Effects of Oxidized and Glycated LDL on Gene Expression in Human Retinal Capillary Pericytes

Abstract

purpose. Modified (oxidized and/or glycated) low-density lipoproteins (LDLs) have been implicated in retinal pericyte loss, one of the major pathologic features of early-stage diabetic retinopathy. To delineate underlying molecular mechanisms, the present study was designed to explore the global effects of modified LDL on pericyte gene expression.

methods. Quiescent human retinal pericytes were exposed to native LDL (N-LDL), glycated LDL (G-LDL), and heavily oxidized-glycated LDL (HOG-LDL) for 24 hours, and gene expression was evaluated by DNA microarray analysis. Several of the gene responses were checked, and in each case confirmed by reverse-transcription real-time PCR.

results. HOG-LDL induced a gene expression pattern markedly distinct from that of N-LDL or G-LDL, whereas G-LDL elicited gene expression similar to that of N-LDL. A comparison of responses to HOG-LDL versus N-LDL revealed 60 genes with expression that varied by ≥1.7-fold. The HOG-LDL-responsive genes included members of functional pathways, such as fatty acid, eicosanoid, and cholesterol metabolism; fibrinolytic regulation; cell growth and proliferation; cell stress responses; the kinin system; and angiogenesis.

conclusions. HOG-LDL elicits gene expression in retinal pericytes that may contribute to pericyte loss and other retinal abnormalities in diabetic retinopathy. Observed proapoptotic and proangiogenic responses to HOG-LDL may be of particular importance in this regard. The genes identified through these studies provide potential therapeutic targets for the prevention and treatment of diabetic retinopathy.

Hyperglycemia enhances nonenzymatic glycation and increases oxidative stress, resulting in modification of proteins, lipids, and DNA by covalent adducts and oxygen free radicals. ^{1 2} Extravasated low-density lipoproteins (LDL) are subject to such modification, and in the arterial intima, their role in the promotion of atherosclerosis is now established. In diabetic retinopathy, a leading cause of blindness among adults, ³ capillary leakage occurs early and results in plasma extravasation. Therefore, the possibility that extravasated, modified LDL may affect retinal pericytes merits investigation. In a prior study, we demonstrated a reduction in the number of cultured pericytes after exposure to glycated and mildly oxidized LDL versus native LDL. ⁴ We also observed that glycated-oxidized LDL stimulates apoptosis in cultured retinal pericytes. ⁵ These findings are consistent with the observation that pericytes are lost by apoptosis in vivo in diabetic retinopathy. ⁶  

In the present study, by using DNA microarray analysis, we investigated the effects of in vitro modified human LDL on gene expression in cultured human retinal pericytes. Native LDL (N-LDL) represented normal LDL as found in plasma in healthy people; glycated LDL (G-LDL) represented mildly modified LDL that has not been damaged by oxidation and is present in plasma of diabetic patients; and heavily oxidized-glycated LDL (HOG-LDL) simulated LDL that has been severely modified by prolonged extravasation. We identified altered expression of genes from several functional pathways, particularly in response to HOG-LDL. These findings may help identify new therapeutic targets for prevention and treatment of diabetic retinopathy. 

Methods

Preparation, Modification, and Characterization of LDL

LDL was isolated from three separate groups of human subjects. Each group comprised four healthy, normolipemic, nondiabetic subjects aged 20 to 40 years who were taking neither prescribed medications nor antioxidant vitamins. The study was approved by the Institutional Review Board at the Medical University of South Carolina (MUSC), and informed consent was obtained from all volunteers, in accordance with the provisions of the Declaration of Helsinki. N-LDL density ([d] = 1.019–1.063 g/mL) was prepared by sequential ultracentrifugation, pooled, and modified in vitro, as described previously. ⁷ Briefly, G-LDL was prepared by incubating N-LDL in 50 mM glucose for 72 hours at 37°C under antioxidant conditions. HOG-LDL was prepared from G-LDL by incubation in 10 μM CuCl₂ for 24 hours at 37°C in air. The protein content of LDL was quantified with a bicinchoninic acid (BCA) assay kit (Pierce, Rockford, IL). Characterization of native and modified LDL by gel electrophoresis (Lipoepg; Beckman, Fullerton, CA), fluorescence at 360 excitation/430 emission (Fluorometer IV; Gilford, Oberlin, OH), and absorbance at 234 nm in a spectrophotometer (model DU650; Beckman) indicated that these products were similar to those previously reported. ⁷  

