LGR5 positivity defines stem-like cells in colorectal cancer

Abstract

Like normal colorectal epithelium, colorectal carcinomas (CRCs) are organized hierarchically and include populations of cells with stem-like properties. Leucine-rich-repeat-containing G-protein-coupled receptor 5 (LGR5) is associated with these stem cells in normal colorectal epithelium; however, the precise function of LGR5 in CRC remains largely unknown. Here, we analyzed the functional and molecular consequences of short hairpin RNA-mediated silencing of LGR5 in CRC cell lines SW480 and HT-29. Additionally, we exposed Lgr5-EGFP-IRES-CreERT2 mice to azoxymethane/dextrane sodium sulfate (AOM/DSS), which induces inflammation-driven colon tumors. Tumors were then flow-sorted into fractions of epithelial cells that expressed high or low levels of Lgr5 and were molecularly characterized using gene expression profiling and array comparative genomic hybridization. Silencing of LGR5 in SW480 CRC cells resulted in a depletion of spheres but did not affect adherently growing cells. Spheres expressed higher levels of several stem cell-associated genes than adherent cells, including LGR5. Silencing of LGR5 reduced proliferation, migration and colony formation in vitro and tumorigenicity in vivo. In accordance with these results, NOTCH signaling was downregulated upon LGR5 silencing. In AOM/DSS-induced colon tumors, Lgr5 high cells showed higher levels of several stem cell-associated genes and higher Wnt signaling than Lgr5 low tumor cells and Lgr5 high normal colon cells. Array comparative genomic hybridization revealed no genomic imbalances in either tumor cell fraction. Our data elucidate mechanisms that define the role of LGR5 as a marker for stem-like cells in CRC.

Summary:

Our functional and molecular findings link the intestinal stem cell marker LGR5 to stem-like colorectal cancer cells. LGR5 silencing reduced proliferation, migration and tumorigenicity, and NOTCH signaling. Primary mouse colon tumors maintained an Lgr5-based stem cell hierarchy

Introduction

Colorectal tumorigenesis is associated with the accumulation of a number of specific genetic changes, which drive the transition from normal epithelium through adenomas to invasive carcinomas. These genetic changes include mutations of specific genes, such as adenomatous polyposis coli (APC) and KRAS, and tumor-specific genomic imbalances (1–3). Similar to normal colorectal epithelium, colorectal tumors consist of heterogeneous cell populations at various levels of differentiation. Although a few years ago all cells within a tumor were considered to be tumorigenic, recent findings suggested that only a subpopulation of tumor cells can regenerate the tumor (4,5). These cells, termed cancer stem cells (CSCs), may also be involved in therapy resistance, tumor relapse and metastasis. Accordingly, the identification and characterization of these cells was the subject of considerable research efforts. With respect to colorectal carcinomas (CRCs), putative CSCs can be identified by leucine-rich-repeat-containing G-protein-coupled receptor 5 (Lgr5; also known as G-protein-coupled receptor 49, Gpr49). Lgr5, a Wnt target gene that acts as receptor for the Wnt agonist R-spondin, is a marker gene for adult intestinal stem cells as revealed by in vivo lineage tracing (6–8). Selective deletion of Apc in the mouse in either Lgr5 positive intestinal stem cells or more differentiated cells revealed that mainly the Lgr5 positive stem cell fraction is capable of forming tumors upon Wnt pathway activation, suggesting Lgr5 positive stem cells as the cells-of-origin of intestinal epithelial tumors (9). Although the cell-of-origin for tumorigenesis and the CSC, which propagates the tumor, need not necessarily be identical, in vivo lineage tracing provides direct evidence for a stem cell activity of Lgr5 positive cells in mouse intestinal adenomas generated by deletion of Apc in Lgr5 positive stem cells (10,11). Resembling the situation in normal intestinal epithelium, adenomas contain a small fraction of Lgr5 positive cells (5–10%) that are able to generate all cell types present within the adenomas, including additional Lgr5 positive cells (11). In human CRC, LGR5 expression is highly enriched in EPHB2 positive cells, which have similar expression profiles to normal intestinal stem cells and—in contrast to EPHB2 negative cells—display reproducible tumorigenic capacity in immunodeficient mice (12). Cataloging the genetic idiosyncrasies of LGR5 positive and negative cells might help to identify the mechanisms that cause these differences in tumorigenic potential. We have therefore investigated the functional and molecular consequences of short hairpin RNA (shRNA)-mediated LGR5 silencing in CRC cell lines SW480 and HT-29. To date, studies on LGR5 in primary CRC samples have been constrained by the lack of a reliable antibody against LGR5. We therefore induced inflammation-driven colon tumors in mice that were engineered to contain one enhanced green fluorescent protein (EGFP)-tagged Lgr5 allele (6). This allowed flow cytometric separation of Lgr5 high and low cells based on GFP expression and thus enabled a genome and transcriptome characterization of these two cell fractions. Our loss-of-function experiments conclusively indicate that LGR5 acts as a marker for stem-like cells in CRC.

