Overexpression of Nrf2 in Renal Proximal Tubular Cells Stimulates Sodium–Glucose Cotransporter 2 Expression and Exacerbates Dysglycemia and Kidney Injury in Diabetic Mice

Under physiological conditions, sodium–glucose cotransporter 2 (SGLT2) in the S1/S2 segments of renal proximal tubules (RPTs) mediates resorption of >90% of the glucose filtered by the glomerulus; SGLT1 in the late RPT (S2/S3 segments) resorbs the remaining glucose (1,2). In diabetes, excessive glucose uptake through SGLT2 may contribute to glucose toxicity, hyperfiltration, and glomerular injury via a tubuloglomerular feedback mechanism (3). Thus, SGLT2 inhibition has the potential to reduce glucose toxicity, hyperfiltration, and renal injury in diabetes. Indeed, the cardio- and renoprotective effects of SGLT2 inhibition have now been documented in large clinical trials in patients with diabetes, irrespective of whether they have chronic kidney disease (CKD) (4–7).

In addition to lowering sodium and glucose resorption, SGLT2 inhibition or knockout attenuates oxidative stress, inflammatory, and fibrotic pathways and improves renal oxygenation and glomerular hyperfiltration in the diabetic kidney (8–13). Elevated SGLT2 expression and activity in RPTs have been reported in preclinical models of diabetes (14,15) and in patients with diabetes (16,17). However, the mechanisms underlying SGLT2 upregulation have not been fully elucidated.

Nuclear factor erythroid 2–related factor 2 (NRF2) functions as a master regulator of redox balance and is important in cellular cytoprotective responses (18). The effect of NRF2 activation, however, is controversial in animals and humans with diabetes (18–23). Studies in diabetic rodents with bardoxolone methyl (BM) analogs (Nrf2 activators) reported antidiabetic effects (19,20), whereas other studies reported that BM analogs increased albuminuria and systolic blood pressure (SBP) in diabetic Zucker obese rats (21) and promoted atherosclerosis and kidney injury in diabetic apoE^−/− mice (22). The Trial To Determine the Effects of Bardoxolone Methyl on eGFR in Patients With Type 2 Diabetes and Chronic Kidney Disease (BEAM) in patients with type 2 diabetes (T2D) with CKD reported reductions in serum creatinine level and slight increases in glomerular filtration rate (GFR) (23), suggesting renoprotection. However, the subsequent phase 3 Bardoxolone Methyl Evaluation in Patients With Chronic Kidney Disease and Type 2 Diabetes (BEACON) in patients with T2D and advanced CKD was discontinued after 9 months because of increased mortality, heart failure rates, and albuminuria; furthermore, BEACON did not result in favorable effects on end-stage kidney disease (24). We reported that global knockout of Nrf2 (Nrf2^−/−) lowered SBP and urinary albumin-to-creatinine ratio (ACR) and inhibited angiotensinogen (Agt) expression in RPTs of Akita Nrf2^−/− mice (25).

In the current study, we hypothesized that Nrf2 overexpression in RPT cells (RPTCs) would stimulate Sglt2 expression and exacerbate hyperglycemia, SBP, and kidney injury in Akita Nrf2^−/−/Nrf2^RPTC transgenic (Tg) mice. We also report the validation of a putative NRF2-responsive element (NRF2-RE) in the mouse and human SGLT2 promoter.

