Diabetic LDL Triggers Apoptosis in Vascular Endothelial Cells

M199, delipidated calf serum, butylated hydroxytoluene, diethylenetriamin-pentaacetic acid, sodium cyanoborohydride, trinitro-benzenesulfonic acid (TNBSA), LPL, Triton X-100, Nonidet P-40, Na₃VO₄, β-glycerophosphate, phenylmethanesulfonyl fluoride, aprotinine, leupeptine, and collagenase were purchased from Sigma Chemical (St. Louis, MO). PBS, l-glutamine, pen/strep, fungizone, and FCS were from Hyclone (Logan, UT). Optimem 1 was from Gibco (Paisley, UK) and glucose solution for infusion (20%) from Mayrhofer Pharmaceuticals (Linz, Austria). KBr, EDTA, NH₄COOH, NaCl, NaN₃, MgCl₂, NaOH, NaHCO₃, Tris(hydroxymethyl)-aminomethan (Tris), and HEPES were purchased from Merck (Darmstadt, Germany); Glyco Gel II boronate affinity columns and Super Signal Substrate from Pierce (Rockford, IL); and the Bioxytech LPO-586 kit from Oxis (Portland, OR). [Methyl-³H]thymidine and HRP-labeled sheep anti-mouse Ig were from Amersham Pharmacia (Buckinghamshire, UK), DNase I from Boehringer Mannheim (Mannheim, Germany), and the TdT-FragEL kit from Oncogene (Boston, MA). SDS and Tween-20 were from Bio Rad (Hercules, CA). Antibodies were obtained from Santa Cruz Biotechnology (Santa Cruz, CA; p53), Transduction Laboratories (Lexington, KY; p21^WAF-1/Cip1, p27^Kip, Cyclin D3, CPP-32, and endothelial NO synthase [eNOS]), and Calbiochem (Cambridge, MA; retinoblastoma protein [pRb], bcl-2, and bak). The caspase inhibitory peptides Z-VAD-FMK and Ac-DEVD-CHO were from Calbiochem (Cambridge, MA) and Bachem (Bubendorf, Switzerland), respectively.

LDL was prepared from fasting EDTA plasma of healthy normolipidemic subjects (n = 17) and diabetic patients (13 men and 4 women, 1 type 1 diabetic patient, 16 type 2 diabetic patients, mean age 60 ± 9 years). EDTA was present throughout the entire isolation procedure. In diabetic patients, mean values were 10.3 ± 1.5% (reference range 4.8–5.2%) for HbA_1c, 229 ± 50 mg/dl (150–200) for cholesterol, 43 ± 10 mg/dl (>45) for HDL, 142 ± 44 mg/dl (<130) for LDL, and 231 ± 133 mg/dl (50–172) for triglycerides.

LDL was isolated by sequential ultracentrifugation in the presence of EDTA using standard techniques (10,12,24). In brief, after density adjustment of plasma to 1.063 g/ml with solid KBr and a centrifugation step (190,000g at 15°C for 18 h, 70.1Ti fixed angle rotor; Beckman Instruments, Fullerton, CA), floating lipoproteins (adjusted to ρ = 1.2 g/ml with KBr) were separated by centrifugation (230,000g at 15°C for 26 h, SW-40Ti swinging bucket rotor; Beckman Instruments) in a linear density gradient (ρ = 1.2–1.01 g/ml). The obtained LDL fraction (ρ = 1.019–1.063 g/ml) was covered with nitrogen to prevent oxidation and dialysed for 3 days at 4°C in the dark against 2 l PBS containing 0.01% EDTA, with daily changes of dialysis solution. Total cholesterol, representing the concentration of isolated LDL, was measured by a standard procedure (CHOD/PAP method).

LDL isolated from the plasma of healthy volunteers was separated into glycated and nonglycated LDL subfractions by Glyco Gel II boronate affinity chromatography.

