Fibroblast growth factor rescues brain endothelial cells lacking presenilin 1 from apoptotic cell death following serum starvation

Presenilin 1 (Psen1) is important for vascular brain development and is known to influence cellular stress responses. To understand the role of Psen1 in endothelial stress responses, we investigated the effects of serum withdrawal on wild type (wt) and Psen1−/− embryonic brain endothelial cells. Serum starvation induced apoptosis in Psen1−/− cells but did not affect wt cells. PI3K/AKT signaling was reduced in serum-starved Psen1−/− cells, and this was associated with elevated levels of phospho-p38 consistent with decreased pro-survival AKT signaling in the absence of Psen1. Fibroblast growth factor (FGF1 and FGF2), but not vascular endothelial growth factor (VEGF) rescued Psen1−/− cells from serum starvation induced apoptosis. Inhibition of FGF signaling induced apoptosis in wt cells under serum withdrawal, while blocking γ-secretase activity had no effect. In the absence of serum, FGF2 immunoreactivity was distributed diffusely in cytoplasmic and nuclear vesicles of wt and Psen1−/− cells, as levels of FGF2 in nuclear and cytosolic fractions were not significantly different. Thus, sensitivity of Psen1−/− cells to serum starvation is not due to lack of FGF synthesis but likely to effects of Psen1 on FGF release onto the cell surface and impaired activation of the PI3K/AKT survival pathway.

embryonic brain, Psen1 deficiency in endothelial cells results in decreased turnover of the extracellular matrix protein fibronectin 14 .
Presenilins and presenilin FAD mutants have long been known to influence stress responses in cells including sensitivity to apoptosis [15][16][17][18][19][20] . To understand the role of Psen1 in endothelial cells, we analyzed the response of embryonic brain endothelial cells to a stress signal generated by serum withdrawal. Serum removal can be used to model apoptosis in endothelial cells [21][22][23][24][25] and causes apoptosis in endothelial cells from various sources including human umbilical vein [26][27][28] , human foreskin microvasculature 29 , and bovine aorta 30 . In the present study, we show that serum starvation of Psen1− /− brain endothelial cells leads to their detachment from a collagen type IV substrate and apoptosis, but does not significantly affect the viability or attachment of wild-type (wt) brain endothelial cells. Using serum-and supplement-free media we show that either acidic or basic fibroblast growth factors (FGFs) are able to rescue brain endothelial cells from apoptotic cell death following serum starvation, whereas vascular endothelial cell growth factor (VEGF) cannot.
Serum deprivation can trigger apoptosis in endothelial cells 26,32 . We tested wt and Psen1− /− brain endothelial cells for their ability to withstand serum deprivation. We found that whereas wt brain endothelial cells could withstand serum starvation, Psen1− /− endothelial cells rapidly underwent apoptosis following serum withdrawal. After 12-18 h of serum deprivation, almost all Psen1− /− cells showed apoptotic features becoming generally rounded and frequently detached from the surface of the culture dish, unlike wt cells which maintained their normal appearance ( Fig. 2A).
Western blot analysis of endothelial cells after 16 h of serum starvation showed the presence of activated (cleaved) caspase 3 as well as the cleaved form of its substrate poly(ADP-ribose) polymerase (PARP) in Psen1− /− cells, but not in wt cells (Fig. 2B). Tunel staining confirmed the presence of Psen1− /− apoptotic cells (Fig. 3D). Nuclei in serum-starved Psen1− /− endothelial cells were frequently fragmented, and nuclear chromatin was often condensed (Fig. 3D). In contrast, wild-type cells showed none of these features (Fig. 3C).
Reduced PI3K/AKT signaling in serum-starved Psen1−/− endothelial cells. The PI3K/AKT pathway is recognized as a key regulator of cell survival and protection against apoptosis in many cell types 33 including endothelial cells 21,[34][35][36] . We examined AKT phosphorylation in endothelial cells after 16 h of serum deprivation. Psen1− /− cells showed significantly less AKT phosphorylation than did wt cells (Fig. 4A). Analysis of the time course of AKT phosphorylation showed that in Psen1− /− cells, pAKT levels rapidly decreased in the first hour and remained low thereafter, whereas pAKT levels in wt cells increased and then remained high (Fig. 4B). p38 MAP kinase also modulates apoptotic cell death in many cell types, and it has been reported that impaired PI3K/AKT signaling leads to activation of p38 and apoptosis 36,37 . Because AKT activation was reduced in Psen1− /− cells, we examined p38 phosphorylation in serum-starved cells. Whereas p38 phosphorylation was barely detectable in wt cells after 16 h of serum starvation, Psen1− /− cells showed elevated levels of phosphorylated p38 consistent with decreased pro-survival AKT signaling and an apoptotic phenotype (Fig. 4C).
FGF but not VEGF prevents apoptotic cell death of Psen1−/− endothelial cells following serum starvation. FGF2 has been reported to protect endothelial cells from the apoptotic effects of serum starvation 22,26 .
We investigated whether FGF could protect Psen1− /− cells from apoptosis induced by serum deprivation. As shown in Fig. 5, basic FGF2 as well as acidic FGF1 at concentrations of 25 ng/ml in serum-free media prevented apoptosis induced by serum-deprivation in Psen1− /− endothelial cells (compare Fig. 5D-F to cells grown without serum in 5B and cells grown in complete endothelial cell growth medium [ECGM] in 5A). FGF1 was inactive when added in the absence of albumin and heparin (Fig. 5C), whereas FGF2 was active in both conditions (Fig. 5C,D). Moreover, Psen1− /− endothelial cells treated with FGF2 in the absence of serum no longer exhibited TUNEL staining and showed no nuclear fragmentation as visualized by DAPI staining (Fig. 3F, compare to cells grown in the absence of serum in Fig. 3D or cells grown in ECGM in Fig. 3B). These results were confirmed by Western blot analysis, in which caspase 3 (Fig. 6A, upper panels), cleaved PARP (Fig. 6A, lower panels), and p38 activation (Fig. 6B) were not detectable in Psen1− /− endothelial cells.
In human umbilical vein endothelial cells (HUVEC), VEGF has been reported to induce a survival response in serum-free medium and to be a primary survival factor via activation of the PI3K/AKT pathway 21 . To determine the effects of VEGF on Psen1− /− cells, AKT phosphorylation was measured following VEGF treatment. As shown in Fig. 6C, the response of Psen1− /− endothelial cells to VEGF was similar to that of wt cells. However, VEGF treatment (25 ng/ml) failed to prevent apoptosis of Psen1− /− brain endothelial cells under serum starvation conditions as indicated by the typical apoptotic phenotype of cells (Fig. 5H) and the presence of cleaved caspase 3, cleaved PARP, and activated p38 (Fig. 6A,B).

