TNFα Inhibits IGFBP-3 through Activation of p38α and Casein Kinase 2 in Human Retinal Endothelial Cells

We recently reported a reciprocal relationship between tumor necrosis factor alpha (TNFα) and insulin-like receptor growth factor binding protein 3 (IGFBP-3) in whole retina of normal and IGFBP-3 knockout mice. A similar relationship was also observed in cultured retinal endothelial cells (REC). We found that TNFα significantly reduced IGFBP-3 levels and vice-versa, IGFBP-3 can lower TNFα and TNFα receptor expression. Since IGFBP-3 is protective to the diabetic retina and TNFα is causative in the development of diabetic retinopathy, we wanted to better understand the cellular mechanisms by which TNFα can reduce IGFBP-3 levels. For these studies, primary human retinal endothelial cells (REC) were used since these cells undergo TNFα-mediated apoptosis under conditions of high glucose conditions and contribute to diabetic retinopathy. We first cultured REC in normal or high glucose, treated with exogenous TNFα, then measured changes in potential signaling pathways, with a focus on P38 mitogen-activated protein kinase alpha (P38α) and casein kinase 2 (CK2) as these pathways have been linked to both TNFα and IGFBP-3. We found that TNFα significantly increased phosphorylation of P38α and CK2. Furthermore, specific inhibitors of P38α or CK2 blocked TNFα inhibition of IGFBP-3 expression, demonstrating that TNFα reduces IGFBP-3 through activation of P38α and CK2. Since TNFα and IGFBP-3 are key mediators of retinal damage and protection respectively in diabetic retinopathy, increased understanding of the relationship between these two proteins will offer new therapeutic options for treatment.


Introduction
Diabetic retinopathy is the leading cause of blindness, with numbers expected to reach epidemic levels by 2030. While a large number of factors are at play in this complicated disease, one that appears to have a major role is the presence of high levels of inflammation/apoptosis mediators such as TNFa [1]. In support of this hypothesis, TNFa knockout mice failed to develop diabetic retinopathy even when treated with a high galactose diet known to trigger manifestations of the disease [2]. Furthermore, TNFa has been suggested to be involved in inducing insulin resistance in adipocytes [3,4], and retinal endothelial cells (REC) [5] through activation of insulin receptor substrate 1 (IRS-1) Ser307 and stimulation of apoptosis. Our work has suggested that additional TNFa pathways may also be involved in apoptosis related to retinopathy. For example, we have recently reported that TNFa can reduce insulin-like growth factor receptor binding protein 3 (IGFBP-3) levels in the retina, leading to increased apoptosis [6]. In the studies reported here, we further examine the link between TNFa and IGFBP-3 cascades and the ways in which they interact to create conditions of retinopathy.
The primary physiological function of IGFBP-3 is to deliver insulin-like growth factor 1 (IGF-1) to cells when it is required [7,8]. However, more recent studies show that IGFBP-3 has additional actions independent of IGF-1 [9,10]. Consistent with this notion, our in vitro findings in retinal endothelial cells indicate that high glucose reduces IGFBP-3 levels, which was correlated with increased TNFa levels [11]. We found similar results in vivo. IGFBP-3 KO mice had significantly increased retinal levels of TNFa [6]. Importantly, we have shown that intravitreal injections of IGFBP-3 into diabetic rat eye led to a significant reduction in TNFa [12]. In all cases, increases in IGFBP-3 are linked to decreases in TNFa; likewise, increases in TNFa are linked to decreases in IGFBP-3. The pathway by which TNFa inhibits IGFBP-3 is unknown and represents the focus of the current study. We know from previous studies that IGFBP-3 levels in retinal endothelial cells can be increased by PKA through activation of DNA-PK-induced phosphorylation of IGFBP-3 on serine 156 [13], and thus lead to reduced apoptosis. In the present study, we analyzed retinal endothelial cells grown in normal or high glucose to determine if TNFa inhibits IGFBP-3 by regulating IGFBP-3 phosphorylation. Our results show that a TNFa-p38a-CK2 pathway is activated under conditions of high glucose and results in a phosphorylation-mediated inhibition of IGFBP-3.  . TNFa increases phosphorylation of p38a to regulate IGFBP-3 levels. Panel A presents ELISA results for IGFBP-3 following treatment with a non-selective p38 inhibitor (SB202190) and TNFa. Panel B shows Western Blot results of the ratio of phospho-p38a to total p38a with a dose-response curve for BIRB796, a p38a selective inhibitor. Panel C shows Western blot results for the ratio of phosphorylation of p38a to total p38 from REC grown in normal glucose (NG), high glucose (HG), high glucose + TNFa (HG+TNFa), high glucose + p38a inhibitor (HG+BIRB796), and high glucose + TNFa + p38a inhibitor (HG+TNFa+BIRB796). Panel D presents ELISA results for IGFBP-3 in cells treated with the same treatments. *P,0.05 vs. NG, #P,0.05 vs. HG, $P,0.05 vs. HG+TNFa. All data are mean 6 SEM. Data is from 3-4 independent experiments for each treatment. doi:10.1371/journal.pone.0103578.g001