Cell Culture and Modified LDL Treatment

Human retinal pericytes (Clonetics, Walkersville, MD) were grown at 37°C with 5% CO₂ in medium containing 5% fetal bovine serum, 0.04% hydrocortisone, 0.4% hFGF-B, 0.1% VEGF, 0.1% R3-IGF-1 (insulin-like growth factor), 0.1% ascorbic acid, 0.1% hEGF (human epidermal growth factor) and 0.1% GA-1000. At 85% confluence, cells were treated in serum-free medium (SFM) for 24 hours to induce quiescence and then for 24 hours with 100 mg/L N-LDL, G-LDL, or HOG-LDL, or they remained in SFM alone. Total RNA was collected from each sample for gene array analysis. This process was repeated four separate times to create a four replicates with each treatment. For additional experiments using RT-PCR to test the combined effects of HOG- and N-LDL, an identical protocol was used, except that HOG- and N-LDL were added together at 100 mg/L each. 

Total RNA Preparation and Microarray Hybridization

Probes for microarray hybridization were prepared as recommended by Affymetrix (Santa Clara, CA). Briefly, total RNA was isolated and purified (RNeasy Mini kit; Qiagen, Valencia, CA). RNA quality was assured by the A260/280 absorbance ratio and the 28s/18s rRNA ratio. Total RNA (10 μg) was converted into double-stranded cDNA with a T7-(dT)24 primer (Genset, San Diego, CA) and a cDNA synthesis kit (Custom SuperScript; Invitrogen, Carlsbad, CA). Biotin-labeled cRNA was synthesized from cDNA by in vitro transcription (Enzo BioArray HighYield RNA Transcript Labeling Kit; Enzo Life Sciences, Farmingdale, NY). After purification (RNeasy columns; Qiagen), labeled cRNA was fragmented and evaluated by agarose gel electrophoresis, to ensure appropriate size distribution (∼35–200 nt). Hybridization of cRNA samples to (HG-U95Av2 GeneChips; Affymetrix), posthybridization washing, fluorescence staining, and scanning were performed at the MUSC DNA Microarray and Bioinformatics Facility. DNA microarray data (raw and normalized) generated by this project are available online through the MUSC DNA Microarray Database (http://proteogenomics.musc.edu/quicksite/projectmanage.php?page=home). 

Real-Time RT-PCR

Triplicate RT-PCR was performed in a thermocycler (SmartCycler II; Cepheid, Sunnyvale, CA), with a PCR kit and master mix (GeneAmp and SYBR Green; Applied Biosystem, Inc. [ABI], Foster City, CA). Primers were selected so that the forward and reverse primer sequences were present on adjacent exons. Single-stranded cDNA was prepared from total RNA (0.5–1.0 μg) with oligo dT primer and MuLV reverse transcriptase. Two-step PCR was conducted with cDNA template (2 μL) and primers (3 picomoles) in a final volume of 25 μL (denaturation and hot start: 95°C for 10 minutes; 45 cycles with melting: 95°C, 15 seconds; and an extension: 60°C, 60 seconds). Melting-curve analysis (0.5°C/s increase from 55°C to 95°C) and agarose gel electrophoresis were performed to ensure correct amplicon size and reaction specificity. Quantification was determined from the threshold cycle (C _T) of the target relative to that of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) in the same cDNA sample. Gene expression was calculated by x-fold change = 2^−Δ(ΔC _T ⁾, where ΔC_T = C_T,target − C_T,GAPDH and Δ(ΔC_T) = ΔC_{T,HOG LDL} − ΔC_T,N-LDL. 

Data Analysis

Microarray results were analyzed using the DNA-Chip Analyzer “dChip” (http://www.dchip.org/, a collaboration of the Wong Laboratory, Stanford University, Stanford, CA, and the Cheng Li Group, Harvard School of Public Health and Dana Farber Cancer Institute, Boston MA). Hybridization intensities were normalized by using the invariant gene set expression model. Model-based expression indices were calculated based on the perfect-match-only model. For genes differentially expressed among all treatments, gene expression levels were filtered with the following criteria: (1) variation across samples between 0.35 and 10; (2) genes scored present in ≥20% arrays; (3) replicate variation between 0 and 0.4; and (4) hybridization intensity ≥30 units in ≥20% of the samples. To identify gene expression differentially induced by HOG- and N-LDL, a change in hybridization intensity of ≥30 units (12.5% of GeneChip mean hybridization values [ABI]) was applied as a threshold to eliminate genes having small proportional changes and/or genes having overall low hybridization intensities. The change (x-fold) cutoffs between 1.5 and 2.0 were tested empirically to determine how they affected the number of genes discovered and the corresponding false discovery rate (FDR), estimated by 100 iterative comparisons of randomized sample groupings (dChip User’s Manual, http://biosun1.harvard.edu/complab/dchip/manual.htm). ^{8 9} As a function of increasing change, the number of genes discovered was significantly reduced (149 genes at 1.5-fold versus 32 genes at 2-fold), whereas FDR was only moderately improved (20% at 1.5-fold versus 12% at twofold). Therefore, the intermediate cutoff of 1.7-fold was selected, identifying 60 genes with an estimated FDR of ∼17% (Tables 1 2) . Unpaired t-tests performed on these genes identified a subset of 41 significant genes (P < 0.05), for which FDR approximated 0.0%, when taking into account all differential expression criteria. 