d-Glucose, trigonelline (a nonspecific inhibitor of Nrf2) (25,26), and oltipraz (a specific Nrf2 activator) (27) were purchased from Sigma-Aldrich Canada Ltd. (Oakville, Ontario, Canada). Normal glucose (NG) (5 mmol/L d-glucose), DMEM (Cat. No. 12320), FBS, and the expression vector pcDNA3.1 were procured from Invitrogen, Inc. (Burlington, Ontario, Canada). HK2 (an immortalized human RPTC line) (Cat. No. CRL-2190) was purchased from ATCC (Manassas, VA). pGL4.20 vector containing a luciferase reporter was obtained from Promega (Sunnyvale, CA). The plasmid, pCI-HA NRF2 containing human NRF2 cDNA was obtained from Dr. Donna D. Zhang (University of Arizona, Tucson, AZ). The NRF2 cDNA was subcloned into pcDNA3.1 plasmid via Kpn1 and Not1 enzyme restriction sites. Mouse Sglt2 promoter (N-1,952/N+684) was amplified by PCR with specific primers (Supplementary Table 1) from pGEM-Sglt2–5pr-mut plasmid (28) (obtained from Dr. Isabelle Rubera, University of Nice-Sophia Antipolis, Nice, France) and then inserted into pGL4.20 plasmid at KpnI and Xho1 restriction sites. Human SGLT2 promoter (N-1,986/N+17) was amplified from HK2 genomic DNA by PCR with specific primers (Supplementary Table 1) as previously described (29). QuikChange II Site-Directed Mutagenesis Kits and LightShift Chemiluminescent Electrophoretic Mobility Shift Assay (EMSA) Kits were procured from Agilent Technologies (Santa Clara, CA) and Thermo Fisher Scientific (Life Technologies Inc., Burlington, Ontario, Canada), respectively. Primer biotin-labeling kits were supplied by Integrated DNA Technologies, Inc. (Coralville, IA).

Chromatin immunoprecipitation (ChIP) assay was performed using the SimpleChIP Plus Sonication Chromatin IP Kit (#56383; Cell Signaling Technologies) with slight modification as previously described (30). Briefly, 70–80% confluent HK2 cells were transfected with or without pCMV-Myc-rat Nrf2 cDNA or pcDNA3.1-human NRF2 cDNA plasmid for 24 h and then crosslinked with formaldehyde. Chromatin was fragmented by sonication and incubated with ChIP-grade anti-NRF2 or anti-Myc or anti-Histone H3 antibody or rabbit IgG (Supplementary Table 2). DNA was purified by spin column and used to amplify the SGLT2 promoter region containing the putative NRF2-RE or the RPL30 exon3 (internal control) by PCR with specific primers (Supplementary Table 1).

Oligonucleotides were synthesized by Integrated DNA Technologies. Scrambled Silencer Negative Control siRNA (sc-37007) and NRF2 siRNAs (309757) were obtained from Santa Cruz Biotechnology (Dallas, TX) and Dharmacon (Ottawa, Ontario, Canada), respectively. Restriction and modifying enzymes were purchased from Invitrogen, Roche Biochemicals, Inc. (Dorval, Quebec, Canada), and GE Healthcare Life Sciences (Baie d’Urfé, Quebec, Canada). The antibodies used are listed in Supplementary Table 2.

Tg mice specifically overexpressing rat Nrf2-Flag in their RPTCs were generated using a strategy similar to the method that we described previously for the generation of Agt^RPTC, Cat^RPTC, hnRNP F^RPTC, and Bmf^RPTC Tg mice (31–34). The non-Akita wild type (WT) and Akita mice were born at and obtained from the same breeders. Akita Nrf2^−/− mice were generated by cross-breeding female homozygous Nrf2^−/− mice with male heterozygous Akita mice. Then, Akita Nrf2^−/−/Nrf2^RPTC Tg mice were generated by cross-breeding female Nrf2^RPTC Tg mice (C57BL/6) with male Akita Nrf2^−/− mice (C57BL/6) (25).

Male (12-week-old) non-Akita WT, Akita, Akita Nrf2^−/−, and Akita Nrf2^−/−/Nrf2^RPTC Tg mice (nine mice per group) were studied. All animals received standard mouse chow and water ad libitum. Animal care and procedures were approved by the Centre de Recherche, Centre Hospitalier de l’Université de Montréal (CHUM) animal care committee and followed the Principles of Laboratory Animal Care (National Institutes of Health Publication No. 85-23, revised 1985, https://grants1.nih.gov/grants/olaw/references/phspol.htm).

SBP was monitored with a BP-2000 tail-cuff pressure monitor (Visitech Systems, Apex, NC) in the morning at least two to three times every 2 weeks for 8 weeks (25,29–34). Each animal was accustomed to the procedure for at least 15–20 min per day for 5 days before the first SBP measurement at week 10. SBP values are presented as mean ± SEM of two to three determinations per mouse per group. GFR was estimated with FITC inulin, as recommended by the Animal Models of Diabetic Complications Consortium (https://www.diacomp.org) with slight modifications (25,33,34).