The nonglycated LDL fraction (diluted to 1 mg/ml total cholesterol) was then incubated (1 week at 37°C in the dark) in PBS containing 2 mg/ml EDTA, 25 μmol/l butylated hydroxytoluene, and 50 μmol/l diethylenetriamin-pentaacetic with either various concentrations of d-glucose (range 30–350 mmol/l; LDL_iv) or without d-glucose added (LDL₀) under nitrogen (10,12). Nonglycated “control” LDL (LDL₀) and LDL_iv samples were covered with N₂, dialysed in the dark (3 days at 4°C against 2 l PBS containing 0.01% EDTA) with daily changes of dialysis solution, stored under N₂ at 4°C in the dark, and used within 1 week.

LDL isolated from the plasma of diabetic patients (LDL_D) was divided into two portions, of which one was applied to Glyco Gel II boronate affinity chromatography columns (as described above) for separation of nonglycated LDL (LDL_N). Both LDL_D and the nonglycated “control” fraction (LDL_N), were covered with N₂, dialysed, and stored as described above.

The degree of glycation was determined using TNBSA, which couples with primary amines and peptides in aqueous solution to give trinitrophenyl derivates (25). In brief, 50 μl LDL_iv and 50 μl LDL₀, both containing 50 μg total cholesterol, were mixed with 50 μl TNBSA (0.1%) and 1 ml NaHCO₃ (4%, pH 8.5). Solutions were covered with nitrogen and incubated at 37°C for 2 h in the dark. Samples were read at 340 nm against a mixture of LDL, NaHCO₃, and distilled H₂O (instead of TNBSA) on a UV/VIS λ2 Spectrometer (Perkin Elmer, Norwalk, CT) using Perkin Elmer computerized spectroscopy software, version 4.0. Relative reduction of free ε-amino groups of l-lysine in glycated components was thus evaluated by comparing the absorbance of LDL_iv with that of (nonglycated) LDL₀.

Oxidative modification of glycated and nonglycated LDL fractions was ruled out by using a sandwich fluorescence immunoassay (using the monoclonal antibody mabOB/04 raised against copper-oxidized LDL as described previously (12).

Human umbilical vein endothelial cells (HUVECs) were isolated and cultured, as previously described (19,26–28), and identified by their “cobblestone” morphology and positive staining for von Willebrand factor antigen. HUVECs were used as individual isolates in primary, or first, subculture. Subconfluent HUVEC cultures were used in experiments determined for Western blot analysis and in situ apoptosis staining, whereas [³H]thymidine apoptosis assays and proliferation assays were performed with growing cultures.

HUVEC cultures were exposed (24 or 48 h) to 100 mg/l (total cholesterol) of in vitro–glycated (LDL_iv) or diabetic LDL (LDL_D) in Optimem 1 supplemented with 5% delipidated bovine serum and antibiotics. For control experiments, each HUVEC culture was exposed to the intraindividual nonglycated LDL control subfraction of the respective healthy donor (LDL₀) or diabetic patient (LDL_N). To allow the binding of added LPL (100 units/l) to endothelial cells, cell culture media did not contain heparin (10). Experimental and control cultures were tested for proliferation, apoptosis, and associated protein expression after 24 and 48 h.

Exponentially growing primary HUVECs were labeled with [³H]thymidine (1 μCi/ml). After reaching confluency (72 h), cultures were trypsinized and seeded into 24-well culture plates (60,000 cells per well). After adherence (4 h) and exposure (24 and 48 h) to LDL (LDL_iv versus LDL₀ or LDL_D versus LDL_N), cells were lysed (20 mmol/l Tris.Cl, pH 7.5, and 0.4% Triton X-100 in PBS) and fragmented (apoptotic) DNA was separated by centrifugation (19,27,28). Radioactivity of fragmented versus total DNA (digested with 180 μg/ml DNase I) was quantitated in triplicates (1900 TR Liquid Scintillation Analyser; Canberra Packard, Meriden, CT). Results are expressed as fragmented DNA per total DNA in relation to the respective intraindividual control cultures (set to 100%).

