Independent of Renox, NOX5 Promotes Renal Inflammation and Fibrosis in Diabetes by Activating ROS-Sensitive Pathways

Diabetes is increasingly recognized as the leading cause of chronic kidney disease. Despite the advent of therapies such as blockade of the renin-angiotensin system and, more recently, sodium–glucose cotransporter 2 (SGLT2) inhibition, a significant number of patients with diabetes still progress to end-stage renal disease requiring costly dialysis and transplantation (1,2). Therefore, there is an urgent need to develop more effective treatments. Enhanced generation of renal reactive oxygen species (ROS), particularly those derived from NADPH oxidases (NOXs), are key mediators of renal inflammation, fibrosis, and mitochondrial dysfunction in diabetic kidney disease (DKD) (3–5). Previous experimental research has focused on the constitutively active NOX isoform NOX4 (Renox), shown to be pathogenic in murine models of DKD (4–9). Although NOX4 is deleterious in experimental DKD, it provides vasculoprotection in the context of atherosclerosis (10–12), reducing the enthusiasm for NOX4 as a sole target of new treatments in diabetic complications. In addition, the only currently available NOX inhibitor, GKT137831, predominantly blocks renal NOX4 (13), shown to be renoprotective in experimental DKD (5,14), but clinical results indicated this drug was ineffective in treating patients with type 2 diabetes and nephropathy (15).

There is increasing evidence that an alternative isoform, NOX5, which is expressed in humans but not in rodents, plays a crucial role not only in DKD (3,16–18) but also in retinopathy (19), cardiovascular (20,21), and cerebrovascular (22) diseases. Indeed, NOX5 is highly upregulated in association with enhanced ROS formation in kidney biopsy specimens of patients with diabetes (3,17). Recently, it has been demonstrated that Nox4 wild-type mice with overexpression of human NOX5 in renal cells, specifically in podocytes and endothelial or smooth muscle cells (SMCs) showed accelerated renal injury in DKD (3,16–18). Both NOX4 and NOX5 are expressed endogenously in humans. It appears that NOX4 is pathogenic in murine models of DKD in the absence of NOX5, whereas NOX5 appears to be relevant in human DKD. However, the role of NOX5, independent of the NOX4 pathway in DKD, has not been determined, because preclinical murine models express Nox4 but not Nox5. Thus, it has been difficult to determine the distinct roles of NOX4 versus NOX5 in DKD. We previously localized the expression of NOX5 in mesangial cells in human kidney biopsy specimens with increased expression in patients with diabetes and nephropathy (17). In the present study, we investigated the role of NOX5 independent of the NOX4 pathway in humanized transgenic (Tg) mice with specific expression of NOX5 in SMCs, which represent mesangial cells in the kidney in the context of genetic deletion or pharmacological NOX4 inhibition, as well as further defining the status of NOX5 and related molecules in human kidney biopsies and cells.

To examine the detrimental role of NOX5 independent of the NOX4 pathway in DKD, we used genetically modified NOX5 Tg mice deficient in Nox4 or NOX4 inhibition by GKT137831 in the presence or absence of diabetes, human mesangial cells silenced for NOX4 or NOX5 exposed to high glucose levels, and human kidney biopsy specimens obtained from individuals with and without diabetes. Sample size was chosen based on results of our previous studies (5,17). In general, 15 mice per group were randomly assigned to the respective in vivo study groups. The animal studies were approved by the Alfred Medical Research & Education Precinct Animal Ethics Committee (no. E1493/2014/B; Melbourne, Victoria, Australia) under guidelines laid down by the National Health and Medical Research Council of Australia.

Human kidney biopsy specimens were obtained from male and female individuals with and without diabetes with nephropathy (estimated glomerular filtration rate, 30–52 mL/min; n = 6/group), and the protocol was approved by the Human Ethics Committee of Austin Health (no. LNR/13/Austin/163; Heidelberg, Victoria, Australia) and complied with the Declaration of Helsinki.

NOX5 Tg mice on an FVB/N background with selective expression of the human NOX5 gene in SMCs, which represent mesangial cells in the kidney, were generated as described previously (3,17,18). Briefly, to generate SMC-specific NOX5 Tg mice, mice of the SM22-tTA-FVB/N strain (SM22; encoded by Sm22 promoter, also known as Tagln) were crossed with mice of the NOX5 β FVB/N strain (Clontech, Mountain View, CA) to produce Sm22⁺NOX5⁺ and Sm22⁺NOX5⁻ mice. A subgroup of Sm22⁺NOX5⁺ mice were crossed with in-house–established Nox4^−/− mice (5,23) on an FVB/N background to generate Sm22⁺NOX5⁺/Nox4^−/− and Sm22⁺NOX5⁻/Nox4^−/− mice.