Blood glucose levels were measured in mice with the Accu-Chek Performa system (Roche Diagnostics, Laval, Quebec, Canada) throughout the study period and with a glucose colorimetric detection kit (Cayman Chemical) at age 20 weeks. The mice were housed individually in metabolic cages for 6 h for urine collection. Urine samples were assayed for albumin and creatinine by albumin Exocell ELISA (Albuwell and Creatinine Companion; Ethos Biosciences, Philadelphia, PA) (25,29–34) and Ang II by ELISA (Bachem Americas, Torrence, CA) as described (35).

Following euthanasia, the kidneys were removed, decapsulated, and weighed. Left kidneys were processed for histology and immunostaining. Right kidneys were harvested for isolation of RPTs by Percoll gradient (25,32–34). Aliquots of freshly isolated RPTs from individual mice were immediately processed for total RNA and protein analysis.

Immunohistochemical staining was performed using the standard avidin-biotin-peroxidase complex method in four to five sections (4 μm thick) per kidney and four to six mouse kidneys per group (ABC Staining System; Santa Cruz Biotechnology). Periodic acid-Schiff (PAS) staining and Sirius Red staining were performed to assess tubulointerstitial fibrosis as previously described (25,32–35). Oxidative stress in RPTs was assessed by dihydroethidium (DHE) staining (Sigma-Aldrich Canada Ltd.) of frozen kidney sections (25,36). Semiquantification of the relative staining was done by using ImageJ software (https://rsb.info.nih.gov/ij).

Immunofluorescence (IF) staining for Sglt2 was performed on 4-μm tissue sections from mouse kidney fixed in formalin and embedded in paraffin followed by staining with Alexa Fluor 594–labeled secondary antibody (Invitrogen). Proximal tubules were identified by fluorescein-labeled lotus tetragonolobus lectin (LTL), a marker of RPT (37) (Vector Laboratories, Burlingame, CA). Image quantification and merge were assessed by ImageJ software. To quantify the amount of Sglt2 expression, the pixel intensity of Sglt2 was divided by LTL intensity. To calculate the average ratio, six sections per mouse, four to six mice per group, were analyzed.

Immortalized human RPTCs (HK2) were cultured as described (29,38). Plasmids pGL4.20 or pGL4.20 containing mouse Sglt2 (N-1952/N+684) or human SGLT2 promoter (N-1,986/N+17) were transiently or stably transfected into HK2 (29). HK2 or stable transformants at 75–85% confluence were synchronized overnight in DMEM containing NG and 1% depleted FBS and then cultured with various concentrations of oltipraz ± trigonelline for the indicated time periods for up to 24 h. SGLT2 expression was assessed with IF in DAPI-stained HK2. In separate experiments, HK2 stable transformants were transiently transfected with pcDNA3.1/NRF2 cDNA.

The mRNA levels of selected genes in RPTs were quantified by real-time quantitative PCR (RT-qPCR) with forward and reverse primers as previously reported (25,29,39–41) (Supplementary Table 1).

Western blotting (WB) was performed, as described previously (25,29,39–41). Relative densities of Agt, Nrf2, Sglt2, and β-actin bands were quantified by computerized laser densitometry (ImageQuant 5.1 software; Molecular Dynamics, Sunnyvale, CA).

Nephrectomy specimens (paraffin sections) for immunostaining were obtained from the Department of Pathology, CHUM. The study was approved by the CHUM clinical research ethics committee. All patients provided written informed consent (general) for the use of their kidney tissue and clinical data in research studies. The clinical characteristics of eight patients (four without diabetes and four with T2D) are shown in Supplementary Table 3, as published previously (41,42). All patients had undergone nephrectomy for kidney cancer.

Data are reported as mean ± SEM. Statistical analysis was performed with Student t test or one-way ANOVA with the Bonferroni correction, as appropriate, using GraphPad Prism 5.0 software (https://www.graphpad.com/prism/Prism.htm). P < 0.05 was taken as statistically significant.

All data sets that were generated or analyzed during the current study are included in the published article (and its online supplementary files).

