A Key Role for ROCK in TNF-α–Mediated Diabetic Microvascular Damage

Abstract

Purpose.: Leukocyte adhesion releases tumor necrosis factor (TNF)–α that contributes to endothelial damage in early diabetic retinopathy (DR). Rho/Rho-kinase (ROCK) signaling mediates retinal endothelial damage in early DR. However, whether ROCK regulates TNF-α–mediated diabetic vascular damage is unknown. Here, the contribution of ROCK to TNF-α–mediated microvascular damage is investigated.

Methods.: In DR patients and nondiabetic control subjects, the levels of membranous (m) TNF-α on neutrophils, soluble (s) TNF-α and its receptors in sera, were measured. In cultured microvascular endothelial cells, phosphorylation of myosin phosphatase target protein (MYPT)-1, a downstream target of ROCK, was investigated with TNF-α or DR sera pretreatment. TNF-α–induced intercellular adhesion molecule-1 (ICAM-1) and endothelial nitric oxide synthase (eNOS) phosphorylation were measured with and without ROCK inhibition by fasudil or ROCK-specific small-interfering RNA (siRNA). In isolated neutrophils from control subjects, MYPT-1 phosphorylation was investigated in the presence of TNF-α. The impact of ROCK inhibition by fasudil on TNF-α–induced integrin (CD18, CD11a, CD11b) and intracellular cytoskeletal changes were investigated.

Results.: The serum levels of mTNF-α, sTNF-α, and its receptors were significantly elevated in DR patients. TNF-α as well as DR sera promoted MYPT-1 phosphorylation in endothelial cells, which was significantly reduced by anti–TNF-α neutralizing antibody. TNF-α–induced ICAM-1 expression, eNOS dephosphorylation, cytoskeletal changes, and CD11b/18 expression in neutrophils were significantly suppressed by fasudil as well as ROCK-specific siRNA.

Conclusions.: ROCK is a key mediator of TNF-α signaling in diabetic microvessels. The important role of TNF-α in early DR provides a new rationale for ROCK inhibition beyond the previously shown mechanisms.

Introduction

Vascular complications are major causes of morbidity and mortality in diabetic patients. ¹ In particular, diabetic retinopathy (DR) is a prevalent microvascular complication of both type 1 and type 2 diabetes. ^2,3 Early DR is characterized by microvascular injury that might remain subclinical for an extended period in diabetic patients. In the later DR stages, hemorrhages from newly formed vessels or tractional retinal detachments cause severe vision loss. Therefore, understanding the molecular changes underlying early DR is critical, to intervene at a stage when visual loss can be prevented. 

Various findings suggest a key role for chronic inflammation in the pathogenesis of early DR. For instance, endothelial damage secondary to increased leukocyte adhesion ^4,5 and upregulation of adhesion molecules, such as intercellular adhesion molecule-1 (ICAM-1) and leukocyte integrins, are characteristic of early DR. ^6,7  

The proinflammatory cytokine, tumor necrosis factor (TNF)–α, plays a pivotal role in various inflammatory conditions. TNF-α has pleiotropic effects such as cell proliferation, differentiation, migration, and apoptosis through activation of diverse signaling cascades. ^8,9 TNF-α is mainly produced by monocytes, macrophages, and neutrophils. It has two isoforms, the 26-kDa membrane-bound form of TNF (mTNF)-α and the 17-kDa soluble form (sTNF)-α. ⁸ There are two known receptors for TNF-α, TNFR1 and TNFR2. The activation of TNFRs initiates the proinflammatory pathway, leading to activation of nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB). sTNF-α binds to TNFR-1 mainly, whereas mTNF-α activates TNFR-2 by cell-to-cell contact. ^10,11 Upon ligation of the TNFRs on target cells by TNF-α, disintegrin and metalloproteases cleave the TNFR release of the soluble fragment, which is an indicator of TNF-α activity. ¹² The soluble TNF-α competes with the signaling through the receptors, limiting excessive TNF-α signaling. ¹³  

High glucose, advanced glycation end products, and serum from patients with type 2 diabetes stimulate monocytes/macrophages and their release of sTNF-α. Serum concentrations of sTNF-α is increased in type 1/type 2 diabetic animals and diabetic patients with microangiopathy. ^14,15 Patients with proliferative DR show significantly higher serum TNF-α compared with that of nonproliferative DR patients. ¹⁵ Moreover, serum and vitreous levels of soluble TNFR (sTNFR) are elevated in DR patients. ¹⁶ TNF-α is elevated approximately 3-fold (both mRNA and protein), and the receptors for TNF-α were increased by 40% in poor glycemic control retina compared with normal rat retina. ¹⁶ TNF-α inhibition reduces leukocyte adhesion in the retina ¹⁷ and the loss of retinal microvascular cells in type 1 and type 2 diabetic rats. ¹⁸ These findings indicate a role for TNF-α in DR pathogenesis and as a potential therapeutic target. However, the details of TNF-α signaling in DR have not been understood. 