Diabetes was induced in 6-week-old male Sm22⁺NOX5⁺/Nox4^−/− and Sm22⁺NOX5⁻/Nox4^−/− mice by five daily intraperitoneal injections of streptozotocin (STZ; Sigma-Aldrich, St Louis, MO) at a dose of 55 mg/kg in citrate buffer, with control mice receiving citrate buffer alone, and the study was followed for 10 weeks. In parallel to genetic deletion, a pharmacological inhibition of NOX4 by GKT137831 was used. A subgroup of male diabetic and nondiabetic control Sm22⁺NOX5⁺ and Sm22⁺NOX5⁻ mice were administered the NOX inhibitor GKT137831 daily by gavage at a dose of 60 mg/kg per day for 10 weeks, commencing with the last injection of STZ.

Mice were placed individually into metabolic cages (Iffa Credo, Lyon, France) for 24 h at week 9 for the collection of urine. Blood glucose, glycated hemoglobin, and systolic blood pressure levels were measured as described previously (5,17). Urinary albumin concentration was measured by using a mouse albumin ELISA quantification kit (Bethyl Laboratories, Montgomery, TX). Urinary and plasma creatinine levels were determined with the Cobas Integra 400 Plus analyzer (Roche Diagnostics, Indianapolis, IN). Creatinine clearance and the urinary albumin to creatinine ratio were determined as previously described (5,17,18). A mouse cystatin C ELISA kit (BioVendor, Brno, Czech Republic) was used to determine plasma cystatin C levels according to the manufacturer’s instructions. After 10 weeks (i.e., at 16 weeks of age), mice were anesthetized by intraperitoneal injection of sodium pentobarbital (100 mg/kg body weight; Euthatal; Sigma-Aldrich, Castle Hill, Australia). Only mice with blood glucose ≥15 mmol/L were included in the experiments; mice with blood glucose <15 mmol/L and with polycystic kidneys were excluded from the study (<5% of the total number of mice).

Total RNA from the renal cortex of mice was extracted with Trizol and analyzed, and cDNA was generated as described previously (5,17). Gene expression, using mouse probes and primers as described in Supplementary Table 2, was analyzed quantitatively and reported relative to the expression of the housekeeping gene 18S (18S ribosomal RNA Taqman Control Reagent Kit) using the Taqman system (ABI Prism 7500; Perkin-Elmer, Poster City, CA). Results were expressed relative to respective nondiabetic control mice without NOX5 expression, which were arbitrarily assigned a value of 1.

Paraffin-embedded kidney sections, 3 μm thick, were stained with periodic acid-Schiff to assess mesangial area, glomerulosclerotic index (GSI) and tubulointerstitial injury (TII), as described previously (5,17). Immunostaining for glomerular nitrotyrosine (rabbit anti-nitrotyrosine, catalog no. ab5411; Millipore, Billerica, MA), collagen IV (goat anti–type IV collagen, catalog no. 1340–01; Southern Biotech, Birmingham, AL), and CD68 (rabbit polyclonal anti-CD68, catalog no. ab125212; LSBio, Seattle, WA) was performed as described previously (5,17). Sections were then examined under a light microscope (Olympus BX-50; Olympus Optical, Tokyo, Japan) with 20 glomeruli being captured per kidney for the assessment of mesangial area and quantification of nitrotyrosine and collagen IV (percentage of glomerular area) using Image-Pro plus 7.0 software (Media Cybernetics, Bethesda, MD). Additionally, 10 tubulointerstitial fields per kidney were captured for the assessment of TII. GSI and TII were graded, as described previously (5,17). For CD68, the number of positively staining macrophages within each glomerulus was counted. All assessments were performed in a blinded manner. Results were expressed relative to respective nondiabetic control mice without NOX5 expression, which were arbitrarily assigned a value of 1.

After dewaxing, human kidney biopsy specimens were fixed in 4% paraformaldehyde for 15 min, permeabilized using 0.1% Triton X-100, and incubated in a blocking buffer (1% BSA, 0.25% Triton X in PBS, pH 7.4). Double immunostaining for NOX5 with nitrotyrosine, SM22-α, EGR-1, PKC-α, TXNIP, and NOX4 in human kidney biopsy specimens was performed by incubation with respective primary antibodies to NOX5 (goat polyclonal, catalog no. NBP1–68862; Novus Biologicals, Centennial, CO), nitrotyrosine (rabbit anti-nitrotyrosine, catalog no. ab5411; Millipore), SM22-α (rabbit polyclonal; catalog no. ab14106 Abcam, Cambridge, MA), EGR-1 (rabbit monoclonal, catalog no. 4154S, Cell Signaling Technology, Beverly, MA, USA), PKC-α (rabbit monoclonal, catalog no. ab32376; Abcam), TXNIP (rabbit monoclonal, catalog no. ab188865; Abcam), and NOX4 (rabbit polyclonal, catalog no. ab154244; Abcam) followed by incubation with secondary antibody Alexa Fluor 488 (donkey anti-goat; Invitrogen, Eugene, OR) and Alexa Fluor 568 (donkey anti-rabbit; Invitrogen). All stained sections were examined and images of 3 glomeruli/section of biopsy specimen were taken using a Nikon eclipse-Ni (Tokyo, Japan) fluorescence microscope. Quantitation of glomerular NOX5 fluorescent intensity (percentage of glomerular area) was performed using Image-Pro plus 7.0 software (Media Cybernetics).