Materials and methods

Cell lines and lentiviral transduction

The six human CRC cell lines (Caco-2, HCT 116, HT-29, SW480, SW620 and T84) were purchased from the American Type Culture Collection (Manassas, VA). All cell lines were cultured in media as recommended by the American Type Culture Collection supplemented with fetal bovine serum (10% v/v), l-glutamine (2mM), penicillin (100U/ml) and streptomycin (100 µg/ml). Lentiviral shRNA transduction of SW480 and HT-29 cells was done using high-titer lentivirus (Clone ID: V3LHS_635055, Open Biosystems, Thermo Fisher Scientific, Lafayette, CO) according to the manufacturer’s instructions.

Mice

Athymic nude mice (strain NCr-nu/nu) were obtained from Frederick National Laboratory for Cancer Research (Frederick, MD). Heterozygous Lgr5-EGFP-IRES-CreERT2 mice [strain B6.129P2-Lgr5tm1(cre/ERT2)Cle/J, henceforth referred to as 'Lgr5-EGFP mice'] were ordered from Jackson Laboratory (Bar Harbor, ME) (6). All mice were bred and housed in a pathogen-free environment and used in experiments in accordance with institutional guidelines at the Center for Cancer Research, National Cancer Institute, National Institutes of Health. All experimental procedures conducted in this study were approved by the Animal Care and Use Committee of the National Institutes of Health.

Tumorigenicity assay

Tumorigenicity assay was performed as described previously (13).

Microarray gene expression profiling of cell lines

Total RNA was isolated from SW480 shLGR5 and control cells and from SW480 spheres and adherent cells using the RNeasy Mini Kit (Qiagen, Hilden, Germany) including DNase I treatment (RNase-Free DNase Set, Qiagen). RNA integrity was assessed by 2100 Bioanalyzer (RNA 6000 Nano LabChip Kit, Agilent Technologies, Santa Clara, CA). Appropriate LGR5 status was confirmed by real-time quantitative reverse transcription–polymerase chain reaction (qRT–PCR). Total RNA (700ng) was labeled using the Quick Amp Labeling Kit, one-color (Agilent) and subsequently hybridized on Human GE 4x44K v2 Microarrays (Agilent) according to the manufacturer’s protocol version 6.5. Slides were scanned with microarray scanner G2565BA (Agilent). Images were analyzed and data were quality controlled using Feature Extraction software version 10.7.1.1 (Agilent).

Carcinogen-induced inflammation-driven colon tumorigenesis model

To induce colon tumors, Lgr5-EGFP mice aged 2–4 months were injected with azoxymethane (AOM, 12.5 µg/g body weight; A5486, Sigma, St Louis, MO) twice and subjected to three cycles of dextrane sodium sulfate (DSS, 2.5%; molecular weight = 36 000–50 000, MP Biomedicals, Solon, OH) in the drinking water (14,15). Tumor growth was monitored by colonoscopies. About 100 days after the first AOM injection, mice were killed.

RNA amplification and microarray gene expression profiling of flow-sorted normal mouse colons and mouse colon tumors

RNA isolated from flow-sorted normal mouse colons and mouse colon tumors was amplified together with spike-ins (1 µl of a 1:50 000 dilution per reaction, One-Color RNA Spike-In Kit, Agilent) using the Ovation Pico WTA System (NuGEN Technologies, San Carlos, CA). The amplified complementary DNA was labeled using the BioPrime® Total Genomic Labeling Module (Invitrogen, Life Technologies, Carlsbad, CA) and subsequently 4 µg of Alexa Fluor® 3 labeled target was hybridized on Whole Mouse Genome Microarrays 4x44K (Agilent) according to NuGEN’s Agilent Solution Application Note #1. Scanning, image analysis and data quality control were done as for cell lines.