The Rho/Rho-kinase (ROCK) pathway plays an important role in vascular diseases. ¹⁹ TNF-α activates the ROCK pathway, which in turn induces stress fiber formation, mediated through actin polymerization in human umbilical vein endothelial cells. ²⁰ Our recent study showed the role of ROCK activation in leukocyte-induced diabetic retinal endothelial injury in early DR. ²¹ However, the role of ROCK in TNF-α–mediated microvascular damage has not been investigated. 

Materials and Methods

Materials

Materials were obtained from commercial suppliers, as follows: anti-CD18 mouse monoclonal antibody (mAb), TNF-α goat polyclonal antibody (pAb; sc-1350), and anti-TNFR1 mouse mAb (Santa Cruz Biotechnology, Santa Cruz, CA); anti-TNFR2 mouse mAb (HyCult Biotechnology, Uden, The Netherlands); polyclonal rabbit antihuman von Willebrand factor (vWF; Dako A/S, Glostrup, Denmark); and Alexa Fluor 488–conjugated antimouse pAb, Alexa Fluor 555-conjugated antimouse and antirabbit pAb, Alexa Fluor 647–conjugated antigoat pAb, and DAPI (4′,6-diamidino-2-phenylindole) (Molecular Probes, Eugene, OR). 

For flow cytometry: antihuman mAbs CD11a, CD11b, and CD18 (BD PharMingen, San Diego, CA); antihuman mTNF-α (eBioscience, San Diego, CA). For immunohistochemistry: antihuman mAbs for CD11a, CD11b, and CD18 (BD PharMingen), and anti-mTNF-α (R&D Systems, Minneapolis, MN). Recombinant human TNF-α (rhTNF-α) and mouse IgG1 (Sigma, Tokyo, Japan). Rho activator lysophosphatidic acid (LPA) (Calbiochem, San Diego, CA); fasudil, a selective ROCK inhibitor (LC Laboratories, Woburn, MA); antihuman TNF-α neutralizing Ab (mouse IgG1) (R&D Systems). The following validated ROCK I/II-specific small-interfering RNA (siRNA) molecules were used: Accell human ROCK1 (6093) siRNA, Accell human ROCK2 (9475) siRNA, and validated Accell nontargeting negative control siRNA (Dharmacon, Inc., Lafayette, CO). 

Preparation of Serum Samples and Enzyme-Linked Immunosorbent Assay (ELISA)

This study was approved by the Institutional Review Board and performed in accordance with the ethical standards of the 1989 Declaration of Helsinki. Written informed consents were obtained from all participants; 116 healthy subjects without diabetes (67 females and 49 males; mean age 64.9 ± 9.2 years) were in the control group, and 110 type 2 diabetic patients with retinopathy (49 females and 61 males; mean age 57.8 ± 13.4 years) who had undergone vitrectomy were in the DR group (Table 1). HbA1c was significantly higher in DR patients (7.7 ± 0.1%) than that in control participants (4.9 ± 0.3%), and the duration of diabetes was 11.97 ± 7.7 year in DR patients (Table 2). 

Table 1. 

 

Baseline Characteristics of All Patients

Characteristic	Control, n = 116	Diabetic Retinopathy, n = 110
Sex
Male	49 (42.2%)	61 (55.5%)
Female	67 (57.8%)	49 (44.5%)
Age, y
Mean ± SD	64.9 ± 9.2	57.8 ± 13.4
Range	35 to 83	30 to 80

 Control: healthy subjects without diabetes; diabetic retinopathy: diabetic subjects with diabetic retinopathy (moderate or severe nonproliferative diabetic retinopathy, proliferative diabetic retinopathy); value depicted as mean ± SD.

Table 2. 

 

Baseline Characteristics of All Patients with Diabetic Retinopathy

Characteristic	Diabetic Retinopathy
HbA1c, %
Mean ± SD	7.7 ± 1.5
Range	5.8 ± 11.3
Duration of diabetes, y
Mean ± SD	11.97 ± 7.7
State of internal clinical diabetic retinopathy
Moderate NPDR	24 (21.8%)
Severe NPDR	32 (29.0%)
PDR	54 (49.0%)

 HbA1c, glycosylated hemoglobin; value depicted as mean ± SD; NPDR, nonproliferative diabetic retinopathy; PDR, proliferative diabetic retinopathy.

Serum concentrations of sTNF-α, sTNFR1, and sTNFR2 were assessed using Human TNF/TNFSF1A, sTNFR1, and sTNFR2 immunoassay ELISA kits (R&D Systems). To normalize these protein levels, total protein concentrations were measured using the bicinchoninic acid kit (Bio-Rad, Hercules, CA). 

