Human retinal endothelial cells (HRECs) and dermal microvascular endothelial cells (HMVECs) were obtained from Cell Systems (Walkersville, MD, USA) and Lonza (Kirkland, WA, USA), respectively. Cell culture conditions used in this study have been previously described (24–26). Briefly, cells were cultured using endothelial basal media-2 (EBM-2, Lonza), supplemented with endothelial growth media-2 (EGM-2) SingleQuots (Lonza Inc.), in 25 cm² tissue culture flasks and maintained in a humidified atmosphere containing 5% CO₂ at 37 °C. At 80% confluence, cells were cultured in serum- and growth factor-free medium overnight before exposure to different D-glucose levels (5 mmol/L, mimicking normal glucose [NG] concentrations, or 25 mmol/L, mimicking high glucose [HG] concentrations). The control media formulation contains approximately 5.5 mM glucose (100 mg/dL), which approximates to normal blood glucose levels in vivo. In cell culture studies, concentrations of glucose approaching 10 mM (180 mg/dL) are considered pre-diabetic and concentrations above 10 mM are analogous to diabetic conditions. To investigate the effect of curcumin analogs, cells were pre-treated with L2H21 or L50H46 for 1 h before exposure to high glucose for 48 h. Concentrations of L2H21 and L50H46 were selected following cell viability assessment. In brief, concentrations that did not reduce cell viability were used. All in vitro experiments were independently repeated at least three times and performed with 6 replicates, unless specified.

Male C57BL/6 mice (8 weeks of age) were obtained from Charles River (Montreal, QC, Canada) and randomly divided into five weight-matched groups (n = 6/group): non-diabetic control (C), diabetic (D), and diabetic mice treated with either curcumin (D-Cur), L50H46 (D-L50H46), or L2H21 (D-L2H21). Mice were made diabetic by intraperitoneal injection of multiple, low-dose streptozotocin (STZ; Sigma-Aldrich, St. Louis, MO, USA; 50 mg/kg, dissolved in citrate buffer, pH = 5.6) (18, 26–28). Non-diabetic control mice received citrate buffer alone. Diabetes was confirmed by blood glucose level >20 mmol/L on two consecutive days. Following induction of diabetes, curcumin (20 mg/kg) or its analogs (10 mg/kg) were administered daily by oral gavage in 1% sodium carboxymethyl cellulose solution (CMC-Na) for 8 and 16 weeks. Untreated diabetic mice and non-diabetic control mice were given the same volume of vehicle alone (CMC-Na). Previous studies testing the effect of curcumin in animal models of diabetes have used doses ranging from 50 to 150 mg/kg (23, 29, 30). However, our pilot studies showed that 20 mg/kg curcumin is effective in suppressing inflammatory cytokine expression in the STZ-induced diabetic mouse model. Hence, this low dose was used as a comparator for the curcumin analogs. The dose for curcumin analogs was selected based on our previous studies utilizing lipopolysaccharide-induced proinflammatory cytokine production (22), which is largely mediated through NF-κB (31). Mice were monitored and recorded for blood glucose, glucosuria, ketonuria, and body weights. Prior to euthanasia, echocardiographic analyses were performed. After 2 or 4 months of diabetes, mice were euthanized by carbon dioxide inhalation, and tissues were harvested including retinal, renal cortical, and left ventricular tissues.

To determine potential adverse effects of L2H21 and L50H46 in mice, toxicity analyses were performed in addition to the regular monitoring of mice. Age-matched mice were divided into six groups: L2H21 high-dose group (1g/kg, n = 7), L2H21 middle-dose group (0.5g/kg, n = 7), L2H21 low-dose group (0.1g/kg, n = 5), L50H46 high-dose group (1g/kg, n = 7), L50H46 middle-dose group (0.5g/kg, n = 7), L50H46 low-dose group (0.1g/kg, n = 6). The analogs were administered only once orally, as a single-dose, and mice were followed-up for 7 consecutive days. Following this time-point, organs were excised, fixed in 10% buffered formalin solution and embedded, and sectioned into 5 μm thick sections. The tissue sections were then stained with hematoxylin and eosin (H&E) for routine histology. A blinded pathologist evaluated the histopathological damage using a light microscope and the images were captured (Nikon, Japan).