Cell Culture
Primary human REC were acquired from Cell System Corporation (CSC, Kirkland, Washington). Cells were grown in M131 medium containing microvascular growth supplements (Invitrogen) (MVGS), 10 mg/ml gentamycin and 0.25 mg/ml amphotericin B. For high glucose conditions, cells were transferred to high glucose (25 mM) (Cell Systems) medium, supplemented with MVGS and antibiotics for 3 days. Only primary cells within passages 6 were used. Cells were quiesced by incubating in high or normal glucose medium without growth factors for 24 hours and used to perform the experiments unless otherwise indicated.

Mutagenesis
The IGFBP-3 NB plasmid DNA was a gift from Maria Grant, MD. Both the phosphorylation Serine 111 site and the phosphorylation Serine 113 site were mutated to Alanine to prevent their phosphorylation using the forward primer: 59-CCAG-GAAATGCTCGTGAGTCGGAGG-39 and the reverse primer 59-CCTCCGACTCACGAGCATTTCCTGG-39 according to the QuikChange II XL Site-Directed Mutagenesis Kit manufacturer. The PCR sample reaction contained 5 ml of 10x reaction buffer, 10 ng dsDNA template, 125 ng of forward primer and 125 ng of reverse primer, 1 ml of dNTP mix, 3 ml QuikSolution, Figure 2. TNFa regulates CK2 levels through p38a. Western blot results for the ratio of phosphorylation of CK2 to total CK2 from REC grown in normal glucose (NG), high glucose (HG), high glucose + TNFa (HG+TNFa), high glucose + p38a inhibitor (HG+BIRB796), and high glucose + TNFa + p38a inhibitor (HG+TNFa+BIRB796). *P,0.05 vs. NG, #P,0.05 vs. HG, $P,0.05 vs. HG+TNFa. All data are mean 6 SEM. Data is from 3-4 independent experiments for each treatment. doi:10.1371/journal.pone.0103578.g002 1 ml of pfuUltra HF DNS polymerase and water to a total volume of 50 ml. PCR conditions used were initial heating at 95uC for 60 seconds to denature cDNA and activate the Taq DNA polymerase, followed by 18 cycles consisting of denaturation at 95uC for 50 seconds, annealing at 60uC for 50 seconds, and extension at 68uC for 300 seconds. After that, the amplification product was digested by DpnI and transformed by XL10-Gold ultracompetent cells. Colonies were selected, and the purified DNA was verified by DNA sequencing (University of Tennessee Health Science Center Molecular Science Core).
Transfection of siRNA and plasmid DNA ON-TARGETplus SMARTpool, human LRP-1 siRNA were purchased from Dharmacon, Inc. We used 4 sets of siRNA, with target sequences of GCUACAAAGUUGACUACGA, GAAAUGCUAGUGAGUCGGA, GCACAGAUACCCAGAA-CUU and GAAUAUGGUCCCUGCCGUA. siCONTROL Non-targeting siRNA #1 (Dharmacon) was used as a nonspecific control. REC were transfected with siRNA at a final concentration of 20 nM using RNAiMAX transfection reagent according to the manufacturer's instructions. Briefly, 60 pmol of RNAi duplex was diluted in 150 ml of OPTI-MEM, and separately 9 ml of Lipofectamine RNAi/MAX was diluted in 150 ml of OPTI-MEM, each in an eppendorf tube. These two solutions were mixed and incubated for 10 min at room temperature. The transfection mixture was applied to REC cells in 60 mm dishes and incubated for 24 h at 37uC in a 5% CO 2 incubator. The cells were used for experiments 24 hours after transfection. The cells were also transfected with IGFBP-3 NB plasmid DNA or its mutant at 1 mg/ml using SuperFect transfect reagent, according to the manufacturer's instructions. Five mg of the plasmid was diluted with M131 medium to a total 150 ml. The transfection mixture was added to the REC in complete growth medium in a 60 mm dish. After 4 hours of incubation, REC were washed once with PBS, followed by fresh cell growth medium. The cells were used for experiment 24 hours after transfection.