Protein extracts were obtained from mouse renal cortex for the measurement of renal MCP-1 level, as described previously (5). The Quantikine Mouse MCP-1 ELISA Kit (R&D Systems, Minneapolis, MN) was used to estimate renal and plasma MCP-1 levels, as per the kit instructions. Renal MCP-1 level was expressed relative to total protein concentration.

Renal protein expression by Western blot analysis was performed as described previously (17,18). Briefly, protein extracts from the renal cortex were electrophoresed on 4–20% Mini-PROTEAN precast gels (Bio-Rad Laboratories, Richmond, CA) under nonreducing conditions. Blots were incubated with primary antibodies to the following: EGR-1 (rabbit monoclonal; Cell Signaling Technology), PKC-α (rabbit monoclonal; Abcam) TXNIP (rabbit monoclonal; Abcam), phospho-p44/42 mitogen-activated protein kinase (MAPK; Erk1/2) (rabbit polyclonal, catalog no. 9101; Cell Signaling Technology), and JNK (rabbit polyclonal, catalog no. ab4821; Abcam) overnight at 4°C followed by incubation with goat anti-rabbit secondary antibodies (Dako Corp., Carpinteria, CA). Membranes were subsequently probed for β-actin (Sigma-Aldrich) or α-tubulin (Cell Signaling Technology) to confirm equal loading of samples. Blots were detected using the ECL detection kit (Sigma-Aldrich) and densitometry was performed using Quantity-One software (Bio-Rad Laboratories).

Glomerular basement membrane (GBM) thickness was measured from electron micrographs, as described previously (7,24). Briefly, 1-mm sections of renal cortex were fixed, dehydrated in acetone, and finally embedded in Spurr’s resin. Sections were then cut and visualized using a Hitachi 7500 transmission electron microscope (Tokyo, Japan). A total of 8–10 electron micrographs of nonoverlapping fields were taken from each glomerulus at a constant magnification of ×10,000. Thickening of GBM was assessed by measuring the thickness of 12–15 evenly distributed points per capillary loop (7,24), using ImageJ software (National Institutes of Health, Bethesda, MD). An average of 120 points per glomerulus was measured, after which the arithmetic mean ± SEM for GBM thickness was determined for each mouse.

Human mesangial cells (17) (Clonetics Mesangial Cell Systems; Lonza, Walkersville, MD) were used for in vitro experiments. Human mesangial cells were grown using mesangial cell medium (Lonza) with 10% FCS in a humidified incubator (5% CO₂ at 37°C). The knockdown (KD) of NOX4 or NOX5 was performed in human mesangial cells using MISSION shRNA–expressing lentivirus vectors (Sigma-Aldrich), as previously described (17). Cells transduced with the MISSION nontarget shRNA control vector particles (Sigma-Aldrich) were used as controls. The KD efficiency in the cells was verified by RT-PCR and was ∼75% for both NOX4 and NOX5. Nontarget and Nox4-shRNA– or Nox5-shRNA–infected cells were then grown in cell media with 5 mmol/L glucose or 25 mmol/L glucose or 5 mmol/L glucose plus 20 mmol/L mannitol and incubated for 24 h at 37°C. Cells were harvested, RNA was extracted by the TRIzol method, and cDNA was synthesized for quantitative RT-PCR, as described previously (5,17). Results were expressed relative to the nontarget control, which was arbitrarily assigned a value of 1. Human probe and primer sequences used for quantitative RT-PCR are listed in Supplementary Table 3.

Luminol analog L-012 was used to detect superoxide in human mesangial cells. Normal, nontarget, and Nox4-shRNA– or Nox5-shRNA–infected cells were resuspended in 200 mL of RPMI medium at a density of 10⁴ cells/well of a 96-well microplate (PerkinElmer) and incubated at 37°C for 24 h. The next day, cells were incubated in a normal glucose concentration (5 mmol/L) or in a high concentration of glucose (25 mmol/L) for 24 h. Each well was washed with Krebs-HEPES, and 100 mL of Krebs-HEPES supplemented with L-012 (100 mmol/L; Wako Chemicals) was subsequently added and incubated at 37°C for 10 min, and plates were read on a luminometer (Berthold Technologies) (5,17). Luminous intensity was measured as relative light units, and results were expressed relative to the nontarget control.

