FGFR2 is required for airway basal cell self-renewal and terminal differentiation

Airway stem cells slowly self-renew and produce differentiated progeny to maintain homeostasis throughout the lifespan of an individual. Mutations in the molecular regulators of these processes may drive cancer or degenerative disease, but are also potential therapeutic targets. Conditionally deleting one copy of FGF receptor 2 (FGFR2) in adult mouse airway basal cells results in self-renewal and differentiation phenotypes. We show that FGFR2 signalling correlates with maintenance of expression of a key transcription factor for basal cell self-renewal and differentiation: SOX2. This heterozygous phenotype illustrates that subtle changes in receptor tyrosine kinase signalling can have significant effects, perhaps providing an explanation for the numerous changes seen in cancer. Summary: During adult airway epithelial homeostasis in mice, FGFR2 signalling is required for self-renewing divisions of basal stem cells and to maintain expression of the key transcription factor SOX2.


INTRODUCTION
Like human airways, the mouse trachea contains three major epithelial lineages (Rock et al., 2010;Teixeira et al., 2013). Basal cells (BCs) are a stem cell population and include slowly dividing stem cells and committed luminal precursors (Mori et al., 2015;Rock et al., 2009;Watson et al., 2015). Luminal secretory cells self-renew and produce terminally differentiated ciliated cells (Rawlins and Hogan, 2008;Rawlins et al., 2007Rawlins et al., , 2009. Multiple studies have shown that SOX2 is a key transcription factor (TF) for the development and maintenance of all airway epithelial cells (Gontan et al., 2008;Hashimoto et al., 2012;Ochieng et al., 2014;Que et al., 2009;Tompkins et al., 2009Tompkins et al., , 2011. Deletion of Sox2 in adult mouse tracheal epithelium caused loss of differentiated cells. Moreover, the Sox2 Δ/Δ BCs were less able to proliferate in vitro or in vivo following injury (Que et al., 2009). SOX2 is thus required for BC self-renewal and luminal differentiation. SOX2 overexpression can be a driver of squamous cell carcinoma, which has a predominantly basal cell phenotype (Correia et al., 2017;Ferone et al., 2016). FGFR2 function has been extensively studied during lung branching where one of its roles is to maintain undifferentiated epithelial progenitors by inhibiting SOX2 expression (Abler et al., 2009;Que et al., 2007;Volckaert et al., 2013). However, at later stages of embryonic development ectopic FGF10 can promote BC differentiation in SOX2 + airway progenitors (Volckaert et al., 2013). The same study expressed a secreted dominant-negative FGFR2 in the late stages of embryogenesis and suggested that there could be a role for FGFR2 signalling in maintenance of airway BCs. We have now specifically tested this hypothesis in the steady-state adult mouse trachea, and show that FGFR2 is required for BC selfrenewal and terminal differentiation. Moreover, FGFR2 signalling maintains SOX2 expression.