Blocking FGF signaling induces apoptosis in wt endothelial cells following serum starvation.
To confirm the essential role of FGF in preventing apoptosis in response to serum deprivation, Psen1 wt cells, which are normally resistant to serum starvation, were treated overnight with 30 μ M SU5402, an irreversible antagonist of the FGFR-specific tyrosine kinase 38 , and serum-starved for 16 h. Pretreatment of wt cells with inhibitor followed by overnight serum starvation resulted in apoptosis as indicated by cell morphology (Fig. 7A,B) and the presence of cleaved PARP (Fig. 7C). Although SU5402 can also inhibit VEGFR 39 , it is likely that apoptosis was caused by inhibition of FGF signaling given that VEGF failed to prevent serum deprivation-induced apoptosis in Psen1− /− endothelial cells.
Blocking γ-secretase activity does not sensitize wt endothelial cells to apoptosis following serum starvation. One of the main biological functions of Psen1 is its role as a component of the γ -secretase complex 1,2 . Therefore, we investigated whether inhibition of γ -secretase activity could affect survival of wt endothelial cells in the absence of serum. As shown in Fig. 7D, pretreatment of wt endothelial cells for 16 h with the γ -secretase inhibitor XXI followed by serum starvation did not result in apoptosis as judged by cleavage of PARP. Accumulation of N-cadherin C-terminal fragment (Fig. 7E) confirmed that the inhibitor blocked γ -secretase activity.