DNA amplification and array comparative genomic hybridization of flow-sorted mouse colon tumors

DNA amplification and array comparative genomic hybridization were performed as described previously and are summarized in Supplementary Materials and methods, available at Carcinogenesis Online (16). Array-based comparative genomic hybridization data have been deposited in Gene Expression Omnibus database (data accession number: GSE46711).

Statistical analysis

Differences between groups were estimated by Student’s t-test, Fisher’s exact test or two-way repeated measures analysis of variance (one factor repetition). P < 0.05 was considered significant.

Microarray gene expression analysis

Log2 intensities were normalized to the 75% percentile according to the manufacturer’s protocol (Agilent). Only probes with intensities higher than 50 were used for the analysis. Unsupervised hierarchical clustering with Euclidean distance and Ward method was performed with Genomics Suite™ software (Partek Incorporated, St Louis, MO). The corresponding functional annotation of differentially expressed genes and their affiliation with specific genetic pathways were interrogated using Ingenuity Pathway Analysis software (Ingenuity Systems, Redwood City, CA). Microarray gene expression data have been deposited in Gene Expression Omnibus database (data accession number: GSE46200).

Results

LGR5 is overexpressed in human CRC cell lines

LGR5 expression is upregulated in primary CRC samples compared with normal colorectal epithelium (17). Using real-time qRT–PCR, we could show that LGR5 expression was also upregulated in CRC cell lines Caco-2, HT-29, SW480, SW620 and T84 (Figure 1A) (18,19). The only cell line without overexpression of LGR5 was HCT 116. This is consistent with recent findings, suggesting that HCT 116 may not be organized hierarchically and may therefore not contain a stem-like cell fraction (13,20,21).

Fig. 1.

Expression levels of LGR5 in CRC cell lines and shRNA-mediated silencing of LGR5 in SW480 and HT-29 CRC cells. (A) LGR5 mRNA levels in CRC cell lines as determined by real-time qRT–PCR. LGR5 mRNA levels were normalized to YWHAZ and are expressed as fold changes relative to normal colorectal epithelial cells. Columns and error bars represent means ± SEM of two independent experiments using triplicate measurements in each experiment. (B) LGR5 silencing efficiency in CRC cell lines SW480 and HT-29 upon lentiviral shRNA transduction as determined by real-time qRT–PCR. LGR5 mRNA levels were normalized to YWHAZ and are expressed as fold changes relative to control. Columns and error bars represent means ± SEM of three independent experiments using triplicate measurements in each experiment. (C) Representative phase-contrast images of SW480 and HT-29 cells upon LGR5 silencing. SW480 is composed of two morphologically distinct subpopulations (spheres and adherent cells), whereas HT-29 shows one phenotype. After LGR5 silencing, spheres in SW480 were completely lost with no apparent changes in HT-29. Scale bars, 100 µm. (D) Expression of LGR5 in spheres versus adherent cells of SW480. Spheres express high levels of LGR5, whereas in adherent cells, virtually no LGR5 expression can be detected by real-time qRT–PCR. LGR5 mRNA levels were normalized to YWHAZ and are expressed as fold changes relative to control. Data represent means ± SEM of three independent experiments using triplicate measurements in each experiment. *P < 0.05, **P < 0.005 and ***P < 0.0005.

LGR5 silencing leads to depletion of a morphologically distinct subpopulation of SW480 CRC cells

To investigate the function of LGR5 in CRC, we transduced two CRC cell lines (SW480 and HT-29), which both showed a similar moderate overexpression of LGR5, with lentiviral shRNA constructs. This significantly reduced LGR5 expression compared with empty vector control (henceforth referred to as ‘control’) cells as assessed by real-time qRT–PCR (Figure 1B). In SW480, silencing of LGR5 led to a remarkable morphologic change, whereas HT-29 did not display an apparent difference (Figure 1C). Unlike HT-29, SW480 typically comprises two morphologically distinct subpopulations, i.e. spheres and adherent cells, when cultured in serum-containing medium. Upon LGR5 silencing, spheres were completely depleted and only an adherent cell layer remained. Spheres, along with LGR5 expression, remained undetectable in SW480 shLGR5 cells over more than 12 months of continuous passage, demonstrating long-lasting effects of the shRNA on LGR5 expression. To exclude that the disappearance of spheres was induced by cell splitting, we next separately analyzed LGR5 messenger RNA (mRNA) expression in spheres and adherent cells of SW480 control cells. Although LGR5 was highly expressed in spheres, adherent cells displayed virtually no LGR5 expression, supporting our conclusion that the morphologic change was caused by silencing of LGR5 (Figure 1D).