RESULTS AND DISCUSSION FGFR2 is required for normal tracheal homeostasis
We detected FGFR2 protein in airway basal cells and at the apical surface of secretory cells (Fig. 1A,B), confirming previous results (Watson et al., 2015). To determine the role of FGFR2 in BCs, we conditionally deleted one copy of Fgfr2 and activated a GFP reporter in adult tracheal BCs using Tg(KRT5-CreER); Rosa26R fGFP/+ ; Fgfr2 fx/+ (Fgfr2 conditional heterozygous, cHet) and control Tg(KRT5-CreER); Rosa26R fGFP/+ mice (Fig. 1C). To test for co-recombination between Fgfr2 fx and the reporter, we isolated GFP + BCs by flow cytometry as GFP + , GSIβ4-lectin + cells at 3 weeks post-tamoxifen (tmx) induction and performed RT-qPCR for Fgfr2 (Fig. 1D). This confirmed that cHet BCs had ∼50% of the control Fgfr2 mRNA level. Hence, we use GFP + cells as a surrogate marker for Fgfr2 Δ/+ cells, being aware that corecombination will not be 100%. Tracheae were harvested at intervals to assess the contribution of GFP + , Fgfr2 Δ/+ BCs to the epithelium during homeostatic turnover ( Fig. 1E). At 1.5 weeks post-tmx, ∼30% of total BCs were GFP + in Fgfr2cHet and control mice. In controls, this percentage increased to ∼60% at 5 weeks post-tmx, before dropping to initial levels by 24 weeks. By contrast, in the Fgfr2cHet tracheae, the percentage of GFP + BCs remained approximately constant at 5 weeks, but decreased to less than 5% of total basal cells by 24 weeks (Fig. 1F). In both genotypes, labelled BCs produced labelled luminal cells. Luminal differentiation initially appeared more rapid in the Fgfr2cHets. However, luminal cell production was not sustained over time, likely due to the loss of GFP + BCs, and by 24 weeks the percentage of labelled luminal cells was significantly lower in the Fgfr2cHet tracheae (Fig. 1G).
This showed that Fgfr2cHet BCs can produce luminal cells, but that mutant basal and luminal cells are gradually lost. One possible reason for the loss of Fgfr2cHet cells is differential fitness and competition with neighbouring wild-type cells (Vivarelli et al., 2012). To test this, we mixed pure populations of Rosa26R tdTomato/+ ; Fgfr2 Δ/+ with unlabelled Fgfr2 +/+ BCs (1:2 ratio) and assessed their ability to compete in vitro at steady-state and following injury. We were unable to find evidence for differential proliferation or survival in the mixed cultures and conclude that it is unlikely that cell competition contributes to the observed loss of mutant cells ( Fig. S1; Movies 1-5).
Fgfr2cHet BCs do not differentiate into fully mature luminal cells We asked whether the loss of Fgfr2cHet cells was due to a decrease in cell division. As expected, proliferation rates were low in all tracheae, but dividing GFP + cells were observed (Fig. S2A). We noted an increase in proliferation of the Fgfr2cHet GFP + cells at 1.5 weeks post-tmx, although this was not statistically significant and the change was not sustained over time (Fig. S2B). Thus, altered proliferation does not explain the phenotype. We also assessed apoptosis using cleaved caspase 3 staining, but did not identify caspase 3 + cells (665 GFP + cells scored in four independent 5 week samples; Fig. S2C,D). We assessed the ability of Fgfr2cHet cells to differentiate by analysing the luminal (KRT8) and basal (KRT5) cytokeratins at 5 weeks post-tmx ( Fig. 2A). A higher percentage of the total GFP + cells co-stained with KRT8 in the mutants, indicating that more cells had begun differentiation to a luminal fate (Fig. 2B). Similarly, plotting the GFP/T1α staining (Fig. 1D) as a percentage of GFP + cells (GFP + , T1α − ) showed more differentiating cells in the mutants (Fig. 2B). Thus, Fgfr2cHet cells exit the basal layer at a greater rate than controls and their descendants take on a luminal KRT8 + , T1α − fate, suggesting a self-renewal defect.
At steady-state, BCs initially differentiate into secretory cells that later produce ciliated cells (Watson et al., 2015). Cell fate analysis at 5 weeks post-tmx showed that both control and Fgfr2cHet BCs produce secretory SCGB1A1 + cells (Fig. 2C,D). Moreover, there were no signs of goblet cell production in the mutants ( Fig. 2C; n=4 MUC5AC lo cells observed from 859 cells counted in 5 Fgfr2cHet individuals). However, analysis of acetylated tubulin-positive cilia (marker of terminal luminal differentiation) at 24 weeks post-tmx showed that the Fgfr2cHet cells never took on a ciliated cell identity (Fig. 2E).
Fgfr2cHet BCs have high levels of β-galactosidase activity in vitro We tested the ability of Fgfr2cHet cells to proliferate and differentiate in vitro using a high dose of an adenovirus containing CMV-Cre (Ad-Cre) to recombine Rosa26R fGFP/+ ; Fgfr2 fx/+ and control Rosa26R fGFP/fGFP BCs grown in self-renewing conditions ( Fig. 3A). When analysed by genomic PCR, this resulted in an almost-pure population of Fgfr2 Δ/+ cells (Fig. S3A,B). Four days after Ad-Cre-mediated deletion, we observed an increased proportion of KRT8 + cells in the Fgfr2cHet cultures ( Fig. 3A-C). This recapitulates the in vivo phenotype and supports the conclusion that Fgfr2cHet BCs have a self-renewal defect. Additional cultures were passaged and grown to confluence before differentiation at air-liquid interface (Fig. S4A-D). The Fgfr2cHet cells survived passaging but did not reach confluence and failed to express markers of ciliated or basal cell differentiation. Moreover, passaged cells were unable to grow in sphere-forming assays (Fig. S4E-H). The passaged Fgfr2cHet cells were somewhat enlarged and flattened, possibly indicating a senescent phenotype (Rodier and Campisi, 2011). We therefore tested for senescence-associated β-galactosidase activity in primary cultures of Fgfr2cHet cells. β-Galactosidase activity was detected in 3/3 Fgfr2cHet cultures and 0/3 controls (Fig. 3D).
Senescence of the Fgfr2cHet cells in vivo could potentially explain why the luminal GFP + cells can express secretory markers, but do not later produce ciliated cells. However, we cannot absolutely exclude a luminal fate choice defect in Fgfr2cHet BCs.