Localization and levels of FGF2 in endothelial cells. FGF2 is present in multiple subcellular locations
depending on the cell type and state of cells, which influences its biological effects 40 . Therefore we investigated FGF2 distribution in wt and Psen1− /− cells in the presence of serum and after 6 h of serum deprivation via immunocytochemical staining. In the presence of serum, FGF2 immunoreactivity in both wt and Psen1− /− cells was largely concentrated in nuclear and cytoplasmic vesicles (Fig. 8A,C). In the absence of serum, FGF2 staining became more diffuse especially in the cytosol, although some vesicles could still be observed (Fig. 8B,D) without any differences between wt and Psen1− /− cells.
As an additional indication of the levels and intracellular distribution of FGF2 cytosolic and nuclear fractions were prepared from control and serum-starved wt and Psen1− /− endothelial cells and FGF2 was captured with heparin-agarose beads. Western blot analysis of the FGF2-enriched fractions did not show any significant differences in the levels of FGF2 in the cytosolic and nuclear fractions of wt and Psen1− /− cells, whether in the presence or absence of serum, when FGF2 levels were normalized to those of β -actin (cytosolic) or lamin B (nuclear) (Fig. 8E).

Discussion
Inhibition of apoptosis and endothelial cell survival is essential during angiogenesis and for maintenance of blood vessel integrity, and is favored by many growth factors as well as integrin-mediated adhesion to the extracellular matrix 34 . Apoptosis of endothelial cells, however, occurs as part of vessel regression and remodeling and is most likely triggered by loss of growth factors 41,42 . In vitro serum deprivation triggers apoptosis in endothelial cells from different sources, but serum deprivation-induced apoptosis can be suppressed by extracellular survival signals promoted by growth factors such as VEGF or FGF 41 .
Our results with brain endothelial cells isolated from wt and Psen1− /− embryonic brains show that Psen1-deficient cells, when deprived of serum, commit to apoptosis, whereas wt cells are resistant to the effects of serum withdrawal. In contrast to wt cells, activation and maintenance of PI3K/AKT-mediated signaling in Psen1− /− cells was impaired and insufficient to withstand serum deprivation. This pathway is considered a major pro-survival pathway for cells as it provides signals to respond to and overcome apoptotic stimuli such as serum deprivation 33 . The possible involvement of Psen1 in PI3K/AKT signaling has been suggested previously in studies which found that cultured fibroblasts and neurons show reduced levels of basal pAKT under normal growth conditions 43,44 . Moreover, Psen1− /− neurons show a progressive reduction in basal pAKT as they mature, which is associated with activation of caspase 3 and apoptosis 44 . In endothelial cells, the presence of Psen1 appears to have protective effects under stress conditions such as serum deprivation. This effect was not associated with Psen1 γ -secretase activity, as treatment of wt cells with a γ -secretase inhibitor did not result in apoptosis. This is consistent with the results of other studies in fibroblasts and neurons showing that Psen1's effect on AKT signaling is γ -secretase-independent 43,44 .
Previous reports found that apoptosis induced by serum deprivation of endothelial cells can be prevented by FGF, VEGF, or phorbol-12, myristate-13 acetate 21,22,26,32,35 . In the case of the Psen1− /− endothelial cells, we found that both FGF1 and FGF2 prevent apoptosis in serum-free conditions. VEGF, however, unlike in other endothelial cells types 21,35 , could not rescue Psen1− /− cells from apoptosis. Studies in HUVECs and bovine aortic endothelial cells have shown that VEGF can have proapoptotic effects when the PI3K/AKT signal is blocked 36 . Because Psen1− /− endothelial cells have impaired PI3K/AKT activation when subjected to serum starvation, it might be that under these conditions VEGF in fact favors apoptosis.
How FGF2 prevents apoptosis is unclear. FGF2 activates PI3K/AKT signaling making one possible mechanism that exogenously added FGF2 stimulates PI3K/AKT signaling sufficiently that heightened growth factor stimulation can overcome deficient PI3/AKT activation in Psen1− /− cells. However, if direct effects on PI3K/ AKT signaling explain the prosurvival actions of FGF2 it is curious that VEGF, which also activates PI3K/AKT signaling 21 , did not rescue. FGF2 activates other signaling pathways including MAPK. In HUVECs, FGF2 also appears to act independently of its mitogenic effects and activation of MAPK to promote cell attachment 22 .
Another study in HUVECs showed the existence of an FGF2 autocrine loop by which cells can become more resistant to serum deprivation by increasing the amount of FGF2 on the cell surface 45 . Under normal conditions, FGF is supplied to endothelial cells in culture by serum and endothelial cell growth supplement, a poorly defined extract from bovine neural tissues known to contain FGF1 and FGF2 46,47 . Therefore, it is possible that under conditions of serum starvation Psen1− /− cells cannot compensate in an autocrine fashion for the loss of the exogenous FGF.
FGF2 is synthesized primarily as its 18-kDa low molecular weight form, and, via alternative initiation codons, higher molecular forms are made that are only intracellular and remain in the cytosol and nucleus 40 . The 18-kDa FGF2 is transported outside the cell by an unconventional secretion pathway that does not involve the endoplasmic reticulum/Golgi-mediated secretion pathway but rather direct translocation across the plasma membrane in a process involving phosphatidylinositol biphosphate (PIP2), ATP1A1, and Tec 1 kinase 48,49 . Translocated FGF2 is not secreted into the medium but retained at the cell surface by binding to heparan sulfate proteoglycan which acts as an extracellular trap and results in FGF2 acting in a paracrine or local manner near the source of its expression 48,49 .
FGF2 signaling is activated by its interaction with FGFR on the cell surface. However, FGF2 can also be taken up by receptor-mediated endocytosis into the cytosol and transported from intracellular vesicles to the nucleus. This translocation appears to be important for eliciting the mitogenic activity of FGF2 40 . Previous studies in HUVECs 45 showed that the 18-kDa form was responsible for protection from apoptosis under serum withdrawal conditions. Therefore, absence of Psen1 could impair the transport of FGF across the membrane. In fact, we have observed that although after 16 h of serum starvation most Psen1− /− cells undergo apoptosis, a few cells can survive. If maintained in the apoptotic conditioned media the surviving cells are eventually able to recover. This supports the idea that in Psen1− /− cells under serum starvation the release of FGF may occur at a much reduced rate, taking longer to build up the FGF2 signaling necessary to activate the PI3K/AKT pathway to promote survival. As for wild type cells, their resistance to serum deprivation is likely due to the fact that they can rapidly compensate for the loss of exogenous FGF by releasing sufficient endogenous FGF to keep the survival pathway active and prevent apoptosis from occurring.

