The Na,K-ATPase acts upstream of phosphoinositide PI(4,5)P2 facilitating unconventional secretion of Fibroblast Growth Factor 2

Abstract

FGF2 is a tumor cell survival factor that is exported from cells by an ER/Golgi-independent secretory pathway. This unconventional mechanism of protein secretion is based on direct translocation of FGF2 across the plasma membrane. The Na,K-ATPase has previously been shown to play a role in this process, however, the underlying mechanism has remained elusive. Here, we define structural elements that are critical for a direct physical interaction between FGF2 and the α1 subunit of the Na,K-ATPase. In intact cells, corresponding FGF2 mutant forms were impaired regarding both recruitment at the inner plasma membrane leaflet and secretion. Ouabain, a drug that inhibits both the Na,K-ATPase and FGF2 secretion, was found to impair the interaction of FGF2 with the Na,K-ATPase in cells. Our findings reveal the Na,K-ATPase as the initial recruitment factor for FGF2 at the inner plasma membrane leaflet being required for efficient membrane translocation of FGF2 to cell surfaces.

Fig. 1. Identification of a sub-domain in the cytoplasmic domain of α1 that is both necessary and sufficient for binding of FGF2.

a Schematic representation of the different GST-tagged α1 constructs used in this study with a comparison to the complete α1 chain of the Na,K-ATPase. Short linkers were introduced between CD1 and CD2 as well as between CD2 and CD3 as described previously. A short linker also exists in the α1-CD3∆sub construct connecting the N- and C-terminal parts of CD3. b Three-dimensional models of the complete α1 chain of the Na,K-ATPase based on the crystal structure of the Na,K-ATPase from Sus scrofa (PDB ID: 3KDP). The different domains were annotated according to topology data from Uniprot (ID P05024, α1 structure from Sus scrofa). Transmembrane helices are shown in cyan, extracellular domains in yellow, and cytoplasmic CD1 and CD2 domains in shades of gray. The CD3 part of the cytoplasmic domain is colored in dark pink with the subCD3 region highlighted in red. c Biochemical pull-down experiments of FGF2 using GST-fusion proteins of various versions of the cytoplasmic domain of α1. Lane 1: FGF2 input (2.5%); lanes 2 and 3: GST control; lanes 4 and 5: GST-α1-CD1-3; lanes 6 and 7: GST-α1-CD3; lanes 8 and 9: GST-α1-subCD3; lanes 10 and 11: GST-α1-CD3∆sub. Bound (33% of each fraction) and unbound material (2.5% of each fraction) was analyzed by SDS-PAGE and Coomassie Blue protein staining. The results shown are representative for five independent experiments. d The intensity of the FGF2 protein bands from SDS-PAGE (panel a) was analyzed with the ImageStudio software package (LI-COR Biosciences). Ratios of bound versus total FGF2 were calculated and normalized to the amounts of FGF2 bound to GST-α1-CD1-3 containing the complete cytoplasmic domain of α1. Data are shown as mean ± SD (n = 7). ***p ≤ 0.001.

Fig. 2. FGF2 binds to α1-subCD3 with sub-micromolar affinity as analyzed by an AlphaScreen protein–protein interaction assay.

a Cross-titration experiments conducted with His-tagged FGF2 and GST-tagged variant forms of the cytoplasmic domain of α1 as indicated. Data were normalized to the signal intensity measured for His-tagged FGF2-wt and GST-α1-CD1-3. The shown heat map is an average from six biological replicates. See Methods for details. b Quantification and statistical analysis of the relative alpha signal intensity from the cross-titration experiment with His-tagged FGF2 and the GST-fusion proteins of the cytoplasmic domain of α1 as indicated in panel a. Data are shown as mean ± SD (n = 6). ***p ≤ 0.001. c Determination of affinity using IC₅₀ values from competition assays. GST-tagged α1 variant forms (CD1-3, CD3, or subCD3) and His-tagged FGF2 proteins were mixed and subjected to a serial dilution of ∆N25-FGF2 used as an untagged competitor. Data are shown as mean ±  SEM (n = 3).

Fig. 3. The basic binding unit between FGF2 and α1-subCD3 is a heterodimer as analyzed by chemical crosslinking experiments.

