|Home | About | Journals | Submit | Contact Us | Français|
Maintenance of adult tissues is carried out by stem cells and is sustained throughout life in a highly ordered manner1,2. Homeostasis within the stem cell compartment is governed by positive and negative feedback regulation of instructive extrinsic and intrinsic signals3,4. ErbB signalling is a prerequisite for maintenance of the intestinal epithelium following injury and tumour formation5,6. As ErbB family ligands and receptors are highly expressed within the stem cell niche7, we hypothesise that strong endogenous regulators must control the pathway in the stem cell compartment. Here we show that Lrig1, a negative feedback regulator of the ErbB receptor family8-10, is highly expressed by intestinal stem cells and controls the size of the intestinal stem cell niche by regulating the amplitude of growth factor signalling. Intestinal stem cell maintenance has so far been attributed to a combination of Wnt and Notch activation and Bmpr inhibition11-13. Our findings reveal ErbB activation as a strong inductive signal for stem cell proliferation. This has implications for our understanding of ErbB signalling in tissue development, maintenance and the progression of malignant disease.
The intestine constitutes an excellent system for studying stem cell function. Intestinal stem cells (ISC) reside at the bottom of crypts, where they are maintained in a multipotent and self-renewing state14. The ISC niche in the small intestine is composed of stem cells above and interspersed between Paneth cells and surrounded by mesenchymal cells7,15. This provides a unique microenvironment that enables constant contribution from stem cells to sustain the high cell turnover of the differentiated compartment16,17. Lgr5 expressing crypt based columnar cells at the bottom of crypts are together with cells located immediately above this region responsible for the life long steady state maintenance of the epithelium14,15,18,19. It is well established how ISCs are maintained and which inductive signals are required for tissue maintenance, however, very little is known with regard to the regulation of these pathways in vivo.
We recently demonstrated that Lrig1 controls stem cell proliferation in the epidermis20. In situ hybridisation now reveals that Lrig1 is highly expressed in the stem cell niche of the small intestine and colon (Figure 1a; Supplementary Figure 1). Lgr5 expressing stem cells can be identified by their high levels of GFP in the Lgr5−eGFP-ires-CreERT2 knock-in mouse14. GFP is subsequently diluted during successive cell divisions of stem cells (GFPhigh) and their early daughter cells (GFPmid and GFPlow). Expression analysis of these different populations demonstrates that Lrig1 levels are highest within the ISCs (GFPhigh vs. GFPmid: 1.6x; GFPhigh vs. GFPlow: 2.2x). Immunofluorescence staining and flow cytometric analysis confirm the overlap of Lrig1 and Lgr5 expression at the protein level and that Lrig1 is expressed in a gradient with highest levels in ISCs and that it is absent from Paneth cells (Figure 1b-c; Supplementary Figure 2a-c). Overall, approximately 1/3 of all cells within the intestinal crypt express Lrig1 (Figure 1d). This comprises the entire CD24low/mid stem cell and progenitor compartment and includes all of the Lgr5+ve ISCs21 (Figure 1e-g). The percentage of Lgr5-GFP+ve cells within the Lrig1 expressing population can unfortunately not be determined due to the mosaic nature of the Lgr5−eGFP-ires-CreERT2 knock-in mouse, where a large fraction of Lgr5 expressing ISCs are GFP negative (http://jaxmice.jax.org/strain/008875.html; Supplementary Figure 1). Characterisation of Lrig1+ve cells isolated by flow cytometry demonstrates that markers, which define ISCs such as Lgr5, Ascl2 and Msi1, are all enriched in this population (Fig. 1h)19,22. We also observe differential expression of multiple transcripts for Lrig1 interaction partners, in particular Egfr (6.2x, Fig. 1h). Expression of Lrig1 is associated with reduced proliferation in other tissues20. Therefore, we test whether Lrig1 expressing cells are mainly quiescent or proliferative. Pulse-chase studies show that 15-20% of the Lrig1 expressing cells incorporate BrdU within 1 hour following one injection and 60% are labelled after a three-day pulse with 7 injections (Figure 1i). The BrdU label is subsequently lost or significantly reduced following a 1-week chase (Supplementary Figure 2d-e). We conclude that Lrig1 is highly expressed by all ISCs, that the majority of Lrig1 expressing cells are proliferating and that especially the expression of Egfr is enriched within the Lrig1+ve population.