For RT-PCR, values were expressed as the mean ± SD. Statistical significance was determined by one-way ANOVA followed by the Dunnett test. P < 0.05 was considered significant. 

Results

Overall Gene Expression Profile

Four replicate DNA microarray experiments were conducted in which quiescent human retinal pericytes were treated in four different ways—that is, for 24 hours with N-LDL, G-LDL, HOG-LDL, or SFM. Analysis of hybridization data indicated that 152 genes were differentially expressed among treatment groups. Hierarchical clustering of the gene expression profiles revealed similarities and differences among the responses to LDL preparations (Fig. 1) . First, N-LDL caused significant changes in pericyte gene expression, inducing upregulation of 58 genes and downregulation of 50 genes by ≥1.7-fold (Supplementary Table S1). Second, G-LDL and N-LDL response profiles were largely the same, suggesting that under these conditions, glycation per se did not act as a gene regulatory stimulus. Third, HOG-LDL induced responses that were either similar or distinct from N-LDL. In instances in which the HOG-LDL response was distinct (Fig. 1 , black bars) the effect was sometimes similar to the SFM state, but in other instances was different from either SFM or N-LDL (Fig. 1 , green bar). Nineteen genes were affected by HOG-LDL ≥1.7-fold in relation to SFM and N-LDL (Supplementary Table S1 and the next section). These results demonstrate the importance of N-LDL in affecting gene regulation, and further indicate that HOG-LDL can cause responses that are distinct from those of either N-LDL or SFM. 

HOG-LDL Induces Specific Changes in Gene Expression

To understand better the effects of HOG-LDL on pericytes, microarray data were analyzed to identify genes differentially regulated by HOG- versus N-LDL. After genes that had low absolute differences (<30 units) were eliminated, 60 genes were found to have expression differing by ≥1.7-fold, 41 of which were significant by unpaired t-test (P < 0.05). Among the 60 genes, 21 were upregulated by HOG-LDL, with mean changes ranging up to 3.3-fold (Table 1) . Two fatty acid desaturases, FADS1 and FADS2, which are involved in the synthesis of long-chain polyunsaturated fatty acids (PUFAs), showed the greatest upregulation. Other lipid-related genes, including the LDL receptor (LDLR) and a fatty acid binding protein (FABP4), also were considerably upregulated. Thirty-nine genes were downregulated by HOG-LDL, with changes ranging to 6.8-fold (Table 2) . Included among the 60 HOG-LDL-regulated genes were the 19 that were also affected by ≥1.7-fold compared with SFM (Tables 1 2 , asterisks), again suggesting that some responses to HOG-LDL are unique and distinct from those to N-LDL or SFM. Overall categorization of the 60 genes is shown in Figure 2 . These genes are implicated in several pathways that can be related to diabetic retinopathy. 

Validation of Microarray Results

To verify the gene expression changes revealed by microarray analysis, 12 responding genes were tested by real-time RT-PCR, confirming the microarray findings in each case. Results for four of these genes are shown in Figure 3 . Of these, FABP4 and LDLR were upregulated by HOG- versus N-LDL, whereas bradykinin receptor B2 (BDKRB2) and insulin-like growth factor binding protein 3 (IGFBP3) were downregulated. Microarray findings were further confirmed for plasminogen activator inhibitor-1 (PAI-1, increased 1.7-fold by microarray) using both RT-PCR (2.8-fold increase) and Northern blot analysis (1.8-fold increase; data not shown). Two lipid-related genes, SDR1 and SC4MOL, which showed a more than 1.7-fold change but were not significantly changed by t-test, were also confirmed by RT-PCR (data not shown). 