Animal Procedure

All experimental procedures on the animals were performed according to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Wistar rats (male, 4 weeks) were obtained from a commercial supplier (Kyudo, Fukuoka, Japan). Each rat received single 65 mg/kg intraperitoneal injections of streptozotocin (STZ; Wako, Tokyo, Japan). Rats with blood glucose levels > 400 mg/dL, 24 hours after STZ injection, were considered diabetic. Animals were euthanized 1 month after diabetes induction. 

Immunofluorescence Microscopy

Deparaffinized sections of diabetic rat retina were incubated with antibodies against TNF-α, TNFR1, and TNFR2. The tissues were from animals that were diabetic for 1 month. After washing, sections were incubated with Alexa Fluor 488–conjugated secondary antibodies (Molecular Probes). vWF and Alexa Fluor 555–conjugated secondary antibody (Molecular Probes) were used for staining endothelial cells. Moreover, sections were incubated with antibodies against CD18, TNF-α, and TNFR2, and then incubated with Alexa Fluor 488/555/647–conjugated secondary antibodies (Molecular Probes). Counterstaining of nuclei was performed with DAPI (Molecular Probes). Staining was observed under a fluorescence microscope (BZ-9000; Keyence Corp., Osaka, Japan). 

Neutrophil Isolation and Flow Cytometry

Neutrophils were isolated from whole blood as previously described. ²⁰ This study was approved by the institutional ethics committees, and the harvested specimens were handled in accordance with the Declaration of Helsinki. The cells were incubated with CD11a mAb (1:50), CD11b mAb (1:50), CD18 mAb (1:50), and mTNF-α mAb (1:25), which were labeled with fluorescein isothiocyanate or phycoerythrin (PE). After washing, surface fluorescence of 10⁵ cells was analyzed by fluorescence activated cell sorting (FACScan; Becton Dickinson, San Jose, CA). Results are expressed as percentage of positive cells or mean fluorescence intensity on a logarithmic scale. In vitro neutrophils were incubated for 1 hour with rhTNF-α (10 ng/mL) or LPA (20 μg/mL), or pretreated with fasudil (20 μM) for 30 minutes before stimulation with rhTNF-α. 

Cell Culture

Human dermal microvascular endothelial cells (HMVECs; Cambrex, East Rutherford, NJ) were cultured as described previously. ²² The cells were starved for 16 hours in Dulbecco's modified Eagle's medium containing 3% calf serum before experiments. 

Silencing of ROCK Expression by siRNAs

Accell siRNA is specially modified for use without a transfection reagent. Cells were trypsinized and incubated at 37°C with 5% CO₂ overnight. The final concentration was 1 μM Accell siRNA (Dharmacon, Inc.) per well in a six-well plate with Accell delivery media (catalog no. B-005000). The growth media was removed from the cells and Accell siRNA was added to each well. Cells were incubated at 37°C with 5% CO₂ for 72 hours. The expression of ROCK I and ROCK II specifically silenced by siRNAs in HMVEC was proven by Western blotting with mouse monoclonal Ab against ROCK1 (1:1000, sc-17794; Santa Cruz Biotechnology) and goat polyclonal Ab (pAb) against ROCK2 (1:1000, sc-1851; Santa Cruz Biotechnology). 

ROCK Activity

HMVECs, neutrophils, and retinal lysates were prepared for Western blotting. The blots were incubated with rabbit phospho-myosin phosphatase target protein (MYPT)-1 (1:1000; Thr 853; Cell Signaling Technology, Danvers, MA) and reblotted with rabbit anti-MYPT-1 (1:2000; Cell Signaling Technology) as previously described. ²¹ The membranes were also reblotted with rabbit anti-MYPT-1 (1:2000; Cell Signaling Technology). Lane loading differences were normalized by GAPDH. ROCK activation was expressed as percentage ratios of phospho–MYPT-1/GAPDH. 

In vitro, HMVECs were starved and then treated with 10 ng/mL of rhTNF-α for indicated times. To examine the potential of ROCK activation by DR serum, HMVECs were incubated with 50% serum from either controls or DR patients. To elucidate the effect of TNF-α in ROCK activation induced by DR serum, serum was pretreated with antihuman TNF-α neutralizing Ab or mouse IgG1 for 1 hour, and then exposed to HMVECs for 3 hours. Furthermore, neutrophils isolated from controls were also resuspended at 10⁶ cells/mL and then treated with 10 ng/mL of rhTNF-α for the indicated times. 

Detection of ICAM-1 Protein and eNOS Activation

To investigate the role of TNF-α for ROCK activation in endothelial cells, HMVECs were incubated for 16 hours with rhTNF-α (10 ng/mL) or Rho activator LPA (20 μg/mL), or pretreated with fasudil for 30 minutes before stimulation with rhTNF-α. Western blotting was performed as described earlier with goat polyclonal Ab (pAb) against ICAM-1 (1:1000, sc-1511; Santa Cruz Biotechnology). 

