Electronegative LDL Impairs Vascular Endothelial Cell Integrity in Diabetes by Disrupting Fibroblast Growth Factor 2 (FGF2) Autoregulation

Cultures of BAECs, maintained in Dulbecoo's modified Eagle's medium supplemented with 10% FBS and antibiotics, were used at 3–5 passages (28). With approval by the Baylor College of Medicine review board, blood samples were obtained from six patients with type 2 diabetes (current A1C >7%, fasting blood glucose >126 mg/dl, mean ± SEM LDL cholesterol 2.99 ± 87 mmol/l [116 ± 18 mg/dl], and age 56 ± 8 years) and six nondiabetic, normolipidemic adult subjects (LDL cholesterol 2.62 ± 0.22 mmol/l [110 ± 11 mg/dl] and age = 53 ± 4 years). LDL particles (d = 1.019–1.063 g/ml) were isolated by ultracentrifugation (28). LDL preparations were fractionated with a fast protein liquid chromatography system (Pharmacia Biotech) through a UnoQ12 anion-exchange column (BioRad). LDL subfractions were eluted with a linear gradient program at a flow rate of 2 ml/min, with concentration monitoring by absorbance at 280 nm (3). Protein concentrations were determined by the Lowry method. oxLDL was prepared by incubating LDL with 5 μmol/l CuSO₄ at 37°C for 24 h. Precautions were taken to prevent all LDL preparations from endotoxin contamination and further oxidation (28).

BAECs were maintained in medium containing 5% serum for 6 h before incubation with various agents for 24 h. Effects of L5 (50 μg/ml), oxLDL (50 μg/ml), unfractionated normolipidemic LDL (50 μg/ml), and PBS (lipoprotein-free control) on BAEC apoptosis were determined. To evaluate their antiapoptotic effects, FGF2 (2–10 ng/ml) or VEGF (10–50 ng/ml; R&D Systems) was added to cells incubated with L5 or oxLDL. To determine the role of Akt in EC survival, cells were treated with the PI3K inhibitor wortmannin (25–200 nmol/l; Sigma-Aldrich) alone or in combination with FGF2 (10–20 ng/ml; R&D Systems) for apoptosis assessment.

Nuclear morphology was examined in treated BAECs stained with Hoechst 33342 (Molecular Probes). Apoptotic nuclei were analyzed with MetaView software (Universal Imaging Corp) in fluorescent images (500 cells/well in three wells) through a Zeiss (Thornwood) inverted microscope (3).

After treatment, activated caspase-3 in BAECs was detected by a caspase-3 detection kit (FITC-DEVD-FMK; Calbiochem) through epifluorescence microscopy. Caspase-3 activity was evaluated by EnzChek Caspase-3 Assay Kit no. 1 (E13183; Invitrogen) according to the manufacturer's instructions.

FGF2 release from the BAECs was examined by a human FGF basic Quantikine enzyme-linked immunosorbent assay kit (DFB50; R&D systems) and analyzed after subtracting the added exogenous FGF2.

FGF2 protein and mRNA levels were assayed by Western and Northern blot analyses, respectively, as previously described (3,28). To assess the state of Akt activation, phosphorylated Akt levels in the cell extracts were determined by immunoblotting with murine anti–phospho-Akt antibody (Ser⁴⁷³ and Thr³⁰⁸; Cell Signaling Technology) (24).

Constitutively active Akt (Akt-CA) plasmid (plncx-HA-myr-Akt) and dominant-negative Akt (Akt-DN) plasmid (plncx-HA-myr-Akt; K179M) were provided by Dr. Marco Marcelli. The constructs were transfected by using FuGENE six (Roche Diagnostics). Briefly, BAECs were plated at a density of 1 × 10⁵ cells/ml in six-well culture dishes and allowed to attach and grow 24 h before transfection. Each transfection mixture was prepared by diluting 3 μl of FuGENE six into 95 μl serum-free Dulbecoo's modified Eagle's medium. A 2-μl aliquot (1 μg) of DNA was gently mixed into the liposomal solution and incubated for 15 min at room temperature. The transfection mixture was slowly added to the cells, which were allowed to recover from the transfection for an additional 24 h before experimental treatments were applied. This transfection method routinely yielded a >60% efficiency.