Effects of Coincubation of N-LDL on HOG-LDL-Induced Gene Alterations

Gene expression responses to HOG-LDL that are distinct from N-LDL could be caused by lipoprotein deprivation (i.e., if modification inactivates the LDL) or as a reaction to a unique signal conferred by oxidation and glycation of LDL. To investigate these possibilities, experiments were performed to test the effects of combined treatment with N-LDL and HOG-LDL. As shown in Figure 3 , HOG-LDL-induced changes in LDLR, BDKRB2, and IGFBP3 expression were reversed by addition of N-LDL, but the increased expression of FABP4 was not altered by adding N-LDL. 

Discussion

Selective loss of pericytes is a major feature of early diabetic retinopathy, but the underlying mechanisms are unclear. Although the causes of vascular changes in the diabetic retina are probably multifactorial, the occurrence of antipericyte autoantibodies in type 2 diabetes suggests that autoimmunity may mediate pericyte loss. ¹⁰ However, the limited incidence of autoantibodies (∼55% of type 2 diabetes with retinopathy at grades from 10 to 53) ¹⁰ indicates that other mechanisms must also contribute to this process. In previous work with cultured bovine retinal pericytes, we reported that mildly glycated and/or oxidized LDL might contribute to pericyte loss, with an effect proportional to the extent of LDL modification. ⁴ Recently, we observed that a heavily modified form of LDL, HOG-LDL, exerts greater toxicity toward human retinal pericytes, stimulating apoptosis. ⁵ In those and the present experiments, HOG-LDL was used to simulate LDL that is severely modified in vivo after prolonged extravasation. This work therefore builds on the hypothesis that severe modification of LDL occurs not only in arteries, where it promotes and accelerates atherosclerosis in diabetes, but also in the diabetic retina. Consistent with this, we found that HOG-LDL induced many changes in pericyte gene expression that were distinct from that caused by N-LDL or G-LDL. Further, many HOG-LDL responsive genes are implicated in key metabolic and signaling functions that influence the pathologic course of diabetic retinopathy. 

The strengths of this study include the use of human cells and lipoproteins, the performance of quadruplicate microarray experiments, the use of three separate LDL pools, and, to date, the confirmation by alternative methods—that is, RT-PCR and Northern blot analysis—of all responses tested. Further, the degree of in vitro modification of LDL and the concentration (100 mg/L) used reflect a conservative estimate of conditions in vivo. Although it is difficult to select a period of exposure to modified LDL that simulates conditions in vivo, 24-hour exposure was chosen, because it does not cause the significant cell loss observed with longer exposure periods and/or higher doses of HOG-LDL. Such treatment aims to induce minor degrees of injury that would be present very early in the development of retinopathy. Therefore, it is not surprising that many of the genes significantly altered by HOG-LDL do not directly contribute to apoptosis. However, it is expected that these changes may eventually lead to this process. Ultimately, there are limitations inherent in using cultured cells to simulate conditions in vivo. 

Effects of HOG-LDL Distinct from Those of N-LDL

Oxidation and glycation of LDL may sufficiently alter the lipoprotein so that it no longer acts as cognate LDL. Alternatively, modification of LDL may confer a novel characteristic to the lipoprotein so that it now acts as a unique signaling entity. The differential expression responses to HOG- versus N-LDL suggest the presence of both mechanisms. For some genes, the HOG-LDL response was unlike that to N-LDL but was similar to the SFM state, indicating that the response is equivalent to that in the absence of N-LDL. However, in other cases, the HOG-LDL response was distinct from responses to either N-LDL or SFM. The existence of both types of genetic responses was confirmed when N-LDL and HOG-LDL were combined. N-LDL eliminated the HOG-LDL effect on LDLR, BDKRB2, and IGFBP3, but not on FABP4. Therefore, the observed changes in gene regulation include at least two components: a specific response to modified LDL and a response to N-LDL deprivation. Both components may be relevant in vivo in diabetes when retinal capillary leakage is present and extravasation of plasma occurs. Pericytes, which are normally outside the blood–retinal barrier, may then be exposed to any combination of native and modified LDL. 