After stimulation for 1 hour, Western blotting was performed with rabbit pAb against phospho-eNOS (1:1000; Ser1177; Cell Signaling Technology). The membranes were reblotted with rabbit anti-eNOS Ab (1:1000; Cell Signaling Technology). Lane loading differences were normalized by GAPDH. eNOS activation was expressed as percentage ratios of phospho-eNOS (Ser1177)/GAPDH. Retinal lysates were prepared as described earlier and analyzed by Western blotting. 

Confocal Laser Scanning Microscope

Neutrophils were fixed and permeabilized with fixative solution and permeabilization buffer as described in F-actin visualization biochem kit (Cytoskeleton, Denver, CO). Cells were incubated with Abs against CD11b (sc-28664; Santa Cruz Biotechnology) and rhodamine-phallloidin, and subsequently Alexa Fluor 488–conjugated secondary Ab (Molecular Probes) was used. Fluorescence images were observed on a confocal laser scanning microscope (Zeiss LSM 510 META; Carl Zeiss, Oberkochen, Germany). 3D images were analyzed by LSM 510 software (Carl Zeiss). 

Statistical Analysis

All results were expressed as mean ± SEM. All statistical analysis was performed with commercial statistical software (JMP version 9; SAS Institute Inc., Cary, NC). Before statistical analysis, normal distribution was tested. Statistical differences were assessed using the nonparametric Kruskal–Wallis ANOVA when distributions were not normal. To adjust for inflated α errors due to multiple comparisons, the corrected significant P value was defined using the Bonferroni correction. Statistical differences between two groups were analyzed by the Mann–Whitney U test. Relationships between variables were determined by multiple linear regression analysis. 

Results

Increased TNF-α and TNF-α Receptors in DR Patients

DR patients were classified according to the International Classification of Diabetic Retinopathy (Table 2). ²³ Briefly, the categories included moderate nonproliferative DR (NPDR) (n = 24), patients with severe NPDR (n = 32), and patients with proliferative diabetic retinopathy (PDR) (n = 54). All patients with moderate and severe nonproliferative DR had severe diabetic macular edema in International Clinical Diabetic Macular Edema Disease Severity Scale. ²³  

To investigate the levels of TNF-α and TNF-α receptors in diabetic patients, we measured these molecules in sera of DR patients and nondiabetic controls. In DR patients, serum sTNF-α (42.2 ± 4.1 pg/mL) was significantly higher compared with that in nondiabetic controls (28.1 ± 1.1 pg/mL). The levels of serum sTNF-α with PDR (55.0 ± 7.2 pg/mL) were significantly higher than those with moderate NPDR (22.0 ± 2.3 pg/mL) or severe NPDR (25.8 ± 3.2 pg/mL) (Fig. 1A). Particularly, in cases with significant angiogenesis and advanced ischemia over the four quadrants very high levels of sTNF-α (>70 pg/mL) were measured. 

Figure 1

 

Concentration of soluble TNF-α and TNFR in sera and expression of membrane form TNF-α on neutrophils. Concentrations of soluble TNF-α (sTNF-α) (A), TNFR1 (sTNFR1) (B), and TNFR2 (sTNFR2) (C) in sera from nondiabetic control subjects (control) and those with DR were measured by ELISA (*P < 0.05, **P < 0.01 versus control; control, n = 116; DR, n = 110). (D) Cell surface expressions of membrane form TNF-α on neutrophils from isotype (thin line), control (dotted line), and DR (thick line) subjects were analyzed by flow cytometry. (E) The percentage of membrane TNF-α–positive cells (*P < 0.05 compared with control; n = 20 each).

The levels of serum sTNFR1 (2051.4 ± 140.0 ng/mL) and sTNFR2 (3637.1 ± 246.5 ng/mL) in DR patients were also significantly higher than the corresponding values in the nondiabetic controls (1349.7 ± 44.6 ng/mL and 2594.3 ± 75.4 ng/mL, respectively). The levels of serum sTNFR1 with severe NPDR (2024.7 ± 292.8 pg/mL) and PDR (2253.3 ± 217.0 pg/mL) were significantly higher than those of nondiabetic controls. However, the levels of serum sTNFR1 with moderate NPDR (1698.7 ± 159.7 pg/mL) did not significantly differ from those of control (Fig. 1B). The levels of serum sTNFR2 with PDR (4035.3 ± 371.4 pg/mL) were significantly higher than those of nondiabetic controls, moderate NPDR (2940.0 ± 249.3 pg/mL), and severe NPDR (3561.6 ± 542.2 pg/mL) (Fig. 1C). The increase in sTNFR1 levels occurred earlier in DR than that in sTNFR2 levels. The levels of sTNFR1 showed a strong positive correlation with sTNFR2 (r ² = 0.80, P < 0.001) (Fig. 1D). sTNF-α did not show a correlation with HbA1c, sTNFR1, or sTNFR2. 