The primer pair used to amplify a 1,754 base pairs of human FGF2 cDNA consisted of 5′-CAGAAAACCCGAGCGAGTAG-3′ (forward strand) and 5′-ACACAGCGGTTCGAGAAGTT-3′ (reverse strand) custom sequences. The amplified full-length FGF2 cDNA was inserted into the expression vector pcDNA 3.1 (Invitrogen) in the sense and antisense orientations, generating pFGF2(+) and pFGF2(−) expression vectors, respectively. For transfection, 10⁶ BAECs were seeded per 100-mm dish and incubated with 100 ng pFGF2(+) or pFGF2(−) plasmid DNA, using the calcium phosphate method (Roche). Stably transfected cells were selected with 800 μg/ml G418 for 14 days. G418-resistant BAECs that harbored pFGF2(+) and pFGF2(−) DNAs were designated BAEC[FGF2(+)] and BAEC[FGF2(−)], respectively. To detect FGF2 expression in transfected cells, G418-resistant BAECs were fixed in 4% paraformaldehyde and incubated with 6% BSA for 20 min, followed by incubation with anti-FGF2 antibody (R&D Systems; 1:100 dilution) for 24 h at 4°C. After incubation in PBS containing 0.2% Triton X-100, cells were treated with AlexaFluor 594 goat anti-rabbit IgG. Immunocytochemical reactions of the samples were analyzed by fluorescence microscopy.

To assess the rate of FGF2 gene transcription, EC nuclei were isolated with NP-40 lysis buffer and stored at −80°C. After thawing, nuclei from 5 × 10⁷ cells were added to 100 μl 2 × reaction buffer containing 100 μCi [α-³²P]UTP at 30°C for 30 min. To remove template DNA after transcription, RNase-free DNase I and 1 mol/l CaCl₂ were added to the mixture and incubated at 26°C for 30 min. To enhance protein digestion, proteinase K and tRNA (Roche) were added to the mixture for an additional incubation period of 30 min at 37°C. The labeled nascent RNA transcripts were extracted with Tri-Reagent (Molecular Research Center) plus chloroform and precipitated with isopropanol. The RNA pellet were then washed with 70% ice-cold ethanol and dissolved in 0.2% SDS. To detect the newly transcribed RNAs, 1 μg each of FGF2 and β-actin cDNA were immobilized onto nitrocellulose, prehybridized in hybridization buffer, and hybridized at 65°C for 24 h with ³²P-labeled nascent RNA transcripts. The membranes were washed twice with 2 × sodium chloride–sodium citrate at 65°C for 20 min and exposed to X-ray film for 24 h.

The significance of the differences between group means was assessed by a modified t test (Bonferroni test) for multiple comparisons. A value of P < 0.05 was considered significant. Results are expressed as means ± SEM.

L5 was present in each of the six diabetic patients but not in any of the nondiabetic, normolipidemic subjects, although the plasma LDL cholesterol concentrations in the six diabetic patients were as low as those in the control group. An example of the elution profile of diabetic L5 resolved by anion-exchange chromatography is shown (Fig. 1).

L5 was potent in inducing EC apoptosis, while L1 has no effect. Both PI3K stimulators, FGF2 and VEGF, were tested to ascertain whether they are equally protective against L5 (50 μg/ml). Unlike FGF2, which completely prevented apoptosis at a concentration of 20 ng/ml, VEGF up to 40 ng/ml only resulted in a mild attenuation (Fig. 2A). After subtracting the added amount, exogenous FGF2 induced FGF2 secretion threefold in the absence of L5. It also preserved FGF2 secretion at the baseline level, which would otherwise have been reduced in the presence of L5 (Fig. 2B). Enhanced caspase-3 detectability and activities in L5-treated BAECs confirmed the apoptotic nature of L5 (Fig. 2C).

The sharp difference between the effects of FGF2 and VEGF prompted us to focus on the role of endogenous FGF2 in EC protection. Further investigations showed that L5 downregulated FGF2 at the transcriptional level, as determined by RT-PCR. This was, again, reversed by the addition of 10 ng/ml FGF2 (Fig. 2D).