LGR5 silencing in SW480 and HT-29 CRC cells reduces proliferation, migration and colony formation in vitro

To evaluate the influence of LGR5 on cell proliferation, we compared the cleavage of the tetrazolium salt WST-1 between shLGR5 and control cells. Upon silencing of LGR5, cleavage of WST-1 was significantly decreased in HT-29 cells (Figure 2A). SW480 cells showed the same tendency although not reaching statistical significance (P = 0.096). This might be attributable to the loss of spheres in SW480 upon silencing of LGR5; spheres usually grow slower than isogenic adherent cells (22). In other words, the proliferating cells of SW480 reside within the LGR5 low adherent cells and not within the LGR5 high spheres. This explains the small effect of LGR5 silencing on proliferation in SW480.

Fig. 2.

Functional effects of LGR5 silencing on SW480 and HT-29 CRC cells. (A) LGR5 silencing decreases proliferation of SW480 and HT-29 CRC cells as quantified based on cleavage of WST-1. Columns and error bars represent means ± SEM of three independent experiments using triplicate measurements in each experiment. (B) Silencing of LGR5 reduces migration in wound healing assays. Columns and error bars represent means ± SEM from one experiment representative of three independent experiments (n = 6 scratches per cell type and time point in each experiment). Scale bars, 100 µm. (C) In line with the results of the wound healing assay, migration of SW480 shLGR5 cells is also decreased in transwell migration assays. Columns and error bars represent means ± SEM from one experiment representative of three independent experiments using triplicate measurements in each experiment. Scale bars, 200 µm. (D) LGR5 silencing substantially reduces colony formation in vitro. Each 250 shLGR5 and control cells were cultured for 2 weeks. The number of colonies was then assessed using crystal violet staining. Columns and error bars represent means ± SEM of three independent experiments using triplicate measurements in each experiment. *P < 0.05 and **P < 0.005.

The effect of LGR5 on migration was assessed by wound healing and transwell migration assays. In the wound healing assay, SW480 shLGR5 cells tended to cover a smaller area of the initial scratch than respective control cells at all time points, reaching statistical significance after 72 h (Figure 2B). To reduce the impact of proliferation as a confounding variable, we additionally monitored migration for each group through transwell membranes and observed a significantly smaller number of migrated shLGR5 cells compared with control cells after incubating for 24 h (Figure 2C). For HT-29, none of the assays allowed a proper assessment of migration dependent on LGR5 status owing to too little migratory activity of HT-29 cells in our experiments.

In the colony formation assay, both cell lines formed significantly fewer colonies when LGR5 was silenced compared with control cells (Figure 2D).

LGR5 silencing reduces tumorigenic capacity of SW480 CRC cells after xenotransplantation

To corroborate our in vitro colony formation results in vivo, we injected CRC cells with differential LGR5 expression levels subcutaneously into the flanks of nude mice and followed the appearance of tumors. SW480 shLGR5 cells were less tumorigenic than control cells and the tumors grew slower (Table I and Supplementary Figure S1A –C, available at Carcinogenesis Online). Reduced tumorigenicity was not observed for HT-29 cells upon LGR5 silencing (Table I). However, tumors derived from HT-29 shLGR5 cells also grew slower than those from control cells (Supplementary Figure S1D, available at Carcinogenesis Online). Sufficient silencing of LGR5 in xenografts was confirmed by real-time qRT–PCR (Supplementary Figure S1E, available at Carcinogenesis Online).

Table I.

Quantification of tumor initiation in nude mice after subcutaneous injection of CRC cells with differential levels of LGR5 expression

Cell line	shLGR5 (no. of tumors/no. of injections)	Control (no. of tumors/no. of injections)	No. of injected cells	P value
SW480	4/10	9/10	2000	0.0052
	2/5	5/5	20 000
HT-29	8/10	10/10	2000	0.47
Cell line	Adherent cells (no. of tumors/no. of injections)	Spheres (no. of tumors/no. of injections)	No. of injected cells	P value
SW480	0/3	2/3	2000	0.40

Lacking a reliable LGR5 antibody, we utilized the phenotypic change of SW480 upon LGR5 silencing to separate LGR5 high cells (spheres) and LGR5 low cells (adherent cells), which we then injected subcutaneously into the flanks of nude mice as a proof-of-principle experiment. Confirming our hypothesis, only LGR5 high cells generated tumors; however, the small number of animals prevented reaching statistical significance (P = 0.40) (Table I).