Cardiac function was assessed in lightly anesthetized mice. Briefly, mice were anesthetized by 1.5% inhaled isoflurane through continuous flow of 2% oxygen for the induction of anesthesia (VetFlo Vaporizer; Kent Scientific, Torrington, CT, USA). Cardiac function was then assessed through echocardiography and Doppler ultrasound using a 40 MHz linear array transducer (MS-550D) and the Vevo 2100 imaging system (VisualSonics, Toronto, ON, Canada). As documented in our previous studies (27, 32), left ventricular ejection fraction (EF) was used as an index of cardiac contractile function, while mitral inflow patterns (E/A ratio) and mitral annulus velocities (E′/A′ ratio) were used to assess diastolic dysfunction. Furthermore, after mice were euthanized, the tibia length and heart weight were recorded to assess the degree of cardiac hypertrophy (33).

Prior to euthanasia (24 h before the end-point), animals were placed in metabolic cages for 24 h and urine samples were collected from all groups. Albumin and creatinine levels were measured using commercial kits (Albuwell and Creatinine companion kit; Exocell, Philadelphia, PA, USA), and albumin-to-creatinine ratios (ACR) were determined (28, 34).

Curcumin analogs were synthesized and characterized as described by us previously (22, 23). Additional information regarding the characteristics of the analogs is provided in Figure S1. Curcumin (Sigma; Oakville, ON, CAN) was used for comparison.

The effect of curcumin and its analogs on cell viability was measured using the water soluble tetrazolium salt-1 viability assay (WST-1, Roche). The assay was carried out as previously described (26, 35).

As previously described (33), HMVECs or HRECs were plated on eight-chamber tissue culture slides and incubated for 48 h after glucose challenge (5 or 25 mmol/L), with or without curcumin or curcumin analog pre-treatment. The cells were fixed with methanol and then stained for an oxidative DNA damage marker, 8-hydroxy-2′-deoxyguanosine (8-OHdG; 1:50, Santa Cruz Biotechnology). Fluorescein isothiocyanate (FITC)-conjugated goat IgG (Vector Laboratories, Burlingame, CA) was used for detection. Slides were mounted in Vectashield fluorescence mounting medium with 4,6-diamidino-2-phenylindole (DAPI; Vector Laboratories) for nuclear staining. Microscopy was performed by an examiner unaware of the identity of the samples using a Zeiss LSM 410 inverted laser scan microscope equipped with fluorescein, rhodamine, and DAPI filters (Carl Zeiss Canada, North York, ON, Canada).

Enzyme-linked immunosorbent assays (ELISAs) were performed to measure tumor necrosis factor-α (TNF-α; R&D Systems, Minneapolis, MN, USA), fibronectin (FN; R&D Systems), collagen-IV alpha 1 chain (COL4A1; Cloud-Clone Corp., Katy, TX, USA), vascular endothelial growth factor (VEGF; R&D Systems), endothelin-1 (ET-1; R&D Systems), and P300 (Cloud-Clone Corp.) in cell or tissue lysates. Briefly, protein concentrations were assessed using the BCA protein assay kit (Pierce, Rockford, IL, USA) and 50 μg of total proteins were used for each assay. Optical density at 450 nm was measured using Multiskan FC Microplate Photometer (Thermo Fisher Scientific, Boston, MA, USA). Readings were corrected using 568 nm absorbance (26–28, 32, 34).