Lower levels of SOX2 expression in the Fgfr2 conditional heterozygous cells
We determined the effects of decreasing FGFR2 signalling on downstream pathways using immunoblotting. There was a 1.5-fold decrease in phosphorylated AKT in the Fgfr2 Δ/+ cells (Fig. 3E,F), BCs rarely make selfrenewing divisions in which a new BC is produced. Mutant BCs are more likely to produce descendants with luminal morphology/markers that are unable to completely differentiate, possibly because they senesce. The result is that GFP + Fgfr2cHet cells are gradually diluted out from both the basal and luminal populations, and the epithelium is sustained by GFP − wild-type BCs.
but no change in phosphorylated ERK1/2 (Fig. S3C,D). These changes are consistent with a decrease in FGFR2 signalling via the PI3K-AKT pathway, which was implicated as the main pathway downstream of FGFR2 in adult small airway secretory cells and the developing trachea (Volckaert et al., 2011(Volckaert et al., , 2013. Most strikingly, there was a twofold decrease in SOX2 in the Fgfr2 Δ/+ cells (Fig. 3E,F; Fig. S3C,D). We confirmed the decrease in SOX2 protein at a cellular level by in vitro immunostaining (Fig. 3G). Similarly, there was consistently lower SOX2 expression in GFP + cells in the Fgfr2cHet tracheae in vivo (Fig. 3H, arrows). As expected from the genetic strategy, in the mutants we also observed GFP + , SOX2 + cells (Fig. 3H, arrowheads) and GFP − , SOX2 − cells (Fig. 3H, yellow arrows), both are likely to have recombined only one floxed allele. Co-immunostaining with FGFR2 confirmed that the GFP + , SOX2 + cells observed in the mutants retained high levels of FGFR2 protein (Fig. 3I).

FGF7 and FGF10 can promote BC colony expansion in vitro
We predicted that if a decrease in Fgfr2 results in loss of BC self-renewal, then activation of FGFR2 in vitro should promote the growth of BC colonies. FGF7 and FGF10 are expressed in homeostatic tracheae (Balasooriya et al., 2016) and are known to activate FGFR2 preferentially in vitro and in vivo (Ornitz et al., 1996). We plated wild-type BCs at low density and added FGF7 or FGF10 on culture day 2 after colonies were established (Fig. 4A). Addition of FGF7 or FGF10 had the opposite effect to decreasing Fgfr2 and significantly increased colony size (Fig. 4B,C). Interestingly, FGF7 and FGF10 had no effect on the level of Sox2 mRNA (Fig. 4D).
In conclusion, our data suggest that a normal function of FGFR2 signalling in adult airway BCs is to promote asymmetric selfrenewing divisions (Fig. 4E). This is consistent with work in the embryonic trachea where ectopic FGF10 was observed to promote BC fate (Volckaert et al., 2013). By contrast, our previous work on FGFR1 in adult BCs showed that FGFR1 is required to inhibit steady-state proliferation and does not change the ability of BCs to self-renew (Balasooriya et al., 2016). Thus, FGFR1 and FGFR2 have independent functions in airway BCs. We cannot exclude the possibility that they also have other overlapping functions.
We also show that steady-state FGFR2 signalling is required, directly or indirectly, to maintain SOX2 protein levels in the adult airway. This is in contrast to the branching lung, where FGFR2 inhibits SOX2 expression at the tips. Interestingly, an FGFR2-SOX2 inductive relationship has been observed in other cell types (Mansukhani et al., 2005). An FGFR2-SOX2 relationship may be maintained in some squamous lung cancers where FGFR2 and SOX2 transcript levels are often correlated (Kim et al., 2016).