After exposure (48 h) of endothelial cells to LDL (LDL_iv versus LDL₀ or LDL_D versus LDL_N), rates of apoptosis were determined using the TdT-FragEL kit according to the manufacturer’s instructions (27). Briefly, cells were fixed with ice-cold methanol and permeabilized with Triton X-100 (0.2% in PBS). After equilibration (TdT-equilibration buffer 1:5 in distilled H₂O), cells were incubated with TdT labeling reaction mixture (TdT enzyme 1:20 in TdT labeling reaction mix; 1 h at 37°C with 95% relative humidity) and subsequently with conjugate solution (conjugate 1:50 in blocking buffer; 30 min at room temperature). Apoptotic cells were then stained with a solution of one tablet each of diaminobenzidine and urea in 1 ml tap water for 15 min at room temperature, resulting in an insoluble brown color. Counterstaining with methyl green was used to quantitate viable (green) and apoptotic (brown) cells by counting at least 1,000 nuclei in multiple fields in a blinded fashion under the microscope (29). Rates of apoptosis are given as the percentage of positively stained nuclei (brown) to total nuclei in relation to the respective intraindividual control cultures (set to 100%).

Confluent monolayers of HUVECs were trypsinized, seeded into 96-well culture plates (10,000 cells per well), and (after 6 h attachment) exposed (48 h) to LDL_iv or LDL_D and the respective nonglycated LDL subfractions as controls (LDL₀/LDL_N) with and without LPL added in the presence of 1 μCi/ml [³H]thymidine. In brief, cells were lysed, harvested (96-well cell harvester; Tomtec, Hamden, CT), and incorporated [³H]thymidine was quantitated (β-plate reader; Wallac Oy, Turku, Finland) (27,28). Results are expressed as percentage of the respective intraindividual control cultures (set to 100%). Samples were tested in quadruplicates.

Endothelial cells exposed (48 h) to either LDL_iv or LDL_D (versus controls treated with LDL₀ or LDL_N) were rinsed with cold PBS and subsequently lysed (27,28). Equivalents of 10 μg total protein were subjected to SDS polyacrylamide gel electrophoresis, transferred onto nitrocellulose membranes, and blocked for unspecific binding sites (27,28). The blots were incubated with primary antibodies (p53 and p27^Kip 1:1,000; p21^WAF-1/Cip1, cyclin D3, CPP-32, and eNOS 1:500; bak 1:100; and pRb and bcl-2 1:50), which were detected with HRP-conjugated anti-mouse IgG using Super Signal Substrate. After exposure of the membranes to Kodak XAR5-Omat films, semiquantitative evaluation of visualized bands was carried out by densitometry (Gel documentation system; MWG Biotech, Ebersberg, Germany) using Gene Profiler 3.56 for Windows.

Data are expressed as means ± SE. Statistical analysis was performed using paired samples t test (SPSS for Windows 7.5.1) with Bonferroni correction for multiple comparisons. Data are considered significant if P < 0.05. If not stated otherwise, six or more different individual HUVEC isolates were used for the experiments.

Mean concentrations of isolated LDL were 5.1 ± 1.1 g/l cholesterol for healthy donors (n = 17) and 3.4 ± 1.3 g/l for diabetic patients (n = 17). Oxidation products were not detectable in any of the LDL preparations (data not shown). Mean glycation of LDL isolated from healthy donors and subsequently glycated in vitro were 0, 5 ± 1, 19 ± 1.6, and 42 ± 3% upon exposure to 0, 30, 100, and 350 mmol/l glucose, respectively (Fig. 1A).

Depending on the degree of glycation, in vitro glycated LDL (LDLiv) increased apoptosis in HUVECs (n = 4) up to 163% compared with LDL₀ (set to 100%, P < 0.05), as determined by DNA fragmentation assays (data not shown) and in situ staining (Fig. 1B–D).

LPL (100 units/l) per se did not affect basal apoptosis of HUVECs (91 ± 6% of control without LPL added, set to 100%), but reduced that triggered by LDL_iv from 163 to 113 ± 9% of control (LDL₀ + LPL set to 100%; n = 4, P < 0.01; data not shown in figures). Diabetic LDL (LDL_D) increased apoptosis in HUVECs after 48 h up to 140% of LDL_N (set to 100%, P < 0.01) (Fig. 2A and B) in the absence of LPL, as documented by both [³H]thymidine assays (Fig. 2A) and in situ staining (Fig. 2B). By adding LPL, LDL_D-triggered apoptosis was further increased to 184% of control (LDL_N + LPL set to 100%, P < 0.001) (Fig. 2A and B) in a time-dependent fashion (Fig. 2A). LDL_D-induced apoptosis was reduced close to baseline (110 ± 9%; n = 5, P < 0.05) by the caspase inhibitors Ac-DEVD-CHO (100 μmol/l) and Z-VAD-FMK (40 μmol/l).