Total RNA from cells and tissues was extracted using TRIzol reagent (Invitrogen, Burlington, ON, CAN), as described previously (11, 12, 17–19, 24–33, 35). Briefly, RNA was extracted with chloroform, followed by centrifugation to separate the sample into aqueous and organic phases. RNA was recovered from the aqueous phase by isopropyl alcohol precipitation and suspended in diethylpyrocarbonate (DEPC)-treated water. Total RNA (1 μg) was used for cDNA synthesis with high capacity cDNA reverse transcription kit (Applied Biosystems, Foster City, CA, USA). The resulting cDNA products were stored at −20°C until further experimentation. Transcript levels were measured using LightCycler (Roche Diagnostics Canada, Laval, QC, CAN) as previously described (11, 12, 17–19, 24–33, 35). For a final reaction volume of 20 μL, the following reagents were added: 10 μL SYBR Advantage qPCR Premix (Clontech, Mountain View, CA, USA), 1 μL each of forward and reverse 10 μmol/L primers (Table S1), 7 μL water, and 1 μL cDNA template. The standard curve method was used to quantify mRNA levels. Standard curves were constructed by using serially diluted standard templates (1:1, 1:10, 1:100, and 1:1,000). The data were normalized to a housekeeping gene, 18S ribosomal RNA or β-actin, to account for differences in reverse transcription efficiencies and the amount of template in the reaction mixtures. In our pilot studies, we observed both 18S ribosomal RNA and β-actin to be stable in HG as well diabetic mouse tissues. In cases where 18S ribosomal RNA was used, cDNA samples were diluted to bring the threshold cycle of house keeping gene to levels similar to β-actin.

Data are presented as mean ± standard error (SEM), unless specified. Statistical significance was determined by using one-way ANOVA, followed by Tukey's post-hoc analysis, for multiple comparisons. A P-value of 0.05 or less was considered to be significant. All calculations were performed with GraphPad Prism 7 (La Jolla, CA, USA).

Our first objective was to establish effective and non-toxic doses of curcumin analogs in cultured endothelial cells. We utilized HMVECs and HRECs as models of microvascular endothelial cells and exposed the cells to varying concentrations of curcumin analogs. Analysis of cell viability by WST-1 assay showed that curcumin and L2H21 reduce viable cell numbers at concentrations equal to or >20 μM (data not shown). Treatment of cells with L50H46 reduced viability at concentrations equal to or >1.25 μM. Based on these results, we selected 10 μM concentration for L2H21 and 0.625 μM concentration for L50H46 for subsequent in vitro experiments (Figure S1D). We also tested and observed no reduction in cell viability following exposure of cells to normal (5 mmol/L) or high levels of glucose (25 mmol/L) for up to 96 h (data not shown).

We administered L50H46 and L2H21 in mice and examined heart, kidney, lung, and liver tissues for evidence of frank toxicity. Our results show that mice administered L50H46 and L2H21, at doses ranging from 100 mg/kg to 1 g/kg, exhibit no observable histological changes in hearts, lungs, kidneys, and liver tissues, as demonstrated in Figure S2. Although we performed toxicity studies using a single dose as done in most preclinical studies, it may not be reflective of any long-term toxicity. Therefore, based on our initial, promising results and the need to perform a long-term study to examine the effects of the analogs on measures of diabetic complications, we elected to use 10 mg/kg dose for the curcumin analogs.

We tested the effect of curcumin analogs in a STZ-induced model of type 1 diabetes. Previous studies have shown that curcumin reduces glycemia in type 2 models (36), and improves insulin resistance (37). Therefore, in type 2 diabetes, curcumin may provide protective effects by reducing the stimulus (hyperglycemia and dyslipidemia) (38). Since our primary goal is to determine whether curcumin analogs prevent hyperglycemia-induced target organ dysfunction, we used the type 1 model. Following the completion of our long-term study, we did not observe any toxic phenotypic effects in the diabetic animals administered curcumin or its analogs. Our results also show that the diabetic mice had significantly increased blood glucose levels and reduced body weights compared to non-diabetic controls, which was expected (Table S2). As desired for our study, blood glucose levels in diabetic mice treated with curcumin or its synthetic analogs were not significantly different compared to untreated diabetic mice at both time-points. In terms of body weights, the values were significantly reduced in diabetic mice treated with curcumin or L50H46 at the 2-month mark when compared to non-diabetic controls; whereas, the body weights of the L2H21-treated diabetic mice were not significantly different. Furthermore, at 4-months, both L2H21 and L50H46-treated diabetic mice demonstrated significantly improved body weights when compared to untreated diabetic and curcumin-treated diabetic mice.