All variables were analyzed by one-way ANOVA using GraphPad Software Prism 7 (San Diego, CA) for multiple comparison of the means followed by Tukey post hoc test or were analyzed by the two-tailed unpaired t test when required. P < 0.05 was considered statistically significant. Results are expressed as mean ± SEM, unless otherwise specified.

The data sets used and/or analyzed during this study are available from the corresponding author on reasonable request.

The pathological relevance of NOX4 has been well characterized in murine models of DKD (5,7), but in the absence of Nox5 expression, because murine models lack the Nox5 gene. NOX5 is expressed in humans but not in rodents and is reported to be pathogenic in human DKD. However, we have recently demonstrated that in Nox4 wild-type mice, overexpression of human NOX5 selectively in SMCs (SMCs represent mesangial cells in the kidney) led to aggravated renal injury in diabetes (17,18). Injury to mesangial cells results in mesangial expansion and glomerulosclerosis, the key features of DKD (4). Thus, to examine the direct causal role of NOX5 in DKD independent of the NOX4 pathway, SMC-specific NOX5 Tg mice deficient in the Nox4 gene or treated with the NOX1/4 inhibitor GKT137831 in the absence or presence of STZ-induced diabetes (STZ-diabetes) were used as preclinical models.

To examine the pathogenic role of NOX5 independent of the NOX4 pathway in DKD, SMC (SM22⁺Nox5⁺)-specific NOX5 Tg mice deficient in Nox4 or treated with the NOX1/4 inhibitor, GKT137831 in the absence or presence of STZ-diabetes were used as experimental models. All groups of diabetic mice, including SM22⁺Nox5⁻/Nox4^−/− and SM22⁺Nox5⁺/Nox4^−/− mice, as well as SM22⁺Nox5⁻/GKT137831 and SM22⁺Nox5⁺/GKT137831 mice, had elevated levels of plasma glucose and glycated hemoglobin, increased food and water intake, enhanced urine output, as well as reduced body weights and increased kidney weight to body weight ratios, when compared with their respective nondiabetic control groups (Tables 1 and 2). Regardless of NOX5 expression, no changes in these metabolic variables were observed in the respective control and diabetic mice (Tables 1 and 2). Systolic blood pressure was similar in all groups (Tables 1 and 2). In addition, we examined plasma cystatin C levels and creatinine clearance to assess renal function. All groups of diabetic mice with and without NOX5 expression had lower levels of plasma cystatin C and increased creatinine clearance than their respective nondiabetic control groups, consistent with hyperfiltration (Tables 1 and 2). NOX5 expression per se did not alter the state of hyperfiltration.

To examine the clinical relevance of NOX5 versus NOX4 in DKD, immunostaining of human kidney biopsy specimens showed increased expression of glomerular NOX5 (Fig. 1A and B) in association with enhanced oxidative stress as reflected by enhanced renal intensity of nitrotyrosine in individuals with diabetes compared with those without diabetes (Fig. 1A). In addition, a positive correlation between HbA_1c and glomerular NOX5 expression was observed in individuals with diabetes (Supplementary Fig. 1 and Supplementary Table 1). The expression level of glomerular NOX4 was not changed in those with diabetes versus those without (Supplementary Fig. 2A). Colocalization of NOX5 was observed in human mesangial cells in kidney biopsy specimens (Fig. 1C). Therefore, in in vitro experiments using human mesangial cells, we examined the relative expression of NOX isoforms in response to glucose. Like human diabetic kidney biopsy specimens, NOX5 expression was upregulated (P < 0.001) in response to high glucose concentration (Fig. 1D), whereas the expression of NOX4 (Fig. 1D) as well as NOX1 and NOX2 isoforms remained unchanged (Supplementary Fig. 2B).

We examined nitrotyrosine as a marker of ROS, previously shown to be increased by diabetes in the glomeruli of wild-type mice with attenuation in Nox4 knockout (KO) mice (5,7). In this study, both genetic deletion and pharmacological inhibition of NOX4 by GKT137831 showed no differences in the renal accumulation of nitrotyrosine in diabetic mice compared with their respective nondiabetic control mice (Fig. 1E–H). However, diabetic mice with NOX5 expression in SMCs had an increased accumulation of glomerular nitrotyrosine compared with diabetic animals not expressing NOX5 (Fig. 1E–H).

Furthermore, we examined the renal expression of Nox1 and Nox2 isoforms. Nox1 expression was almost undetectable. Consistent with glomerular nitrotyrosine, in the absence of NOX5 expression, renal upregulation of the Nox2 gene was not seen in diabetic mice with Nox4 deletion or inhibition by GKT137831 (Fig. 1I–J). However, in the presence of NOX5 expression, diabetic mice had upregulation of Nox2 in the kidney when compared with diabetic mice not expressing NOX5 (Fig. 1I–J). In addition, we examined urinary 8-isoprostane levels as a marker of systemic ROS. Unlike glomerular nitrotyrosine, urinary 8-isoprostane levels remained elevated in diabetic mice despite Nox4 deletion (Fig. 2K) or its inhibition by GKT137831 (Fig. 2L). Interestingly, expression of NOX5 in SMCs led to a further increase in urinary 8-isoprostane levels in Nox4-deficient diabetic mice (P < 0.05; Fig. 2K).

