal antibodies raised against Lgr5 identify CSCs in CRC.

Although Lgr5 has been suggested to serve as a CSC marker in a range of different studies [4, 16, 23], these data represent the first formal proof that Lgr5 is expressed on CSCs and can be used to isolate them. Lgr5 represents a CSC marker for CRC, which is not just a biomarker but probably plays an important role in maintaining CSCs undifferentiated. The function of Lgr5 was long undetermined, but recently, we and others described R-spondins as ligands for this G protein-coupled receptor. Binding of R-spondins to Lgr5 occurs in concert with the E3 ligases ZNFR3 and/or RNF43, which then leads to internalization of these Wnt receptor inhibitors and thus facilitation of Wnt signal transduction. All four family members of the R-spondins family can serve as secreted ligands for Lgr5. As R-spondins significantly enhance Wnt signaling in the intestine [23], this Lgr5/R-spondin/E3 ligases interaction likely determines intestinal stem cell maintenance. In agreement, in vitro murine intestinal organoid cultures require R-spondin supplements [24]. Additionally, in vivo, R-spondins can provide stimulating proliferating signals for intestinal crypts cells (which are Lgr5+) by enhancing the Wnt signaling [25–27]. Similarly, deletion of the two E3 ligases in the intestinal stem cell results in Wnt-dependent adenoma formation, indicating that the mere deletion of the inhibitors is sufficient to drive expansion of the cells as long as Wnt ligands are provided (Nature, in press). The source of R-spondin in vivo is probably the mesenchyme, which was already identified as producer of this ligand in the fetal mouse intestine [28]. As Wnt signaling itself is important in maintaining stemness in normal intestinal stem cells [29], the presence of Lgr5 and its ligand R-spondin might be essential components in mediating this Wnt activity.

Not only normal stem cells but also CSCs are regulated by Wnt signaling. We showed that high Wnt activity can mark the tumorigenic and clonogenic fraction in CRC [7]. Interestingly, this Wnt signal is regulated by the microenvironment, as surrounding myofibroblasts that express high levels of hepatocyte growth factor (HGF) could activate Wnt signaling in the tumor cells [7]. R-spondin and its receptor Lgr5 appear to maintain CSCs in an undifferentiated stem cell state. That is, overexpression of Lgr5 enhances clonogenicity, while knockdown of Lgr5 prevents clonogenic growth of the CSCs. In normal intestine, Wnt activity in stem cells is induced by Wnt3 produced by neighboring Paneth cells, which provide a niche for Lgr5+ stem cells [30]. Although CSCs carry activating Wnt pathway mutations, our previous results indicate that this does not imply that exogenous triggers cannot modulate the pathway. As such, it would be possible that R-spondins and Wnt ligands can still enhance stemness in the tumor cells. In this light, it is important to mention that our gene expression data show R-spondin mRNA in LS174T cells (J.P. Medema, unpublished data). Whether such autocrine stimulation would be sufficient to drive clonogenicity is not clear but it could be the underlying reason for Lgr5-mediated control of clonogenicity. In vivo, such a stimulation could be the result of Paneth-like cells that appear to exist in the tumor or alternatively, the myofibroblasts may, beside HGF, also produce R-spondins to regulate Wnt activity. To identify the nature of the Lgr5-driven cancer stemness, further studies are needed. Our current data, however, firmly establish Lgr5 as a functional CSC marker for CRC.

CONCLUSION

This study shows that three newly developed antibodies against Lgr5 can identify the CSC fraction in CRC. FACS staining of primary CRC cells for Lgr5 resulted in the identification of a small population of Epcam+Lgr5+ cells. Spheroid cultures, derived from primary CRC, are known to be enriched for CSCs and were found to express high levels of Lgr5, while cellular differentiation reduced Lgr5 expression. The Lgr5high CRC cells are more clonogenic and tumorigenic than Lgr5low CRC cells. Finally, Lgr5 overexpression resulted in higher clonogenic growth, indicating that Lgr5 is a new, functional marker for CRC CSCs.

Acknowledgements

We thank Felipe de Sousa E Melo, M.Sc. for helpful comments and critical reading of the manuscript. K.K. was supported by an AMC graduate school scholarship. J.P.M. is supported by a VICI grant of NWO. P.R.P. and J.P.M. are supported by Dutch Cancer Society grant UvA 2009-4416.

DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST

The authors indicate no potential conflicts of interest.

Ancillary