Haploinsufficiency of Fgfr2 in conditionally deleted adult cells
We were surprised that our Fgfr2cHet BCs displayed striking phenotypes when germline Fgfr2 Δ/+ animals are viable and fertile (Yu et al., 2003). We therefore looked for subtle epithelial defects in germline-deleted Fgfr2 Δ/+ tracheae compared with wild-type siblings, but were unable to find any abnormalities (Fig. S5). Fgfr2 is haploinsufficient in several organs, including the lacrimal and salivary glands (Shams et al., 2007). We suggest that in mouse embryos heterozygous for Fgfr2, genetic compensation operates in most tissues. However, conditional heterozygous deletion in the adult by-passes such mechanisms. This is very similar to recent findings from zebrafish genetics where genetic compensation has been found to operate in germline mutants, but not in acute knockdowns (Rossi et al., 2015). It raises the possibility that many genes that the mouse developmental community assume are uninteresting/redundant based on lack of germline knockout phenotypes do play important roles in development/homeostasis.

Tamoxifen
Adult (>8 week) animals were injected intraperitoneally four times, every other day, with 0.2 mg/g body weight tamoxifen.

Tracheal epithelial cell culture
Tracheal cells were isolated following published methods (Rock et al., 2009). Briefly, cells were incubated in Dispase II (Gibco, 16 U/ml) for 20 min at room temperature. Epithelial sheets were dissociated using 0.1% trypsin/EDTA. Unless otherwise stated, 5×10 4 cells in 0.5 ml MTEC/+ media (You et al., 2002) were plated on collagen-coated 12-well tissue culture inserts (BD Falcon,353180). For tracheospheres, cells were passaged into 50% matrigel (Becton Dickinson). Adeno-Cre (University of Iowa, Gene Transfer Vector Core) was incubated at MOI 2500; vector pfu 1×10 6 for 8 h. Recombinant mouse FGF7 and FGF10 (R&D Systems) were used at 100 ng/ml. For competition assays, mixed populations of cells were grown to confluence and then imaged every 4 h for 10 days in a Nikon Biostation. Alternatively, confluent cultures were scratched and imaged every 2 h for 5 days. In vitro experiments were preformed in triplicate.

Immunostaining
Tracheae were fixed in 4% paraformaldehyde at 4°C for 4 h; washed PBS, sucrose protected, embedded in OCT (Optimum Cutting Temperature Compound, Tissue Tek) and sectioned at 6 μm. Airway culture inserts were washed in PBS, fixed for 10 min in 4% paraformaldehyde at room temperature and permeabilized with 0.3% Triton X-100. Primary antibodies are listed in Table S1. Alexa Fluor-conjugated secondary antibodies (1:2000) were from Life Technologies (Table S1). DAPI and fluoromount were from Sigma. X-gal staining was performed using Senescence βgalactosidase staining kit (Cell Signaling, 9860).

Microscopy and image scoring
Slides were imaged on a Zeiss AxioImager compound, or a Leica Sp8/Sp5 confocal microscope. Cells were scored manually in Fiji. For cryosections, every epithelial cell along the entire proximal to distal length of a longitudinal section from the centre of the trachea was scored. For cultured cells at least three random fields of view from each insert were scored. Raw cell counts are available in Fig. S6.

Immunoblots
Cells were collected in Cell Extraction Buffer (Invitrogen, FNN0011) with protease inhibitor (Roche 04693116001) and PMSF (Sigma, P7626). Proteins were separated on 10% or 12% SDS-PAGE gels before being transfer onto Millipore Immobilon-P PVDF Membrane (Merck Millipore, IPVH00010). Primary antibodies are listed in Table S1. Detection with HRP-conjugated secondaries (Abcam, 1:10,000) and enhanced chemiluminescense (Thermo Scientific, PI-32109) was carried out. Quantitation is based on protein from three biological replicates separated on the same polyacrylamide gel. Band intensity was analysed in Fiji normalised to the loading control.

Statistics
P-values were obtained using an unpaired two-tailed student's t-test with unequal variance.