Genes Involved in Fatty Acid, Eicosanoids, and Inflammation

The genes most upregulated by HOG-LDL are involved in fatty acid metabolism. Among these, FADS1 and FADS2 govern rate-limiting steps in long-chain PUFA synthesis. ¹¹ FABP4 enhances fatty acid turnover, and a knockout of FABP4 inhibits atherosclerosis in animals. ¹² Enhanced long-chain PUFA synthesis could reflect compensatory replacement of PUFAs (perhaps lost by free radical oxidation), or could reflect injurious changes in the synthesis of eicosanoids, mediators of inflammation, and thrombosis. ¹³ Other evidence indicates that oxidized LDL modulates arachidonate and eicosanoid pathways. Oxidized LDL activates cytosolic phospholipase A₂, releases arachidonate, and induces apoptosis in macrophages and fibroblasts. The apoptosis is attenuated by phospholipase A₂ inhibition. ¹⁴ Oxidized LDL also stimulates phospholipase A₂ activity and prostaglandin E₂ and thromboxane B₂ production in rat mesangial cells. ¹⁵ In addition, oxidized LDL induces cyclooxygenase-2 protein expression and prostaglandin E₂ release in endothelial cells. ¹⁶ In the present study, HOG-LDL decreased expression of prostaglandin E synthase by 1.9-fold. Therefore, further delineation of the eicosanoid pathway in mediation of HOG-LDL toxicity is required, especially since chronic inflammation has been implicated in diabetic retinopathy. ^{17 18 19}  

Additional evidence that HOG-LDL induces inflammatory responses in pericytes is the upregulation of the acute-phase gene, PTX3, consistent with induction of PTX3 by oxidized LDL in human vascular smooth muscle cells ²⁰ and its strong expression in atherosclerotic lesions. ²¹ Further, TNFAIP6, which is secreted by various cells on proinflammatory cytokine stimulation and which exhibits anti-inflammatory effects, ²² was downregulated. A knockout of TNFAIP6 exacerbates inflammation in mice ²³ ; and, hence, reduced expression may indicate impaired ability of pericytes to combat local inflammation. 

Genes Involved in Cholesterol Metabolism

HOG-LDL modulated several genes involved in cholesterol metabolism. IDI1 and SC4MOL, key enzymes in cholesterol biosynthesis, ^{24 25} were both upregulated. These changes may increase cholesterol synthesis in pericytes exposed to HOG-LDL. Also, cholesterol accumulation is suggested by the upregulation of LDLR ²⁶ and the downregulation of ABCA6, a transporter mediating cholesterol efflux. ²⁷ Consistent with this, by direct measurements, modified LDL is known to increase intracellular cholesterol accumulation in human macrophages and smooth muscle cells. ^{28 29} LDL from diabetic patients compared with that of healthy subjects induces an increase in intracellular cholesteryl ester in monocyte-macrophages ³⁰ and aortic smooth muscle cells. ³¹  

Genes in the Fibrinolytic Pathway

PAI-1, a primary regulator of fibrinolysis, is increased in serum, vitreous, subretinal fluid, and retinal microvasculature of diabetic patients (Brignole F, et al. IOVS 1994;35:ARVO Abstract 2041). ³² Overexpression in mouse retinal microvessels thickens basal membranes and increases the endothelial cell/pericyte ratio, reminiscent of diabetic retinopathy. ³³ PAI-1 upregulation by HOG-LDL could partially explain the elevation of PAI-1 in diabetes and is in agreement with findings that UV-oxidized LDL enhances PAI-1 expression, both mRNA and protein, in human umbilical vein endothelial cells. ³⁴ Apart from modulating fibrinolysis and extracellular matrix proteolysis, ³⁵ PAI-1 induces apoptosis, ³⁶ although antiapoptotic effects have also been observed. ³⁷  

Genes Involved in Cell Growth and Proliferation

HOG-LDL generally upregulated pericyte growth inhibitors while downregulating growth-promoting genes. Among upregulated genes, P8 mediates growth arrest and apoptosis ^{38 39} and is over-expressed in response to apoptotic signals. ⁴⁰ DDK1 antagonizes cell growth induced by WNT proteins, ⁴¹ interference with which is implicated in apoptotic degeneration in retinitis pigmentosa. ⁴²  

Of downregulated genes, IGFs and IGFBPs play key roles in the balance between cell proliferation and apoptosis. ⁴³ Reduced expression of IGFBP3/5 after HOG-LDL treatment suggests a role for IGF signaling in pericyte loss. This agrees with findings that oxidized LDL modulates IGF1, IGF1 receptors, and IGFBPs in rat aortic vascular smooth muscle cells, contributing to apoptosis. ⁴⁴ Decreased serum IGFBP3 has been found in patients with diabetic retinopathy. ⁴⁵ IGFBP3 is thought to sensitize cells to mitogenic IGF1 stimuli, whereas IGFBP5 stimulates growth of human retinal endothelial cells. ⁴⁶  

Other downregulated survival-promoting genes that are of interest include AREG (growth factor); TGFB1 (growth factor); and ALDH1A2, RARRES1, and RAI3 (retinoic acid synthesis and response genes). 

