Lrig1 is a haploinsufficient tumor suppressor gene in malignant glioma

Recently, a genome-wide association study showed that a single nucleotide polymorphism (SNP) —rs11706832—in intron 2 of the human LRIG1 (Leucine-rich repeats and immunoglobulin-like domains 1) gene is associated with susceptibility to glioma. However, the mechanism by which rs11706832 affects glioma risk remains unknown; additionally, it is unknown whether the expression levels of LRIG1 are a relevant determinant of gliomagenesis. Here, we investigated the role of Lrig1 in platelet-derived growth factor (PDGF)-induced experimental glioma in mice by introducing mono-allelic and bi-allelic deletions of Lrig1 followed by inducing gliomagenesis via intracranial retroviral transduction of PDGFB in neural progenitor cells. Lrig1 was expressed in PDGFB-induced gliomas in wild-type mice as assessed using in situ hybridization. Intriguingly, Lrig1-heterozygous mice developed higher grade gliomas than did wild-type mice (grade IV vs. grade II/III, p = 0.002). Reciprocally, the ectopic expression of LRIG1 in the TB107 high-grade human glioma (glioblastoma, grade IV) cell line decreased the invasion of orthotopic tumors in immunocompromised mice in vivo and reduced cell migration in vitro. Concomitantly, the activity of the receptor tyrosine kinase MET was downregulated, which partially explained the reduction in cell migration. In summary, Lrig1 is a haploinsufficient suppressor of PDGFB-driven glioma, possibly in part via negative regulation of MET-driven cell migration and invasion. Thus, for the first time, changes in physiological Lrig1 expression have been linked to gliomagenesis, whereby the SNP rs11706832 may affect glioma risk by regulating LRIG1 expression.


Introduction
Both adult and pediatric diffuse gliomas are devastating diseases with considerable morbidity and poor cure rates. Although these diseases share clinical and molecular features, they are also distinct. Gliomas frequently display dysregulated growth factor signaling; it is estimated that 88% of adult glioblastomas harbor mutations that affect the receptor tyrosine kinase (RTK)/RAS/PI(3)K signaling axis 1 . For example, 35% of proneural glioblastomas have focal amplification of PDGFRA (which encodes plateletderived growth factor receptor alpha, PDGFRA) whereas 11% have mutations in PDGFRA; 97% of classical glioblastomas have high-level EGFR (encoding epidermal growth factor receptor, EGFR) amplification whereas 55% have mutations in EGFR; and 37% of mesenchymal glioblastomas have a deletion of NF1 (encoding neurofibromin 1, NF1) 2 . Indeed, the PDGFA gene was recently shown to drive all non-CpG island methylator phenotype glioblastomas 3 . In contrast to adult diffuse gliomas, which typically originate in cerebral white matter, most pediatric diffuse gliomas arise in the pons region of the brainstem (diffuse intrinsic pontine glioma, DIPG) 4 . The mutational profiles of DIPGs are different from those of adult diffuse gliomas. For example, DIPGs frequently harbor mutations in histone H3 (typically p.Lys27Met mutations in H3F3A or HIST1H3B in 88% of cases) 5 and activating mutations in ACVR1 (encoding ACVR1/ALK4, a bone morphogenetic protein type I receptor; 20-32% of cases) [5][6][7] . However, similar to many adult diffuse gliomas, DIPGs frequently show focal gains in PDGFRA (36-40% of cases) 6,8 , including activating point mutations of the gene (5% of DIPG cases) 9 . Thus, although there are important differences in diffuse gliomas between adults and children, the frequent activation of growth factor receptor signaling in general, particularly PDGFR signaling, appears to be common in gliomas that manifest in both populations.
Recently, a genome-wide association study revealed that a single-nucleotide polymorphism (SNP) in intron 2 of LRIG1 (encoding leucine-rich repeats and immunoglobulin-like domains (LRIG)-1) influences the risk of occurrence of diffuse glioma 10 . LRIG1 is an integral membrane protein belonging to the LRIG family [11][12][13] and negatively regulates various oncogenic RTKs, including EGFR 14,15 , EGFRvIII 16 , hepatocyte growth factor receptor (MET) 17 , RET proto-oncogene product (RET) 18 , and PDGFRA 19 . Lrig1 knockout mice have hyperproliferative epidermal and intestinal stem cells [20][21][22][23] . Moreover, LRIG1 expression is a good prognostic indicator of a variety of human cancers 24 . The non-physiological overexpression of LRIG1 has been shown to inhibit the proliferation of certain glioma cell lines in vitro 16,25,26 . Additionally, a soluble form of the LRIG1 ectodomain can inhibit EGFR signaling in trans as well as suppress the proliferation of glioma cells in vitro and the growth of human glioma xenografts in vivo [26][27][28] . However, the physiological role of LRIG1 in gliomagenesis has not been experimentally investigated to date.
In the present study, we investigated the role of physiological Lrig1 expression in PDGFB-induced glioma in mice and analyzed the effects of forced LRIG1 overexpression on human glioblastoma xenografts in vivo and on human glioblastoma cells in vitro.