We have previously reported that overexpression of human NOX5 in Nox4 wild-type mice was associated with accelerated renal injury and albuminuria in diabetes (3,17,18). Therefore, in the present study, we examined the direct causal role of NOX5 in the absence of NOX4. We measured both 24-h urinary albumin excretion (Fig. 2A and B) and the albumin to creatine ratio (Fig. 2C and D). Despite genetic deletion or pharmacological inhibition of NOX4, the level of albuminuria was higher (P < 0.001) in all groups of diabetic mice after 10 weeks of diabetes when compared with the respective control groups (Fig. 2A–D). Interestingly, there was further significant increase in albuminuria with NOX5 expression in diabetic mice (Fig. 2A–D). NOX5 expression in SMCs per se did not show any differences in albuminuria in nondiabetic mice (Fig. 2A–D).

Both mesangial expansion and GSI remained increased in diabetic mice despite Nox4 deletion or inhibition by GKT137831, in comparison with their nondiabetic counterparts (Fig. 3A–F). In addition, diabetic mice with NOX5 expression in SMCs had accelerated mesangial expansion and glomerulosclerosis (Fig. 3A–F). Furthermore, despite Nox4 deletion or inhibition, diabetic mice had renal upregulation of the cell proliferation markers Pcna and Ki67 with further upregulation in the presence of NOX5 expression in SMCs (Fig. 3G–J). Moreover, assessment of the glomerular ultrastructure showed increased thickening of the GBM in diabetic mice, which was further increased by NOX5 expression (P < 0.05; Fig. 3K and L). Unlike GSI, TII remained the same in all respective groups of mice with and without NOX5 expression in the presence or absence of diabetes (Supplementary Fig. 3A–D).

In DKD, activation of growth factors is associated with enhanced renal cell proliferation, mesangial expansion, as well as increased extracellular matrix (ECM) accumulation as a result of excessive synthesis of ECM proteins (4). Indeed, consistent with GSI and mesangial expansion, gene expression of classical ECM molecules, including renal collagen IV (Fig. 4A and B) and fibronectin (Fig. 4C and D), α-SMA (Fig. 4E and F), and the growth factor Ctgf (Fig. 4G and H), as well as glomerular deposition of the collagen IV protein (Fig. 4I and L) remained increased in diabetic mice despite Nox4 deletion or inhibition by GKT137831. In contrast, all of these ECM molecules and fibrotic markers were further increased by SMC-specific NOX5 expression and diabetes (Fig. 4A−L). NOX5 expression in SMCs, per se, did not appear to result in any differences in the aforementioned fibrotic factors in nondiabetic mice (Fig. 4A−L).

DKD is associated with inflammation, with renal ROS shown to be activating inflammatory cascades (4,25). Similar to the oxidative stress marker nitrotyrosine, a range of proinflammatory markers, including gene expression of renal Tlr-4 (Fig. 5A and B), expression of renal Mcp-1, both at the gene (Fig. 5C and D) and protein (Fig. 5E and F) level, as well as the number of glomerular CD68⁺ cells (Fig. 5G and H and Supplementary Fig. 4A and B) remained unchanged in diabetic mice deficient in Nox4 or treated with GKT137831, when compared with their respective nondiabetic control mice. In contrast, all of these proinflammatory markers were increased by SMC-specific NOX5 expression (Fig. 5A−H). Notably, GKT137831-treated groups of mice with NOX5 expression had upregulated renal Tlr-4 even in the absence of diabetes, when compared with control mice not expressing NOX5 (P < 0.05; Fig. 5B). Similar to Tlr-4 and CD68, expression of markers characteristic of M1 macrophages, including Cd80 and Tnf-α, were upregulated in the presence of NOX5 expression in diabetic mice (Fig. 5I−J). However, markers characteristic of M2 macrophages such as Cd204, IL-10, and arginase-1 (Arg-1), were not altered by the presence of NOX5 expression and diabetes in Nox4 KO mice (Supplementary Fig. 4C–E).

To further establish the link between NOX5 and important regulatory elements considered relevant to human DKD, we examined the expression of the transcription factor EGR-1; NOX-sensitive PKC-α, a signaling molecule implicated in diabetes associated albuminuria (7,26); and a key metabolic gene involved in redox balance, TXNIP, in the human kidney biopsy specimens as well as in the experimental animals. In humans, we found increased fluorescent intensity of EGR-1, PKC-α, and TXNIP in association with increased NOX5 expression in the kidneys of individuals with diabetes versus those without (Fig. 6A−C).