Genes Related to Stress Responses

Several stress-responsive genes were modulated by HOG-LDL. ATF3, a stress-inducible transcriptional repressor highly expressed in cells undergoing apoptosis in atherosclerotic lesions, ^{47 48} was upregulated. It has been found that oxidized LDL induces ATF3 and increases its DNA binding, and that inhibition of ATF3 suppresses oxidized LDL-induced cell death. ⁴⁸ This implicates ATF3 as an important mediator of apoptosis. 

Thioredoxin-interacting protein (TXNIP) is an endogenous inhibitor of the redox protein thioredoxin. ⁴⁹ Overexpression of this gene increases oxidative stress, while its silencing restores thioredoxin activity in hyperglycemia. ⁵⁰ Consequently, a reduced expression of TXNIP may reflect enhanced thioredoxin antioxidative responses. This finding differs from that observed in aorta smooth muscle cells where hyperglycemia inhibits thioredoxin by inducing TXNIP. ⁵¹ TXNIP is also induced by glucose incubation in fibroblasts ⁵¹ and mesangial cells. ⁵²  

Metallothioneins (MTs) are ubiquitous metal-binding proteins involved in metal homeostasis, free radical scavenging, cell proliferation, and stress-induced apoptosis. ⁵³ Decreased retinal MTs are associated with senescence ⁵⁴ and macular degeneration, ⁵⁵ whereas increased MTs protect against oxidative stress and apoptosis. ⁵⁶ A reduction in expression of MT1B and MT2A in pericytes may reflect an impaired antioxidant defensive mechanism. 

Bradykinin Receptors

Bradykinin’s activity (vasodilation, blood flow, inflammatory responses) is mainly mediated by B₂ kinin receptors in the retina. ⁵⁷ Reduced expression of B₂ receptors in pericytes by HOG-LDL may disturb retinal hemodynamics. Recent studies have shown a likely shift of vasodilatory functions from B₂ to B₁ receptors in retinal microvessels in streptozotocin-diabetic rats, ⁵⁸ implicating the kinin system in diabetic retinopathy development. 

Genes Related to Angiogenesis

HOG-LDL altered the expression of diverse genes related to angiogenesis. Cysteine-rich angiogenic inducer 61 (CYR61) and angiomotin-like 2 (AMOTL2), both with proangiogenic activity, were upregulated. CYR61 stimulates the migration of human microvascular endothelial cells, induces neovascularization in rat cornea, and regulates genes participating in wound healing. ^{59 60} AMOTL2 is related to angiomotin and belongs to the motin family. ⁶¹ Angiomotin increases endothelial cell migration by binding to angiostatin, an inhibitor of angiogenesis. ⁶²  

In contrast, antiangiogenic genes, including STC1, TIMP3, TGFB1, and COL15A1 were downregulated by HOG-LDL. STC1 is markedly upregulated in angiogenesis models, ^{63 64} where it inhibits hepatocyte growth factor-induced angiogenesis by interfering with pathways downstream of its receptor. ⁶⁴ TIMP3 inhibits angiogenesis both in vitro ⁶⁵ and in vivo in a mouse model of retinopathy ⁶⁶ by inhibiting VEGF receptor binding and downstream signaling. ⁶⁷ TGFB1, secreted by pericytes, inhibits endothelial cell growth and migration. ^{68 69} Its downregulation may promote endothelial cell proliferation and contribute to neovascularization in diabetic retinopathy. ⁷⁰ Type XV collagen α1 (COL15A1) is the precursor of restin, which is homologous to endostatin and exhibits antiangiogenic effects. ⁷¹  

Thus, HOG-LDL appears to shift the balance in favor of angiogenesis by upregulating proangiogenic genes and downregulating antiangiogenic genes. This surprising finding goes against the notion that oxidized LDL impairs endothelial cell growth and generally inhibits angiogenesis. ^{72 73} However, studies have shown that oxidized LDL upregulates VEGF expression in endothelial cells. ⁷⁴ The pericyte–endothelial cell relationship is complex and dynamic, in that pericytes may be both negative and positive regulators of endothelial cell proliferation and neovascularization. ^{75 76 77 78 79 80} Therefore, our findings suggest that, besides causing apoptosis, modified LDL at sublethal concentrations may have proangiogenic effects through alteration of gene expression in retinal pericytes. 

²  Contributed equally to the work and therefore should be considered equivalent authors.

 Supported by a Research Grant from the American Diabetes Association (TJL) and National Heart, Lung, and Blood Institute Grant HL52813 (WSA). The MUSC DNA Microarray Facility is supported by Grant R24CA95841 from the National Cancer Institute.