Lrig1 was expressed in PDGFB-induced mouse gliomas
To address the role of LRIG1 in PDGF-driven gliomas, we used the RCAS/Ntv-a system to induce glioma in mice with different Lrig1 genotypes via the intracranial transduction of neural progenitor cells with PDGFB-encoding RCAS viruses. In this glioma model, most PDGFBtransduced mice develop gliomas within 12 weeks of age; these lesions present either oligodendroglial or glioblastoma-like morphology 19,29 . We have previously shown that these tumors express Lrig2 19 . To investigate whether these tumors also express Lrig1, we performed in situ hybridization (Fig. 1a) and observed that the PDGFB-induced gliomas as well as normal mouse brain tissue expressed Lrig1. Quantification of the in situ hybridization signals for Lrig1 from three normal brains, three grade II-III tumors, and three grade IV tumors showed pronounced intra-group variability and a lack of a consistent difference among the groups (Fig. 1b).

Generation and characterization of Lrig1-deficient Ntv-a mice
To further explore the role of Lrig1 in PDGFB-driven malignant gliomas, Lrig1-deficient mice were generated by knocking out exon 1 of Lrig1 (Fig. 2a). The Lrig1deficient mice (B6.129-Lrig1 tm1.1Hhed ) showed no signs of illness or any other overt phenotype other than psoriasislike lesions on their tails as previously described in another Lrig1-deficient mouse strain 20 . The quantitative reverse transcription polymerase chain reaction (RT-PCR) results demonstrated that compared to the wild-type mice, 1-week-old Lrig1-deficient mice presented no detectable Lrig1 transcripts in their brains, and mice heterozygous for this deficiency presented reduced levels (Fig. 2b). The Western blot results confirmed the complete absence of full-length Lrig1 protein in the brains of Lrig1-deficient mice and reduced levels in heterozygous mice compared to the levels in wild-type mice (Fig. 2c, d).
Malignancy of PDGFB-induced gliomas was enhanced by the ablation of one Lrig1 allele The Lrig1-deficient mouse strain was crossed with transgenic Ntv-a mice followed by backcrossing of those offspring with the Ntv-a mice; this husbandry setup generated mice homozygous for Ntv-a and heterozygous for Lrig1 (Ntv-a+/+; Lrig1+/−). These mice were intercrossed, and their offspring were injected with PDGFBencoding RCAS virus-producing DF-1 cells at birth and analyzed at 12 weeks of age. The mice developed no tumors, low-grade tumors (grade II-III), or glioblastomalike high-grade tumors (grade IV) (Fig. 3). The overall tumor incidence was similar among the genotypes; however, the incidence of high-grade tumors was significantly higher in the Lrig1-heterozygous mice than that in the wild-type mice (19.6 vs. 1.6%, respectively; Pearson's test, p = 0.002). The differences in the incidence of low-grade and high-grade tumors between the wild-type and knockout mice or between the heterozygous and knockout mice were not statistically significant (Pearson's χ 2 test; p = 0.173 and p = 0.204, respectively).

