Lrig1 expression identifies airway basal cells with high proliferative capacity and restricts lung squamous cell carcinoma growth

Background Lung squamous cell carcinoma (LUSC) accounts for a significant proportion of cancer deaths worldwide, and is preceded by the appearance of progressively disorganised pre-invasive lesions in the airway epithelium. Yet the biological mechanisms underlying progression of pre-invasive lesions into invasive LUSC are not fully understood. LRIG1 (leucine-rich repeats and immunoglobulin-like domains 1) is downregulated in pre-invasive airway lesions and invasive LUSC tumours and this correlates with decreased lung cancer patient survival. Methods and results Using an Lrig1 knock-in reporter mouse and human airway epithelial cells collected at bronchoscopy, we show that during homeostasis LRIG1 is heterogeneously expressed in the airway epithelium. In basal airway epithelial cells, the suspected cell of origin of LUSC, LRIG1 identifies a subpopulation of progenitor cells with higher in vitro proliferative and self-renewal potential in both the mouse and human. Using the N-nitroso-tris-chloroethylurea (NTCU)-induced murine model of LUSC, we find that Lrig1 loss-of-function leads to abnormally high cell proliferation during the earliest stages of pre-invasive disease and to the formation of significantly larger invasive tumours, suggesting accelerated disease progression. Conclusion Together, our findings identify LRIG1 as a marker of basal airway progenitor cells with high proliferative potential and as a regulator of pre-invasive lung cancer progression. This work highlights the clinical relevance of LRIG1 and the potential of the NTCU-induced LUSC model for functional assessment of candidate tumour suppressors and oncogenes.


Introduction
Lung cancer is the main cause of cancer-related deaths worldwide with 2.1 million new cases diagnosed each year [1]. The majority of patients present with late-stage incurable disease [2], therefore, devising strategies for early detection and treatment is key to improving lung cancer outcomes. genetic, epigenetic and transcriptional changes in pre-invasive lesions [9][10][11][12][13]. However, the biological relevance of these changes in lung cancer development is not understood.
Leucine-rich repeats and immunoglobulin-like domains 1 (LRIG1) is a transmembrane protein that acts as a negative regulator of epidermal growth factor receptor (EGFR) signalling [14,15]. LRIG1 is located in chromosome 3p14, a region that is frequently affected by copy number alterations in pre-invasive lung lesions and NSCLC [16,17]. Loss of heterozygosity of LRIG1 is seen in 75% of human lung cancer cell lines and low levels of LRIG1 expression have been correlated with decreased overall survival in patients with NSCLC [18][19][20]. We have shown previously that both transcript and protein levels of LRIG1 are lower in pre-invasive CIS lung lesions compared to donor-matched healthy epithelial tissue [18], suggesting an early role for LRIG1 during lung carcinogenesis. Here, we investigate the expression of LRIG1 in the normal airway epithelium and examine consequences of its loss-of-function during LUSC development.

Materials and methods
Mouse husbandry and experimentation Animal work was approved by the University College London biological services review committee and carried out in compliance with the UK Home Office procedural and ethical guidelines. The C57BL/6 Lrig1::eGFP-IRES-CreERT 2 murine line [21] was backcrossed twice to FVB/N. Mice were maintained in a mixed C57BL/6 and FVB/N background in individually ventilated cages, on a 12-h day/night cycle with access to food and water ad libitum. Littermates were distributed in the appropriate experimental or control groups.
Human tissue samples Ethical approval was obtained through the National Research Ethics Committee (REC reference 06/Q0505/ 12). Bronchial samples were taken during autofluorescence bronchoscopy from areas of normal bronchial epithelium. Bronchial brushes were used directly for flow cytometry and tissue biopsies frozen in optimal cutting temperature (OCT) compound (Tissue-Tek 4583).