 Submitted for publication December 20, 2004; revised March 30, 2005; accepted April 12, 2005.

 Disclosure: W. Song, None; J.L. Barth, None; Y. Yu, None; K. Lu, None; A. Dashti, None; Y. Huang, None; C.K. Gittinger, None; W.S. Argraves, None; T.J. Lyons, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Table 1.

 

Genes Upregulated ≥1.7-fold by HOG- Versus N-LDL in Cultured Human Retinal Pericytes

Gene	Gene Symbol	Accession Number	Change (x-fold)	P
Cholesterol, lipid, fatty acid
Fatty acid desaturase 2	FADS2	AL050118	2.9	0.003
Fatty acid desaturase 1*	FADS1	AF035284	2.8	0.048
Low-density lipoprotein receptor (familial hypercholesterolemia)	LDLR	L00352	2.7	0.002
Fatty acid-binding protein 4, adipocyte*	FABP4	AA128249	2.5	0.012
Sterol C4-methyl oxidase-like*	SC4MOL	AI535653	2.0	0.091
Isopentenyl-diphosphate delta isomerase*	IDI1	X17025	1.7	0.018
Transcription
p8 protein (candidate of metastasis 1)	P8	AI557295	1.9	0.005
Activating transcription factor 3*	ATF3	L19871	1.7	0.020
Signal transduction
Regulator of G-protein signaling 20*	RGS20	AF060877	1.9	0.028
Related RAS viral (r-ras) oncogene homolog 2	RRAS2	AI365215	1.7	0.003
Atherosclerosis
Cysteine-rich, angiogenic inducer, 61	CYR61	Y11307	2.6	0.029
Pentaxin-related gene, rapidly induced by IL-1 beta	PTX3	M31166	1.9	<0.001
Plasminogen activator inhibitor-1	PAI-1	J03764	1.7	0.008
Other
Asparagine synthetase*	ASNS	M27396	2.7	0.045
Actin, gamma 2, smooth muscle, enteric*	ACTG2	D00654	2.0	0.083
Angiomotin like 2	AMOTL2	AB023206	1.8	0.017
Poliovirus receptor-related 3	PVRL3	AL050071	1.8	0.022
Dickkopf homolog 1 (Xenopus laevis)	DKK1	AB020315	1.8	0.118
Deafness, autosomal dominant 5	DFNA5	AF073308	1.7	0.016
DnaJ (Hsp40) homolog, subfamily B, member 4	DNAJB4	U40992	1.7	0.044
Unknown
Hypothetical protein LOC339005*	LOC339005	AL046961	1.8	0.180

 Genes denoted by asterisks were also altered by ≥1.7-fold by HOG- LDL versus SFM and may therefore be regulated by modified LDL in a distinct or unique manner. Only genes exhibiting a mean change in hybridization intensity of ≥30 units between treatments are included in the table. P denotes results of unpaired t-tests. See text for estimated false-discovery rates.

Table 2.

 

Genes Downregulated ≥1.7-fold by HOG- Versus N-LDL in Cultured Human Retinal Pericytes