A schema for generating the Nrf2 Tg mice is depicted in Fig. 1A. PCR analysis confirmed selective expression of the Nrf2-Flag Tg in the kidney and RPTs of male Nrf2^RPTC Tg mice but was undetectable in the kidneys and other organs of non-Tg mice (Fig. 1B) or Akita mice (Supplementary Fig. 1). The Nrf2-Flag Tg was detected in isolated RPTs of Akita Nrf2^−/−/Nrf2^RPTC Tg mice but not in WT, Akita, or Akita Nrf2^−/− mice (Fig. 1C). A mutated ins2 gene was detected in Akita, Akita Nrf2^−/−, and Akita Nrf2^−/−/Nrf2^RPTC Tg mice but not in WT mice (Fig. 1D).

Nrf2 expression was significantly higher in Akita versus WT and was nondetectable in Akita Nrf2^−/− mice and partially restored in Akita Nrf2^−/−/Nrf2^RPTC Tg mice by WB (Fig. 1E) and immunostaining (Fig. 1F). NAD(P)H:quinone oxidoreductase 1 (NQO-1) is a downstream target gene of Nrf2 activation (18). NQO-1 immunostaining was higher in RPTCs from Akita than WT mice and Akita Nrf2^−/− mice and partially restored in Akita Nrf2^−/−/Nrf2^RPTC Tg mice (Fig. 1G). In contrast, Kelch-like ECH-associated protein 1 (Keap1) immunostaining did not differ in these groups (Fig. 1H). These observations were confirmed by semiquantitation (Fig. 1I–K, respectively).

We detected significantly higher serum glucose levels in Akita than in WT mice, lower glucose levels in Akita Nrf2^−/− mice versus Akita mice, and elevated glucose levels in Akita Nrf2^−/−/Nrf2^RPTC Tg versus Akita Nrf2^−/− mice at age 20 weeks by colorimetric kit (Table 1). No significant differences of blood glucose were detected among Akita, Akita Nrf2^−/−, and Akita Nrf2^−/−/Nrf2^RPTC Tg mice by a glucometer (Supplementary Fig. 2A). Longitudinal (Supplementary Fig. 2B) and cross-sectional SBP measurements (Table 1) documented significantly higher SBP in Akita than in WT mice, whereas SBP was significantly lower in Akita Nrf2^−/− than in Akita mice. SBP was slightly higher in Akita Nrf2^−/−/Nrf2^RPTC Tg mice than in Akita Nrf2^−/− mice, without reaching statistically significant differences. Body weight (BW) of WT mice was significantly higher than that of Akita, Akita Nrf2^−/−, and Akita Nrf2^−/−/Nrf2^RPTC Tg mice at the age of 20 weeks, with no statistically significant differences among the latter three groups (Table 1 and Supplementary Fig. 2C). Kidney weight (KW) and the KW/tibial length and GFR/BW ratios were also significantly increased in Akita mice versus WT, whereas these parameters were lower in Akita Nrf2^−/− mice. GFR/BW was significantly higher in Akita Nrf2^−/−/Nrf2^RPTC Tg than in Akita Nrf2^−/− mice (Table 1). Akita mice had significantly elevated urinary ACR levels versus WT and Akita Nrf2^−/− mice. Nrf2 overexpression significantly increased ACR in Akita Nrf2^−/−/Nrf2^RPTC Tg versus Akita Nrf2^−/− mice (Table 1). Fractional excretion of glucose was the highest in Akita Nrf2^−/− mice among the four groups (Table 1).

Double IF of kidney sections with an anti-Sglt2 antibody and LTL-FITC antibody confirmed significantly higher Sglt2 expression in RPTs from Akita mice than from WT mice (Fig. 2A). Akita mice showed higher expression of Agt in RPTs than WT mice (Fig. 2B). In contrast, Sglt2 and Agt expression was significantly lower in RPTs from Akita Nrf2^−/− mice than Akita mice, and this was reversed in Akita Nrf2^−/−/Nrf2^RPTC Tg mice. WB for Sglt2 and Agt (Fig. 2C and D) and RT-qPCR of their respective mRNAs from isolated RPTs (Fig. 2E and F) confirmed these changes.

PAS staining revealed more pronounced proximal tubular cell atrophy and tubular luminal dilatation with accumulation of cell debris in Akita mice than in WT mice and confirmed by semiquantitation of morphological changes (Fig. 3A and Table 1). These abnormalities were attenuated in Akita Nrf2^−/− mice and partially reversed in Akita Nrf2^−/−/Nrf2^RPTC Tg mice.

