**3. Results**

#### *3.1. HDAC6 Expression and Activity are Increased in the Diabetic Retina*

HDAC6 expression and activity were measured in human DR using postmortem human retinas from diabetic and non-diabetic donors and STZ-rats compared to normoglycemic age-matched control. As shown in Figure 1A,B, Western blotting analysis showed a 2.5-fold increase in HDAC6 expression in retinas of postmortem diabetic donors as compared to retinas of non-diabetic donors (*p* < 0.003; *n* = 8). We then measured the expression and retinal tissue distribution of HDAC6 in STZ-rats (8 weeks of hyperglycemia) compared to normoglycemic age-matched control rats. Western blotting analysis (Figure 1C,D) showed a 2.2-fold increase of HDAC6 protein levels in retinas of STZ-rats at 8 weeks of hyperglycemia in comparison to age-matched normoglycemic control rats (*p* < 0.006; *n* = 6). Further, HDAC6 enzymatic activity, measured with a fluorimetric assay, was significantly increased in retinas of STZ-rats when compared to normoglycemic age-matched control rats (*p* < 0.001; *n* = 6) (Figure 1E). Finally, immunohistochemical analysis of normal and diabetic rat retinal sections (Figure 1F), confirmed HDAC6 increased expression in diabetic rat retinas and showed its immunolocalization in several retinal layers, particularly in the inner nuclear layer (INL), retinal pigmented epithelium (RPE), and around retinal blood vessels in the ganglion cell layer (GCL) (white arrows in Figure 1F).

#### *3.2. Tubastatin A Decreases the Expression and Activity of HDAC6 in the Diabetic Retina*

Next, we determined the effect of the HDAC6 specific inhibitor Tubastatin A (TS), on diabetes-induced increase in HDAC6 expression and activity in the retina of diabetic rats. STZ-rats were treated with 10 mg/kg of TS, administered intraperitoneally every other day starting two weeks after the onset of diabetes and prolonged for another 6 weeks (total 8 weeks of diabetes). As shown in Figure 2A, TS treatment resulted in a marked reduction of HDAC6-specific immunoreactivity in comparison to untreated STZ-rats (DB). Western blotting analysis confirmed these data by showing a significant reduction in HDAC6 protein levels in retinas of TS-treated STZ-rats (DB + TS) in comparison with untreated STZ-diabetic rats (DB) (*p* < 0.05; *n* = 6) (Figure 2B,C). As expected, we also observed

a significant decrease in HDAC6 enzymatic activity (Figure 2D) in retinas of TS-treated STZ-rats (DB + TS) in comparison to untreated STZ-rats (DB) (*p* < 0.01; *n* = 6).

**Figure 1.** Histone deacetylase 6 (HDAC6) expression in the diabetic retina. (**A**) Western blotting analysis measuring HDAC6 protein levels in human postmortem retinas from diabetic and non-diabetic donors (control). (**B**) Bar histograms representing relative optical densities from the immunoblotting shown in (A) and normalized versus the loading control actin. Values are expressed as mean ± SEM for *n* = 8. \* *p* < 0.01 vs. control. (**C**) Western analysis of HDAC6 protein expression in retinas of streptozotocin-induced diabetic rats (STZ-rats) (DB) at 8 weeks of hyperglycemia and age-matched normoglycemic control rats (control). (**D**) Bar histograms representing densitometric quantification of HDAC6 protein levels normalized to actin. (**E**) HDAC6 activity measured, by a fluorimetric assay, in retinas of STZ-rats and control normoglycemic rats. (**F**) Representative microimages of immunohistochemical analysis of HDAC6 (green) in retinas of STZ-rats at 8 weeks of hyperglycemia and of age-matched normoglycemic control rats. Nuclei were stained with 4-,6-diamidino-2-phenylindole (DAPI). White arrows indicate areas of increased immunoreactivity. Scale bar, 50 μm. Values are expressed as mean ± SEM for *n* = 6. \* *p* < 0.01 vs. control.

**Figure 2.** Effects of Tubastatin A on HDAC6 expression and activity. (**A**) Representative images of immunohistochemical analysis of HDAC6 (green) of retinal cryosections of STZ-rats (DB) (8 weeks of hyperglycemia), age-matched normoglycemic control rats, and STZ-rats receiving TS 10 mg/kg (DB + tubastatin A (TS)). Nuclei were stained with DAPI. Scale bar, 50 μm. (**B**) Western analysis assessing HDAC6 protein levels in STZ-rats after 8 weeks of hyperglycemia (DB), STZ-rats treated with 10 mg/kg TS (DB +TS) and age-matched normoglycemic rats (control). (**C**) Bar histograms representing densitometric values of HDAC6 protein expression measured in the different experimental groups and normalized to actin. (**D**) HDAC6 activity measured, by fluorimetric assay, in the three experimental groups (control, DB and DB + TS). Values are mean ± SEM for *n* = 6. \* *p* < 0.05 vs. control and # *p* < 0.05 vs. DB.