Invasion of TB107-derived malignant glioma was suppressed by LRIG1 overexpression in vivo
Because ablating one copy of Lrig1 resulted in a decrease of Lrig1 expression and an increase in the tumor grade (grade IV vs. grade II-III), we surmised that there exists a relationship between the reduced expression of Lrig1 and the increased malignancy of tumors. Therefore, we tested the complementary hypothesis, i.e., whether increasing LRIG1 expression in high-grade glioma cells would reduce their malignancy. To this end, we transduced two human high-grade glioblastoma cell lines, TB101 and TB107, with lentiviral vectors expressing either LRIG1 (pLVX-LRIG1) or empty control vector (pLVX-Puro) and assessed the behavior of the transduced cells in immunocompromised mice in vivo. The transduced cells were orthotopically transplanted into the brains of immunodeficient BALB/cA nude mice. Four months after transplantation, the mice were sacrificed and analyzed. The incidence of tumors among the mice transplanted with TB101 Puro , TB101 LRIG1 , TB107 Puro , and TB107 LRIG1 were 11/12, 9/11, 10/10, and 12/12, respectively. Unfortunately, the primary tumors of the mice transplanted with TB101 cells were very small with few satellite colonies; these lesions were difficult to delineate. Thus, no further analyses of the TB101 Puro and TB101 LRIG1 groups were performed. By contrast, the primary tumors in the mice transplanted with TB107 cells were easily identifiable. LRIG1 overexpression resulted in the significantly reduced invasion of TB107 LRIG1 tumors Fig. 1 In situ hybridization of Lrig1 in normal brain tissue and PDGFB-induced glioma in mice. Newborn Ntv-a mice were transduced with PDGFB-expressing avian retroviruses. At 12 weeks of age, the mice were sacrificed, and their brains were dissected and analyzed via in situ hybridization. a Micrographs showing images of in situ hybridized sections of normal brain (upper row), grade II glioma (middle row), and grade III-IV glioma (lower row). The left panel shows sections hybridized with the negative control probe, whereas the right panel shows sections hybridized with the Lrig1 probe. Scale bar, 100 μm. b Quantification of the total number of signals (single spots and clusters) per area in normal brain (n = 3), grade II glioma (n = 3), and grade III-IV glioma (n = 3). There was no significant difference in Lrig1 expression (in situ hybridization signals per area) among the groups into the surrounding brain tissue compared to that of the TB107 Puro tumors (Fig. 4a, b). Thus, the number of tumor satellites was significantly lower in the LRIG1overexpressing TB107 LRIG1 tumors than that in the TB107 Puro tumors (Student's t-test, p = 0.02). In vitro, compared with the control TB107 Puro cells, the LRIG1overexpressing TB107 LRIG1 cells did not exhibit any changes in their proliferation rate (Fig. 4c); however, migration was reduced in the LRIG1-overexpressing TB107 LRIG1 cells (Fig. 4d). Intriguingly, the TB101 groups showed an inverse pattern, i.e., TB101 LRIG1 cells exhibited a lower proliferation rate than the TB101 Puro cells and an unaltered migration rate (Fig. 4c, d).

Cell migration was suppressed by ectopic human LRIG1 expression in a context-dependent manner
To further analyze the effects of LRIG1 overexpression on the migration of glioma cells, we transduced TB101 and TB107 cells with a doxycycline-inducible LRIG1 allele and analyzed their migration with or without the induction of ectopic LRIG1 expression. We focused on cell migration because it represents the invasive capacity of tumor cells. The induction of ectopic LRIG1 expression suppressed the migration of TB107 DoxLRIG1 cells but did not affect the migration of TB101 DoxLRIG1 cells (Fig. 5a-c). Both exon 1 and the PKG-neo cassette were flanked by loxP sites; thus, both regions could be deleted in a single step by mating the mice with OzCre mice to generate the disrupted Lrig1 allele. b Relative Lrig1 mRNA expression levels in mouse brains from wild-type (WT), Lrig1-heterozygous (HZ), and Lrig1-deficient (KO) mice. Total RNA was prepared from the brains of 1-week-old mice, and the Lrig1 expression levels were determined with quantitative RT-PCR. The Lrig1 levels were normalized to those of the internal control gene Rn18s. An arbitrary scale was established such that the mean Lrig1 levels in the mouse reference RNA had a value 1. In the Lrig1-deficient mice, Lrig1 could not be detected in any of the samples analyzed. Error bars indicate ±standard deviations. †p = 6.6 × 10 −7 , Student's ttest. c Western blot analysis of brain lysates from 1-week-old wild-type (WT), Lrig1-heterozygous (HZ), and Lrig1-deficient (KO) mice using antibodies against Lrig1 or actin. d Quantification of Western blots of brain lysates from wild-type mice, Lrig1-heterozygous mice, and Lrig1-knockout mice. The mean Lrig1/actin ratios are shown on an arbitrary scale. Error bars indicate ±standard deviations. *p = 0.025, **p = 0.0045, and ***p = 0.00088, Student's t-test