Work with Lrig1 knock out (KO) mice has until now been carried out on an outbred genetic mouse background20,23. To eliminate the contribution of genetic variability in the interpretation of the effects of Lrig1, we backcrossed Lrig1 KO mice to an Fvb/N background for more than 7 generations. On this background, no differences are observed between heterozygous and wildtype littermates. Yet, Lrig1 KO animals are smaller than their littermates from postnatal day 8 onwards and have to be sacrificed around postnatal day 10 due to the severity of the phenotype. Whilst the sizes of most organs are in relative proportion to the reduced body weight, the relative size of the intestine is grossly enlarged in Lrig1 KO mice (Figure 2a-d; circumference: 7.3±0.5mm for WT/Het (n=5) and 11.0±1.0mm for KO (n=4); p=0.01).
Histological examination of the intestine reveals that loss of Lrig1 causes a dramatic increase in crypt size along the entire length of the small intestine (Figure 2e-f; Supplementary Figure 3a). This is associated with an increased number of proliferating cells as measured by expression of phospho-histone H3 (Figure 2g-h, pHistoneH3/crypt: 3.9±0.9 fold KO/ctrl, p=0.05). Phenotypically, crypts from KO mice and littermate controls are indistinguishable until postnatal day 6 (Supplementary Figure 3b). This coincides with stem cell specification in the neonatal intestinal epithelium24. Crypt hyperplasia can be observed from this time-point onwards in Lrig1 KO animals. Gene expression profiling reflects the phenotypic change associated with loss of Lrig1 as gene set enrichment analysis (GSEA) shows overrepresentation of crypt signature genes within the up-regulated genes (p=1.5×10−177)25.
The enlarged crypt morphology can be explained by a specific increase in the transit-amplifying compartment, by a disproportionate increase in individual crypt cell lineages, or by increased numbers of ISCs with a proportionate increase in transit-amplifying and Paneth cells. The major intestinal cell lineages can be identified by their differential expression of CD24 and binding of Ulex Europaeus Agluttinin (UEA-I) by secretory cells (Supplementary Figure 4). Loss of Lrig1 causes a profound increase in the proportion of the CD24low/mid/UEA-Ineg stem cell and progenitor compartment (D) and CD24high/UEA-Ipos Paneth cells (B), whereas the population comprising mature enterocytes is proportionally decreased (C) (Figure 3a-b). Goblet (A) and enteroendocrine (E) cell populations show no significant change. In addition to their increased numbers, we observe increased expression of ISC markers within the CD24low/mid/UEA-Ineg population supporting an expansion of the stem cell compartment upon loss of Lrig1 (Figure 3c). To confirm this finding, we perform in situ hybridisation for Olfm4, a marker of ISCs22, and Cryptdin6, a marker of Paneth cells26. Starting from postnatal day 6 we observe an increased number of Olfm4 positive cells in Lrig1 KO samples and from postnatal day 10 an increased proportion of Cryptdin6 positive Paneth cells (Figure 3d). The dynamic expansion of the stem cell compartment is supported by flow cytometric analysis of dissociated cells from the intestinal epithelium (Supplementary Figure 5a-e). The expression domain of the independent stem cell marker Msi1 is also expanded upon loss of Lrig1 (Supplementary Figure 5f-g). Taken together, this supports a direct effect of Lrig1 on ISCs, which subsequently affects the size of ISC compartment and the crypt.
Lrig1 interacts with the ErbB family, cRet and cMet in vitro and reduces signalling strength by negatively regulating both protein levels and the activity of the growth factor receptors8-10,27,28. Epithelial cells in the intestine express detectable levels of Egfr, ErbB2, ErbB3 and cMet (Figure 1h and and4a).4a). ErbB signalling is mediated via ligand stimulation of Egfr and ErbB3, which upon homo- and heterodimerisation with each other or with the orphan receptor ErbB2 activates downstream signalling cascades29. cMet on the other hand is activated upon HGF stimulation. Loss of Lrig1 causes a pronounced increase in the protein levels of all of these interaction partners irrespective of transcript levels (Figure 4a-b). This is associated with a concomitant increase in the activity of the ErbB family, but with no detectable effect on the activation of cMet (Figure 4a and data not shown). The increased ErbB activation is reflected by strong up-regulation of MAPK signalling and increased levels and activity of the mitogen-induced transcription factor cMyc (Figure 4c-f; GSEA of the Myc dependent gene network30: p=3.0×10−18). In order to address whether the increased levels of ErbB signalling reflects merely the expansion of the crypt or if loss of Lrig1 alters signalling dynamics we utilise two independent approaches. Firstly, phosphorylation of Egfr (pEgfr) is a direct measure for receptor activation in vivo. At postnatal day 10 phosphorylation can be observed on UEA-1pos secretory cells and progenitor cells (Supplementary Figure 6). In control animals, pEgfr levels are low at the crypt bottom and peak around cell position 8 (Figure 4g,i). Of note, this is the reverse expression pattern of Lrig1. In the Lrig1 KO, Egfr activation is uniform within the crypt (Figure 4h,i). Secondly, expression analysis of isolated cell populations signifies the increased activation of Myc within ISCs and progenitors upon loss of Lrig1 (Figure 4j). We conclude that loss of Lrig1 affects ErbB signalling within the ISC compartment.