#### *3.3. Tubastatin A Preserves Retinal Structural Morphology and Reduces Vascular Leakage in Diabetic Retina*

Morphological and morphometric analyses were conducted evaluating retinal cryosections stained with hematoxylin and eosin to assess the effects of TS treatment on retinal histopathology (Figure 3A,B). Figure 3B shows that total retinal thickness was significantly reduced in diabetic rats after 8 weeks of hyperglycemia (DB) in comparison to control age-matched normoglycemic rats (control) (*p* < 0.03). Treatment of diabetic rats with TS (DB + TS) normalized the morphology of the retinal layers as shown by a significant preservation of total retinal thickness (Figure 3B) (*p* < 0.05) compared to untreated diabetic rats (DB).

**Figure 3.** Effects of Tubastatin A on retinal histopathology and vascular leakage. (**A**) Hematoxylin and eosin (H&E) staining of retinal cryosections assessing retinal morphology of STZ-rats (DB), STZ-rats receiving TS (10 mg/kg) (DB + TS) and normoglycemic control rats (control). Scale bar, 50 μm. (**B**) The bars represent retinal thickness values measured in H&E retinal cryosections obtained from the different treatment groups. (**C**) Western analysis of albumin extravasation in retinas of diabetic STZ-rats (DB), STZ-rats receiving 10 mg/kg TS (DB + TS) and normoglycemic control rats (control). (**D**) Bar histograms representing optical density of albumin normalized to actin. Values are mean ± SEM for *n* = 6. \* *p* < 0.05 vs. control and # *p* < 0.05 vs. DB.

Blood–retinal barrier (BRB) dysfunction, measured as an increase in vascular permeability, is an important evidence of diabetes-induced retinal vascular abnormalities [19,20]. To determine the effect of TS on hyperglycemia-induced vascular leakage in the diabetic retina, we assessed albumin extravasation after perfusion in retinas of control, DB and DB + TS rats by Western blotting. As shown in Figure 3C,D, extravascular albumin levels were significantly higher in retinas of STZ rats (DB) when compared to normoglycemic age-matched rats (control), whereas treatments with TS (DB + TS) significantly reduced albumin leakage in diabetic rats (*p* < 0.02 vs. control and *p* < 0.05 vs. DB; *n* = 6).

#### *3.4. Tubastatin A Decreases the Levels of Senescence Markers in the Diabetic Retina*

We previously showed that diabetes promotes retinal vascular senescence and this e ffect is associated with loss of the NAD + -dependent histone deacetylase sirtuin 1 (SIRT1) and up-regulation of senescence markers [4,7]. We, therefore, determined whether inhibition of HDAC6 by TS a ffected this mechanism in the diabetic retina.

Expression of SIRT1 was analyzed by Western blotting in retinas of rats from the di fferent experimental groups (control, DB and DB + TS rats). As shown in Figure 4A,B, we found that SIRT1 expression was significantly decreased in the diabetic group (DB) compared to normoglycemic control and treatments of the diabetic rats with TS partially rescued it (*p* < 0.05 vs. control and *p* < 0.01 vs. DB; *n* = 6). Furthermore, immunohistochemical analysis of the senescence marker the phosphorylated form of H2A histone family member X ( γH2AX) showed (Figure 4C) increased immunoreactivity in the diabetic rat retinas as compared to control group, particularly in the INL and in GCL (white arrows, Figure 4C). However, in TS-treated diabetic retinas, γH2AX-specific immunofluorescence was markedly decreased compared to STZ-rat retinas (Figure 4C).

**Figure 4.** Effects of Tubastatin A on senescence in diabetic retina. ( **A**) Immunoblot showing protein expression levels of SIRT1 measured in retinal extracts of STZ-rats (DB), STZ-rats receiving TS (10 mg/kg) (DB + TS) and age-matched normoglycemic control rats. (**B**) Bar histograms representing densitometric quantification of SIRT1 immunoblotting normalized to the loading control actin. ( **C**) Representative images of immunohistochemical analysis of γH2AX (red) in retinas of STZ-rats (DB), STZ-rats treated with TS (DB + TS) and age-matched normoglycemic control rats (control). Nuclei were stained with DAPI. Scale bar, 50 μm. Values are mean ± SEM for *n* = 6. \* *p* < 0.05 vs. control and # *p* < 0.01 vs. DB.