LRIG1 overexpression inhibited the phosphorylation of MET but not of EGFR in TB107 cells
To investigate the mechanism by which LRIG1 overexpression inhibits migration in TB107 DoxLRIG1 cells but not in TB101 DoxLRIG1 cells, we analyzed the levels of phosphorylated RTK (pRTK) using a phospho-RTK array in TB101 DoxLRIG1 and TB107 DoxLRIG1 cells with endogenous or ectopic LRIG1 expression (Fig. 5d). In the noninduced TB101 DoxLRIG1 cells, EGFR, the insulin receptor, IGF1R, and AXL were phosphorylated (Fig. 5d, left, upper panel). Following the induction of LRIG1 overexpression for 48 h, the phosphorylation levels of these proteins remained largely unaffected (Fig. 5d, left, lower panel). In the non-induced TB107 cells, EGFR, MET, the insulin receptor, IGF1R, AXL, RET, and VEGFR2 were phosphorylated ( Fig. 5d, right, upper panel). Following the induction of LRIG1 overexpression for 48 h, the phosphorylation of MET and RET was strongly inhibited, whereas the levels of phosphorylated EGFR remained unaffected (Fig. 5d, right, lower panel). To confirm these observations in TB107 cells and to investigate the kinetics of the changes in pRTK levels, immunoblotting and enzyme-linked immunosorbent assay (ELISA) were performed. Immunoblotting confirmed that MET phosphorylation was reduced when LRIG1 was overexpressed, whereas EGFR phosphorylation was not affected (Fig. 5e). The reduction in MET phosphorylation was observed as early as 12 h after the induction of ectopic LRIG1 expression. However, we were unable to confirm any effect of LRIG1 on the phosphorylation of RET using a phosphorylated RET (p-RET)-specific ELISA (Fig. 5f). Based on these data, LRIG1 overexpression inhibits the phosphorylation of MET but not that of EGFR in TB107 DoxLRIG1 cells.

LRIG1-induced suppression of TB107 cell migration was partially dependent on MET inhibition
Because LRIG1 overexpression reduced phosphorylated MET (p-MET) levels in TB107 DoxLRIG1 cells, we investigated whether LRIG1-induced suppression of TB107 DoxLRIG1 cell migration was mediated through MET inhibition. Migration was analyzed in TB107 cells with endogenous or ectopic LRIG1 expression and in the presence or absence of the MET-specific kinase inhibitor PHA-665752 (Fig. 5g). Both treatment with PHA-665752 and LRIG1 overexpression suppressed the migration of TB107 DoxLRIG1 cells; furthermore, the observed suppression of migration was more pronounced in cells with LRIG1 overexpression than in cells treated with PHA-665752. In the former case, the suppressive effect was not further enhanced by treatment with PHA-665752. LRIG1 overexpression partially inhibited MET phosphorylation, whereas 500 nM PHA-665752 completely abolished all detectable MET phosphorylation (Fig. 5h, i). Thus, the migration of TB107 DoxLRIG1 cells was partially dependent on MET activity; however, the LRIG1-mediated suppression of migration of TB107 cells may also involve mechanisms other than those mediated by MET.