ELISA Analysis
An ELISA for IGFBP-3 levels was performed using an IGFBP-3 ELISA assay kit according to the manufacturer's instructions to evaluate the IGFBP-3 levels following the treatments. Equal amount of protein was loaded onto the IGFBP-3 coated microplate. After washing away unbounded substances, an enzyme-linked polyclonal antibody specific for IGFBP-3 was added to the wells. A substrate solution was added to the wells and color developed in proportion to the amount of IGFBP-3 bound in the initial step. The intensity of the color was measured at 450 nm (vs. 540 nm reference).
Western Blot Analysis-After appropriate treatments and rinsing with cold phosphate-buffered saline, REC were scraped into lysis buffer containing protease and phosphatase inhibitors. Equal amounts of protein from the cell extracts were separated on the pre-cast tris-glycine gel (Invitrogen, Carlsbad, CA) and blotted onto a nitrocellulose membrane. After blocking in TBST (10 mM Tris-HCl buffer, pH 8.0, 150 mM NaCl, 0.1% Tween 20) containing 5% (w/v) BSA, the membrane was treated with phospho-p38a, p38, phospho-CK2 and CK2 antibodies followed by incubation with HRP conjugated secondary antibodies. The antigen-antibody complexes were detected using chemilluminescence reagent kit (Thermo Scientific).

Statistics
All the experiments were repeated a minimum of three times, and the data are presented as mean 6 SEM. Data was analyzed by Kruskal-Wallis test, following by Dunn's testing with p values , 0.05 were considered statistically significant. In the case of Western blotting, one representative blot is shown. The control was normalized to 1, and compared treatment to control based on fold change.

Results
TNFa significantly increased phosphorylation of P38a, leading to a decrease in IGFBP-3 We have previously published that TNFa can significantly reduce IGFBP-3 levels in REC [6]. Others have reported that TNFa can regulate IGFBP-3 through activation of p38 in mammary epithelial cells [14]. To determine if p38 was involved in the TNFa regulation of IGFBP-3, we first grew REC in normal glucose (NG), high glucose (HG), or treated cells in HG with TNFa or a p38a inhibitor (SB202190 (non-selective) or BIRB796 (p38a selective)). We initiated this work using inhibition of all p38 isoforms using SB202190 and found the non-selective p38a and p38b inhibitor did not alter the role of TNFa in IGFBP-3 inhibition ( Figure 1A). Since p38a is more commonly associated with inflammation and apoptosis, we then chose to use the p38a selective inhibitor, BIRB796. It is a highly selective inhibitor of p38a, with a Kd of 100 pM. BIRB796 can inhibit JNK2 but at 330-fold less than p38a. Because TNFa selectively activated p38a ( Figure 1B), we next demonstrated that inhibition of p38a with BIRB796+TNFa treatment increased IGFBP-3 compared HG+ TNFa only ( Figure 1C & D). To verify we were using the best dose of BIRB796 in REC, we performed a dose-response curve for BIRB796 actions in inhibiting TNFa induced phospho-p38a activation and found BIRB796 had a role of p38a inhibition starting at 5 mM. To insure complete inhibition we used 10 mM BIRB796 for all subsequent experiments.