In pilot experiments (n = 2; data not shown in figures), exposure (24 h) of individual HUVEC cultures to high LDL_D concentrations (500 mg/l) increased rates of apoptosis to 125% (99% at 100 mg/l) (compare Fig. 2A) in the absence of LPL and to 155% (140% at 100 mg/l) (compare Fig. 2A) in the presence of LPL, each compared with 500 mg/l LDL_N, suggesting some concentration dependency of LDL_D action.

In vitro–glycated LDL (LDL_iv) reduced proliferation of HUVECs in both the absence (−34%, P < 0.01) and presence (−37%, P < 0.001) of LPL (100 units/l) in a glycation-dependent fashion, when compared with LDL₀ (Fig. 3A). LPL (100 units/l) per se did not affect HUVEC proliferation (98 ± 7% of control without LPL added, set to 100%).

Diabetic LDL (LDL_D) slightly reduced proliferation of HUVECs in the absence (−9%, P < 0.01), but not the presence, of LPL (Fig. 3B) when compared with LDL_N action.

In vitro–glycated LDL (LDL_iv) increased protein expression of the apoptosis promotors bak (Fig. 4A and C) and CPP-32 (Fig. 4B and D) in a glycation-dependent fashion to 188 and 149%, respectively (P < 0.05 compared with LDL₀, set to 100%). These effects were blunted by LPL (data not shown), as was apoptosis. In contrast to bak and CPP-32, cellular expression of bcl-2, a prominent inhibitor of apoptosis, remained unaffected by LDL_iv (111 ± 17% of LDL₀; data not shown within figures).

Diabetic LDL (LDL_D) likewise upregulated (P < 0.05) bak (18%) (Fig. 5A) and CPP-32 (11%) (Fig. 5B), which were somewhat further increased upon addition of LPL (bak 33% and CPP-32 22%, P < 0.05 compared with control [= LDL_N + LPL]). However, LDL_D, like LDL_iv, did not affect bcl-2 expression in HUVECs (110 ± 7% of LDL_N set to 100%; data not shown within figures).

HUVECs exposed to in vitro glycated LDL (LDL_iv) reduced (P < 0.05) cyclin D3 expression (Fig. 6A) in both the absence (−27%) and presence of LPL (−34%) in a glycation-dependent manner. In parallel, hypo- (−LPL: −22 ± 6%; +LPL: −27 ± 8%; data not shown within figures) and hyperphosphorylated (−LPL: −53%; +LPL: −61%) (Fig. 6C) forms of the pRb were reduced by glycated LDL (P < 0.05), while the tumor suppressor p53 and the inhibitors of cyclin-dependent kinases p21^WAF-1/Cip1 and p27^Kip remained unaffected (data not shown).

Diabetic LDL (LDL_D) reduced (P < 0.05) expression of cyclin D3 (−24%, Fig. 6B) and of both hypo- (−19 ± 3%; data not shown within figures) and hyperphosphorylated (−22%) (Fig. 6D) forms of pRb in the absence of LPL, but not in its presence (Fig. 6B and D; data of hypophosphorylated pRb are not shown within figures). Expression of the cell-cycle checkpoint molecules p53, p21^WAF-1/Cip1, and p27^Kip was not affected by LDL_D (data not shown).

eNOS expression was reduced in HUVECs by both LDL_iv (−30%, P < 0.05) (Fig. 7A) and LDL_D (−23%, P < 0.001) (Fig. 7B) compared with the respective control cultures exposed to LDL₀ or LDL_N (set to 100%), respectively. This effect was augmented by addition of LPL for LDL_iv (−51%, P < 0.001 vs. LDL₀ + LPL) (Fig. 7A) but not for LDL_D (−23%, P < 0.05 vs. LDL_N + LPL) (Fig. 7B).