Exposure to high levels of glucose induces pro-inflammatory (26–28, 39), pro-angiogenic (11, 17, 25, 35, 40), and pro-fibrogenic (12, 24, 27, 32, 33) phenotypes in vascular endothelial cells. We selected the most prominent markers of these processes and determined the expression levels in HRECs and HMVECs under basal and high glucose conditions. In keeping with our previous studies (11, 12, 17, 24–28, 32, 33, 35, 39, 40), both cell types demonstrated significant elevations of TNF-α, IL-6, FN, COL4α1, VEGF-A, ET-1, and p300 transcripts following 48-h high glucose (HG) culture (Figures 1A–G). Conversely, the administration of L2H21 and L50H46 significantly reduced these transcripts in cells exposed to HG. Specifically, in HG-treated HRECs and HMVECs, both curcumin analogs reduced IL-6 (Figures 1A,H), TNF-α (Figures 1B,I), FN (Figures 1C,J), COL4α1 (Figures 1D,K), VEGF-A (Figures 1E,L), ET-1 (Figures 1F,M), and p300 (Figures 1G,N) levels compared to HG. Curcumin, used as a comparator, showed the same results.

Although unique RNA alterations were observed in the endothelial cells following L2H21 or L50H46 treatment, we wanted to delineate whether similar alterations exist at the protein level. Therefore, we selected specific protein markers for examination (FN, COL4α1, VEGF-A, ET-1, and p300) in HRECs and HMVECs. Exposure of cells to HG showed that all proteins are increased in HRECs and HMVECs. Pre-treatment of HRECs with L2H21 or L50H46 prior to HG exposure prevented the increases in FN (Figure 2A; L2H21 = 36% reduction, L50H46 = 44% reduction), COL4α1 (Figure 2B; L2H21 = 40%, L50H46 = 58%), VEGF-A (Figure 2C; L2H21 = 59%, L50H46 = 71%), ET-1 (Figure 2D; L2H21 = 42%, L50H46 = 33%), and p300 (Figure 2E; L2H21 = 66%, L50H46 = 58%). Interestingly, curcumin was only able to reduce COL4α1 (34% reduction), VEGF-A (46%), and p300 (60%) proteins. Similarly, in HMVECs, L2H21 treatment significantly downregulated FN, COL4α1, VEGF-A, ET-1 and p300 proteins by approximately 39, 43, 75, 32, and 55%, respectively, as compared to HG (Figures 2F–J). L50H46 treatment also significantly decreased FN (33%), COL4α1 (50%), VEGF-A (53%), and p300 (78%) in HMVECs compared to HG. On the other hand, when curcumin-treated HMVECs were compared to HG, only protein markers for COL4α1, VEGF-A, and p300 were significantly reduced. Furthermore, we found statistically significant differences between L50H46 and curcumin treatments for FN expressions in HRECs (P = 0.0419); while, in HMVECs, differences existed between L2H21 and curcumin treatments for VEGF-A (P = 0.0464) and between L50H46 and curcumin treatments for p300 (P = 0.0368) proteins.

To investigate whether curcumin analogs modulate oxidative stress, we measured oxidative DNA damage by 8-OHdG immunofluorescence staining. As shown in Figure 3, HG increased oxidative DNA damage in both HRECs and HMVECs. Interestingly, pre-treatment of cells with curcumin, L2H21, or L50H46 prior to HG challenge attenuated increases in 8-OHdG immunoreactivity. These results suggest that curcumin analogs dampen glucose-induced oxidative DNA damage.