Staining for DHE (Fig. 3B) was significantly increased in the kidneys from Akita mice versus WT, but it did not differ from that in Akita Nrf2^−/− and Akita Nrf2^−/−/Nrf2^RPTC Tg mice. Increased Sirius Red (Fig. 3C) staining and transforming growth factor-β1 (TGF-β1) immunostaining (Fig. 3D) were noted in tubules near glomeruli in Akita mice versus WT but were lower in Akita Nrf2^−/− mice and similar to Akita Nrf2^−/−/Nrf2^RPTC Tg mice. These changes were confirmed by semiquantification of staining of DHE (Fig. 3E) and Sirius Red (Fig. 3F) and of TGF-β1 mRNA expression by RT-qPCR (Fig. 3G). These changes were also associated with significant increases in fibronectin 1 (Fn1) mRNA expression (Fig. 3H) in Akita Nrf2^−/−/Nrf2^RPTC Tg mice versus Akita Nrf2^−/− mice. Cat mRNA expression was significantly lower in Akita mice than in WT mice, whereas Nrf2 overexpression failed to increase Cat expression in Akita Nrf2^−/−/Nrf2^RPTC Tg mice versus Akita Nrf2^−/− mice (Fig. 3I). In contrast, Nox4 mRNA expression was higher in Akita than in WT mice but did not differ in Akita Nrf2^−/− and Akita Nrf2^−/−/Nrf2^RPTC Tg mice.

Oltipraz increased SGLT2 expression in HK2, and this was inhibited by trigonelline (Fig. 4A and B). Oltipraz treatment also increased expression of both NRF2 and SGLT2 mRNA in a concentration-dependent manner (Supplementary Fig. 3A and B, respectively). Transfection with NRF2 siRNA inhibited the stimulatory effect of oltipraz on NRF2 and SGLT2 mRNA expression (Fig. 4C and D), whereas scrambled siRNA had no effect. Furthermore, oltipraz stimulated mouse and human SGLT2 promoter activity, and its stimulatory effects were inhibited by trigonelline (Fig. 4E and F). Transient transfection of the plasmid pcDNA-NRF2 cDNA significantly stimulated NRF2 and SGLT2 mRNA expression (Fig. 4G and H) as well as mouse and human SGLT2 promoter activity in HK2 (Fig. 4I and J).

To validate the effects of the putative Nrf2-RE on the mouse Sglt2 promoter (N-1,527/N-1,516, 5′-CTGACACTGCT-3′) and human SGLT2 promoter (N-1,316/N-1,305, 5′-GTGACACAGCA-3′) (Note: the putative mouse and human NRF2-RE motifs are homologous to the consensus Nrf2 motif [a/gTGACt/ac/aAGCA] [44]), different lengths of Sglt2 promoters or SGLT2 promoters were transiently transfected into HK2 and then cultured ± oltipraz in NG medium. Sglt2 promoter (N-1,952/N+684) (Fig. 5A) and SGLT2 promoter (N-1,986/N+17) (Fig. 5B) exhibited 29-fold and 28-fold increases, respectively, compared with control plasmid pGL4.20 in HK2. Deletion of nucleotides N-1,952 to N-1,248 in Sglt2 promoter and N-1,986 to N-1,285 in SGLT2 promoter respectively reduced the promoter activity to 24-fold and increased it 14-fold versus pGL4.20. Further deletion of nucleotides N-1,952 to N-235 in Sglt2 promoter and N-1,986 to N-194 in SGLT2 promoter respectively reduced the promoter activity to 16-fold and increased it 4-fold versus pGL4.20. Interestingly, the activity of Sglt2 promoter and SGLT2 promoter was further increased by 1.7-fold and 1.4-fold, respectively, in HK2 in the presence of oltipraz (Fig. 5C and D). Oltipraz did not increase the promoter activity of other fusion genes. Furthermore, deletion of the putative NRF2-RE, N-1,527 to N-1,516 in Sglt2 promoter (Fig. 5E) and N-1,316 to N-1,305 in SGLT2 promoter (Fig. 5F) abolished the stimulatory effect of oltipraz.