NOTCH signaling is downregulated upon LGR5 silencing

To examine whether differential LGR5 expression levels in CRC cells would be reflected in specific changes to the cellular transcriptome, we performed microarray gene expression profiling of SW480 shLGR5 and control cells and of SW480 spheres and adherent cells. As LGR5 silencing in SW480 resulted in a profound morphologic change, we reasoned that genes regulated by LGR5 might overlap with genes differentially expressed between spheres and adherent cells. Indeed, gene expression profiling revealed a significant overlap of deregulated genes (false discovery rate < 1.0E-103; Figure 3A). These genes included known oncogenes (e.g. MYB, MYCN) and certain drug efflux genes (e.g. ABCB1, ABCC2), whose expression was positively correlated with LGR5 expression. The expression differences between spheres and adherent cells were more pronounced than between shLGR5 and control cells, yet unsupervised hierarchical clustering showed a clear separation of all four cell fractions (Figure 3B). Ingenuity Pathway Analysis revealed that the NOTCH signaling pathway was downregulated when LGR5 was silenced. Conversely, it was upregulated in LGR5 high spheres. Apart from LGR5, spheres also expressed higher levels of several other stem cell-associated genes including SOX2, ALDH1A1 and SMOC2 when compared with adherent cells (12,23–25). Our gene expression results were exemplarily validated by immunohistochemistry against two differentially expressed genes, cleaved NOTCH1 and SOX6 (Figure 3C).

Fig. 3.

Gene expression profiling of shLGR5 versus control cells and adherent cells versus spheres of SW480. (A) Venn diagram showing a significant overlap of differentially expressed genes when comparing shLGR5 versus control cells and adherent cells versus spheres of SW480 (false discovery rate < 1.0E-103). (B) Unsupervised hierarchical clustering based on gene expression profiles of shLGR5 and control cells as well as adherent cells and spheres from SW480. Samples cluster by cell type first. The different LGR5 fractions are clearly separated. (C) In line with our gene expression data, immunohistochemistry against both cleaved NOTCH1 and SOX6 also shows a lower expression in SW480 shLGR5 than in SW480 control cell derived xenograft tumors. Scale bars, 20 µm.

Lgr5 is overexpressed in AOM/DSS-induced mouse colon tumors but expression is, like in normal colon epithelium, restricted to a small percentage of cells

We next aimed to determine a detailed molecular characterization of Lgr5 positive and negative cells in primary colon tumors. As the lack of a reliable LGR5 antibody prevented these analyses in primary human CRC, we exposed heterozygous Lgr5-EGFP mice, harboring one EGFP-tagged Lgr5 allele, to a carcinogen-induced inflammation-driven colon tumorigenesis model based on AOM and DSS (Supplementary Figure S2A, available at Carcinogenesis Online) (6,14). All AOM/DSS-induced tumors were restricted to the colon, predominantly located in the rectosigmoid colon, sometimes forming multiple lesions; no lymph node or hematogenous metastases were detectable (Supplementary Figure S2B, available at Carcinogenesis Online). Histologically, AOM/DSS-induced colon tumors in Lgr5-EGFP mice resembled well-differentiated human tubular adenocarcinomas, mainly of the intramucosal type (Supplementary Figure S2C, available at Carcinogenesis Online).