Gene	Gene Symbol	Accession Number	Change (x-fold)	P
Signal transduction
Retinoic acid receptor responder 1	RARRES1	AI887421	−6.8	0.013
Bradykinin receptor B2	BDKRB2	X86163	−3.0	<0.001
Insulin-like growth factor binding protein 3	IGFBP3	M35878	−2.9	0.001
Chemokine (C-X-C motif) ligand 12	CXCL12	L36033	−2.2	0.001
Insulin-like growth factor binding protein 5*	IGFBP5	L27560	−2.0	0.020
Tumor necrosis factor, alpha-induced protein 6*	TNFAIP6	M31165	−2.0	0.058
Stanniocalcin 1	STC1	U25997	−2.0	0.012
Prostaglandin E synthase	PTGES	AF010316	−1.9	<0.001
Protein kinase, cAMP-dependent, regulatory, type II, beta	PRKAR2B	M31158	−1.8	0.003
Transcription
Inhibitor of DNA binding 1*	ID1	X77956	−4.2	0.040
Basic transcription element binding protein 1	BTEB1	D31716	−1.8	0.124
Nuclear transcription factor Y, beta	NFYB	AA621555	−1.8	0.153
Thioredoxin interacting protein	TXNIP	S73591	−1.8	0.036
Jun B proto-oncogene	JUNB	X51345	−1.7	0.010
Stress response
Metallothionein 1F (functional)*	MT1F	H68340	−1.7	0.044
Metallothionein 1B (functional)	MT1B	M13485	−1.7	0.043
Cell cycle
Retinoic acid induced 3	RAI3	AF095448	−2.0	0.073
Transforming growth factor, beta-induced, 68kDa	TGFB1	M77349	−1.9	0.005
CDC28 protein kinase regulatory subunit 2	CKS2	X54942	−1.7	0.181
Kinetochore associated 2	KNTC2	AF017790	−1.7	0.198
Extracellular matrix
Tissue inhibitor of metalloproteinase 3*	TIMP3	U14394	−2.6	0.004
Collagen, type XV, alpha 1	COL15A1	L25286	−2.1	0.006
Enzyme
Proprotein convertase subtilisin/kexin type 1*	PCSK1	X64810	−3.3	0.001
Short-chain dehydrogenase/reductase 1	SDR1	AF061741	−2.3	0.134
Sulfatase 1	SULF1	AB029000	−2.2	0.002
Aldehyde dehydrogenase 1 family, member A2	ALDH1A2	AB015228	−2.1	0.013
Flavin containing monooxygenase 3*	FMO3	M83772	−2.0	0.215
Transglutaminase 2	TGM2	M55153	−1.9	0.022
Sulfotransferase family, cytosolic, 1B, member 1	SULT1B1	D89479	−1.7	0.001
Transporter
ATP-binding cassette, sub-family A (ABC1), member 6*	ABCA6	AI651024	−2.2	0.002
Aquaporin 9	AQP9	AB008775	−1.9	0.009
Other
Topoisomerase (DNA) II alpha 170kDa*	TOP2A	L47276	−1.8	0.050
Secretogranin II (chromogranin C)	SCG2	M25756	−2.5	0.092
Amphiregulin (schwannoma-derived growth factor)	AREG	M30704	−2.4	0.019
Statherin	STATH	M18078	−2.0	0.157
Vascular cell adhesion molecule 1*	VCAM1	M30257	−1.9	0.111
ZW10 interactor	ZWINT	AF067656	−1.7	0.266
Unknown
PN001_AH_H08.r yodnorm mRNA sequence	—	AI547258	−1.9	0.108
KIAA0101	KIAA0101	D14657	−1.7	0.229

 See Table 1for key.

Figure 1.

 

Hierarchical clustering of genes differentially regulated in response to experimental LDL treatments. With the dChip analysis tool, hierarchical clustering was performed on the 152 genes differentially regulated between the control (SFM) and experimental LDL samples. Expression levels among replicate treatments were averaged before standardization and subsequent distance calculations. Red: relative high expression; blue: relative low expression; white: intermediate expression (see scale bar at the bottom of the cluster). Black vertical bars: sets of genes for which the response to N-LDL and G-LDL was similar but for which the response to HOG-LDL was distinct; green vertical bar: a group of genes that appeared to be specifically affected by HOG-LDL. Genes presented in the heat diagram are listed in Supplementary Table S1.

Figure 2.

 

General categorization of the 60 genes with expression that was significantly altered by HOG- versus N-LDL (100 mg/L, 24 hours) in cultured human retinal pericytes. Genes were identified as differentially regulated when average expression levels changed by ≥1.7-fold compared with N-LDL treatment. Categorization was based on information gathered from a literature search and online databases (i.e., OMIM; Online Mendelian Inheritance in Man; http://www.ncbi.nlm.nih.gov/Omim/ provided in the public domain by the National Center for Biotechnology Information, Bethesda, MD; and The Nucleotide Database, a collaboration among the tripartite, international collaboration of sequence databases: GenBank, the National Institutes of Health [NIH], Bethesda, MD; European Molecular Biology Laboratory [EMBL], Hinxton Hall, UK; and DNA Database of Japan [DDBJ], Mishima, Japan).

Figure 3.

 

Verification of gene expression in response to HOG-LDL by real-time RT-PCR and evaluation of the effect of N-LDL on HOG-LDL-elicited alterations of gene expression in cultured human retinal pericytes. Data represent the mean ± SD of results in triplicate experiments. *P < 0.05; **P < 0.01, versus N-LDL. Consistent with gene array findings, HOG-LDL (100 mg/L, 24 hours) significantly upregulated FABP4 and LDLR, and downregulated BDKRB2 and IGFBP3. Co-incubation with N-LDL (100 mg/L, 24 hours) reversed the HOG-LDL-induced changes in expression of LDLR, BDKRB2, and IGFBP3, but not of FABP4.

Supplementary Materials

Supplementary Table S1 - 217 KB 

The authors thank the MUSC DNA Microarray Facility, a shared-resources facility, for services provided during the study.