EMSA revealed that the double-strand DNA fragment nucleotides N-1,535 to N-1,505 containing the core Nrf2-RE (N-1,527 to N-1,516) of Sglt2 promoter (Fig. 6A) and N-1,323 to N-1,296 containing the core NRF2-RE (N-1,316 to N-1,305) of SGLT2 promoter (Fig. 6B) bind to nuclear proteins from HK2 cells, which can be displaced by respective WT DNA but not by mutated DNA fragments. ChIP assays were used to test whether endogenous NRF2 interacts with the NRF2-RE of the SGLT2 promoter in vitro. Figure 6 displays the PCR product of pulled down DNA by anti-NRF2 antibody with primers specific to the SGLT2 promoter in HK2 cells transfected with or without pCMV-Myc-rat Nrf2 cDNA (Fig. 6C) or pcDNA3.1/NRF2 cDNA (Fig. 6D). An ∼224-base pair (bp) DNA fragment was generated in naive HK2 (lane 2), but no similar DNA fragment was generated with pulldown by anti-Histone H3 (lane 3) and rabbit IgG (lane 4). The ∼224-bp DNA fragment was further enhanced in HK2 when transiently transfected with pCMV-Myc-rat Nrf2 cDNA or pcDNA3.1/NRF2 cDNA (lane 6). Again, no similar DNA fragment was generated with pulldown by anti-Histone H3 (lane 7) and rabbit IgG (lane 8). On the other hand, ∼161 bp of hRPL was generated with DNA pulldown by anti-Histone H3 (lanes 3 and 7) but not by rabbit IgG (lanes 4 and 8).

We detected more pronounced immunostaining for NRF2 and IF for SGLT2 in normal areas from nephrectomy specimens from patients with kidney cancer and diabetes compared with patients without diabetes (Fig. 7A and B). KEAP-1 expression appeared to be similar in kidney specimens from patients with and without diabetes (Supplementary Fig. 4A). These observations were confirmed by semiquantitation of NRF2, SGLT2, and KEAP1 staining (Fig. 7C and D and Supplementary Fig. 4B, respectively).

Our results demonstrate that selective overexpression of Nrf2 in RPTCs of Akita Nrf2^−/− mice effectively upregulates Sglt2 expression, resulting in elevation of blood glucose levels, GFR, ACR, and tubulointerstitial fibrosis. Consistently, in cultured HK2, pharmacological stimulation of NRF2 with oltipraz or transfection with NRF2 cDNA stimulated SGLT2 expression and its promoter activity, and their effects were reversed by trigonelline and NRF2 siRNA. NRF2 binds to NRF2-RE in the SGLT2 promoter as revealed by EMSA and ChIP assay. Furthermore, specimens from kidneys of patients with diabetes exhibited higher expression of NRF2 and SGLT2 in RPTs than from kidneys of patients without diabetes. Our findings identify a link by which NRF2 activation by oxidative stress (secondary to hyperglycemia) stimulates SGLT2 expression and activation, leading to further elevations in blood glucose, GFR, and progression of kidney injury in diabetes.

Akita mice, an autosomal dominant model of spontaneous type 1 diabetes (T1D) with a mutated Ins2 gene, closely mimic human T1D, with renal and cardiac morphological changes characteristic of early to moderately advanced human T1D (45,46). We found significantly lower blood glucose levels in Akita Nrf2^−/− mice compared with Akita mice. However, there were no differences between Akita mice and Akita Nrf2^−/−/Nrf2^RPTC Tg mice at 20 weeks of age, suggesting that the presence of Nrf2 in the proximal tubules prevented this amelioration.

We previously reported that Nrf2 stimulates RPTC Agt transcription through binding to an NRF2-RE in the Agt promoter (47). While the higher RPTC Agt expression in Akita Nrf2^−/−/Nrf2^RPTC Tg compared with Akita Nrf2^−/− mice was associated with a modest increase in SBP (3.5 mmHg), it did not reach statistical significance. Similarly, we detected higher, though statistically nonsignificant, urinary Ang II levels in Akita Nrf2^−/−/Nrf2^RPTC Tg compared with Akita Nrf2^−/− mice (Table 1), consistent with lack of significant effect on SBP. However, additional studies are needed to clarify why SBP and urinary Ang II levels did not significantly differ between Akita Nrf2^−/−/Nrf2^RPTC Tg and Akita Nrf2^−/− mice.