FGF2 and α1-subCD3 were mixed at a total concentration of 20 µM at molar ratios of 1:1 (+/+), 1:2 (+/++), and 2:1 (++/+) (FGF2/α1–subCD3). As chemical crosslinkers, bismaleidomethane (BMOE) and disuccinimidyl glutarate (DSG) were used at the molar ratios provided in Methods. Samples were analyzed by SDS-PAGE followed by either western blotting (a, c) or total protein staining using Coomassie. FGF2 and α1-subCD3 migrate with an apparent molecular weight of 18 and 27 kDa, respectively. The crosslinking product with an apparent stoichiometry of 1:1 shows a migration behavior that corresponds to a molecular weight of about 45 kDa as indicated. Lane 1: molecular weight standards; lane 2: input FGF2; lane 3: input α1-subCD3; lane 4: FGF2 plus α1-subCD3 without crosslinker; lane 5: FGF2 plus crosslinker; lane 6: α1-subCD3 plus crosslinker; lanes 7–9: FGF2 plus α1-subCD3 at 1:1, 2:1, and 1:2 stoichiometries in the presence of crosslinker. a BMOE: immunoblots using anti-FGF2 and anti-α1 primary antibodies. c DSG: immunoblots using anti-FGF2 and anti-α1 primary antibodies. b BMOE: total protein visualized by Coomassie staining. d DSG: total protein visualized by Coomassie staining.

Fig. 4. NMR analysis of the interaction between FGF2-C77/95S and α1-subCD3.

a Shown are two regions of the overlay of ¹H-¹⁵N-HSQC spectra of 77 µM ¹⁵N-labeled FGF2-C77/95S in the absence (gray) or presence of 70 µM (orange) or 138 µM α1-subCD3 (blue). b Signal-to-noise ratios (SN ratio) for indicated peaks of the spectra shown in a are plotted here with SN(+70 µM α1-subCD3)/SN(-α1-subCD3) in orange and SN(+138 µM α1-subCD3)/SN(-α1-subCD3) in blue.

Fig. 5. K54 and K60 are required for efficient binding of FGF2 to α1-subCD3 as analyzed by AlphaScreen protein–protein interaction experiments.

a Cross-titration experiments conducted with various forms of His-tagged FGF2 and GST-tagged α1-subCD3. For each biological replicate, data were normalized to the signal measured for His-FGF2/GST-α1-CD1-3. As indicated by the color legend, data were represented as a heat map with the highest signal set to 100% (displayed in red) and the lowest signal set to 0% (displayed in blue). The shown heat map is an average from five biological replicates. For details, see Methods. b Quantification and statistical analysis of relative alpha signal intensities from the cross-titration experiment shown in panel a. Data are shown as mean ± SD (n = 5). ***p ≤ 0.001.

Fig. 6. Characterization of the molecular interface between FGF2 and α1-subCD3 based on in silico docking studies and atomistic molecular dynamics simulations.

a Representative structure of the WT1 cluster for in silico docking of FGF2 and α1-subCD3. b Average structure of the FGF2/α1–subCD3 interface. c Critical residues in the FGF2/α1–subCD3 interface including K54 and K60. d Pairwise contact map with all residues for the FGF2/α1–subCD3-wt interface. e Pairwise contact map with all residues for the FGF2/α1–subCD3-D560N interface. f Pairwise average interaction energy map for the FGF2/α1–subCD3-wt interface. g Pairwise average interaction energy map for the FGF2/α1–subCD3-D560N interface. In panel a, the most representative structure of the WT1 cluster is illustrated. The human α1-subCD3 domain was aligned to residues T380–V597 of the crystal structure of the α1-subunit of the Na,K-ATPase from Sus scrofa (PDB ID: 3KDP). The Na,K-ATPase is represented as an orange surface with the α1-subCD3 domain highlighted using a darker shade. FGF2 is shown as a violet surface. The PI(4,5)P₂ and phosphatidylcholine membrane lipids are represented using van der Waals spheres with red and gray colors, respectively. In panel b, an average structure of the FGF2/α1–subCD3 interface is shown illustrating the contribution to the interaction for each residue individually. It is defined as the sum of the probabilities of contacts for each residue and it is represented as a colored surface using the RGB color scale. Panel c highlights critical residues responsible for the FGF2/α1–subCD3 interaction including the FGF2 residues K54 and K60. Panels d and e show a pairwise contact map with all residues between FGF2/α1–subCD3-wt and FGF2/α1–subCD3-D560N, respectively. As a threshold, a probability of contact of more than 50% was set. Panels f and g show the average interaction energy (electrostatic and van der Waals contributions) of each residue pair between FGF2/α1–subCD3-wt and FGF2/α1–subCD3-D560N, respectively.

Fig. 7. Structural elements in α1-subCD3 with relevance for the binding interface with FGF2.

a Cross-titration experiments conducted with His-tagged FGF2 and variant forms of GST-tagged α1-subCD3. A representative intensity map is shown with a color code for binding efficiency as indicated. b Quantification and statistical analysis of the relative alpha signal intensities comparing the interaction of His-tagged FGF2 with the variant forms of GST-tagged α1-subCD3 indicated. Data were normalized based on the signal intensity for FGF2/α1–subCD3. Standard deviations are shown. For details, see Methods. c Analysis of protein folding measuring thermal stability of α1-subCD3 variant forms. Protein samples of 10 µl at a final concentration of 3 mg/ml were analyzed by differential scanning fluorimetry (nanoDSF). Data are shown as mean ± SD (n = 4). For details, see Methods.