TNFa-dependent phosphorylation of P38a stimulated casein kinase 2 (CK2) levels
It has been shown that p38 can regulate CK2 levels in HeLa cells [15] and that CK2 can regulate IGFBP-3 in fibroblasts [16]. We wanted to determine if a similar p38 -CK2 pathway exists in REC and if so, whether it was regulated by TNFa. Figure 2 demonstrates that TNFa significantly increased CK2 phosphorylation, which was inhibited by pre-treatment with the selective P38a inhibitor, BIRB796. Taken with Figure 1 data, the data suggests that TNFa increased phosphorylation of both P38a and CK2 and that p38a regulated CK2 activity.

CK2 inhibitor blocked TNFa inhibition of IGFBP-3 expression
Because others have shown that CK2 regulates IGFBP-3 a transformed fibroblast cell line [16], we wanted to examine this pathway in REC. Figure 3 shows that 10 mM TBB inhibited phospho-CK2 significantly. TNFa significantly reduced IGFBP-3 levels, and this effect was blocked when the CK2 inhibitor (TBB) was combined with TNFa treatment (Figure 3). While TBB can potentially affect other pathways, it is a highly selective cellpermeable inhibitor of CK2 that functions through ATP/GTPcompetitive actions. These data show that CK2 acts downstream from TNFa and is required for the subsequent inhibition of IGFBP-3.

TNFa activation of P38a and CK2 led to inhibition of IGFBP-3 production through phosphorylation of IGFBP-3 Ser111/Ser113
After demonstrating that TNFa increased phosphorylation of P38a and CK2, and ultimately caused a decrease in IGFBP-3 levels, our next step was to determine if specific phosphorylation sites on IGFBP-3 were targeted by CK2. We wanted to ascertain whether TNFa regulated IGFBP-3 levels through phosphorylation of Serine 111 and Serine 113 on IGFBP-3, as these are key sites for CK2 phosphorylation [17]. Unfortunately, no phospho-specific antibodies exist for these two sites. To answer our question without antibodies, we first performed site directed mutagenesis for Serine 111 and Serine 113 on IGFBP-3 NB plasmid to create IGFBP-3 NB Ser-Ala . We then wanted to verify that IGFBP-3 NB Ser-Ala plasmid was able to increase total IGFBP-3 levels even in the presence of TNFa (Fig. 4), which include a control to verify successful reduction of IGFBP-3 with IGFBP-3 siRNA. As expected, levels of endogenous IGFBP-3 were decreased in response to high glucose and were further decreased when TNFa was added. In contrast, cells transfected with IGFBP-3 NB Ser-Ala plasmid and treated with HG and TNFa had significantly higher levels of total IGFBP-3, which would be expected to include any remaining endogenous IGFBP-3 as well as mutated IGFBP-3 NB Ser-Ala . In order to distinguish between endogenous IGFBP-3 and mutated IGFBP-3 NB Ser-Ala , we pretreated cells with IGFBP-3 siRNA to block transcription of endogenous IGFBP-3. Under these conditions, levels of IGFBP-3 (presumably only IGFBP-3 NB Ser-Ala ) remained significantly elevated even in the presence of TNFa, suggesting that the absence of active serine sites reduced the inhibitory effects of TNFa. Some may expect that transfection with the IGFBP-3 mutant would have reduced IGFBP-3 levels compared to wildtype IGFBP-3. However, we found that transfection with either the mutant or wildtype plasmid increased total IGFBP-3 levels. Others have reported this for other proteins [18]. The difference between mutant and wildtype IGFBP-3 levels would like be noted in the phospho-IGFBP-3 levels rather than total IGFBP-3 (measured here). Unfortunately, antibodies are not available to investigate this. Therefore, we then compared the HG+TNFa+BP3M with the HG+TNFa+BP3M+BP3siRNA, which should limit endogenous IGFBP-3 (wildtype IGFBP-3). When you compare these 2 bars, there is a significant difference. Therefore, we feel that these 2 data points suggest that TNFa does regulate IGFBP-3 through Serine111/113. Based on these findings, we conclude that TNFa acts on Serine 111 and Serine 113 phosphorylation sites to decrease IGFBP-3. A more direct test of this hypothesis will require production of site-specific antibodies.