Next, we functionally test in a defined 3D intestinal culture system whether ErbB inhibition is the main function of Lrig1. ISC self-renewal and proliferation require stimulation of the Egfr and the Wnt pathways, and inhibition of the Bmpr pathway31. In wild type samples, stimulation of ErbB signalling with exogenous ligands is a prerequisite for the maturation from spheres into budding organoids. Interestingly, Lrig1 KO spheres mature into budding organoids in the absence of exogenous ErbB ligands (Figure 4k-l). Nevertheless, ErbB signalling, like stimulation of the Wnt pathway and inhibition of BMPR signalling, is an obligate requirement for growth of both WT and KO organoids, as inhibition of ErbB signalling with a number of ErbB inhibitors rapidly affects cell survival (data not shown). Importantly, loss of Lrig1 does not cause increased expression of ErbB ligands (Figure 4m). We conclude that upon loss of Lrig1, endogenous ErbB ligands are sufficient to support organoid growth due to reduced ErbB inhibition in ISC. Moreover, as the phenotype observed in vivo can be recapitulated in vitro in a purely epithelial culture, the function of Lrig1 is likely to be crypt autonomous.
To address in vivo whether activation of the ErbB family is responsible for the observed phenotype, neonatal mice were treated daily with the ErbB inhibitor Gefitinib. Following administration of inhibitor in vivo, crypt proliferation as well as ISC and Paneth cell numbers is restored to normal levels in the Lrig1 KO animals (Figure 5a-l). Moreover, treatment with Gefitinib reduces pEgfr levels (Figure m-p). Gefitinib-treated KO animals are still significantly smaller than control animals, and we hypothesise that Lrig1 has important non-reversible functions in other tissues such as the stomach, where Lrig1 is expressed by a small sub-population of cells within the pyloric glands (M.P. and K.B.J. Unpublished observations). The proposed mechanism of Lrig1 as an inhibitor of ErbB function in vivo is further substantiated by genetic rescue of the phenotype by crossing Lrig1 KO animals to hypomorphic Egfrwa-2 mice (Figure 5q-t). The Egfrwa-2 mouse strain harbours a missense mutation in the kinase domain of Egfr, which compromises the activity of the receptor upon ligand stimulation32. Homozygosity for Egfrwa-2, when crossed with Lrig1 KO mice, is associated with perinatal lethality irrespective of Lrig1 status. Analysis of the Egfrwa-2 heterozygous animals at postnatal day 10 demonstrates that the phenotypical changes associated with loss of Lrig1 are rescued in approximately 40% of the mice (n=15 out of 37, p=0.0095). Thus, the phenotypical changes observed upon loss of Lrig1 can be rescued by both pharmacological and genetic modulation of endogenous ErbB activity, hereby demonstrating ErbB activity as the main target of Lrig1 in vivo.
Our data support a model whereby Lrig1 regulates proliferation within the ISC niche by inhibiting ErbB signalling. Expression of Lrig1 enables stem cells and progenitors to fine-tune their cellular response upon ligand stimulation and ensures that tissue homeostasis is maintained (Figure 5u). Proliferation and maintenance of the ISC compartment have until now been attributed to Wnt and Notch activation in combination with BMPR inhibition11-13. We show that ErbB signalling is a strong mitotic signal for ISCs, and when uninhibited leads to a rapid expansion of the stem cell compartment. The prominent role of ErbB signalling in the ISC compartment is supported by recent evidence from Drosophila, where signalling via Egfr, the only ErbB receptor in Drosophila, is required for intestinal maintenance33,34. Moreover, the natural response to injury and tumour initiation in the mouse intestine, which includes stem cell activation, is severely compromised upon loss of Egfr and ErbB35,6. Analogous to our observations, eliminating one functional copy of the Egfr significantly alters the injury response to sub-lethal doses of radiation35. In the case of injury models it remains to be shown whether Egfr and ErbB3 function is required for expansion of the stem cell compartment or as a general mitotic stimuli.