Flow cytometry
Flow cytometry was performed on a Fortessa cell analyser (BD Biosciences) and cell sorting on a FACSaria (BD Biosciences). For cell-cycle analysis, live cells were incubated with 1 μg·mL −1 Hoechst 33342 for 30 min at 37°C prior to antibody staining. Reagents used are indicated in supplementary table S1. Data was analysed using FlowJo 10.0.6 (Tree Star).

NTCU-induced LUSC model
The dorsal fur of 6-week-old female mice was shaved and 75 μL of 0.013 M N-nitroso-tris-chloroethylurea (NTCU) (Santa Cruz sc-212265) diluted in acetone applied twice weekly for 12 weeks. Controls received only acetone. Mice were monitored for a further 11 weeks and weighed twice weekly. On sacrifice, lungs were insuflatted with 4% paraformaldehyde/PBS and fixed overnight at 4°C before paraffin embedding.

Histology and immunofluorescence
Haematoxylin and eosin staining was performed on an automated staining system (Tissue-Tek). Immunofluorescence and immunohistochemical staining were performed using standard protocols. Antigen retrieval methods, antibodies, reagents and equipment are detailed in the supplementary methods.

Bioinformatic analyses
Analyses of the Human Lung Cell Atlas single-cell RNA sequencing (scRNAseq) dataset [22] were conducted as detailed in the supplementary methods. We assessed LRIG1 expression in two datasets of human pre-invasive squamous cell lung cancer lesions obtained at bronchoscopy downloaded from the Gene Expression Omnibus. The first dataset, GSE33479, contained 122 samples of pre-invasive lesions, from normal epithelium to invasive cancer, profiled using Agilent microarrays [10]. The second, GSE94611 and GSE108082, contained laser-captured epithelial samples from 51 CIS lesions ( progressive and regressive) [11]. Data were analysed in the R statistical environment (version 3.5.0; www.r-project.org) using Bioconductor version 3.7.

Statistical analyses
Statistical analysis was performed on Prism 7 (GraphPad). Tests and sample sizes are indicated in figure legends; p<0.05 was considered to be statistically significant.