Discussion
Here, we showed that Lrig1 was a haploinsufficient tumor suppressor of PDGFB-induced experimental diffuse glioma in mice. To the best of our knowledge, this is the first demonstration of haploinsufficiency of any Lrig gene. The loss of a functional Lrig1 allele, which resulted in reduced mRNA and protein expression of Lrig1, promoted the progression of gliomas from grade II/III to grade IV. Our findings are particularly intriguing in the light of the recent demonstration that the SNP rs11706832 in intron 2 of LRIG1 constitutes a diffuse glioma susceptibility locus 10 . However, Melin and colleagues did not address the molecular function of rs11706832 in their study. Therefore, our current demonstration that the LRIG1 expression level is a determinant of the aggressiveness of PDGFB-driven gliomas may be highly relevant to the etiology of human gliomas.
Because the reduced Lrig1 expression increased the malignancy of low-grade diffuse glioma, we tested the complementary hypothesis, i.e., whether forced LRIG1 overexpression in human high-grade glioma cells (i.e., glioblastoma cells) can reduce their malignancy. Accordingly, LRIG1 overexpression in the primary human glioma cell line TB107 suppressed glioma invasion in vivo and cell migration in vitro. In vivo, LRIG1-overexpressing TB107 tumors exhibited a markedly lower intracranial dissemination than the control TB107 tumors. In vitro, the LRIG1-overexpressing TB107 cells did not display any change in the proliferation rate compared with the control cells; however, the LRIG1-overexpressing TB107 cells exhibited reduced migratory activity. Intriguingly, the other human glioblastoma cell line, TB101, exhibited contrary behaviors when LRIG1 was overexpressed, i.e., its migration rate was unaltered whereas its proliferation rate was reduced. Thus, LRIG1 may function as a tumor suppressor in glioma by suppressing cellular invasion and Effects of LRIG1 expression on the invasion, proliferation, and migration of glioma cells. TB101 and TB107 cells were transduced with either control vector, to produce TB101 Puro and TB107 Puro cells or a stable LRIG1 expression vector to produce TB101 LRIG1 and TB107 LRIG1 cells. a Nude mice were intracranially injected with TB107 Puro or TB107 LRIG1 cells. Four months after injection, the mice were killed, and their brains were analyzed using immunohistochemistry. Micrographs showing representative images of mouse brains transplanted with TB107 Puro cells or TB107 LRIG1 cells and stained for human vimentin to visualize the tumor cells. b Representation showing the quantification of the number of satellite tumors per section from mice with TB107 Puro (n = 10) and TB107 LRIG1 tumors (n = 12). Horizontal lines indicate the medians of the respective group. The difference in the number of satellite tumors surrounding the TB107 Puro and TB107 LRIG1 primary tumors was significant (Student's t-test, p = 0.02). c Proliferation rates of the respective cell lines. Error bars indicate ± standard deviations of the means from triplicate samples. One of three similar experiments is shown. The difference in proliferation rate between TB101 Puro and TB101 LRIG1 cells was significant (**p < 0.01, Student's t-test). d Migration of the respective cell lines. Cells were seeded in transwell inserts with serum-free medium in the upper chamber and medium containing 5% FBS in the lower chamber. After the cells were incubated 24 h, the inserts were fixed and stained with crystal violet, and the cells that migrated across the membrane were counted. The mean number of migratory cells per ten high-power fields in one of three similar experiments is shown; standard deviations of triplicate samples are indicated with error bars. ***p < 0.001, Student's t-test migration as well as cellular proliferation in a cell contextdependent manner. The LRIG1-mediated suppression of TB107 cell migration was accompanied by a reduction in MET phosphorylation, and treatment with the MET-specific kinase inhibitor PHA-665752 showed that TB107 cell migration was partially dependent on MET kinase activity. This outcome mirrors the results of a previous study, which showed that LRIG1 overexpression inhibited the migration of the breast cancer cell lines MDA-MB-157 and MDA-MB-231, at least in part, by inhibiting MET 30 . Intriguingly, TB101 cells neither expressed detectable phospho-MET, nor exhibited LRIG1-mediated suppression of migration. Although PHA-665752 completely inhibited MET kinase activity in TB107 cells, it could not suppress cell migration to the same degree as LRIG1 overexpression. Furthermore, despite LRIG1 overexpression strongly suppressing TB107 cell migration, this ectopic expression only partially inhibited MET kinase activity. Taken together, these results suggest that LRIG1 suppressed the migration of TB107 cells through a mechanism that is partially dependent on MET inhibition. The molecular mechanisms responsible for METindependent LRIG1-mediated suppression of migration in TB107 cells still requires elucidation. Nevertheless, our results show that LRIG1 suppresses the migration of glioma cells in a cell context-dependent manner partially via MET inhibition. LRIG1 has also been reported to negatively regulate RET 18 and EGFR. However, our results regarding the possible effects of LRIG1 on RET in TB107 cells were ambiguous; the phospho-RTK array showed a reduction in p-RET upon LRIG1 induction, but the p-RET-specific ELISA did not show such a change. Therefore, we could not definitively conclude whether LRIG1 overexpression could suppress RET in TB107 cells. Intriguingly, we did not detect any changes in the phosphorylation levels or protein expression of EGFR upon LRIG1 overexpression in TB107 cells. These results contradict those of several previous studies reporting that LRIG1 negatively regulates EGFR [14][15][16]22,23,[31][32][33] . However, our inability to detect the inhibition of EGFR phosphorylation in TB107 cells does not exclude the possibility that signaling downstream of EGFR was still impaired by LRIG1 expression. In a previous study, we found that LRIG2 affects transcriptional responses to PDGF stimulation even in the absence of detectable effects on PDGFR phosphorylation 19 .
In summary, our novel finding that Lrig1 is a haploinsufficient tumor suppressor gene in glioma provides a possible functional link between the SNP rs11706832 and the etiology of diffuse glioma. More broadly, the results also suggest that other modes of LRIG1 gene regulation, including promoter methylation, histone acetylation, and gene copy number variations, could be highly relevant in gliomagenesis and glioma progression both in children and in adults. Thus, it will be important to establish whether rs11706832 truly affects LRIG1 expression and/or whether LRIG1 is regulated by other mechanisms in glioma. Additionally, further elucidating the interactions between LRIG1 and other genes and mutations that regulate the initiation and progression of human diffuse One of three independent experiments is shown. In each experiment, triplicate wells were analyzed for each treatment, and the error bars indicate standard deviations. *p < 0.05, Student's t-test. c Immunoblots showing the induction of LRIG1 by doxycycline at various time points; actin is shown as the loading control. d TB101 DoxLRIG1 (left panel) or TB107 DoxLRIG1 (right panel) cells were incubated in the presence (Dox+) or absence (Dox−) of doxycycline for 48 h followed by lysis, and the phospho-RTK levels were analyzed using a phospho-RTK array. The positions and identities of detectable spots are indicated with arrows and include p-EGFR, p-MET, p-INSR, p-IGF1R, p-AXL, p-RET and p-VEGFR2. e TB107 DoxLRIG1 cells were incubated in the presence or absence of doxycycline for the indicated times followed by lysis, and an immunoblot analysis for LRIG1, p-EGFR (Tyr1086), total EGFR, p-MET (Tyr1234/1235), and total MET was conducted with actin as the loading control. f TB107 DoxLRIG1 cells were incubated in the presence or absence of doxycycline for 48 h followed by lysis, and total RET and p-RET (panTYR) were analyzed using specific ELISA kits. One of three similar experiments depicting the mean of triplicate samples is shown; standard deviations are indicated by error bars. g-i Pharmacologic and LRIG1-induced suppression of TB107 cell migration. LRIG1 expression was induced by culturing TB107 DoxLRIG1 cells with doxycycline for 48 h. Then, the cells were seeded in transwell inserts in the presence or absence of the MET inhibitor PHA-665752 for 24 h. To enhance transwell migration via chemotaxis, the medium in the upper chamber was serum-free, whereas the medium in the lower chamber contained 5% FBS. g At the end of the experiment, the transwell inserts were fixed and stained with crystal violet, and the number of migratory cells was counted. The mean number of migratory cells per ten high-power fields of three wells from one of three independent experiments is shown. Standard deviations of triplicate wells are indicated by error bars. *p < 0.05 and **p < 0.01, Student's t-test. h In parallel experiments, cells were seeded in six-well plates but were otherwise treated in the same manner as the cells in the transwell chambers. After 12 or 24 h of incubation, the cells were lysed and subjected to immunoblot analyses for LRIG1, p-EGFR (Tyr1086), total EGFR, p-MET (Tyr1234/1235), and total MET with actin as the loading control. i Quantification of the p-MET/actin ratios 12 h (left) and 24 h (right) after the induction of LRIG1. The mean ratios of three independent experiments are shown with standard deviations indicated by error bars. *p < 0.05, **p < 0.01, and ***p < 0.001, Student's t-test glioma is critical in better understanding the progression of this disease.