Furthermore, to check the level of mTNF-α in DR, we harvested neutrophils from DR patients and normal controls, and measured the level of mTNF-α in neutrophils. We examined neutrophils from DR patients that had a significant amount of angiogenic vessels in the four quadrants, suggested to be a high level of serum sTNF-α. The percentage of mTNF-α positive DR neutrophils (25.8 ± 4.4%: thick line) was significantly higher than that in the nondiabetic controls (12.3 ± 4.5%: dotted line, *P < 0.05 compared with control; n = 20 each, Fig. 1E, thin line: mouse isotype control). These results establish significantly higher sTNF-α, sTNFR1, and sTNFR2 in serum and mTNF-α in neutrophils of DR patients compared with that in nondiabetic controls. 

Expression and Localization of TNFRs with TNF-α in Diabetic Retinal Vessels

To investigate the expression and localization of TNFRs in diabetes, we performed immunohistochemistry in retinas of STZ-treated rats. Both TNFR1 and TNFR2 were mainly expressed in vWF-positive retinal endothelial cells (Figs. 2A, 2B). TNF-α was mainly expressed in CD18(+) leukocytes (Fig. 2C). The CD18 (+) leukocytes were found in the vicinity of TNFR2-expressing vessels. These data suggest that the mTNF-α on the CD18(+) leukocytes colocalized and possibly bound to TNFR2(+) diabetic retinal endothelium. 

Figure 2

 

Localization of soluble and membrane-bound TNF-α in retinal vessels. Paraffin-embedded sections of diabetic rat retinas were immunohistochemically analyzed with TNFR1 or TNFR2 antibody, and TNF-α antibody. Endothelial cells were stained with vWF. (A) TNFR1 (green) was mainly localized in retinal vessels (red). (B) Immunohistochemically analyzed with TNFR2 antibody (green) and vWF (red). TNFR2 was mainly localized in retinal vessels. (C) Immunohistochemically analyzed with TNF-α antibody (red). Adhering leukocytes were stained with CD18 antibodies (green). In diabetic retinas, CD18-positive leukocytes (green) with TNF-α expression (red) were bound with TNFR2-expressing retinal vessels (blue). Yellow color (white arrowhead) indicates double-stained area (CD18 and TNF-α).

Involvement of TNF-α in Endothelial ROCK Activation during Diabetes

To examine whether TNF-α regulates ROCK activation in vascular endothelium, we measured MYPT-1 phosphorylation in TNF-α–treated endothelium. MYPT-1 phosphorylation in HMVECs significantly increased by 30 minutes after TNF-α stimulation and remained high for at least 360 minutes (Figs. 3A, 3B). Moreover, MYPT-1 phosphorylation was significantly promoted by 100 pg/mL TNF-α stimulation (Figs. 3C, 3D). To investigate whether serum from DR patients affects endothelial ROCK signaling, we treated HMVECs with serum from advanced DR patients that have high levels of sTNF-α or nondiabetic control serum. MYPT-1 phosphorylation after 3-hour incubation with control sera did not differ from untreated cells (1.2-fold), whereas stimulation with DR sera upregulated MYPT-1 phosphorylation by 2.06-fold compared with untreated cells (n = 9 each, Figs. 3E, 3F). Pretreatment with anti–TNF-α neutralizing Ab significantly reduced DR sera-induced MYPT-1 phosphorylation (52.6% reduction), whereas control IgG1 did not affect MYPT-1 phosphorylation (Figs. 3I, 3J). In contrast, pretreatment with anti–TNF-α neutralizing Ab (84.4 ± 8.7% versus control, n = 4 each) or control IgG1 (116.7 ± 4.4% versus control, n = 4 each) did not affect the baseline MYPT-1 phosphorylation from the treatment with nondiabetic control sera (Figs. 3G, 3H). These data indicate that soluble TNF-α in the DR serum is a key regulator for ROCK activation in DR. 