Conclusions
Serum starvation is routinely used as a tool to study molecular mechanisms involved in protein degradation, cellular stress responses, autophagy, and apoptosis under well-defined normal and pathological conditions. In the present study, we found that serum starvation of brain endothelial cells deficient for Psen1 induces cellular apoptosis and is associated with impairment of pro-survival PI3K/AKT signaling. In contrast, cells expressing Psen1 were insensitive to serum starvation and displayed normal PI3K/AKT signaling. The protective effect of Psen1 expression upon serum starvation was independent of its γ -secretase function. We also showed that addition of FGF1 or FGF2 in the serum-free medium entirely prevents apoptosis in Psen1− /− brain endothelial cells. Analysis of FGF2 by immunostaining and Western blot showed that FGF2 levels and distribution in endothelial cells did not differ significantly in the presence or absence of serum in either cell type. This suggests that the sensitivity of Psen1− /− endothelial cells to serum deprivation is not due to changes in the intracellular FGF2 levels or localization within the cell.
FGF2 may rescue Psen1− /− endothelial cells from apoptosis through its ability to activate PI3K/AKT signaling. FGF2 is also an unusual protein that is not secreted directly into the medium but released through a non-canonical mechanism that involves translocation through the cell membrane and retention at the cell surface mainly by heparan sulfate proteoglycans. If Psen1 plays a role in this translocation process in the absence of serum Psen1− /− cells may fail to release sufficient endogenous FGF to the cell surface, a defect that can be compensated for by the provision of exogenous FGF.