Wortmannin, a PI3K inhibitor, decreased FGF2 protein levels in a concentration-dependent (25–200 nmol/l) manner. At 200 nmol/l, it reduced FGF2 protein by more than one-half in comparison with that reduced by β-actin (Fig. 3A). Supplementation with 20 ng/ml FGF2 increased FGF2 expression in cells treated with 0–50 nmol/l wortmannin. It was able to preserve FGF2 expression to the baseline level, where no exogenous FGF2 or wortmannin was given, in the presence of 50 nmol/l wortmannin. However, FGF2 failed to effectively counteract 100–200 nmol/l wortmannin. Because wortmannin blocks and FGF2 stimulates Akt phosphorylation, the competition between them suggests that FGF2 expression depends on Akt activity and that when PI3K is adequately suppressed, FGF2 is unable to stimulate its own expression.

Since FGF2 protects EC from spontaneous apoptosis, we examined whether wortmannin alone can induce EC apoptosis. Incubation of EC with wortmannin resulted in spontaneous apoptosis within 24 h in a concentration-dependent manner (Fig. 3B). At lower wortmannin concentrations (25 and 50 nmol/l), wortmannin-induced apoptosis could be completely prevented by 10 ng/ml FGF2. At 100 nmol/l, wortmannin induced apoptosis in 75% of cells. FGF2 expression began to succumb to 100 nmol/l wortmannin and failed to attenuate wortmannin-associated apoptosis. At 200 nmol/l of wortmannin, nearly all cells underwent apoptosis, and FGF2 was unable to overcome the effect of wortmannin.

To confirm the findings made with wortmannin, we transiently transfected BAECs with an Akt-CA or Akt-DN expression vector to establish BAEC(Akt-CA) and BAEC(Akt-DN) cells. Compared with empty vector transfected BAEC(ev), the baseline phosphorylated Akt levels were unchanged in BAEC(Akt-CA) and significantly decreased in BAEC(Akt-DN) (Fig. 4A). L5 suppressed Akt phosphorylation in BAEC(ev) but had no significant effect on BAEC(Akt-DN) or BAEC(Akt-CA) transfectants. When 10 ng/ml FGF2 was added simultaneously with 50 μg/ml L5, it completely prevented L5-mediated phosphorylated Akt downregulation in BAECs. In BAEC(Akt-DN) and BAEC(Akt-CA), the phosphorylated Akt levels remained unchanged with or without FGF2 supplementation.

Expression of Akt-CA or Akt-DN markedly altered FGF2 expression in opposite directions. To determine whether elevated FGF2 mRNA levels in BAEC(Akt-CA) may be negatively regulated by L5, cells were treated with L5 for 24 h and FGF2 mRNA levels determined by Northern blotting. As shown in Fig. 4B, L5 treatment significantly decreased FGF2 mRNA in BAEC(ev), while it had no effect on FGF2 mRNA levels in BAEC(Akt-CA) or BAEC(Akt-DN) transfectants.

The above findings indicate that endogenous FGF2 is essential to both FGF2 transcription and Akt phosphorylation. The significance of this positive feedback mechanism was further assessed in BAEC(Akt-DN) and BAEC(Akt-CA) exposed to 50 μg/ml L5. In BAEC(Akt-DN), spontaneous apoptosis was intensified and could not be rescued by FGF2 (Fig. 4C). In contrast, L5 was unable to induce apoptosis in BAEC(Akt-CA), which is strengthened by its artificially reinforced Akt-FGF2 survival loop.

Because oxLDL also induces EC apoptosis, we examined whether oxLDL mimics L5 in inhibiting Akt phosphorylation. At 24 h, oxLDL markedly decreased Akt phosphorylation. In contrast, FGF2 increased the phosphorylated Akt level by almost twofold and completely prevented oxLDL-mediated reduction in phosphorylated Akt levels (Fig. 5A). Compared with PBS and LDL, oxLDL markedly reduced FGF2 mRNA levels, but the inhibition was prevented by 10 ng/ml FGF2 (Fig. 5C). OxLDL also induced apoptosis in ∼50% of treated BAECs (Fig. 5D). Simultaneous addition of FGF2 attenuated oxLDL-induced apoptosis in a concentration-dependent manner. At 10 ng/ml FGF2, very few oxLDL-treated cells were apoptotic. In contrast, 50 ng/ml VEGF failed to attenuate oxLDL-induced apoptosis. These results illustrate that oxLDL, though known to be different from L5, possesses similar cytotoxic properties by disturbing the Akt-FGF2 pathway.