Mouse strains and animal husbandry
The housing and care of the mice as well as all experiments were performed in accordance with the European Communities Council Directive (86/609/EEC). The experimental protocols were approved by the Regional Ethics Committee of Umeå University, Umeå, Sweden (registration no. A5-2010, A41-10, and A42-10). The animals were housed under controlled conditions with a 12-h day/night cycle and provided water and standard chow pellets (cat. no. 801730, Special Diets Services, NOVA-SCB Sweden, Sollentuna, Sweden) ad libitum. The construction of the Lrig1 targeting vector and the generation of mice with floxed (B6.129-Lrig1 tm1Hhed ) and deleted (B6.129-Lrig1 tm1.1Hhed ) Lrig1 alleles were performed at Ozgene (Bentley DC, WA, Australia) using the same protocol as previously described for the generation of the Lrig2-deficient mice 19 . A schematic representation of the wild-type, conditional, and knockout Lrig1 alleles are shown in Fig. 2a. Ntv-a transgenic mice that express the Tv-a avian retrovirus receptor under the control of the Nestin (Ntv-a mice) promoter were a kind gift from Lene Uhrbom (Uppsala University, Sweden) and Eric Holland (Memorial Sloan-Kettering Cancer Center, NY, USA). Seven-week-old BALB/cA nude mice were obtained from Taconic Europe A/S (Ry, Denmark).