Fig. 8. FGF2-GFP recruitment at the inner leaflet depends on direct interactions with the cytoplasmic domain of α1.

a FGF2-wt versus FGF2-K54/60E (FGF2 mutant deficient in binding to α1). b FGF2-wt versus FGF2-K54/60E in a K127Q/R128Q/K133Q background (FGF2 mutant deficient in binding to PI(4,5)P₂). c Direct comparison between FGF2-K54/60E and FGF2-K127Q/R128Q/K133Q following GFP background subtraction. Quantification of FGF2-GFP membrane recruitment at the inner leaflet of intact cells for all wild-type and mutant forms of FGF2 shown in panels a and b of Supplementary Fig. 5. Time-lapse TIRF movies with a total of 100 frames (100 ms/frame) were analyzed using the Fiji plugin TrackMate. The number of GFP particles were normalized for both surface area and the relative expression levels of each FGF2 fusion protein in the corresponding cell line. In panel a, FGF2 mutants defective in binding to α1 are shown (K54E, K60E, and K54/60E). In panel b, the same mutants were combined with mutations in the PI(4,5)P₂ binding pocket of FGF2 (K127Q/R128Q/K133Q). The mean values of each condition are shown in brackets with the wild-type form of FGF2-GFP set to 1. Data are shown as mean ± SD (n = 4). ***p ≤ 0.001. In panel c, the most important conditions were directly compared following GFP background subtraction and are shown as bar graphs. Data are shown as mean ± SD (n = 6). P value for (a) was 0.0106; p value for (b) was 0.0115.

Fig. 9. Efficient secretion of FGF2 from cells is facilitated by its interaction with α1-subunit of the Na,K-ATPase.

a Cell surface biotinylation experiments were conducted as described in Methods using stable CHO-K1 cell lines expressing either FGF2-wt-GFP, FGF2-K54E-GFP, FGF2-K60E-GFP, FGF2-K54/60E-GFP, or FGF2-C77/95A-GFP in a doxycycline-dependent manner. Aliquots from the total cell lysate (1.6%) and from the biotinylated fraction (33.3%; corresponding to the cell surface population of proteins) were subjected to SDS-PAGE and western blotting. Anti-GFP antibodies were used to detect the various FGF2-GFP fusion proteins indicated. Anti-GAPDH antibodies were used to detect intracellular GAPDH as a control for cell integrity during cell surface biotinylation. Primary antibodies were detected by fluorophore-labeled secondary antibodies and quantified using the Odyssey® CLx Imaging System (LI-COR Biosciences). b The efficiency of FGF2-GFP secretion of each variant form shown in panel a was quantified and normalized to the wild-type form that was set to 100%. Data are shown as mean ± SD (n = 4). P value for a was 0.0161. ***p ≤ 0.001. c Stable CHO-K1 cell lines expressing either FGF2-wt-GFP, FGF2-K127Q/R128Q/K133Q-GFP, FGF2-K54E/K127Q/R128Q/K133Q-GFP, FGF2-K60E/K127Q/R128Q/K133Q-GFP, FGF2-K54/60E/K127Q/R128Q/K133Q-GFP, or FGF2-C77/95A-GFP in a doxycycline-dependent manner were analyzed by cell surface biotinylation as described in the legend to panel a. d The efficiency of secretion of each variant form of FGF2-GFP shown in panel c was quantified and normalized as described in the legend to panel b. Data are shown as mean ± SD (n = 4). ***p ≤ 0.001.

Fig. 10. Ouabain inhibits proximity events between the α1-subunit of the Na,K-ATPase and FGF2 in a cellular context.

DuoLink assays (PLA®; Sigma-Aldrich) were conducted to quantify proximity of the α1 chain with FGF2 in a cellular context as described previously. Cells were incubated with ouabain for 2 h at the concentrations indicated. Cells were fixed with acetone. Nuclei were labeled with Sytox green (Life Technologies) and cells were imaged by confocal microscopy. For further details, see Methods. a Representative example of mock-treated HeLa cells. Scale bar = 10 µm. b Selected area of panel a at a higher magnification. DuoLink proximity events (red dots) in the vicinity of the plasma membrane are labeled with white arrowheads. c Representative example for cells treated with 50 µM ouabain. d Quantification, normalization, and statistical analysis of α1/FGF2 proximity events in the absence and presence of ouabain at the concentrations indicated. Images were analyzed with the DuoLink Image Tool Software (Olink Bioscience). Background signals were subtracted and data were normalized relative to the mock control. Data are shown as mean ± SD (n = 5). P value for (a) was 0.1124; p value for (b) was 0.0381. ***p ≤ 0.001.