Because other growth factors can also activate PI3K, we examined whether endogenous FGF2 by itself is adequate to regulate its own expression. Stably transfected BAEC lines that over- or underexpress FGF2 were established and designated as BAEC[FGF2(+)] and BAEC[FGF2(−)], respectively. As expected, immunocytochemical staining of BAEC[FGF2(+)] revealed significantly increased levels of intracellular FGF2 compared with parental BAECs (Fig. 6A). In contrast, almost no FGF2 staining was detected in BAEC[FGF2(−)], demonstrating the effectiveness of the antisense FGF2 RNA in inhibiting FGF2 protein expression. Western blot analyses also showed that intracellular FGF2 protein was greatly increased in BAEC[FGF2(+)] and markedly reduced in BAEC[FGF2(−)] compared with BAECs or BAECs transfected with empty vector (Fig. 6B). Nuclear run-on assays confirmed that rates of FGF2 gene transcription were greatly increased in BAEC[FGF2(+)] compared with BAECs and BAECs transfected with the empty vector. The antisense FGF2 RNA in BAEC[FGF2(−)] was expected to reduce FGF2 protein levels only by inhibiting FGF2 translation; therefore, it is important to note that FGF2 transcription was also severely impaired in these cells. The results indicate that FGF2 transcription is dependent on sufficient supply of endogenously produced FGF2.

We also examined whether reducing endogenous FGF2 alone can impair Akt phosphorylation. At 24 h, unfractionated, normolipidemic LDL (50 mg/ml) slightly elevated the phosphorylated Akt level compared with PBS treatment (Fig. 7A). In contrast, L5 caused a significant suppression in Akt phosphorylation. Of importance, increased FGF2 transcription in BAEC[FGF2(+)] was accompanied by a twofold increase in phosphorylated Akt level. In sharp contrast, Akt phosphorylation was severely suppressed in BAEC[FGF2(−)], a change similar to that seen in L5-treated BAECs. Most importantly, FGF2 augmentation in BAEC[FGF2(+)] protected Akt phosphorylation in the presence of L5. Knowing that exogenous FGF2 preserves BAEC survival, we characterized how the endogenous FGF2 transfectants performed in our apoptosis assay, especially when challenged with L5. BAECs transfected with empty vector, [FGF2(−)], or [FGF2(+)] were exposed to 50 μg/ml LDL or L5 for 24 h. As expected, BAEC(ev) control subjects were healthy after LDL administration but highly apoptotic in the L5 group. BAEC[FGF2(−)] transfectants struggled to survive whether they exposed to LDL or L5, while BAEC[FGF2(+)] transfectants were resistant to L5-induced apoptosis (Fig. 7B).

Type 2 diabetes is equivalent to cardiovascular disease for increasing cardiovascular risk even in the absence of elevated LDL cholesterol (2). However, LDL cholesterol remains the primary target for cardiovascular risk reduction in all patients (1). This implies that diabetic LDL, irrespective of plasma LDL cholesterol concentrations, may exhibit certain properties similar to those of nondiabetic, hypercholesterolemic LDL. Here, we demonstrate that the L5 circulating in diabetic patients without elevated LDL cholesterol is similar to hypercholesterolemic L5, as previously described by us (3). In essence, diabetic L5 induces EC apoptosis by disrupting a novel FGF2 autoregulatory mechanism.

Exogenous FGF2 stimulates proliferation and prevents apoptosis induced by various factors in ECs and many cell types (17,29). Because some other growth factors exhibit similar effects (30), it was not appreciated how critical endogenous FGF2 is in securing EC survival. Our findings indicate that L5 inhibits FGF2 transcription with resultant caspase-3 activation and EC apoptosis, which can be prevented by restoring the FGF2-Akt pathway.

Exogenous FGF2 is known to induce expression of many genes (17), but how FGF2 expression is regulated has not been fully described. FGF2 exerts most of its effects by activating the PI3K-Akt pathway (31). Our findings indicate that FGF2 is autoregulated in an autocrine manner through a positive feedback mechanism that involves Akt activation. L5 disrupts this FGF2 autoregulatory mechanism, resulting in EC apoptosis. In early atherosclerosis, minimal apoptosis of vascular EC occurs at sites with low shear stress (32,33). In advanced atherosclerosis, apoptotic ECs contribute to thrombosis by activating tissue factor (34). Thus, protection of this FGF2-PI3K-Akt autoregulatory loop is important for the prevention or deceleration of atherothrombosis.