Retroviral transduction of mice
Newborn mice obtained from cross-breeding Lrig1heterozygous Ntv-a mice received an intracranial injection 2 μl (2 × 10 5 cells) RCAS-PDGFB-HA-producing DF-1 chicken fibroblasts into the right frontal hemisphere with a 10-μl Hamilton syringe as previously described 19 . The mice were euthanized at 12 weeks after injection or earlier if they showed symptoms of disease. The mouse brains were collected, fixed in 4% paraformaldehyde, embedded in paraffin, and analyzed for tumors by a neuropathologist (W.W.) who was blinded to the treatments.

In situ hybridization
In situ hybridization for Lrig1 and the control genes was performed using an RNAScope assay (Advanced Cell Diagnostics, Milano, Italy) as previously described 19 . An RNAScope probe set targeting Lrig1 (probe-Mm-Lrig1, Mouse: 310521) was custom designed by and purchased from Advanced Cell Diagnostics. The RNAScope positive control probe set (probe-Mm-Ppib: 313911) was used to verify the integrity of the RNA, whereas the RNAScope dapB (310043) control probe set was used to assess non-specific signals. The hybridized and stained slides were scanned using a Pannoramic 250 Flash II scanner (3DHistech, Budapest, Hungary) at 400× magnification, and the images were downscaled to 100× and imported into ImageJ software. The in situ hybridization signals were identified via color thresholding with hue = 180-255, saturation = 35-255, and brightness = 0-210, followed by noise reduction using a median filter with radius = 0.5 px. After the signals were dichotomized into single spots or clusters based on a cutoff area of 16 pixels, they were counted. Nuclei were identified via color thresholding with hue = 125-175, saturation = 41-255, and brightness = 0-235, followed by the application of a maximum filter with radius = 1.0 px and a watershed filter. Spots smaller than 16 pixels were excluded.

Immunoblotting and phospho-RTK assay
Whole-cell lysates were prepared in a modified RIPA buffer as previously described 34 (Fig. 2) or in a "cell extraction buffer" (Thermo Fisher Scientific, Gothenburg, Sweden) supplemented with either a complete-mini protease inhibitor cocktail or, for phospho-protein analysis, an EDTA-free protease inhibitor cocktail (Roche Diagnostics Scandinavia AB, Bromma, Sweden). Protein concentrations were determined using Pierce BCA Protein Assay Reagent (Thermo Fisher Scientific). Equal amounts of protein were separated via polyacrylamide gel electrophoresis using NuPAGE Novex 3-8% tris-acetate gels (Thermo Fisher Scientific) and electrotransferred onto a polyvinylidene difluoride membrane (Thermo Fisher Scientific). The membrane was blocked with 5% fetal bovine serum (FBS) for phospho-protein analysis and with 5% non-fat milk for all other analyses. Thereafter, the membrane was incubated with primary antibody, washed, and incubated with horseradish peroxidase-conjugated secondary antibody. Immunoreactive proteins were detected and quantified using chemiluminescence with an Amersham ECL Select Western blotting detection reagent (GE Healthcare, Uppsala, Sweden) and a ChemiDoc XRS imaging system with its associated Image Lab software (Bio-Rad Laboratories AB, Solna, Sweden). To screen for changes in phospho-RTK levels, samples containing 150 μg of whole-cell lysates were analyzed using a Human Phospho-RTK Array kit (R&D Systems Europe Ltd., Abingdon, UK) according to the manufacturer's instructions.

RNA extraction and quantitative real-time RT-PCR
RNA was prepared from tissues using an RNAqueos kit (Thermo Fisher Scientific) followed by digestion of contaminating DNA using a TURBO DNA-free kit (Thermo Fisher Scientific) according to the manufacturer's instructions. Quantitative RT-PCR was performed as previously described 13 . Triplicate samples of 20 ng of total RNA were analyzed using a qScript 1-Step qRT-PCR Kit (Quanta Biosciences, Gaithersburg, MD, USA) according to the manufacturer's instructions on a Bio-Rad CFX96 apparatus (Bio-Rad Laboratories AB). A Lrig1-specific TaqMan gene expression assay kit Mm00456116_m1 was obtained from Applied Biosystems (Thermo Fisher Scientific). Rn18s (also known as 18 S rRNA) was used as the internal control for normalization of the Lrig1 expression data. The Rn18s-normalized Lrig1 levels were divided by the corresponding level obtained from qRT-PCR amplification of QPCR Mouse Reference Total RNA (Agilent Technologies, Santa Clara, CA, USA); the Lrig1 level in the reference RNA was set to a value of 1.