Results
LRIG1 is heterogeneously expressed in the airway epithelium Using immunofluorescence, we have shown previously that LRIG1 is expressed in the murine upper airways [18]. To characterise its expression within the different cell subpopulations of the airway epithelium we used an Lrig1 knock-in reporter mouse. In this model, a cassette encoding eGFP-IRES-CreER T2 is inserted downstream of the Lrig1 start codon (figure 1a) [23]. Enhanced green fluorescent protein (eGFP) expression occurs from the endogenous Lrig1 promoter and results in a loss-of-function allele. As shown previously for the skin and intestine [23], antibody staining confirmed expression of eGFP recapitulates endogenous LRIG1 in the upper airways (figure 1b).
In both murine skin and intestinal epithelia, LRIG1 localises to defined stem cell compartments where it regulates stem cell activity [24,25]. In contrast, Lrig1::eGFP expression was evident throughout the upper airway epithelium. eGFP was detected in basal (KRT5 + ), ciliated (acetylated-tubulin (ACT + )) and, to a lesser extent, in club (SCGB1A1 + ) epithelial cells (figure 1b). There was no obvious enrichment of eGFP within the KRT5 + cell subpopulation of the submucosal glands, where a reservoir of stem cells that contribute to the regeneration of the surface epithelium after severe injury reside [26,27] (figure 1b). Whole-mount immunostaining in the adult murine lung demonstrated Lrig1::eGFP expression extended throughout the bronchial tree, including the bronchi and bronchioles (figure 1c).
To determine the contribution of basal, ciliated and secretory cells to the cellular population expressing Lrig1::eGFP in upper airway epithelium we used flow cytometry. Tracheal epithelial cells were isolated from mice carrying one copy of the Lrig1::eGFP reporter allele. The immune and endothelial populations were eliminated through negative selection for CD45 and CD31, respectively. Lrig1 + epithelial cells were identified by dual expression of epithelial cell adhesion molecule (EPCAM) and eGFP. Griffonia simplicifolia isolectin B4 (GSI-B4), which selectively binds to a cell surface carbohydrate found on airway basal cells [28,29], was used to identify the basal cell population. CD24 and SSEA1 expression were used to separate ciliated and secretory cells, respectively [30] (figure 1d). Basal cells constituted mean±SEM 32.2± 3.4% of the Lrig1::eGFP + population, whereas ciliated and secretory cells made up the 63.9±2.7% and 1.3±0.3%, respectively (figure 1e). The distribution of cell types did not change significantly with selection of the brightest eGFP + cells (supplementary figure S1), indicating that Lrig1 expression is not enriched within the basal cell compartment. Only 50.8±4.1% of the total airway basal cell population was Lrig1::eGFP + .
Murine and human basal airway epithelial cells expressing Lrig1 have increased in vitro self-renewal capacity As only a subpopulation of airway basal epithelial cells expressed Lrig1, we investigated whether the properties of the Lrig1-expressing basal cells differ from the Lrig1-negative subpopulation. EPCAM + GSI-B4 + airway basal epithelial cells were isolated from murine tracheas heterozygous for the Lrig1::eGFP allele. Cell-cycle analysis revealed the Lrig1::eGFP + basal cell subpopulation contained a higher proportion of cells in G2/M compared to the Lrig1::eGFPfraction (t-test p=0.0019) (figure 2a and b). This agrees with previous skin and stomach findings where Lrig1 expression identifies more proliferative stem/progenitor cell populations [21,23,24].
To determine whether Lrig1-expressing basal cells display increased self-renewal potential, we assessed their clonogenic potential in two-dimensional culture. Freshly isolated single Lrig1::eGFP + and Lrig1:: eGFPbasal cells from mice heterozygous for the reporter allele were sorted into 96-well plates and assessed for colony-forming ability at day 10. Basal cells expressing Lrig1::eGFP formed significantly more clones than those not expressing Lrig1 (t-test p=0.013) (figure 2c). When seeded into Matrigel, basal airway epithelial cells give rise to three-dimensional organoids called "tracheospheres" [31]. Under these conditions, Lrig1::eGFP + basal cells generated significantly more tracheospheres than Lrig1::eGFPcells (t-test p=0.002) (figure 2d and e), indicating that Lrig1-expressing murine basal epithelial cells have higher in vitro propagation potential.