This study demonstrates that both in vitro (LDL_iv) and in vivo glycated (LDL_D [diabetic]) LDLs reduce proliferation and trigger apoptosis in HUVECs. Glycated LDL’s proapoptotic activity already occurs at a concentration (100 mg/l) corresponding to ∼10% of that seen in normal human plasma and further increases with higher concentrations, as shown for 500 mg/l (half of the normal plasma concentration). From this, it appears that glycated LDL could contribute to endothelial dysfunction in vivo and thus could be critically involved in the acceleration and pathogenesis of diabetic angiopathies (13,14,23), as reported for other inducers of endothelial apoptosis, such as high glucose (13,19), oxidized LDL (30,31), free fatty acids (22), and homocysteine (21). Endothelial cells undergoing apoptosis impair the intact endothelial monolayer and barrier function, trigger plaque erosion as well as rupture, and can lead to coronary thrombosis due to provision of proadhesive and procoagulatory activity (14,18,32).

The present study was performed in HUVECs that are readily available and represent an established in vitro model for endothelial cells. Although preliminary data obtained at our laboratory suggest LDL_iv’s antiproliferative and LDL_D’s proapoptotic activity to also relate to other types of human endothelial cells (adult venous and microvascular endothelial cells; M.A., unpublished observations), it cannot be excluded that endothelial cells originating from other vascular beds might show a different response.

Since oxidation products were not detectable in any of the LDL preparations, the observed action of glycated LDL is not attributable to LDL oxidation. Notably, although diabetes-associated oxidative stress is assumed to account for oxidation of lipoproteins, recent data suggest that neither diabetic nor in vitro–glycated LDL per se show any significant degree of oxidation (12,33,34). These observations could relate to the presence of 1) defense systems as antioxidants and metal ion binding proteins present in plasma and 2) the antioxidants EDTA and butylated hydroxytoluene throughout the isolation and in vitro glycation procedures. In vivo, LDL’s oxidative modification appears to predominantly occur in the arterial intima, being in line with the presence of oxidized LDL in atherosclerotic lesions (35). Since LDL_D and LDL_iv differ in their action against the endothelium, one could speculate that such differences relate to their different oxidizability in vascular cells.

Compared with LDL_iv preparations, which were glycated up to 42%, glycation of LDL in diabetic patients, previously shown to correlate with HbA_1c (1,4–6), is lower and ranges from 5 to 10% only (1,5,6). Thus, in relation to its degree of glycation, LDL_D exhibits a 2.7-fold higher proapoptotic activity in endothelial cells than LDL_iv.

Apoptosis induced by LDL_iv, as well as by LDL_D, is associated with increased expression of the apoptosis promotors bak and CPP-32 (synonym for caspase-3) (Table 1). These proteolytically activated cystein proteases are responsible for the destruction of the cellular architecture during apoptosis and for detachment and subsequent clearance of apoptotic cells from the surrounding tissue (36,37). The major importance of caspases in glycated LDL-evoked endothelial cell death is confirmed by the finding that HUVECs are rescued from LDL_D-induced apoptosis by caspase inhibitors such as Ac-DEVD-CHO (inhibiting caspase-3, -6, -7, -8, and -10) and Z-VAD-FMK (inhibiting caspase-1, -3, -4, and -7) in the presence and absence of LPL.

Since LDL_D’s proapoptotic activity exceeds that of LDL_iv by 2.7-fold (in the absence of LPL) when corrected for the degree of glycation, LDL_D-induced apoptosis is associated with relatively lower bak and CPP-32 expression (Table 1) when compared with LDL_iv. LDL_D-triggered apoptosis could thus involve additional, yet unidentified, mechanisms. However, neither LDL_iv nor LDL_D affected the expression of bcl-2, a prominent inhibitor of apoptosis (28). The resulting increased ratio of bak/bcl-2 favors a proapoptotic state and is thus in line with the observed phenomena.

LDL_iv-induced apoptosis (163% of control) is reduced (113% of control) by addition of LPL, as is the associated expression of bak and CPP32. However, in the presence of LPL, LDL_D’s proapoptotic activity (352.8% when corrected for the degree of glycation) clearly exceeds (27.1-fold) that of LDL_iv.