Reciprocal actions between TNFa and IGFBP-3
The search for key proteins involved in the retinal damage associated with diabetic retinopathy is an ongoing quest. In previous work using b-adrenergic receptor knockout mice [19,20] or b-adrenergic receptor agonists [11,21] under diabetic conditions, we found that TNFa is associated with both decreased badrenergic receptor signaling and diabetic-like changes. Importantly, TNFa knockout mice do not develop diabetes even when given a high galactose diet known to cause diabetes in control mice [1,2]. While much is known about TNFa receptor downstream signaling in the pro-apoptosis pathway [22], we question whether TNFa may have broader effects on other pathways that indirectly or directly influence apoptosis. We have previously reported that another protein, IGFBP-3, is reduced in response to diabetes, in correlation with the observed increase in TNFa [11]. IGFBP-3 is of particular interest because it has a significant protective effect in retina. We wanted to determine if increased levels of TNFa could directly inhibit IGFBP-3 expression and thereby eliminate the protective actions of IGFBP-3 in the retina.
Based on earlier studies by Coverley et al. [17], we examined whether TNFa could change IGFBP-3 levels by triggering the CK2-mediated serine phosphorylation of IGFBP-3. While TNFa can regulate CK2 itself, we found that TNFa preferentially phosphorylates p38a, leading to activation of CK2. The use of the p38a-specific inhibitor demonstrated that p38a is upstream of CK2. When p38a is inhibited, TNFa can no longer reduce IGFBP-3. These findings suggest that the pathway employed by TNFa to inhibit IGFBP-3 is TNFaRp38aRCK2-| IGFBP-3 ( Figure 5).

Possible role of TNFa and IGFBP-3 in insulin resistance
TNFa and IGFBP-3 may both be involved in insulin resistance, as we previously demonstrated that TNFa is key to activation of an insulin resistance phenotype in REC, noted by increased IRS-1 Ser307 and IR Tyr960 , with increased apoptosis of REC [5]. Additional studies demonstrate that treatment with IGFBP-3 restores normal insulin signal transduction in diabetic rats when administered by intravitreal injection [12]. We have also demonstrated that IGFBP-3 can regulate monocyte adhesion through altered TNFa actions [23]. Taken together, these findings of the interaction of TNFa and IGFBP-3 suggest that TNFa regulation of IGFBP-3 will alter insulin signaling in the diabetic retina. This will be a focus of future studies.
TNFa is a key player in diabetic retinopathy, as well as other ocular pathologies. In this study, we demonstrate the cellular signaling by which TNFa can regulate IGFBP-3. Since diabetes leads to increased TNFa and decreased IGFBP-3, increased understanding of the interrelationship of TNFa and IGFBP-3 will likely lead to refined therapies that can promote IGFBP-3 actions, while inhibiting TNFa activities. Since b-adrenergic receptor agonists appear to promote both of these pathways to protect the diabetic retina, reduction of TNFa to promote IGFBP-3 levels in the retina through b-adrenergic receptor stimulation may prove to be a new therapeutic option.