Next, we investigated whether LRIG1 expression identifies a more proliferative basal cell population within the human airway epithelium. Examination of publicly available single-cell transcriptomic data of the human airway epithelium (www.lungcellatlas.org) [32] showed that similarly to the mouse, LRIG1 is heterogeneously expressed by basal, ciliated and club cells (supplementary figure S2). Using the scRNAseq dataset from the Human Lung Cell Atlas [22], we assessed LRIG1 expression within the different clusters of basal cells present in the human bronchial epithelium. This revealed that LRIG1 expression enriches for "proliferating basal cells". The "basal cell" cluster, which is quiescent, was enriched in the LRIG1-negative fraction. Cells with a "proximal basal" signature were present in both Expression of LRIG1 within the KRT5 + basal cell compartment of the normal human bronchial epithelium was confirmed by double immunofluorescence (figure 3d). Normal human basal cells were isolated by flow cytometry from endobronchial brushings obtained during autofluorescence bronchoscopy. Following elimination of CD45 + , CD31 + and dead cells, co-expression of the cell-surface proteins integrin α-6 (ITGA6) and nerve growth factor receptor (NGFR), which are enriched in airway basal cells [31] (supplementary figure S4), was used to identify this cell population. Basal cells were then separated based on LRIG1 immunoreactivity (figure 3e). Live-cell cell-cycle analysis demonstrated an increased proportion of cells in G2/M in human basal cells expressing LRIG1 compared to the LRIG1-negative fraction ( paired t-test p=0.015) (figure 3f ). When colony-forming ability was assessed, LRIG1 + cells gave rise to colonies with a higher efficiency than their LRIG1counterparts ( paired t-test p=0.0067) (figure 3g). Together, our results show that expression of LRIG1 within the basal cell compartment identifies a more proliferative the lung airways. Antibody staining for SCGB1A1 and ACT was used for identification of club and ciliated cells, respectively. d) Immunophenotypic characterisation of tracheal epithelial cells expressing Lrig1::eGFP by flow cytometry. Live tracheal epithelial cells (P1) were identified by negative selection of cells expressing CD31, CD45 (endothelial cells and lymphocytes) and 4′,6-diamidino-2-phenylindole (DAPI), followed by positive selection of epithelial cell adhesion molecule (EPCAM) + eGFP + cells (P2). Griffonia simplicifolia isolectin B4 (GSI-B4) labelling was used to distinguish basal cells (P3) within the EPCAM + eGFP + population. Ciliated and secretory cells were selected within the GSI-B4-negative fraction by expression of CD24 (P5) and SSEA1 (P6), respectively. e) Distribution of cell types in the eGFP + tracheal epithelium (mean±SEM, n=7).  cellular subpopulation with increased in vitro propagating potential in both murine and human airway epithelium.
LRIG1 functions as a negative regulator of EGFR signalling and thus LRIG1 + cells showing a greater proliferation seems paradoxical. Therefore, we hypothesised that LRIG1 expression is a checkpoint for proliferation employed by progenitor basal cells, but not needed in a low-proliferating population (LRIG1-negative). To test this, we knocked-down LRIG1 in human basal cells and indeed noted enhanced cell population expansion (supplementary figure S5). Hence, LRIG1 marks a progenitor population of basal cells and functions as a checkpoint to proliferation.
Progression of pre-invasive lung cancer lesions to invasive LUSC is associated with decreased LRIG1 expression To better understand the role of LRIG1 in LUSC formation, we examined expression of LRIG1 at different stages of human LUSC development. We analysed a published dataset including patient biopsies from normal bronchial tissue, six morphologically distinct grades of pre-invasive disease, ranging from hyperplasia to CIS, and LUSC [10]. This revealed decreased LRIG1 expression in pre-invasive samples that varied from metaplasia to severe dysplasia, relative to normal tissue ( p<0.05) (figure 4a). Further decline in LRIG1 levels was observed in LUSC samples when compared to normal and hyperplastic tissue ( p=2×10 −5 and p=0.0026, respectively) (figure 4a). Additionally, assessment of LRIG1 expression in a longitudinally characterised cohort of CIS lesions that either progressed to invasive LUSC or regressed [11] showed significantly lower LRIG1 levels in the progressive group ( p=0.0086) (figure 4b). Together, these data indicate that a decrease of LRIG1 expression is associated with progressive severity of pre-invasive disease through to invasive lung cancer.