We first examined expression, distribution and frequency of Lgr5 in AOM/DSS-induced colon tumors. Real-time qRT–PCR analysis using RNA isolated from histologically defined tumor and non-tumor regions revealed a significant overexpression of Lgr5 in AOM/DSS-induced tumors compared with normal colon mucosa (Figure 4A and Supplementary Figure S3A, available at Carcinogenesis Online). Immunohistochemistry against GFP showed that, consistent with previous reports, Lgr5-EGFP expressing cells in normal colons were located at the crypt bottoms (Supplementary Figure S3B, available at Carcinogenesis Online) (6). In AOM/DSS-induced tumors, Lgr5-EGFP expression was found to be restricted to small populations of scattered cells (Figure 4B). Quantification by flow cytometric analysis revealed that normal colons contained on average 3.8% of Lgr5-EGFP high cells. This percentage was similar to that of AOM/DSS-induced tumors, which harbored on average 3.4% of Lgr5-EGFP high cells (Figure 4C). Supported by the results from our immunohistochemistry-based analysis of Lgr5-EGFP expression, this suggests that a stem cell hierarchy is preserved in AOM/DSS-induced tumors (9,11). However, there were 3 out of 14 (21.4%) AOM/DSS-induced tumors without detectable Lgr5-EGFP expressing cells based on GFP expression, indicating that occasionally also Lgr5-EGFP low cells could acquire the capacity for tumor initiation and/or maintenance (10). This is consistent with observations in human CRC samples, in which LGR5-expressing cells were not detectable in one-third of the analyzed samples, arguing for a frequent stem cell and a rare non-stem cell driven carcinogenesis (12). We confirmed by real-time qRT–PCR that GFP high flow-sorted cells had significantly higher Lgr5 expression levels than GFP low flow-sorted cells (Figure 4D).

Fig. 4.

Expression, distribution and frequency of Lgr5 in AOM/DSS-induced mouse colon tumors. (A) Varying degrees of overexpression of Lgr5 in AOM/DSS-induced colon tumors as determined by real-time qRT–PCR. Lgr5 mRNA levels were normalized to Gapdh and are expressed as fold changes relative to normal adjacent colon. Horizontal lines represent means. Triplicate measurements were used for each data point. (B) Lgr5-EGFP expression in AOM/DSS-induced colon tumors is restricted to small populations of scattered cells. Scale bars, 50 µm. (C) Quantification of Lgr5-EGFP high cells by flow cytometric analysis revealed that both normal colons and AOM/DSS-induced tumors contain similar small percentages of Lgr5-EGFP high cells. N, normal colon; T, tumor. (D) Lgr5 expression is enriched in flow-sorted Lgr5-EGFP high normal colon or colon tumor cells compared with respective Lgr5-EGFP low normal colon or colon tumor cells as determined by real-time qRT–PCR. Lgr5 mRNA levels were normalized to Gapdh. Horizontal lines represent means. Triplicate measurements were used for each data point. *P < 0.05 and **P < 0.005.

Transcriptome profiles of Lgr5 high and low epithelial cells from AOM/DSS-induced mouse colon tumors are clearly distinct

To examine whether the differential Lgr5 expression levels in normal mouse colons and AOM/DSS-induced mouse colon tumors would be reflected in specific gene expression profiles, we performed microarray gene expression profiling of normal colons and AOM/DSS-induced tumors flow-sorted into Lgr5 high and low fractions based on GFP expression. Reassuringly, the previously described intestinal stem cell-specific genes Lgr5 and Smoc2 were upregulated in Lgr5 high normal and tumor cells compared with Lgr5 low normal and tumor cells (12,25). In addition, Lgr5 high tumor cells showed increased expression of EphB2, which is known to be coexpressed with Lgr5 (12). Unsupervised hierarchical clustering showed a clear separation between Lgr5 high and low tumor cells; however, separation according to Lgr5 expression levels was not as clear in normal cells (Figure 5A). Ingenuity Pathway Analysis revealed that the Wnt signaling pathway was upregulated in Lgr5 high tumor cells compared with Lgr5 low tumor cells and also with Lgr5 high normal colon epithelial cells. Gene expression results were exemplarily validated by immunohistochemistry against Sox6 (Figure 5B).

Fig. 5.

Gene expression profiling of normal mouse colons and AOM/DSS-induced mouse colon tumors flow-sorted for Lgr5-EGFP. (A) Unsupervised hierarchical clustering based on gene expression profiles of Lgr5 high and low fractions from normal colons and AOM/DSS-induced tumors. Samples cluster by entity (normal colon or tumor) first. In tumors, the different Lgr5 fractions are clearly separated; however, in normal colons, they are separated by both Lgr5 and individual mice. (B) In line with our gene expression data, Sox6 expression could also be demonstrated by immunohistochemistry (positivity in a small population of scattered cells in AOM-/DSS-induced tumors). Scale bar, 20 µm.