While diabetic LDL has higher proapoptotic potential compared with in vitro glycated LDL, the antiproliferative activities of LDL_iv and LDL_D are comparable considering the latter’s lower degree of glycation (Table 1). Since diabetic LDL exhibits its antiproliferative activity in the absence but not the presence of LPL, it remains elusive whether its antiproliferative activity has a role in the development of vascular endothelial dysfunction in vivo, where LPL is bound to the endothelium.

Reduction of cyclin D3 and hypo- and hyperphosphorylated pRb is exclusively seen in association with the antiproliferative activity exerted by LDL_iv (with or without LPL) and LDL_D (without LPL) in HUVECs. Such findings suggest that the antiproliferative activity of both glycated LDL species relates to reduced expression of the growth factor sensor cyclin D3 and of pRb, particularly of its hyperphosphorylated form. It remains, however, unknown why the LDL_D-associated reduction of cyclin D3 and hypophosphorylated pRb exceeds that induced by LDL_iv (3.7- and 3.6-fold, respectively) (Table 1) when corrected for the degree of glycation. Reduction of cyclin D3 is rate limiting for S-phase entry (38), and underphosphorylation of pRb results in diminished liberation and availability of E2F (39), a transcription factor that triggers cell progression through the cell cycle by induction of target genes (e.g., cyclins E and A and several DNA replication enzymes) (40,41). In contrast to other models of apoptosis and associated growth arrest (42), in subconfluent HUVECs, the tumor suppressor p53 and the inhibitors of cyclin-dependent kinases p21^WAF-1/Cip1 and p27^Kip are affected by neither LDL_iv nor LDL_D.

eNOS has been shown to be a prerequisite of vascular endothelial growth factor–induced endothelial proliferation (43) and to be reduced in association with endothelial apoptosis and growth arrest (27,44,45). The present study shows that although reduced by both LDL species, LDL_D-induced reduction of eNOS is more pronounced in both the absence (3.2-fold) and presence (1.9-fold) of LPL when corrected for the degree of glycation.

Abrogation by LPL of LDL_iv’s proapoptotic and LDL_D’s antiproliferative activity, but not of the respective reduction in eNOS expression (Table 2), however, excludes a regulatory role of eNOS in glycated LDL’s proapoptotic or antiproliferative responses of endothelial cells. Since reduction of eNOS expression could diminish NO bioavailability (12), LDL_D could well account for the impaired endothelial vasodilation seen in hyperglycemic states (46,47).

The effects elicited in the vascular endothelium by LDL_D are more pronounced than those induced by LDL_iv, with respect to apoptosis (−LPL: 2.7-fold; +LPL: 27.1-fold) and reduction of eNOS expression (−LPL: 3.2-fold; +LPL: 1.9-fold). In addition, the proapoptotic and antiproliferative activity of LDL_iv and LDL_D differs completely in response to LPL (Table 2). These findings suggest that the effects of LDL_iv and LDL_D in vascular endothelial cells are caused by the molecules’ different structure or their different cellular uptake, which is currently poorly characterized. In this context, it is worth noting that nonglycated LDL and LDL_iv are taken up by either the LDL receptor–dependent pathway (nonglycated LDL) or by nonclassical pathways that are independent of LDL receptor and LDL receptor–related protein (10). Such nonclassical pathways could involve LPL (10), surface glycosaminoglycans (10,48), or yet uncharacterized interactions of glycated LDL with the respective cell surface in the absence of LPL. The observation that binding, uptake, and degradation of moderately compared with highly in vitro–glycated LDL preparations are different (10) suggests that such differences might also exist for binding, oxidizability, and catabolism of in vivo–glycated diabetic LDL.

In conclusion, the present study shows that in the presence of LPL, the activity of LDL_D (isolated from the plasma of diabetic patients) clearly exceeds that of in vitro glycated LDL_iv with respect to apoptosis (27.1-fold) when corrected for the degree of glycation. From this it is apparent that in vitro–glycated LDL cannot generally be used as a model substance for diabetic LDL, which, due to its more pronounced activity, carries a higher potential for endothelial cell damage than in vitro–glycated LDL preparations.

This study was supported by grants from the Austrian National bank (6568 to S.M.B.-P.) and the Austrian Science Foundation (SFB 714 to W.F.G.).

The assistance of Dr. Claudia Ludwig is greatly acknowledged.