Establishment of a lung squamous cell carcinoma murine model
To investigate the consequences of decreased LRIG1 levels during LUSC development, we used a chemically induced murine model of LUSC [33]. Topical application of NTCU to young-adult mice twice weekly for 12 weeks resulted in LUSC formation over the next 11 weeks (figure 5a). Histological analysis protein C (SPC). Rare KRT5 + cells expressing low levels of SCGB1A1 were detected within NTCU-induced lesions ( figure 5c and d).
Lrig1 loss of function leads to increased tumour size in a murine LUSC model The Lrig1::eGFP-IRES-CreER T2 knock-in allele leads to Lrig1 loss of function (figure 6a). To establish whether decreased Lrig1 gene dosage impacts LUSC formation, we compared effects of NTCU in Lrig1-null, -heterozygous and wild-type mice. Animals were monitored for weight change and there was a marked separation between the treatment and control arms from week 15 (two-way ANOVA, p<0.05) (figure 6b).
KRT5 + basal cells are restricted to the upper airway murine epithelium [34]. Thus, expression of KRT5 beyond the trachea and the divisions of the right and left mainstem bronchi is abnormal. Therefore, we used KRT5 immunostaining to assess the presence of pre-invasive lesions in the bronchial tree following NTCU treatment. The proportion of total bronchial tree affected by pre-invasive disease in each mouse was calculated (figure 6c) and compared between Lrig1-null, -heterozygous and wild-type mice. Lrig1 genotype had an effect on the extent of pre-invasive disease (Kruskal-Wallis test, p=0.033), with the Lrig1-null (Lrig1 −/-) group displaying lower overall bronchial epithelial surface with abnormal KRT5 expression, when compared to Lrig1 heterozygous (Lrig1 +/-) mice (Dunn's multiple comparisons test, p=0.049) (figure 6d). Areas of pre-invasive disease were divided according to severity into flat atypia, low-grade dysplasia and high-grade dysplasia (table 1 and figure 6e and f ). There were comparable proportions of both low-grade and high-grade lesions across groups (figure 6e). However, the presence of flat atypia, the earliest feature of pre-invasive disease, was significantly lower in the Lrig1-null animals when compared to both wild-type and Lrig1 +/mice (two-way ANOVA followed by Tukey's multiple comparisons tests, p<0.001) (figure 6e).
Next, we assessed invasive tumours in the three groups. There was a trend towards increased incidence of LUSC as Lrig1 gene dosage decreased (Chi-squared test for trend, p=0.03) (table 1). When we compared the frequency and size of tumours between animals with different Lrig1 genotypes, Lrig1 gene dosage had no significant effect on the number of tumours per individual (figure 6g). However, the absence of LRIG1 led to significantly larger tumours than those in mice with active LRIG1 expression (Kruskal-Wallis test, p=0.0002) (figure 6h). Larger tumours, alongside the lower frequency of flat atypia in the Lrig1-null lungs, suggests loss of Lrig1 facilitates and/or accelerates pre-invasive disease progression into invasive LUSC.
To determine whether LRIG1 regulates LUSC progression by modulating cell fate or promoting proliferation, we investigated expression of P63 and Ki67 within the different spectrum of lesions. KRT5 + lesions were classified according to their severity and the fraction of cells expressing each marker (regardless of expression levels) were scored. The overall proportion of P63 + cells in NTCU-induced lung lesions was not significantly affected by Lrig1 genotype or lesion grade (two-way ANOVA, p>0.05) (figure 7a). However, when comparisons were made only among lesions of the same grade, we found that flat atypia lesions contained a significantly higher proportion of P63 + cells in Lrig1-null mice than in their wild-type counterparts (Kruskal-Wallis test followed by Dunn's multiple comparison test, p=0.04). Assessment of the fraction of Ki67 + cells revealed differences between distinct lesion grades (two-way ANOVA, lesion grade effect, p=0.0007). When we evaluated genotype effects within lesions of the same grade, we found that in comparison to wild-type mice, mice lacking Lrig1 also exhibited an increased fraction of Ki67 + cells in flat atypia lesions (Kruskal-Wallis test followed by Dunn's multiple comparison test, p=0.04) (figure 7b). This suggests that LRIG1 activity restrains LUSC formation by restricting progenitor cell proliferation in early pre-invasive disease.