In animals, both renal gene and protein expression of EGR-1 (Fig. 6D−F), PKC-α (Fig. 6G−I), and MAPKs, particularly phosphorylated extracellular signal-regulated kinases 1 and 2 (pERK1/2) (Fig. 6M−O), remained unchanged in Nox4-KO diabetic mice without NOX5 expression compared with the nondiabetic control mice. However, these ROS-sensitive factors were significantly upregulated in the kidney in the presence of NOX5 expression in the SMCs (Fig. 6D−I and 6M−O) of diabetic mice compared with diabetic mice not expressing NOX5. Similar to PKC-α, PKC-ε was upregulated in NOX5-expressing diabetic mice (Fig. 6P). The expression of renal TXNIP remained elevated (P < 0.05) in diabetic mice despite Nox4 deletion and was further increased with NOX5 expression (P < 0.01) (Fig. 6J−L). In addition, gene expression of Egr-1, Pkc-α, Txnip, Erk1, and Pkc-ε was increased in nondiabetic mice with NOX5 expression in comparison with control mice without NOX5 expression (Fig. 6D, G, J, M, and P). The protein expression of phosphorylated c-Jun N-terminal kinase (pJNK1) was elevated in both groups of diabetic mice compared with the respective control mice and was not changed in the presence or absence of NOX5 expression (Fig. 6Q and R). Similar findings were observed in GKT137831-treated mice with regard to the gene expression of Egr-1, Pkc-α, and Txnip (Supplementary Fig. 5A–C), with expression of NOX5 in SMCs upregulating these key genes. Interestingly, consistent with these ROS-sensitive factors, renal gene expression of Vegf was not altered by diabetes alone and was upregulated in the presence of NOX5 expression and diabetes (Supplementary Fig. 5D and E).

Consistent with the in vivo findings linking NOX5 to proinflammatory, ROS-sensitive factors and profibrotic pathways, in vitro, silencing of NOX5 in human mesangial cells had a more pronounced attenuating effect on ROS formation induced by high glucose levels (Fig. 7A) as well as on downregulation of proinflammatory genes TLR4 and MCP-1 (Fig. 7B and C) in comparison with NOX4 silencing. Furthermore, silencing of NOX5 showed more robust attenuation of upregulation of ROS-sensitive genes EGR-1, PKC-α, and TXNIP induced by high glucose concentrations when compared with NOX4 silencing (Fig. 7D−F). In contrast, silencing of NOX5, but not NOX4, attenuated upregulation of profibrotic genes including Ctgf and collagen I and III (Fig. 7G−I) induced by high glucose levels, consistent with the dominant role of the pro-oxidant enzyme NOX5 versus NOX4 on renal fibrosis in the context of human DKD. Moreover, we observed that silencing of NOX5 in mesangial cells showed downregulation of NOX4 but not vice versa (Fig. 7J and K).

Intrarenal oxidative stress plays a central role in the pathogenesis of DKD (4,27); results of studies of rodents suggest that the NOX family of enzymes, particularly the NOX4 isoform, is a prominent source of renal ROS in both physiological and pathological conditions such as DKD (4,5,7–9). The NOX5 isoform, present in humans but absent from the genome of rodents, increasingly has become recognized for its role in DKD and other diabetic complications (3,16–18). To our knowledge, this is the first comprehensive study in which the pro-oxidant enzyme NOX5 has been identified as a key contributor to DKD, independent of the NOX4 pathway. We demonstrated colocalization of NOX5 in mesangial cells, with increased expression of NOX5 in kidney biopsy specimens of individuals with diabetes, in association with enhanced ROS formation and NOX5-dependent, ROS-sensitive pathways, including EGR-1, PKC-α, and TXNIP. Furthermore, in human renal cells, we identified upregulation of NOX5 in response to high glucose concentration and found that silencing of NOX5 exhibited pronounced attenuation of markers of inflammation, fibrosis, and ROS-sensitive factors via a reduction in ROS formation. Moreover, it appears that NOX4 is downstream of and potentially regulated by NOX5.

In the in vivo setting, we evaluated the pathogenic role of NOX5 independent of the NOX4 pathway. We previously demonstrated that the renoprotective effects of genetic Nox4 deletion were also conferred by administration of the specific NOX1/NOX4 inhibitor GKT137831 in murine models of DKD (5,14). However, these rodent models do not express NOX5 naturally (28), and GKT137831 was shown to have, at best, a very weak inhibitory effect on NOX5 (13). Thus, in parallel to genetic deletion of Nox4, we examined the effect of this inhibitor, GKT137831, in our NOX5 Tg mouse models.