Figure 3

 

Role of TNF-α in endothelial ROCK activation during diabetes. Phosphorylated (Thr853) and total MYPT-1 in HMVECs were detected by Western blot analysis. (A) HMVECs were stimulated with recombinant TNF-α (10 ng/mL) for the indicated time periods (*P < 0.05, **P < 0.01 compared with time 0; n = 4 each). (C) HMVECs were stimulated with recombinant TNF-α (3 hours) for the indicated concentrations (*P < 0.05, **P < 0.01 compared without TNF-α; n = 4 each). (E) HMVECs were stimulated with and without 50% sera from nondiabetic control subjects (control) and those with DR for 3 hours (**P < 0.01, N.S., not significant; n = 9 each). (G, I) After pretreatment of anti–TNF-α neutralizing Ab (0 or 10 μg/mL) or mouse IgG1 Ab (10 μg/mL) for 1 hour, HMVECs were stimulated with 50% sera from control (G) or DR (I) for 3 hours (**P < 0.01, N.S., not significant; n = 9 each). Lane loading differences were normalized by reblotting the membranes with an Ab against GAPDH. ROCK activity was expressed as the ratio of phospho-MYPT-1 to GAPDH. Average signal intensities quantified and expressed as percentage ratio of time 0 (B), without stimulation (D), without sera stimulation (F), without any pretreatment (H, J).

ROCK Involvement in TNF-α–Mediated Microvascular Endothelial Damage

To investigate the role of ROCK signaling in TNF-α–induced vascular endothelial injury, we used fasudil to inhibit ROCK activation and measured ICAM1-1. TNF-α–induced ICAM-1 expression was significantly reduced by fasudil starting at 10 μM (Figs. 4A, 4B). The Rho activator LPA as well as TNF-α significantly increased ICAM-1 expression in HMVECs (LPA, 1.79-fold and TNF-α, 3.05-fold) compared with control (Figs. 4C, 4D), suggesting that ROCK regulates ICAM-1 expression. To further elucidate the role of ROCK signaling in TNF-α–induced microvascular damage, we checked whether ROCK inhibition by fasudil affects TNF-α–induced ICAM-1 expression and eNOS phosphorylation (Ser1177) at a known activation site. ²⁴ Pretreatment with fasudil (20 μM) significantly reduced TNF-α–induced ICAM-1 expression by 49.4% (Figs. 4C, 4D). Both LPA and TNF-α treatment significantly reduced eNOS phosphorylation compared with control by 27.5% and 19.3%, respectively. Pretreatment with fasudil almost completely reversed TNF-α–induced eNOS dephosphorylation (Figs. 4E, 4F). 

Figure 4

 

Impact of TNF-α on microvascular endothelial cells mediated through ROCK pathway. (A) HMVECs were pretreated with fasudil (30 minutes) for the indicated concentrations before stimulation with recombinant TNF-α (10 ng/mL, 16 hours) (**P < 0.01 compared without pretreatment; n = 4 each). (C) After pretreatment with or without 20 μM fasudil for 1 hour, HMVECs were stimulated with LPA (20 μg/mL) or TNF-α (10 ng/mL) for 16 hours. ICAM-1 was detected by Western blot analysis (**P < 0.01, N.S., not significant; n = 11 each). (E) HMVECs were treated for 1 hour with or without fasudil in the same manner as in (C). Phosphorylated (Ser1177) and total eNOS in HMVECs were detected by Western blot analysis (*P < 0.05, **P < 0.01, N.S., not significant; n = 7 each). Lane loading differences were normalized by reblotting the membranes with an Ab against GAPDH. Average signal intensities quantified and expressed as percentage of the ratio of control without any stimulation (B, D, F).

Since fasudil inhibits both ROCK isoforms as well as PRK2, MSK1, and MAPKAP-K1b, ²³ we further used a ROCK-specific siRNA to inhibit ROCK signaling. We confirmed that the siRNA reduced both ROCK1 and ROCK2 expression, as previously reported ²⁴ (Fig. 5A). The ROCK-specific siRNA significantly inhibited TNF-α–induced ICAM-1 expression (Figs. 5B, 5C). These data confirmed that ROCK inhibition effectively reduces TNF-α–induced endothelial ICAM-1 expression. 

Figure 5

 

Impact of ROCK inhibition using siRNA on TNF-α–activated endothelial cells. (A) Silencing of ROCK expression by siRNAs. ROCK1 and 2 in HMVEC without or with control siRNA or ROCK1/2-specific siRNA were detected by Western blot analysis. (B, C) After pretreatment with 1 μM Accell siRNA (nontargeting negative control siRNA, ROCK1/2 siRNA) or 20 μM fasudil for 1 hour, HMVECs were stimulated with TNF-α (10 ng/mL) for 16 hours. ICAM-1 was detected by Western blot analysis (*P < 0.05, N.S., not significant; n = 4 each). Lane loading differences were normalized by reblotting the membranes with an Ab against GAPDH. Average signal intensities quantified and expressed as percentage of the ratio of control without any stimulation (C).