Lgr5 high and low epithelial cells from AOM/DSS-induced mouse colon tumors are both chromosomally stable

Genomic imbalances influence gene expression patterns (26). To exclude that the transcriptional differences between Lgr5 high and low tumor cells were imposed by distinct patterns of chromosomal aberrations in the two cell fractions, we additionally performed array comparative genomic hybridization from the flow-sorted AOM/DSS-induced mouse colon tumors. All eight tumors were chromosomally stable, and thus, no difference between Lgr5 high and low cells could be detected (Supplementary Figure S4, available at Carcinogenesis Online) (27). This indicates that other mechanisms, for instance epigenetics, may be the driving force in AOM/DSS-induced mouse colon tumors (28). In conclusion, these data conclusively indicate that LGR5 marks stem-like CRC cells and defines a cell compartment in which proliferating, migrating and tumorigenic CRC cells are enriched. This is consistent with the observed upregulation of stem cell-related signaling pathways such as NOTCH or Wnt in LGR5 high CRC cells.

Discussion

Here, we studied the functional and molecular consequences of LGR5 silencing in CRC cell lines and identified LGR5 as a marker for stem-like cells in CRC. Based on Lgr5 expression, we defined a gene expression signature for stem-like cells in CRC using an inflammation-driven mouse colon tumorigenesis model based on AOM and DSS, which mimics sporadic CRC development.

To understand the role of LGR5 in colorectal tumorigenesis, which is controversial, we silenced its expression in two CRC cell lines (SW480 and HT-29) (18,29–34). Silencing of LGR5 in these cell lines resulted in reduced proliferation, migration and colony formation in vitro as well as reduced tumorigenicity in vivo. This is consistent with previous studies targeting LGR5 in various cancer entities including basal cell carcinoma, gastric cancer, glioblastoma and CRC (18,29,31–34). In these cancers, expression of LGR5 was associated with increased cell proliferation, migration and invasion and decreased apoptosis. In turn, silencing of LGR5 in these cancers decreased proliferation, colony formation and tumorigenicity and enhanced apoptosis. Upregulation of LGR5 in non-tumorigenic NIH3T3 fibroblasts and in HaCat keratinocytes induced colony formation and promoted tumorigenicity (18,29).

For reasons that remain to be understood, all those results are in contrast to the findings by Walker and colleagues in CRC cell lines LIM1215 and LIM1899. Upon silencing of LGR5, they reported increased migration, colony formation and tumorigenicity, and opposite phenotypes when LGR5 was overexpressed (30).

The reduced tumorigenic capacity of SW480 in vivo upon LGR5 silencing, along with the decreased colony formation in vitro, suggests that LGR5 expression is associated with stem-like properties in this cell line. The observation that shLGR5 and control HT-29 cells were equally tumorigenic upon xenotransplantation is somewhat contradictory but might be either explained by the lower silencing efficiency or because tumorigenicity of HT-29 does not depend on LGR5 expression. Also, HT-29 has much lower Wnt activity (35). The number of injected cells might have been too high to reveal differences since HT-29 is a highly tumorigenic cell line. Alternatively, the reduction of tumor growth rate but not tumor incidence upon LGR5 silencing in HT-29 might suggest that LGR5 silencing affects tumor propagation rather than tumor initiation in this cell line. The slower tumor growth rate upon silencing of LGR5 for both SW480 and HT-29 cells is consistent with previous data showing that Lgr5 positive cells are actively proliferating (6).

Our functional analyses were complemented by studying global gene expression levels of cell fractions with differential LGR5 expression. Spheres (in contrast to adherent cells) of SW480 expressed several stem cell-associated genes including LGR5 and showed an upregulation of NOTCH signaling. Consistently, spheres appeared to be more tumorigenic than adherent cells when xenografted in nude mice. Silencing of LGR5 resulted in the depletion of spheres. In line with our findings, LGR5 is upregulated in spheroid cultures of colorectal CSCs and, conversely, becomes downregulated during in vitro differentiation of these CSCs (32). Taken together, these findings suggest that LGR5 identifies a stem-like cell compartment in CRC, as it does in normal colorectal epithelium. This has recently been shown directly by in vivo lineage tracing in mouse intestinal adenomas (11).

NOTCH signaling was downregulated in SW480 CRC cells upon LGR5 silencing, whereas genes involved in NOTCH signaling were overexpressed in LGR5 high SW480 spheres. Consistently, Lgr5 deficiency in the mouse during intestine development also seems to have an inhibitory effect on the Notch signaling pathway (36). In addition, downregulation of NOTCH signaling in CRC cell lines and primary CRC samples via reduction of NOTCH1 or RBPJk decreases proliferation, colony formation and tumorigenicity, whereas upregulation of NOTCH signaling via NOTCH1 results in opposite changes (3,37,38). This has also been demonstrated for other tumor entities including pancreatic, lung or breast cancer (39–41). Consistent with our gene expression analysis, it appears most likely that the reduced proliferation and migration after LGR5 silencing in CRC cells are a consequence of reduced NOTCH signaling (3).