Due to the reported anti-proliferative effect of Lrig1 in human and murine epidermis9,20, we have assessed the proliferation status of Lrig1 expressing cells. As predicted from the overlap in Lrig1 and Lgr5 expression, we observe that the majority of Lrig1 expressing cells like Lgr5+ve cells are highly proliferative. We cannot rule out that mitotic heterogeneity exists within the Lrig1+ve population; however, it is evident that Lrig1 does not define a population of quiescent cells in the intestine. The prominent effect on the stem cell compartment upon loss of Lrig1 does however highlight that ISCs are exposed to multiple signals that promote proliferation and that distinct inhibitory factors like Lrig1 are required to control stem cells.
We unambiguously demonstrate that Lrig1 is a key regulator of tissue homeostasis in the intestinal epithelium. As multiple tissue-specific stem cells rely on Lrig1 to regulate their behaviour, we propose that Lrig1 provides a general mechanism to control tissue homeostasis. Based on the study described here, and the non-overlapping expression pattern of Lrig1 and Lgr5 in stem cells of the epidermis20,36, we anticipate that distinct cellular microenvironments, which can either converge as in the intestine or diverge as in the epidermis, specify stem cell behaviour in vivo.
FVB/N mice were obtained from Charles River. Lrig1 KO mice23, Egfrwa-2 and Lgr5−EGFP-IRES-CreERT2 knock in mice14 have been described. EgfrWa-2 mice crossed to Fvb/N for 1 generation were bred with Lrig1 KO. The offspring was subsequently intercrossed in order to obtain Lrig1 KO mice on the hypermorphic Egfrwa-2 background. For DNA labelling experiments mice received single or repeated intraperitoneal injections of 100μL 10mg/mL BrdU (10mg/mL) every 12-hour. For experiments with Gefitinib (Cambridge Bioscience, Cambridge, UK), neonatal mice received daily intraperitoneal injections of 20μg/g bodyweight of Gefitinib. All in vivo experiments were performed under the terms of a UK Home Office license.
Intestinal samples were fixed and embedded in paraffin using standard protocols. Analysis of Lgr5-GFP/Lrig1 and UEA-1/pEgfr was carried out as recently described22.
The following antibodies were used rabbit anti-β-catenin (Santa Cruz, sc-7199, 1:250), rabbit anti-Phospho-histone H3 (ser10) (Cell Signaling #9701, 1:400), rabbit anti-phospho-MEK1/2 (Ser221) (Cell signalling, #2338, 1:500), mouse anti-BrdU (Cell signaling, Bu20a, 1:250), goat anti-Lrig1(R and D Systems, AF3688, 1:100), rabbit anti-Myc (Millipore, 06-340, 1:100), mouse anti-Ki67 (Monosan, Clone MM1, 1:1000), PE-conjugated anti-Lrig1 (R and D systems, FAB3688P, 10μL/106 cells), Pacific blue and Alexa647 conjugated Rat anti-CD24 (Biolegend, M1/69, 5μL/106 cells), rat anti-Msi1 (MBL, D270-6, 1:10000), rabbit anti-phospho Egfr (Y1068) (Abcam, EP7774Y, 1:250), rabbit anti-Egfr (Cell Signaling, #4267 clone D38B1, 1:250), rabbit anti-phospho ErbB2 (Y1248) (Abcam, ab47755), mouse anti-ErbB2 (Millipore, 05-1130, Clone N3D10, 1:500), rabbit anti-phospho ErbB3 (Y1289) (Cell Signaling, #4791, Clone 21D3, 1:250), mouse anti-ErbB3 (Millipore, 05-390, Clone 2F12, 1:500), rabbit anti-cMet (Santa Cruz, sc-162, 1:500), mouse anti-β-actin, (Sigma A5316, clone AC-74, 1:5000) and Atto-488 conjugated UEA-1 (Sigma, 10μL/106 cells for flow cytometry, 1:1000 for near-native sections) according to manufacturer’s instructions. For confocal microscopy samples were imaged using a LSM700 confocal microscope or Leica SP5 TCS. Z-stacks were acquired at optimal stack distance and at 1024×1024 dpi resolution. Maximum intensity projections of Z-stacks were generated in Image J.
One cm pieces of the jejunum from postnatal day 10 Lrig1 KO and control littermates were rinsed with ice cold PBS and snap frozen on dry ice. RNA was isolated from homogenised intestine or flow-sorted cells and either used directly for microarray experiments or for qPCR as described20.
Gene specific expression assays (Applied Biosystems) or Sybr green analysis (Invitrogen) with optimised primer-pairs were used for qPCR on an Applied Biosystems 7500HT RealTime PCR System (Applied Biosystems, Foster,USA). Samples were normalised using the ΔCt method. For clustering analysis average expression levels from at least three independent biological replicates were converted to z-scores (z=(value-average value)/stdev) and plotted using heatmap.2′ function from the Bioconductor ‘gplots’ library.
