A reciprocal regulatory circuit between CD44 and FGFR2 via c-myc controls gastric cancer cell growth

Despite their suggested importance, the mechanistic roles of FGFR2 and gastric cancer stem cell (GCSC) marker CD44 remain unclear. We investigated cross talk between CD44 and FGFR2. FGFR2 and CD44 positively regulate each other's expression. While FGFR2 suppresses c-Myc transcription, CD44 activates it. c-Myc in turn augments FGFR2 transcription. CD44 knockdown (KD) depleted FGFR2 and other GCSC markers, decreased c-Myc and Sox2 expression, and suppressed tumor growth, whereas CD44 activation led to FGFR2 induction. FGFR2 KD decreased most GCSC marker expression, including CD44, but increased c-Myc and Sox2 expression and attenuated tumor growth. FGFR2 kinase inhibitor and FGFR2 neutralizing antibody decreased the CD44+/hi GCSC fraction. Conversely, FGFR2 overexpression increased CD44 and accelerated tumor growth in mice. FGFR2 was co-expressed and colocalized diffusively with CD44, EpCAM, and LGR5. In contrast, phospho-FGFR2 colocalized densely with CD44, forming an aggregated signaling complex that was prevented by FGFR2 inhibition. The c-Myc KD depleted FGFR2 but not CD44. Similarly to CD44+/hi phenotypes, sorted FGFR+/hi cells had larger volumes, formed more tumor spheres, grew faster in vivo with bigger tumor mass, and expressed more CD44, EpCAM, and HER2. These findings suggest that FGFR2+/hi cells have stemness properties. Moreover, in situ FGFR2 expression in patient-derived gastric cancer tissue correlated with tumorigenic potential in a xenograft model. In conclusion, CD44 and FGFR2 maintain stemness in gastric cancer by differentially regulating c-Myc transcription.


Expression constructs and shRNA-mediated knock down
Full length FGFR2IIIb (epithelial type transcript variant) was PCR-cloned into pcDNA3.1(+)/myc-His A (Invitrogen) using the indicated primer set (see Table  S2). Mutant FGFR2 with a deleted extracellular domain (E) from a.a. 22 to 377 (R2ΔE) or a deleted intracellular cytoplasmic domain (C) from a.a. 399 to 821 (R2ΔC) were generated by PCR deletion methods using specific primer sets (Table S2). MKN45 GC cells that express CD44 but not FGFR2, and hence display weak tumorigenicity in mice, were transfected with FGFR2-myc wild type (WT) by AMAXA program L-029 and were selected with G418 (neomycin; 400 μg/m). In KD experiments, at least two different small hairpin RNA (shRNA) sets selective for each FGFR2 (five sets), CD44 (two sets), and c-Myc (five sets) were used for cloning into constitutive or doxycycline (Dox)-inducible vectors (see Table S2). Three sets of c-Myc siRNA (sets 6, 7, 8) were employed wherever necessary. The plasmid was transfected into the cells with the AMAXA kit using solution V and protocol O17. Conditional sub-cell lines in which FGFR2 or CD44 is subjected to depletion by Dox treatment were generated by the standard antibiotic selection protocol.
Tumor size at 4 weeks post injection was measured. Where necessary, SNU-16 cells were stably labeled with firefly luciferase in order to monitor tumor growth and spreading real time in situ.

Confocal microscopy, immunofluorescence (IF) staining, and flow cytometry (FC)
Cells were smeared on the IF slide glass, dried quickly, fixed with 4% paraformaldehyde, permeabilized with 0.25 % Triton X-100 (where necessary), and incubated with 5 % (w/v) BSA in 1X PBS for 30 minutes at room temperature. The cells were then stained with diluted primary antibody (either dye non-conjugated or directly conjugated with FITC, PE, APC or PECy5), followed by incubation with the corresponding secondary antibody conjugated to Alexa Fluor ® 568 (red) or Alexa Fluor ® 488 (green) where necessary (see Table S1 for lists of antibodies). For confocal microscopy, the stained cells were mounted with Prolong Gold Antifade Reagent with DAPI (Invitrogen), and were examined using a Leica TCS SP5 spectral confocal system (LSM 710). For FC, cells in single cell suspension were incubated with the indicated dye conjugated or non-conjugated primary antibody. Where necessary, the indicated secondary antibody reactive to the non-conjugated primary antibody was used to label the cells. When necessary, the cell fractions positive for specific markers were then sorted by FACSAria (Beckton) and were subjected to secondary assays including tumor sphere, cell growth, real time quantitative PCR, or tumorigenesis.

Immunoprecipitation (IP) and western blotting
Cells were lysed in 1 ml of 1% Triton X-100, 150 mM NaCl, 1 mM EDTA, 50 mM Tris HCl pH 7.4 supplemented with 0.1% leupeptin and 0.01% phenylmethylsulfonyl fluoride (PMSF), and 200 nM sodium vanadate (NAV)/1 M sodium fluoride (NaF), where necessary. The lysates were cleared by centrifugation at 12,000 rpm for 5 min at 4°C and quantitated by Bradford method [2]. For IP, the lysate was precleared by protein A/G agarose and subjected to IP at 4°C overnight as previously described [3]. An equal amount of protein or equal volume of immune precipitate was separated on SDS-denaturing polyacrylamide gel, transferred to nylon membrane, and reacted to appropriate antibody.

Tumor sphere assay and patient-derived GC xenograft tumorigenic assay
To test xenograft tumorigenic assays of tumor samples from patient specimens, fresh GC tissues from 105 GC patients were minced and 0.2 cm 3 tissues were engrafted in nude mice (n ≥3 per each GC patient) for 12-16 weeks. Tumors were passaged in vivo to new mice from tumor-bearing mice without culturing in vitro until at least three passages were completed. Also, the indicated cell numbers of indicated cell lines were subcutaneously or orthotopically implanted in nude mice or NSG mice (wherever indicated) and tumor growth was determined by measuring diameters or luminescence (for the Luc-labeled cells). For in vitro cultures, tumor tissues were minced and cultured in flasks in RPMI 1640 supplemented with 10% FCS. For the tumor sphere assay, cells were plated at a density 100 cells per well in ultra low cluster 96 well plates (Costar, #3474) containing TeSR™ 2 media (STEMCELL, #05860), cultured for 14 days, and scored for size and number of spheres. In detail, the growth of SNU-16 cells was monitored in real time by stably labeling the cells with firefly luciferase as previously reported [4]. The luminescence intensity of luciferase labeled SNU-16 was 3.78 RLU/cell, sufficient to monitor the bioluminescence in mice. Luciferase-labeled SNU-16 cells were comparable to the mother SNU-16 cells in their ability to form tumors in vivo in a dose-and time-dependent manner. All mice (n = 5) formed s.c. tumors when injected with 100, 1,000, and 10,000 cells. In 1-and 10-cell injected groups, 2/5 mice formed tumors of considerable size with viable bioluminescence. Strikingly, mice from the 1 cell-injected group developed tumors with an average size of 2.93 ± 0.94 cm 3 at 6 weeks post injection, which was approximately half the average size observed in mice injected with 10,000 cells (5.38 ± 3.28 cm 3 ) ( Figure S1 increased in FGFR2-transfected cells (red arrow) relative to pcDNA-transfected cells (red arrow). E. CD44 activation by hyaluronic acid (HA) results in an initial decrease and subsequent increase in CD44, which was colinear to p-FGFR2 and p-ERK, downstream of both CD44 and FGFR2 pathways [5]. HA treatment over 12 hr reduced CD44 levels (data not shown).