#### *3.5. Tubastatin A Decreases Hyperglycemia-Induced Oxidative*/*Nitrative Stress in Retina*

Increased oxidative/nitrative stress has been shown to be a key pathogenic hub for the development of DR [4,21]. To understand the potential role of HDAC6 in this process, we investigated TS effects on hyperglycemia-induced oxidative/nitrative stress by measuring retinal levels of superoxide, by dihydroethidium (DHE) staining and nitrotyrosine (NT) and 4-hydroxynonenal (4-HNE) by dot-blot analysis. Retinal cryosections probed with DHE fluorescent staining, showed increased fluorescence intensity in the diabetic rat retinas (DB) compared to normoglycemic group (control) (Figure 5A). This effect was markedly reduced by treatment of the diabetic rats with TS (Figure 5A). Quantification of fluorescence intensity confirmed our staining data (*p* < 0.01 vs. control and *p* < 0.01 vs. DB; *n* = 6) (Figure 5B). Accordingly, dot blot analysis of retinal levels of NT and 4-HNE showed that TS treatment prevented the increase of both these markers that we observed in diabetic rats (*p* < 0.05 vs. control and *p* < 0.05 vs. DB for NT and *p* < 0.01 vs. control and *p* < 0.01 vs. DB for 4-HNE; *n* = 6) (Figure 5C–E).

**Figure 5.** Effects of Tubastatin A on retinal redox homeostasis. (**A**) Representative images of retinal cryosections from the different experimental groups (control, DB and DB + TS) probed with dihydroethidium (DHE) to detect superoxide. Scale bar, 50 μm. (**B**) Quantification of relative fluorescence intensity of DHE staining. Values are mean ± SEM for *n* = 6. Results are presented as a fold change of control. \* *p* < 0.01 vs. control and # *p* < 0.01 vs. DB. (**C**) Dot blot analysis assessing levels of nitrotyrosine (NT) and 4-hydroxynonenal (4-HNE) in retinas of three experimental groups (control, DB and DB + TS rats). (**D**,**E**) Quantification of optical density of NT and 4-HNE immunoblotting normalized versus actin. Values are mean ± SEM for *n* = 6. \* *p* < 0.05 vs. control and # *p* < 0.05 vs. DB for NT. \* *p* < 0.01 vs. control and # *p* < 0.01 vs. DB for 4-HNE. (**F**–**I**) mRNA levels of heme oxygenase-1 (HO-1), NAD(*p*)H dehydrogenase quinone 1 (NQO1), glutamate-cysteine ligase regulatory subunit (GCLM) and glutamate-cysteine ligase (GCLC) evaluated by qPCR and normalized to mRNA for hypoxanthine phosphoribosyltransferase 1 (HPRT-1). Values are mean ± SEM for *n* = 6. \* *p* < 0.05 vs. control and # *p* < 0.05 vs. DB.

#### *3.6. Tubastatin A Restores Antioxidant Activity in the Diabetic Retina*

Redox stress in the diabetic retina could result from increased oxidase activities, but also from reduced endogenous antioxidant activities. Nuclear factor erythroid-2-related factor 2 (Nrf2) is a master regulator of endogenous antioxidants gene expression [22], therefore, we tested the effect of TS on the regulation of Nrf2-dependent antioxidant signaling, by monitoring, the expression levels of well-established Nrf2–dependent gene targets. QPCR analysis revealed that diabetes promoted a significant reduction in the expression levels of the Nrf2-dependent genes: Heme oxygenase-1 (HO-1), NAD(*p*)H dehydrogenase quinone 1 (NQO1), glutamate-cysteine ligase regulatory subunit (GCLM), and glutamate-cysteine ligase (GCLC) (Figure 5F–I) However, treatment of diabetic rats with TS restored the mRNA levels of all these genes (*p* < 0.005 (HO-1), *p* < 0.01 (NQO1), *p* < 0.02 (GCLM), *p* < 0.02 (GCLC) vs. control and *p* < 0.002 (HO-1), *p* < 0.01 (NQO1), *p* < 0.02 (GCLM), *p* < 0.03 (GCLC) vs. DB; *n* = 6), thus, suggesting that TS restored Nrf2-dependent signaling in the diabetic retina.