Discussion
Studies of the homeostatic upper airway epithelium in mouse and human highlight phenotypic and functional heterogeneity within the basal cell compartment. Current evidence indicates basal cells comprise a population of multipotent stem cells and committed progenitors [32,[35][36][37]. Here we find that Lrig1 is heterogeneously expressed in the airway epithelium and its expression identifies a more proliferative subpopulation of basal cells with increased in vitro self-renewal potential. These findings are in line with previous studies in the epidermis and forestomach, where LRIG1 expression marks proliferative epithelial stem/progenitor cells, and in glandular gastric epithelial cells, where Lrig1-expressing cells have higher organoid-forming ability than their Lrig1-negative counterparts [21,23]. This suggests a conserved role for LRIG1 in regulation of progenitor cell activity in distinct epithelia.
In contrast to the skin and intestine, where Lrig1 is enriched in the stem cell compartment [24,25], in the murine gastric epithelium Lrig1 is expressed by both progenitor and differentiated cells [23,38]. Similarly, we find that in the upper airways, expression of Lrig1 is detected in basal, ciliated and secretory cells. In the mouse distal airways, scRNAseq analyses have shown that Lrig1 is enriched in SCGB1A1 + club cells and expressed at moderate levels by SPC + SCGB1A1 + bronchioalveolar stem cells [39]. Analogous studies of the human airways show that LRIG1 is heterogeneously expressed across the different epithelial cell subpopulations [32]. This raises the possibility that LRIG1 may exert functional roles in airway epithelial stem cells, lineage-restricted progenitors and differentiated cells. In future studies it would be important to determine how LRIG1 dysregulation affects these different populations using lineage-specific loss-or gain-of-function approaches.
Dysregulation of LRIG1 is seen across different tumour types [40]. Decreased LRIG1 expression has been associated with poor prognosis in a range of epithelial tumours, including lung cancer [19,20,41]. We now show that downregulation of LRIG1 occurs early during LUSC evolution, before development of high-grade lesions and invasive tumours. Primary human pre-invasive squamous lung cancer has not been grown in culture, limiting functional investigation of the molecular alterations occurring at this stage of LUSC development. Here we find that LRIG1 knockdown enhances in vitro propagation of primary basal cells isolated from the human normal airway epithelium, supporting a potential regulatory role in pre-invasive disease development. Recently, air liquid interface cultures of immortalised human bronchial epithelial cells with loss of TP53 and SOX2 overexpression have been shown to recapitulate bronchial dysplasia features [42]. This model may be useful to assess the consequences of loss of LRIG1 in pre-invasive squamous cancer in future studies.
In an NTCU-induced murine model of LUSC, we found a trend towards increased LUSC development as Lrig1 gene dosage decreased. The absence of Lrig1 led to decreased presence of flat atypia and formation of larger invasive tumours. A correlation between lower LRIG1 expression and increased tumour size has been identified in hepatocellular carcinoma [43]. Gain-and loss-of-function studies in cancer cell lines from lung, breast and liver indicate that LRIG1 modulates cancer cell proliferation [43][44][45]. Assessment of Ki67 expression in NTCU-induced lesions showed that Lrig1 depletion leads to abnormally high proliferative rates in early flat atypia lesions, but not in dysplasia or invasive tumours. These observations suggest the larger tumours in the Lrig1-null lungs result from accelerated disease progression, rather than from increased cell proliferation in established tumours.
We have shown that LRIG1 exerts a tumour suppressive role in LUSC development. Therefore, LRIG1 loss may be used as a biomarker of poor outcome of pre-invasive lesions and lead to a lower threshold for intervention. With the loss of the break of LRIG1 in EGFR signalling contributing to the development of pre-invasive disease, our work highlights a potential role for EGFR antagonists in LUSC prevention. Other data suggest a key role in the suppression of LUAD. As a negative regulator of EGFR signalling, high LRIG1 expression is associated with a dramatic 2.8-year improved survival in LUAD patients [20]. Decreased LRIG1 transcript levels have been reported in LUAD cell lines, especially those with mutant  EGFR [44]. Transfection of EGFR-mutant LUAD cells with LRIG1 led to decreased proliferation, invasion and migratory potential [44].
To date, functional analyses of the role of LRIG1 in NSCLC have relied on in vitro studies and xenograft models using LUAD cell lines [18,44]. Here, we have used a chemically induced model of LUSC that develops endogenous cancer lesions recapitulating the histopathological features of the human disease, including step-wise progression through pre-invasive stages. RNA sequencing analyses of NTCU-induced tumours indicate that their transcriptional landscape mimics human LUSC [46], supporting the relevance of this model for understanding LUSC progression. We show that using NTCU-induced carcinogenesis in the context of a transgenic murine model enables assessment of gene function during different stages of LUSC development and provides a system for validating candidate tumour suppressors, oncogenes and therapeutic targets for the prevention and treatment of LUSC.