To determine whether similar transcript level reductions occur at the tissue level after L2H21 or L50H46 treatment, we examined tissues from diabetic mice treated with the analogs. When compared to non-diabetic controls, diabetic mice demonstrated increases in the panel of selected transcripts in the retina, heart, and kidney tissues in a time- and tissue-dependent manner (Figures 4–6). In the retina at both end points, most of the transcripts appeared to be downregulated following treatment with curcumin or its analogs when compared to untreated diabetic mice (Figures 4A–N). Specifically, at 2 months, significant decreases were observed in IL-6 (Figures 4A), FN (Figure 4C), COL4α1 (Figure 4D), and ET-1 (Figure 4F) levels following curcumin or L2H21 treatment; whereas, L50H46 treatment significantly decreased IL-6, FN, and COL4α1 transcript levels. Moreover, at the 4-month time point, significantly lower TNF-α (Figure 4I), FN (Figure 4J), COL4α1 (Figure 4K), ET-1 (Figure 4M), and p300 (Figure 4N) levels were noted in the retinal tissues of L50H46-treated mice when compared to untreated diabetic controls. At this time point and compared to the same controls, L2H21-treated group only showed significant reductions in FN (59%), COL4α1 (46%), ET-1 (47%), and p300 (30%); while, curcumin significantly reduced TNF-α (49%) and COL4α1 (47%) transcript levels. Curcumin and its analogs were relatively comparable in impact for most of the transcripts at 2 and 4 months, with significances mainly observed for IL-6 (curcumin vs. L2H21, P = 0.0443) and FN (L50H46 vs. curcumin, P = 0.0019) at 2 months, and TNF-α (L50H46 vs. L2H21, P = 0.0191, and curcumin vs. L2H21, P = 0.0316) at 4 months.

We then performed similar analyses and examined the effects of curcumin and its analogs on cardiac tissues. Administration of curcumin, L2H21, or L50H46 appeared to downregulate majority of the transcripts in cardiac tissues of diabetic mice (Figures 5A–L, N). In fact, when compared to untreated diabetic mice at 2 months, L2H21 and L50H46-treated mice had significantly lower IL-6 (Figure 5A), TNF-α (Figure 5B), FN (Figure 5C), VEGF-A (Figure 5E), and p300 (Figure 5G) levels; while, curcumin-treated mice also produced similar statistically significant reductions in the above markers, with the exception of p300. Furthermore, L2H21 treatment for a 4-month duration significantly diminished cardiac transcripts encoding FN (Figure 5J, 52% reduction), COL4α1 (Figure 5K, 47% reduction), VEGF-A (Figure 5L, 54% reduction), and p300 (Figure 5N, 25% reduction) when compared to untreated diabetic mice. While significant reductions in transcripts, compared to untreated diabetic mice, were observed for TNF-α (Figure 5I, ~54%) and COL4α1 (~36%) following curcumin treatment and for COL4α1 (~29%) following L50H46 treatment at 4 months. Interestingly, cardiac tissues from mice treated with curcumin or its analogs provoked the upregulation of ET-1 at 4 months (Figure 5M)—with significant differences observed for L2H21 treatment as compared to untreated diabetic mice. Our results also show that curcumin and its analogs are comparable in the reduction of cardiac transcripts at 2 months. However, at the 4-month time point, significant differences in effects existed between curcumin and L2H21 treatments for TNF-α (P = 0.0129), and between L2H21 and L50H46 treatments for FN (P = 0.0402).

Lastly, we extended our analyses to kidney tissues. We found that when compared to untreated diabetic controls, IL-6 (Figure 6A), TNF-α (Figure 6B), FN (Figure 6C), VEGF-A (Figure 6E), ET-1 (Figure 6F), and p300 (Figure 6G) transcripts were significantly dampened in renal cortical tissues after L2H21 or L50H46 treatment at 2 months of diabetes; while, curcumin treatment only demonstrated significant downregulations in IL-6, FN, VEGF-A, and p300 transcripts. Moreover, at 4 months, L50H46 treatment significantly decreased FN (Figure 6J) and VEGF-A (Figure 6L) expressions by approximately 40 and 51%, respectively, when compared to untreated diabetic mice. As well, p300 RNA levels were significantly reduced by curcumin (34%), L2H21 (33%), and L50H46 (38%) treatments at 4 months (Figure 6N). Of note, slight elevations in COL4α1 transcripts were detected at 4 months following treatment with curcumin or its analogs (Figure 6K); however, these changes were not significant. L2H21 and curcumin treatments generated different response levels for IL-6 (P = 0.0416) and TNF-α (P = 0.0082) at 2 months, while significances existed between treatments in ET-1 expressions at 4 months (L50H46 vs. L2H21, P = 0.0001, and L50H46 vs. curcumin, P = 0.0001; Figure 6M).