Impact of TNF-α–Induced ROCK Activation on Integrin Expression and Cytoskeletal Rearrangement in Neutrophils

To examine whether ROCK is involved in TNF-α–induced activation of neutrophils, we harvested neutrophils from patients and treated them with TNF-α. TNF-α significantly increased MYPT-1 phosphorylation in neutrophils by 30 minutes after stimulation. The phosphorylation reached its maximum by 180 minutes and remained elevated at least 360 minutes after stimulation (Figs. 6A, 6B). To investigate the influence of ROCK activation for cytoskeleton and integrin expression, we stimulated neutrophils by the Rho activators, LPA and TNF-α, with or without fasudil, and analyzed the cells with flow cytometry and confocal microscopy. LPA as well as TNF-α remarkably increased integrin CD18 (LPA, 1.35-fold and TNF-α, 1.36-fold) and CD11b (LPA, 3.26-fold and TNF-α, 3.93-fold) surface expression on neutrophils compared with control. Pretreatment with fasudil significantly reduced the TNF-α–induced increase in CD18 and CD11b surface expression on neutrophils by 119% (CD18) and by 62.3% (CD11b). However, CD11a expression in neutrophils was not affected by these treatments (Figs. 6C–H). 

Figure 6

 

Impact of TNF-α on neutrophils mediated through ROCK pathway. Phosphorylated (Thr853) and total MYPT-1 in neutrophils were detected by Western blot analysis. (A) Neutrophils were stimulated with recombinant TNF-α (10 ng/mL) for the indicated periods of time (*P < 0.05, **P < 0.01 compared with time 0; n = 7 each). (B) Lane loading differences were normalized by reblotting the membranes with an Ab against GAPDH. Average signal intensities quantified and expressed as percentage ratio of time 0. (C–H) After pretreatment with or without 20 μM fasudil for 1 hour, neutrophils were stimulated with LPA (20 μg/mL) or TNF-α (10 ng/mL) for 1 hour. Cell surface expressions of CD18 (C), CD11a (E), and CD11b (G) on neutrophils were analyzed by flow cytometry (thin line, LPA stimulation; thick line, TNF-α stimulation without fasudil; dotted line, TNF-α stimulation with fasudil). MIF of CD18 (D), CD11a (F), and CD11b (H) were measured and expressed as percentage of the control without any stimulation (*P < 0.05, N.S., not significant; n = 5 each).

Confocal microscopy showed that the CD11b integrin was stored within the cells in control neutrophils (Fig. 7A). In comparison, LPA as well as TNF-α induced F-actin polymerization and CD11b translocation to the cell surface (Figs. 7B, 7C), and fasudil suppressed both TNF-α–induced F-actin polymerization and integrin translocation (Fig. 7D). The quantitation of the CD11b localization showed that LPA as well as TNF-α remarkably decreased the ratio of internal expression area/surface expression area of CD11b (LPA, 44.4% reduction and TNF-α, 71.5% reduction). Pretreatment with fasudil almost completely recovered the TNF-α–induced reduction in CD11b expression on neutrophils (*P < 0.05, N.S., not significant, Fig. 7E). The CD18 subunit was also mobilized to the cell surface, which was also suppressed by fasudil (data not shown). In contrast, CD11a was constitutively expressed on normal cells, and its expression could not be significantly affected by LPA or TNF-α (see Supplementary Material and Supplementary Figs. S1A–E). These data indicate that ROCK and TNF-α regulate CD11b and CD18 but not CD11a expression in neutrophils. 

Figure 7

 

Influence of TNF-α for cytoskeleton and integrin localization in neutrophils. After pretreatment with or without 20 μM fasudil for 1 hour, neutrophils were stimulated with LPA (20 μg/mL) or TNF-α (10 ng/mL) for 1 hour. Cells were permeabilized and immunohistochemically analyzed with F-actin (red). CD11b was detected by green fluorescence. (A) CD11b was stored in neutrophils (white arrowhead), and F-actin was nearly unpolymerized. (B, C) After stimulation with LPA or TNF-α, F-actin was significantly polymerized, and CD11b was translocated to the cell surface, coinciding with the location of F-actin polymerization (white arrowhead). (D) Fasudil suppressed TNF-α–induced F-actin polymerization and CD11b translocation to cell surface (white arrowhead). (E) Percentage ratio of internal expression/surface expression in CD11b (*P < 0.05, N.S., not significant; n = 4 each).

Discussion

To meet the needs for the alarmingly rising prevalence of diabetic complications, in-depth understanding of the pathophysiology is essential. Recently we introduced the involvement of the Rho/ROCK pathway in retinal microvascular damage of early DR. ²¹ Several studies showed an important role for TNF-α in diabetic retinal microvascular damage. ^17,18 The present study shows the crucial role of ROCK pathway in TNF-α–mediated diabetic retinal microvascular damage. 