Cell culture
Cells were grown in Dulbecco's modified Eagle's medium containing 10% FBS and supplemented with 50 µg/ml gentamicin, MEM non-essential amino acids, and 50 µM 2-mercaptoethanol and cultured at 37°C in an environment containing 5% CO 2 . The cell culture reagents were obtained from Gibco (Thermo Fisher Scientific). The glioma cell line TB101 and its culture conditions have been previously described 35 . The primary glioma cell line TB107 was established from a primary human glioblastoma specimen as previously described for TB101 35 . The TB101 and TB107 cell lines were obtained from glioblastoma patients with informed consent, and the study was approved by the local ethics committee. Lenti-X 293 cells were obtained from Clontech Laboratories, Inc. (Mountain View, CA, USA).

Xenograft transplantations, tissue handling, and immunohistochemistry
Eight-week-old BALB/cA nude mice were intracranially implanted with the indicated cells. Under general anesthesia, either 100 000 TB101 or 50 000 TB107 cells in a total volume of 2 μl of DMEM/F12 were stereotactically injected into the frontal lobe of the mice. Four months after tumor implantation, the mice were euthanized with carbon dioxide. Immediately after the mice were sacrificed, the brains were removed and cut into four coronal sections that were fixed in 4% paraformaldehyde for 24 h and then transferred to 70% ethanol. Thereafter, the fixed tissues were embedded in paraffin. The paraffin blocks were sliced into 4-µm sections on a microtome and then mounted onto poly-L-lysine-coated glass slides. The tissue sections were baked for 2 h at 60°C, de-waxed in xylene, rehydrated with distilled water, and subjected to heat-induced epitope retrieval using a citrate buffer at pH 6.0 [Dako, Glostrup, Denmark (now Agilent Technologies)] followed by incubation with anti-vimentin antibody (1:100) overnight at 4°C. The vimentin-bound antibodies were visualized using an EnVision kit (Dako).

Plasmids
The plasmids pLVX-TRE3G and pLVX-Tet3G were purchased from BioNordika Sweden AB; pLVX-LRIG1 has been previously described 37 ; and pLVX-LRIG1-TRE3G was constructed by ligating PCR-amplified LRIG1 into a BamHI-MluI-digested pLVX-TRE3G vector using an Infusion HD Cloning Kit (BioNordika Sweden AB). The integrity of all the vector constructs was confirmed with DNA sequencing.

Gene transduction
To produce lentiviral particles, Lenti-X 293 cells were co-transfected with the relevant expression vectors and the Lenti-X HTX packaging mix using Xfect transfection reagent (Clontech Laboratories) according to the manufacturer's instructions. Glioma cells were transduced with lentiviral particles according to the manufacturer's protocol. The cells transduced with either pLVX-LRIG1 or pLVX-Puro were selected with 1 μg/ml puromycin. Cells that had been co-transduced with pLVX-LRIG1-TRE3G and pLVX-Tet3G were selected with 1 μg/ml puromycin and 1 mg/ml G418. When indicated, LRIG1 expression was induced in the pLVX-LRIG1-TRE3G/pLVX-Tet3G cotransduced cell lines by the addition of 1 μg/ml doxycycline to the cell culture medium.

Migration assay
Cell migration was analyzed using a modified Boyden chamber assay. The cells were incubated in the presence or absence of 1 μg/ml doxycycline for 48 h and then plated in the upper chamber of transwell inserts with a pore size of 8 µm (Corning B.V. Life Sciences, Amsterdam, The Netherlands or Becton Dickinson AB, Stockholm, Sweden) in serum-free medium (Thermo Fisher Scientific) in the presence or absence of 1 μg/ml doxycycline. The lower chamber contained medium supplemented with 5% FBS either with or without 1 μg/ml doxycycline. In some experiments, the MET kinase-specific inhibitor PHA-665752 (500 nM final) was added to the upper and lower chambers. After 24 h, the transwell inserts were fixed and stained with 0.1% crystal violet, and the migratory cells were counted in 5 or 10 microscopic fields using a 20× or 40× objective.

Statistical analyses
Statistical analyses were performed using SPSS 22 software (IBM Corporation, Armonk, NY, USA). ANOVA was used to compare the means of multiple groups. The significance level was set to p < 0.05.