To build on our in vitro results, we selected one marker each from the signaling pathways (TNF-α, COL4α1, ET-1, and p300) and examined the levels of these proteins in retinal, cardiac, and kidney tissues of mice. Similar to our findings in vitro, all proteins showed higher levels in untreated diabetic mice (Figures S3–S5). However, analysis of the effects of curcumin analogs showed differential responses at the 2 and 4-month time points. For example, in retinal tissues, significant decreases were observed in p300 expressions after L2H21 treatment at 2 months (Figure S3D; 48% reduction), and in TNF-α (Figure S3E; 89%) and COL4α1 (Figure S3F; 71%) expressions after L50H46 and curcumin treatments, respectively, at 4 months. Additionally, inter-treatment differences existed in retinal TNF-α (L50H46 vs. curcumin, P = 0.0315, and L50H46 vs. L2H21, P = 0.0023) and COL4α1 (curcumin vs. L50H46, P = 0.0480) at 4 months. Overall, treatment with curcumin, L2H21 or L50H46 had relatively comparable effects on the expression of these proteins in diabetic retinal tissues.

Moreover, compared to untreated diabetic mice, cardiac tissues of mice treated with curcumin or its analogs showed significantly reduced protein levels of TNF-α (Figure S4A), ET-1 (Figure S4C), and p300 (Figure S4D) at 2 months, and TNF-α (Figure S4E), COL4α1 (Figure S4F), ET-1 (Figure S4G), and p300 (Figure S4H) at 4 months. L2H21 and L50H46 treatments also showed differences when compared to curcumin for TNF-α at 2 months (L2H21 vs. curcumin, P < 0.0001, and L50H46 vs. curcumin, P < 0.0001); while, at 4 months, L50H46 treatments showed differences when compared to curcumin for TNF-α (P = 0.0010) and ET-1 (P = 0.0160). Nevertheless, similar to the results of retinal tissues, administration of curcumin or its analogs produced comparable reductions of proteins in cardiac tissues.

We next examined the protein levels in renal cortical tissues. When compared to kidneys harvested from diabetic mice, L2H21-treated mice only showed significant reductions in p300 proteins at 2 months (Figure S5D; 35% decrease); while, L50H46-treated mice showed decreases in TNF-α (Figure S5E; 70%) and p300 (Figure S5H; ~37%) proteins at 4 months. Alternatively, inter-treatment significance existed between L50H46 and L2H21 treatments for TNF-α at 4 months (P = 0.0141), whereas, the remaining markers did not show any differences between treatments.

Overall, at the protein level, our in vitro and in vivo findings suggest that intervention with curcumin or its analogs can induce general downregulations of key protein mediators involved in inflammation, fibrosis, and angiogenesis.

Our next objective was to explore whether L50H46 or L2H21 treatments can improve the functional deficits characteristic of diabetic heart and kidney diseases. At 2 months, echocardiographic examination did not reveal decreases of E/A and E′/A′ ratios in untreated diabetic mice compared to non-diabetic controls (Figures S6A,B). While, at 4 months, reduced E/A ratios (27% reduction, P = 0.0040) were observed in diabetic mice (Figures S6F,G). However, when diabetic animals were treated with curcumin analogs, E/A and E′/A′ ratios were partially improved compared to untreated diabetic mice; in which, L2H21 and L50H46 treatments exhibited significant improvements in E/A ratios at 4 months (L2H21 improved E/A ratio by 35%, P = 0.0063, and L50H46 improved E/A ratio by 32%, P = 0.0083, when compared to untreated diabetic mice). Furthermore, increases in fractional shortening were observed in diabetic animals and treatment with curcumin, L2H21, or L50H46 showed partial reductions in fractional shortening at the time points (Figures S6C,H). As for quantifying cardiac hypertrophy, we used heart weight/tibia length measurements. Our findings demonstrated that increases in heart weight/tibia length ratios from the heart of diabetic animals were prevented by treatment with curcumin analogs, L50H46 and L2H21, after both 2 and 4 months (Figures S6D,I).

We further evaluated renal albumin excretion. Diabetic mice displayed significantly elevated urinary albumin excretion at 2 (P < 0.0001) and 4 (P = 0.0002) months compared to non-diabetic animals (Figures S6E,J). Treatment of mice with curcumin analogs showed reductions in urinary albumin excretion at both time points.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.