We evaluated the soluble and the membrane-bound TNF-α and TNF-α receptors in patients with diabetic retinopathy. The concentrations of soluble forms of both TNF-α and TNFRs in DR sera were elevated and higher than the previously reported levels in diabetic patients without microangiopathy. ¹⁵ We further found that serum TNF-α is significantly higher in PDR patients compared with moderate or severe NPDR patients. ²⁵ Particularly, sTNF-α (>70 pg/mL) was very high in patients with major angiogenesis and advanced ischemia in all quadrants, which underlines a role for TNF-α in DR pathogenesis, in line with prior experimental reports. ^17,18  

The level of mTNF-α on neutrophils in DR patients was significantly higher than that of nondiabetic controls. In vivo, the number of adhering leukocytes was increased in diabetic retinal vessels. ²¹ Adherent leukocytes affect the surface molecules of the endothelial cells, because both TNFR1 and TNFR2 are expressed in diabetic retinal vessels. mTNF-α on CD18 positive leukocytes bound to TNFR2 in diabetic retinal vessels. Most of the mTNF-α activities are specific to the TNFR1, including ICAM-1 expression, which is exclusively mediated through TNFR1 signaling. It is suggested that the extracellular part of TNFR2 captures TNF-α and delivers it to the TNFR1, resulting in an enhanced response to TNF-α. ²⁶ Another point of distinction is that TNFR-1 signaling activates the classical NF-κB pathway, whereas the mTNF-α–mediated TNFR2 signals through an alternative NF-κB pathway. ^10,11 Overall, TNFR1 and TNFR2 function synergistically, which might also be the case in DR. The increase in sTNFR1 starts earlier in DR than in sTNFR2, which might be through an early effect of sTNF-α on retinal endothelial cells. Then, with the increase in the number of adhering leukocytes in diabetic retinal vessels in later stages of DR, the mTNF-α on the adherent leukocyte facilitates TNF-α–mediated signaling in diabetic retinal endothelial cells. This study introduces the key role of TNFR activation in retinal endothelium through mTNF-α from adherent leukocytes in DR. 

To check the involvement of ROCK activation in TNF-α–induced microvascular damage, we investigated whether ROCK inhibition by fasudil as well as siRNA affects TNF-α–induced ICAM-1 expression or eNOS phosphorylation of an activation site (Ser1177). Our results indicated that TNF-α rapidly activated the ROCK signaling pathway in neutrophils, and increased integrin CD18 and CD11b surface expression on neutrophils, likely through release of cytoplasmic storage vesicles, which were significantly reversed by fasudil. In vivo, CD18 and CD11b but not CD11a mediate leukocyte adhesion to diabetic retinal vasculature. ²⁷ Rho/ROCK signaling is associated with ICAM-1 clustering, ²⁸ which jointly form the anchoring structures for leukocyte integrins. ^29,30 Activation of the Rho/ROCK pathway causes firm adhesion of leukocytes to the microvasculature by promoting the higher affinity state of integrins ³¹ and their assembly ³² mediated through phosphorylation of myosin regulatory light chain. The Rho/ROCK pathway is involved in cytoskeleton reorganization in neutrophils. ³³ The cytoskeletal actin filaments and microtubules regulated Mac-1 surface expression on neutrophils. Our data support the fact that TNF-α–induced rapid externalization of CD18/11b is mediated through cytoskeletal reorganization, actin polymerization, and stress fiber formation, all of which were reduced by fasudil. 

Our investigations with fasudil provide novel molecular insights into mechanisms of TNF-α signaling in leukocyte-mediated microvascular injury. However, fasudil does not exclusively inhibit ROCK. ³⁴ Among others, fasudil also inhibits PKC and MLCK, although likely not at the concentrations used in this study. ³⁵ Thus, the results obtained with fasudil in this study are largely consistent with the role of ROCK. Furthermore, our data with ROCK-specific siRNA confirmed the data from the fasudil studies. 

Our results show that TNF-α inhibition significantly reverses diabetes-induced reduction and inactivation of eNOS in retinal endothelium associated with the ROCK pathway. There is an apparent discrepancy between the beneficial effects of eNOS downregulation in the previous reports ^17,18 and the beneficial effects of eNOS upregulation and activation in the present work, stimulating further research and clarification. At first, eNOS expression is transiently increased with hyperglycemia, and in the subsequent 2 weeks it gradually decreases. ³⁶ eNOS phosphorylation is also reduced in diabetic mouse aortas. ³⁷ The vasculoprotective effects of eNOS-generated nitric oxide (NO) have been established. ^38,39 Physiologic levels of eNOS-generated NO prevent apoptosis and prevent caspase-3 activation. ⁴⁰ Our findings showing that the beneficial role of eNOS in diabetic retinopathy is in line with and supported by existing reports. ^38,39  

Supplementary Materials

pdf S1 

Footnotes

 Disclosure: R. Arita, None; S. Nakao, None; T. Kita, None; S. Kawahara, None; R. Asato, None; S. Yoshida, None; H. Enaida, None; A. Hafezi-Moghadam, None; T. Ishibashi, None