NOX5 primarily generates superoxide, which interacts with nitric oxide to form peroxynitrite, in turn resulting in nitrotyrosine formation. Consistent with our previous observations (5,8,14), genetic deletion or pharmacological inhibition of NOX4 did not appear to result in increased ROS formation, as evidenced by a lack of glomerular nitrotyrosine accumulation in diabetic mice. However, NOX5 expression resulted in enhanced glomerular accumulation of nitrotyrosine in the presence of diabetes. This suggests that NOX4 blockade is insufficient to halt ROS generation in the context of NOX5 expression, pinpointing NOX5 as the likely source of this increase in intrarenal ROS. This finding confirms and extends in vitro research suggesting that GKT137831 has no major inhibitory effect on NOX5 (13), thereby suggesting that this drug is not sufficiently effective in blocking NOX5-derived ROS in vivo.

There is a bidirectional interplay between renal ROS and inflammation, thereby perpetuating renal injury in DKD (4). TLR4 and the chemokine MCP-1 are modulated by ROS, leading to macrophage infiltration and inflammation in chronic diseases such as DKD (16,18,29–31). Indeed, TLR4 has also been linked to both NOX4 (32,33) and NOX5 (18) in DKD and nondiabetic inflammatory models (16). In the present study, both genetic deletion and pharmacological inhibition of NOX4 in diabetic mice mitigated renal expression of Tlr4 and Mcp-1 as well as glomerular CD68⁺ macrophage deposition and markers characteristic of M1 macrophages, including Cd80 and Tnf-α. Conversely, despite Nox4 deficiency, NOX5 expression in diabetic mice enhanced these proinflammatory cascades. Thus, independent of NOX4, NOX5 appears to enhance renal macrophage infiltration and inflammation in DKD. Furthermore, silencing of NOX5 in human renal cells showed a pronounced attenuation of increased expression of TLR-4 and MCP-1 induced by high glucose concentrations. NOX4 blockade completely abrogated mRNA levels of the phagocytic NOX isoform Nox2 in diabetic mice in the absence of NOX5. By contrast, Nox2 was upregulated in the presence of diabetes and NOX5. This observation indicates there is a feedback mechanism activated by NOX5-derived enhanced ROS that augments inflammation in DKD.

To link NOX5-derived, ROS-mediated renal inflammation and fibrosis in DKD, we examined key signaling pathways, including the transcription factor EGR-1; protein kinases PKC-α, PKC-ε, and MAPK (ERK1/2); and the metabolically sensitive gene TXNIP, and their modulation by NOX5. EGR-1 is linked to renal inflammation and fibrosis, with Egr-1–deficient mice exhibiting improved renal function and reduced structural injury (34). EGR-1 has also been reported to be a potential transcription activator of NOX4 in DKD (34–36). The involvement of PKC-α in albuminuria development in DKD was modulated by both NOX4 and NOX5 (7,9,17,18). Similar to PKC-α, an association between oxidative stress and increased expression of renal PKC-ε has also been reported in a rat model of DKD (37). In addition, activation of the MAPK pathway including ERK1/2 is widely reported in DKD (38).

In our study, genetic deletion or pharmacological inhibition of NOX4 attenuated the increase in renal Egr-1, Pkc-α, Pkc-ε, and Erk1 expression seen in diabetic mice. However, upregulation of Egr-1, Pkc-α, Pkc-ε, and Erk1 occurred in the presence of NOX5 expression, suggesting a direct role for NOX5 in modulating their expression in DKD. Consistent with these findings, NOX5 KD in human renal cells led to downregulation of EGR-1 and PKC-α in response to high glucose concentration. The metabolic gene TXNIP increases ROS by binding and attenuating the antioxidant action of thioredoxin (39–41) in DKD. Increased renal TXNIP has been reported in human diabetic kidneys (39), whereas Txnip-deficient diabetic mice had reduced albuminuria and renal fibrosis via decreased ROS and Nox4 expression (41). Interestingly, despite Nox4 deletion, Txnip was upregulated in NOX5-expressing mice, in the presence or absence of diabetes, suggesting a direct role for NOX5 in regulating TXNIP in DKD. Indeed, our study findings indicate that both NOX4 and NOX5 regulate these key factors and that blockade of NOX4 alone cannot attenuate these molecules in the presence of NOX5 expression. This is clearly relevant in a human context where both NOX4 and NOX5 are expressed endogenously.

In DKD, renal ROS increase cellular proliferation and differentiation via modulation of genes, including the cell-proliferation markers Ki-67 and PCNA (4,42). Moreover, the growth factor CTGF enhances cell migration, proliferation, and ECM remodeling in DKD (43,44). In addition, activation of both CTGF (45) and α-SMA (46) in the kidney induces epithelial to mesenchymal transition, a process that may contribute to renal fibrosis, including in DKD (47). In this study, despite Nox4 deletion or inhibition, expression of NOX5 enhanced the renal upregulation of Ki67, Pcna, Ctgf, and a-SMA in diabetic mice, suggesting a key role for NOX5 in both renal cell proliferation and fibrosis. Furthermore, independent of NOX4, diabetic mice expressing NOX5 had more pronounced renal structural injury, with a further increase in mesangial area, glomerular collagen IV deposition, and enhanced glomerulosclerosis than what we observed in diabetic mice without the NOX5 isoform. These overt structural findings seen in NOX5-expressing diabetic mice occurred as a result of increased expression of genes implicated in fibrosis, including collagen IV and fibronectin. Moreover, silencing of NOX5, but not NOX4, attenuated upregulation of profibrotic genes induced by high glucose concentrations, including Ctgf and Collagen I and III, consistent with the dominant role of the pro-oxidant enzyme NOX5 versus NOX4 in the context of human DKD.