Furthermore, we assessed the expression and activity of the endogenous antioxidant Trx-1 (Figure 6A,B). As previously reported [8], Trx-1 expression was significantly increased in retinas of STZ-rats (DB) in comparison to normoglycemic control rats (Figure 6A,B). Treatment of the diabetic rats with TS, however, significantly decreased Trx-1 expression in diabetic rats (*p* < 0.0051; *n* = 6). Trx-1 activity, measured with a fluorimetric assay, was found to be significantly lower in retinas of STZ-rats (DB) than in normoglycemic age-matched rats (control) (Figure 6C). However, treatment of diabetic rats with TS rescued/normalized this antioxidant enzymatic activity (*p* < 0.01 vs. DB; *n* = 6) (Figure 6C).

**Figure 6.** Effect of Tubastatin A on thioredoxin-1 expression and activity. (**A**) Western analysis of thioredoxin-1 (Trx-1) protein expression measured in retinas of STZ-rats (DB), STZ-rats receiving TS (10 mg/kg) (DB + TS) and normoglycemic control rats (control). (**B**) Quantification of optical density of Trx-1 immunoblotting normalized versus actin. (**C**) Trx-1 enzymatic activity measured in STZ-rats (DB), STZ-rats treated with TS (10 mg/kg) (DB + TS). Values are mean ± SEM for *n* = 6. \* *p* < 0.01 vs. control and # *p* < 0.01 vs. DB.

#### *3.7. E*ff*ect of High Glucose and Tubastatin A on HDAC6 Expression and Activity in Human Retinal Endothelial Cells*

To determine the specific impact of HDAC6 on retinal endothelial cells and microvascular dysfunction, we performed experiments in vitro using HuREC exposed to different glucose levels. First, we confirmed that HDAC6 mRNA expression levels, measured in HuREC by qPCR analysis, were significantly increased when the cells were treated with high glucose concentrations (HG, 25 mM) as compared to cells treated with the osmotic control l-glucose (LG) or exposed to normal glucose containing media (NG, 5.5 mM) (*p* < 0.01 vs. NG; *n* = 3) (Figure 7A). Accordingly, HDAC6 protein expression (Figure 7B) was also significantly augmented in HG-treated HuREC in comparison with LG or NG (*p* < 0.01; *n* = 3). Parallel to HDAC6 protein up-regulation, we also found that HG increased HDAC6 activity 48 h post-treatment in comparison to NG and LG controls (Figure 7C) (*p* < 0.005; *n* = 3). Moreover, treatment of HuREC with TS (5 μM, 6 h pre-treatment + 48 h in combination with HG) significantly down-regulated the activity of HDAC6 in HuREC exposed to HG (*p* < 0.01; *n* = 3) and had no significant effects on cells exposed to normal glucose control (NG) (Figure 7D).

**Figure 7.** Effects of high glucose and Tubastatin A on HDAC6 expression and activity in HuREC. (**A**) HDAC6 mRNA expression, measured by qPCR, in HuREC exposed to different glucose levels (NG = 5.5 mM d-glucose, HG = 25 mM d-glucose) and the osmotic control l-glucose (25 mM) for 48 and 72 h. (**B**) Immunoblotting showing HDAC6 protein levels measured 48 h after exposure of HuREC to HG or the controls NG or LG. (**C**) HDAC6 activity measured in HuREC by fluorimetric assay at 48 h exposure of the cells to HG, NG or LG. (**D**) HDAC6 activity measured in HuREC by fluorimetric assay after 48 h of exposure of the cells to HG or HG plus TS (5 μM starting 6 h before HG treatment) and compared to the controls NG or LG. Values are mean ± SEM for *n* = 3. \* *p* < 0.01 vs. NG and # *p* < 0.01 vs. HG.

#### *3.8. E*ff*ects of HDAC6 Inhibition on Oxidative*/*Nitrative Stress and Endogenous Antioxidants in HuREC*

To explore the potential contribution of HDAC6 to HG-induced redox imbalance in HuREC, we assessed the formation of ROS from cellular sources by determining CellROX deep green fluorescence

intensity in HuREC exposed to NG or HG for 48 h with or without TS (5 μM) (Figure 8A). We found that HG increased superoxide-dependent fluorescence intensity in HuREC as compared to NG group, however, this effect was largely blocked by TS (Figure 8A).