RNA was quality controlled for concentration, purity and integrity using spectroStar omega (BMG labtech) and Bioanalyser (Agilent). The amplification was performed using the TotalPrep 96-RNA Amplification kit (Ambion). Total RNA (~300ng) was reverse transcribed into cDNA and amplified by in vitro transcription to generate biotin-labelled cRNA. cRNA (1500 ng) was hybridised to whole genome bead arrays (MouseWG-6 v2.0 Expression BeadChip) according to the direct hybridisation assay from Illumina and scanned using an Illumina BeadArray scanner. Bead level data from all hybridizations was background corrected, using default parameters of the RMA algorithm37, and summarised using the beadarray package for the Bioconductor suite (http://www.bioconductor.org) for the R statistical programming environment (http://www.r-project.org). Processed sample expression profiles were quantile normalised using the limma package for Bioconductor prior to analysis of differential regulation between sample groups with the moderated t-statistic of the same package38. In order to reduce errors associated with multiple hypothesis testing, the significance p-values obtained were converted to corrected q-values using the false discovery rate (FDR) method39, as implemented in the qvalue package for Bioconductor. Differential regulation between two sample groups was deemed significant at a threshold of q < 0.01 (FDR 1%, Supplementary Table 1).
In order to generate a reference gene universe for gene set enrichment analysis we defined the intestinal transcriptome as genes detected in three replicates from KO and control samples (13,298 unique genes, 2,575 up-regulated genes, 2,553 down-regulated genes; Supplementary Table 2). For GSEA using Fisher’s exact test we obtained a crypt gene signature25 (1,692 genes in reference list, 802 genes up-regulated; Supplementary Table 3) and a Myc dependent gene network30 (374 genes in reference list, 144 genes up-regulated; Supplementary Table 4).
For protein isolation, intestinal epithelium was extracted as for organoid cultures and lysed in RIPA buffer.
DIG in situ hybridisation was carried out essentially as described before using IMAGE clones40.
Intestinal epithelial cells were isolated as described22. To generate a single cell suspension, cells were incubated with trypsin22 or with 0.25mg/mL Thermolysin for 1 minute at 37°C in PBS supplemented with 1%BSA for trypsin sensitive antigens such as Lrig1. Cell sorting was carried out using a FACSAria (BD Biosciences) for isolation of cells based on CD24 and UEA-1, and a MoFlo (Dako Cytomation) for isolation of Lrig1 and Lgr5-eGFP expressing cells. Flow cytometric analysis was carried out on a CyAn ADP analyzer (Dako Cytomation), and data was processed in FlowJo.
The BrdU labelling analysis was performed with the APC-BrdU flow cytometry kit (BD biosciences).
Primary crypts were cultured according to Sato et al.31 using reduced concentrations of murine recombinant R-spondin1 (500ng/mL – R&D Systems) and varying concentrations of EGF (Peprotech). Organoid structures were imaged at day 6. Cell proliferation was measured by BrdU incorporation by incubation with 20μM BrdU (Roche) for 1 hour at 37°C before fixation.
Confocal images of intestinal samples stained for pEgfr were analysed in ImageJ to determine the levels of membrane-localised pEgfr. Intra-crypt intensities were normalised to the average intensity of position 6-10. This corresponds to the peak in control samples.
Organoids in matrigel were observed under phase contrast using an Axiovert 200M microscope (Zeiss) equipped with an AxioCam MRc (Zeiss). Images were interactively analyzed using ImageJ to determine perimeter and area of individual organoids and calculate the branching coefficient (1-4π(Area/perimeter2)) of the formed structures. Images were acquired from a minimum of 8 organoids per condition derived from independent KO and control samples,
Statistical significance of quantitative data was determined by applying a two-tailed Student’s t-test to raw values or to the average values obtained from analysis of independent organoid experiments. A two-tailed Fisher’s exact test was used to analyse the significance of the genetic rescue of the Lrig1 KO phenotype, and to determine the significant overlap between different gene lists.
We thank Erwin Wagner, Michaela Frye, Stuart Yuspa, Bill Otto, Rachael Walker, Rosemary Jeffery, Harry Begthel, Margaret McLeish, the WTCSCR Biofacility and Cambridge Genomics Services for reagents, technical assistance and advice, and Robert Fordham and Robert Williams for critical comments on the manuscript. We gratefully acknowledge support from the MRC and Wellcome Trust.
Author information: The authors declare no competing financial interests.
Microarray data have been deposited at the EBI ArrayExpress under accession number: E-MTAB-378.