Hence, our gene expression results substantiate our findings from functional assays upon LGR5 silencing in CRC cell lines at a molecular level, suggesting LGR5 as a potential therapeutic target. For instance, Honokiol inhibits CRC growth by targeting NOTCH signaling in colorectal CSCs (42).

Pursuing stem cells as therapeutic targets might help to overcome some of the frustrations associated with current cancer treatment regimens. Cataloging gene expression data from these stem cells serves as a first step in understanding the molecular features distinguishing these cell types from the bulk of tumor cells or from normal adult tissue stem cells (43–46).

Despite the advantages of in vitro cultures to analyze features of stemness potential, such as ease of propagation, there are concerns that molecular changes might be induced when cells are cultured in the absence of their physiological context (13,47). Thus, we extended our gene expression profiling to ex vivo isolated cells. Normal mouse colons and AOM/DSS-induced mouse colon tumors from Lgr5-EGFP mice were flow-sorted into Lgr5 high and low epithelial cell fractions. Global gene expression analyses of flow-sorted fractions revealed an overexpression of the Wnt signaling pathway in Lgr5 high tumor cells compared with both Lgr5 low tumor cells and Lgr5 high normal cells. This supports previous findings showing that Wnt activity defines colorectal CSCs (48). Although Horst et al. (49) found that Wnt activity alone might not be sufficient to convey stem-like potential, our findings suggest that the combination of both Wnt signaling and Lgr5 might help to determine cells with stem-like properties in CRC.

The association of LGR5 and NOTCH signaling seen in cell lines could not be recapitulated in vivo when flow sorting AOM/DSS-induced mouse colon tumors for Lgr5. Also, in contrast to flow-sorted Lgr5 high and low cells from AOM/DSS-induced tumors, the Wnt signaling pathway was not significantly altered in our loss-of-function experiments in cell lines, though tending to be higher in LGR5 high spheres compared with LGR5 low adherent cells (P = 0.16). The biological difference of cell lines being cultured without their physiological context might explain these findings. Increasing evidence indicates that not only intrinsic factors but also extrinsic factors like the microenvironment can influence the CSC phenotype. For instance, Vermeulen et al. (48) showed in CRC that stromal myofibroblasts surrounding CSCs not only can maintain Wnt signaling activity in CSCs but also can activate Wnt signaling in more differentiated tumor cells and thereby induce the CSC phenotype. Furthermore, there is a methodological difference that might contribute to the heterogeneity of our findings in cell lines and ex vivo isolated tumor cells. In the ex vivo model, flow-sorted Lgr5 high and low cells were compared, whereas in the cell line experiment, LGR5 was silenced actively via shRNA and compared with the original cell population. Overall, the observed heterogeneity across the different lines and models reflects one of the major problems in CSC research.

In summary, our comprehensive functional and molecular analysis of LGR5 in CRC cell lines and AOM/DSS-induced mouse colon tumors conclusively links LGR5 to stem-like cells in CRC. LGR5 did not only serve as a marker for these stem-like CRC cells but was also of functional relevance for CRC cells, thus representing a potential therapeutic target, in particular as conditional deletion of Lgr5 in mouse guts does not seem to negatively affect normal intestinal epithelium (8). To further specify the role of LGR5 in human CRC, studies of LGR5 in primary human CRC specimens will be needed in the future and, as a prerequisite, the development of a reliable LGR5 antibody.

Supplementary material

Supplementary Material and methods, Table S1 and Figures S1–S4 can be found at http://carcin.oxfordjournals.org/

Funding

Intramural Research Program, National Institutes of Health, National Cancer Institute; German Academic Exchange Service (D.H.).

Glossary

Abbreviations:

AOM

azoxymethane

APC, adenomatous polyposis coli; CRC

colorectal carcinoma

CSC

cancer stem cell

DSS

dextrane sodium sulfate

EGFP

enhanced green fluorescent protein

LGR5

leucine-rich-repeat-containing G-protein-coupled receptor 5

mRNA

messenger RNA

qRT–PCR

quantitative reverse transcription–polymerase chain reaction

shRNA

short hairpin RNA.