**Figure 8.** Effects of Tubastatin A on cellular and mitochondrial oxidases activities in HuREC. (**A**) CellROX fluorescent assay showing superoxide formation (green) in HuREC exposed to HG for 48 h or to HG in the presence of 5 μM of TS (HG + TS) and compared to HuREC cultured in NG conditions in the absence (NG) or in presence of 5 μM TS (NG + TS). (**B**) Images of MitoSOX assay showing superoxide formation from mitochondria oxidase (red) in HuREC exposed to HG for 48 h or HG in the presence of 5 μM of TS (HG + TS), also for 48 h, and compared to HuREC cultured in NG conditions in the absence (NG) or in the presence of 5 μM TS (NG + TS). In A and B blue fluorescence show cell nuclei counterstained with DAPI. Scale bar, 50 μm.

Same results were obtained while monitoring the effects of HG and TS on superoxide production from mitochondrial sources (Figure 8B). Analysis of mitochondrial oxidases activity by MitoSOX, showed that exposure of HuREC to HG for 48 h increased mitochondrial superoxide-dependent reactivity; however, TS prevented this effect (Figure 8B). In all cases, TS treatment did not affect the response of the cells to NG (NG + TS).

Furthermore, dot blot analysis showed that the levels of the oxidative/nitrative stress markers NT and 4-HNE, were increased by HG, however treatment of the cells with TS halted this effect of HG

(*p* < 0.0001 vs. NG and *p* < 0.0001 vs. HG; *n* = 3) (Figure 9A–C). To ascertain whether TS was also able to normalize endogenous antioxidants, we determined the effects of HG in presence and/or absence of TS on Trx-1 activity (Figure 9D). Similarly, to what was observed in the diabetic rats, glucidic stress (HG) significantly decreased Trx-1 activity in HuREC and this was rescued by TS (*p* < 0.001 vs. NG and *p* < 0.01 vs. HG; *n* = 3) (Figure 9D).

**Figure 9.** Effects of Tubastatin A on high glucose-induced redox imbalance in HuREC. (**A**) Representative images of dot blot analysis demonstrating nitrotyrosine (NT) and 4-hydroxynonenal (4-HNE) formation in HuREC exposed to HG for 48 h or HG in the presence of 5 μM of TS (HG + TS), also for 48 h, and compared to HuREC cultured in NG conditions in the absence (NG) or presence of 5 μM TS (NG + TS). (**B**,**C**) Quantification of optical density of NT and 4-HNE immunoblotting normalized versus actin. Values are mean ± SEM for *n* = 3. \* *p* < 0.0001 vs. NG and # *p* < 0.0001 vs. HG. (**D**) Fluorimetric assay results representing Trx-1 activity in HuREC assessed after 48 h of exposure to different glucose levels (NG = 5.5 mM d-glucose, HG = 25 mM d-glucose) in the presence or absence of TS (5 μM). Values are mean ± SEM for *n* = 3. \* *p* < 0.01 vs. NG and # *p* < 0.01 vs. HG.

#### *3.9. E*ff*ects of HDAC6 Inhibition on HG-induced HuREC Senescence*

Finally, we examined the effects of HDAC6 inhibition by TS on HG-induced HuREC senescence. Assessment of SA-β-Gal activity in HuREC exposed to different glucose conditions showed increased number of positive cells in the HG treatment group compared to the NG control (*p* < 0.001; *n* = 3) (Figure 10A,B). However, treatment of HuREC with TS prevented the increase of SA-β-Gal–reactive cells in HG conditions (*p* < 0.005; *n* = 3) (Figure 10A,B).

**Figure 10.** Tubastatin A effects on HG-induced senescence markers in HuREC. (**A**) Representative images of senescence-associated β-Galactosidase (SA-β-Gal) reactivity assay in HuREC exposed to HG for 48 h or HG plus 5 μM of TS (HG + TS), also for 48 h, and compared to HuREC cultured in NG conditions in the absence (NG) or presence of 5 μM TS (NG + TS). Positive cells develop the blue color. Scale bar, 50 μm. (**B**) Quantification of SA-β-Gal-positive cells. Values are number of positive cells per well versus total number of cells expressed as a percent. *n* = 3. \* *p* < 0.001 vs. NG and # *p* < 0.001 vs. HG. (**C**) Western blotting analysis showing protein levels of the histone deacetylase SIRT1 in HuREC treated in the same experimental conditions as described in (**A**). Bar histograms represent optical density values of the blots normalized for the loading control actin. Values are mean ± SEM for *n* = 3. \* *p* < 0.01 vs. NG and # *p* < 0.01 vs. HG.

Moreover, Western analysis of protein levels of the histone deacetylase SIRT1 showed that this was significantly down-regulated in HG-treated HuREC in comparison to NG. Treatment of the cells with TS significantly reduced the effects of HG by rescuing SIRT1 protein levels (*p* < 0.005 vs. HG; *n* = 3) (Figure 10C).