Electronegative LDL, including L5 and LDL(−), is mildly oxidized in vivo. Their presence strongly supports the role of LDL modification in atherosclerosis (35). Recently, electronegative LDL has also been located in atherosclerotic lesions (36). Under in vitro settings, minimally modified LDL may closely resemble L5 (37). The effects of L5 could be reproduced by oxLDL. Thus, it is not the degree of oxidation but the altered properties of these lipoproteins that are responsible for the adverse effects of modified LDL.

It should be noted that other growth factors, including VEGF, platelet-derived growth factor, epidermal growth factor, and nerve growth factor, can also activate PI3K (25,29,38). Thus, L5 may also act by downregulating these other growth factors. However, our data show that correcting the deficit in FGF2 alone provided complete protection against L5. Failure of VEGF to prevent L5-induced apoptosis suggests that VEGF is unable to substitute for FGF2 in regulating all FGF2-dependent effects. In contrast, FGF2 enhancement may indirectly activate VEGF-regulated bioactivities by stimulating VEGF expression (39).

Treatment of BAECs with wortmannin decreased FGF2 expression, leading to spontaneous apoptosis in the absence of L5. When the effect of wortmannin is overcome by FGF2, FGF2 expression is restored, strongly suggesting that FGF2 expression is an Akt-dependent process. Phosphorylated Akt levels were markedly increased in BAEC(Akt-CA) and severely decreased in BAEC(Akt-DN). The changes in phosphorylated Akt level led to changes in FGF2 expression in the same direction. Thus, although wortmannin may also inhibit other kinases (40), the role of Akt activation in FGF2 expression is confirmed. In BAEC(Akt-DN), apoptosis was not prevented by FGF2 because there was almost no Akt available to convey the FGF2 signaling. In contrast, the Akt-FGF2 signaling was so strong in BAEC(Akt-CA) that L5 was unable to induce apoptosis.

While Akt can also be activated by other growth factors, our findings in the BAEC[FGF2(+)] and BAEC[FGF2(−)] settings proved that endogenous FGF2 plays an essential role in this process. Baseline expression of FGF2 was high in BAEC[FGF2(+)] and almost nonexistent in BAEC[FGF2(−)]. Antisense RNA should inhibit translation but not transcription of the targeted gene. It is therefore important to emphasize that FGF2 transcription was indeed markedly suppressed in BAEC[FGF2(−)]. A major function of FGF2 as an autocrine is to induce its own expression (17). Under our mitogen-free culture conditions, FGF2 appears to be the primary stimulator for its own transcription. In BAEC[FGF2(−)], reduction in intracellular FGF2 protein leads to eventual reduction in FGF2 transcription. Downregulation of FGF2 was always accompanied by a reduction in Akt phosphorylation. Likewise, augmented FGF2 expression was always accompanied by enhanced Akt phosphorylation. Thus, by manipulating Akt and FGF2 expression, we have proven that Akt phosphorylation depends on a continuous supply of endogenous FGF2 and that FGF2 transcription is dependent on Akt phosphorylation.

We previously reported that the apoptotic signals of L5 are transduced by the PAF receptor and that hydrolysis of PAF-like lipids contained in L5 abolishes the apoptotic activity (3). Combining our previous and current findings, we postulate that FGF2 transcription is regulated by a novel FGF2-PI3K-Akt autoregulatory loop and that L5 impairs EC survival by disrupting this signaling loop (Fig. 8). Our findings suggest that L5 may play a role in diabetes-associated atherosclerosis. By disrupting the FGF2-PI3K-Akt loop, L5 can potentially impair the FGF2-dependent reendothelialization and collateral formation essential for tissue perfusion (19,20). A prospective study is warranted to determine the clinical significance of L5 or electronegative LDL in patients with diabetes.

This work was supported by research grants 9630095N and 1-04-RA-13 from the American Diabetes Association (ADA), an Atorvastatin Research Award, a grant from the Philip Morris External Research Program, an Independent Medical Grant from Pfizer Pharmaceuticals (to C.-H.C.), a fellowship from the American Heart Association Texas Affiliate (to J.P.W.), grant AR38858 from the National Institutes of Health (NIH) (to W.S.L.L.), grant HL63364 from the NIH, research grant 7-03-RA-108 from ADA (to C.-Y.Y.), and grants NTUH 92A14 and 93A02 from National Taiwan University Hospital, Taiwan (to P.-Y.C. and Y-.T.L.).

We thank Kerrie C. Jara for editorial assistance.