Previously, we demonstrated that blockade of NOX4 attenuated albuminuria in diabetic mice (5,7). In addition, Tg mice expressing NOX5 in podocytes (3), SMCs (17,18), and endothelial cells (18) led to enhanced albuminuria in diabetes. In this study, despite Nox4 deletion or pharmacological inhibition, albuminuria remained elevated in diabetic mice, with NOX5 expression further exacerbating albuminuria in diabetes. Because NOX4 blockade failed to reduce albuminuria in the presence of NOX5, these findings emphasize the importance of NOX5 over NOX4, which is a particularly relevant finding in regard to human DKD with expression of both these isoforms. In DKD, renal vascular endothelial growth factor (VEGF) is positively correlated with albuminuria and endothelial dysfunction (48,49). We previously demonstrated a link between NOX4 and VEGF in DKD (5,7). Recently, we also found an association between NOX5 and VEGF, albeit in diabetic retinopathy (19). In this current study, NOX5 expression modulated Vegf expression in the kidney independently of NOX4. In addition, correlations between PKC-α activation and VEGF expression, along with albuminuria, have also been reported in DKD in studies in which diabetic Pkc-α KO mice had reduced albuminuria and less renal VEGF expression (50,51). It appears that NOX5 is directly involved in modulating both PKC-α and VEGF expression, which perhaps explains the association between NOX5 and enhanced albuminuria in our models. An association between NOX4 and thickening of the GBM, a classical ultrastructural abnormality in DKD (52), has been demonstrated with attenuation of GBM thickening in podocyte-specific Nox4-deficient diabetic mice (7). Furthermore, Holterman et al. (3) showed increased GBM thickening in podocyte-specific NOX5 Tg mice with diabetes. Importantly, independent of NOX4, we found a significant increase in GBM thickness in NOX5-expressing diabetic mice.

In conclusion, independent of the NOX4 pathway, the progression of DKD, as evidenced by enhanced renal ROS, inflammation, elevated glomerulosclerosis, and albuminuria in association with ROS-sensitive key factors, EGR1, MAPK (ERK1/2), PKC-α/ε and TXNIP, is accelerated in the presence of NOX5 expression (Supplementary Fig. 6). Hence, in this study, we provide key evidence for a predominant role for NOX5 in DKD. In addition, we highlight that the only clinically evaluated NOX inhibitor, GKT137831, did not effectively inhibit NOX5-mediated renal injury in DKD. Thus, we clearly demonstrate the predominant role of NOX5 as a prooxidant enzyme in human DKD and provide the impetus for the development of an NOX5-specific inhibitor to alleviate the increasing burden of DKD.

Acknowledgments. The authors thank Professor George Jerums (Department of Endocrinology, Austin Health, University of Melbourne) and Professor Richard J. MacIsaac (Departments of Endocrinology and Diabetes, St Vincent's Hospital and University of Melbourne, Australia) for helping us obtain human kidney biopsy specimens; and E. Lastavec and M. Lyttle (Department of Diabetes, Central Clinical School, Monash University) for experimental animal handling and technical support.

Funding. This work was supported by National Health and Medical Research Council of Australia (NHMRC) project grant APP100585 as well as by Juvenile Diabetes Research Foundation (JDRF) grant 2-SRA 2014-259-Q-R. K.J.-D. and M.E.C. received Senior Research Fellowships from the Australian NHMRC (grants APP1059124 and APP1078808), and J.C.J. received Early Career Research Fellowships from Australian NHMRC (APP1126169) and JDRF (201302918). Genkyotex provided the NOX1/4 inhibitor GKT137831.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. J.C.J. and K.J.-D. conceived and designed the research. J.C.J., A.D., J.G., A.C., S.U., J.A.Ø., J.O., and C.E.H. performed the experiments and analyzed the data. A.S., D.A.P., E.I.E., and M.T.C. collected human kidney biopsy samples and gave scientific input. J.C.J. and K.J.-D. wrote and reviewed the manuscript. H.H.H.W.S., M.E.C., R.M.T., and C.R.K provided intellectual input and reviewed the manuscript. J.C.J. and K.J.-D. are the guarantors of this work and, as such, had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. This study was presented in oral form at the 56th European Association for the Study of Diabetes Annual Meeting 2020 (virtual), Vienna, Austria, 21–25 September 2020.