Methods
Coating of culture surfaces. Polystyrene Nucleon plasticware (6-and 12-well dishes) and Permanox chamber slides (Nunc, Roskilde, Denmark) were coated with a solution of 10 μ g/ml mouse collagen type IV (BD Biosciences, Bedford, MA) in 50 mM HCl (0.14 ml/cm 2 ) and air dried for 48-72 h in a sterile biological safety cabinet. Collagen type IV-coated plasticware was stored in sealed bags at − 20 °C until use. Preparation of embryonic cerebral vascular endothelial cells. Endothelial cells were isolated from E13.5-14.5 embryonic telencephalic and mesencephalic brain regions and grown on collagen type IV-coated plasticware in ECGM, which consisted of Dulbecco's Modified Eagle Medium/Ham's F12 (DMEM-F12) supplemented with 10% heat-inactivated horse serum, 10% heat-inactivated fetal calf serum, 100 μ g/ml endothelial cell growth supplement, and 100 μ g/ml heparin 31 . Cells were always kept at high densities and passaged at a 1:2 dilution. The endothelial phenotype of mouse embryonic endothelial cells was verified by immunocytochemical staining with antibodies against PECAM-1, fibronectin, and laminin as previously described previously 31 .

Serum starvation, growth factors, and inhibitors. Cultured endothelial cells were rinsed 3 times
with Hanks balanced salt solution (HBSS) and serum-starved in either DMEM alone or DMEM containing 0.1% bovine serum albumin (BSA, fraction V) and 100 μ g/ml heparin. FGF1, FGF2, and VEGF (R&D Systems, Minneapolis, MN) were added individually at 25 ng/ml in DMEM-BSA-heparin media along with the γ -secretase inhibitor XXI (compound E, 10 μ m, EMD Millipore, Billerica, MA) or the FGFR antagonist SU5402 (10 μ m, Santa Cruz Biotechnology, Santa Cruz, CA). In some experiments, FGF1 and FGF-2 were added in DMEM alone without BSA and heparin.

Western blotting. Endothelial cells (attached and unattached cells in the case of apoptotic cells) were lysed in
10 mM sodium phosphate, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% Triton, 0.5% Na deoxycholate, and 0.5% sodium dodecyl sulfate (SDS) supplemented with a protease and phosphatase inhibitor cocktail (Pierce, Rockford, IL). Protein concentration was determined with the BCA reagent (Pierce). Aliquots of the lysates (10 μ g protein) were separated by SDS-polyacrylamide gel electrophoresis (PAGE) and blotted onto polyvinyl difluoride (PVDF) membranes. Membranes were blocked for 1 h in Tris buffer saline/1% Tween-20 (TBST) containing 0.5% non-fat dry milk and incubated overnight at 4 °C with the appropriate primary antibody diluted in the above blocking solution or in TBST/5% BSA. After washing with phosphate-buffered saline (PBS), the membrane was incubated with anti-rabbit IgG horseradish peroxidase-conjugated secondary antibodies (GE Healthcare, Piscataway, NJ) and the bands visualized with the ECL Prime reagent (GE Healthcare). The membranes were exposed to film or imaged with the Amersham Imager 600 (GE Healthcare). Quantitation was performed with Image Quant software (GE Healthcare). Statistical analysis was performed as previously described 14  For FGF2 analysis EC cells were serum-starved for 6 h and harvested along with control cells that were not serum-starved by scraping with a rubber policeman. Cytosol and nuclear fractions were isolated using the N-Per reagent (Pierce) as described by the manufacturer's instruction. Protein concentrations were determined by the  Immunocytochemistry and TUNEL assay. Endothelial cells grown on collagen IV-coated slides were fixed with acetone/methanol or 4% paraformaldehyde in phosphate-buffered saline (PBS). The slides were blocked for 1 h in TBS containing 0.3% Triton X-100 and 10% normal goat serum and incubated overnight at room temperature with primary antibodies diluted in the above blocking solution. Slides were washed with PBS and incubated for 2 h with the appropriate secondary antibodies conjugated to Alexa Fluor 488 or 562 (Molecular Probes-ThermoFisher, Grand Island, NY) before being washed with PBS and mounted with Fluorogel mounting medium (EMS, Hatfield, PA,). The following antibodies were used: rabbit polyclonal anti-laminin (1:100) and rabbit polyclonal anti-fibronectin (1:400; both from Sigma-Aldrich), rat monoclonal anti-PECAM (CD-31, 1:100; BD Biosciences, San Jose, CA), and rabbit polyclonal anti-FGF2 (1:300; Abcam). Terminal deoxynucleotidyl transferase dUTP (TUNEL) staining was performed using the In Situ Cell Death Detection Kit (Roche Life Science, Indianapolis, IN) according to the manufacturer's instructions.