In the current study, we detected increased Nrf2 and NQO-1 expression in RPTs of 20-week-old Akita mice compared with WT mice. These were associated with marked increases in DHE staining. DHE staining, however, was not different in Akita Nrf2^−/− mice compared with Akita Nrf2^−/−/Nrf2^RPTC Tg mice. We do not presently understand why Nrf2 overexpression did not lead to attenuation of oxidative stress in Akita Nrf2^−/−/Nrf2^RPTC Tg mice compared with Akita Nrf2^−/− mice. Previous studies from our group (39,47,48) and others (49–51) consistently showed enhanced NADPH oxidase activity and Nox4 expression with reduced Cat expression and activity in diabetic rodents, indicating that hyperglycemia would alter relative expression and activity of Nox4 and Cat, thereby enhancing reactive oxygen species generation in the kidney (47,48).

Akita mice exhibited increased tubulointerstitial fibrosis compared with WT mice, which was less apparent in Akita Nrf2^−/− mice. Increased tubulointerstitial fibrosis was also observed in Akita Nrf2^−/−/Nrf2^RPTC Tg mice compared Akita Nrf2^−/− mice. The functional relation between Nrf2 and tubulointerstitial fibrosis is incompletely understood. One possibility is that Nrf2 stimulates Agt expression and Ang II production, which, in turn, would stimulate TGF-β1 and subsequently enhance the expression of extracellular matrix proteins and profibrotic genes in RPTCs. Indeed, we detected higher Agt, TGF-β1, and Fn1 expression in RPTs of Akita mice than in WT. Most recently, Rush et al. (52) reported that genetic or pharmacologic Nrf2 activation increases proteinuria and kidney injury in several mouse models of CKD, consistent with our observation that Nrf2 overexpression or activation may exacerbate kidney injury in diabetic mice.

The mechanism by which NRF2 overexpression leads to upregulation of renal SGLT2 gene expression in diabetes remains unclear. We demonstrated that oltipraz and NRF2 cDNA transfection increased SGLT2 promoter activity. Deletion of the putative NRF2-RE markedly reduces oltipraz upregulation of mouse and human SGLT2 promoter activity in HK2. Moreover, biotin-labeled mouse and human NRF2-RE specifically binds to nuclear proteins. Importantly, our ChIP assays confirmed NRF2 interaction with SGLT2 promoter loci. Taken together, these data demonstrate that NRF2 binds to NRF2-RE to stimulate SGLT2 transcription in vivo.

Our results may have clinical implications that also may help to explain the harmful effects of NRF2 activation observed with BM in patients with T2D and CKD in the BEACON trial (24). We speculate that this adverse action may be attributed to SGLT2 and AGT upregulation by NRF2 in RPTs, leading to hyperfiltration, hypertension, and exacerbation of nephropathy and to adverse cardiac effects. Thus, the critical question of whether NRF2 activation may be harmful in patients with T2D patients and CKD warrants further investigation. Of note, the safety and efficacy of BM are currently being tested in ongoing phase 2/3 clinical trials on T1D/T2D (53).

In summary, our findings demonstrate that selective Nrf2 overexpression in RPTCs leads to upregulation of renal Sglt2 and Agt expression with subsequent increases in blood glucose and fibrotic gene expression, with resultant kidney injury in mice with diabetes. The results imply an important role of oxidative stress (hyperglycemia)–induced NRF2 expression and SGLT2 activation in the exacerbation of hyperglycemia and renal injury in diabetes. Our observations raise the possibility that selective targeting of NRF2 might provide a potentially novel approach toward prevention and reversal of diabetes-associated nephropathy.

Funding. This work was supported, in part, by grants from the Canadian Institutes of Health Research (MOP-84363 and MOP-142378 to J.S.D.C., PJT173512 to S.L.Z., and MOP-97742 to J.G.F.), the Kidney Foundation of Canada (KFOC170006 to J.S.D.C.), and the Natural Sciences and Engineering Research Council of Canada (RGPIN-2017-05615 to S.L.Z.). K.N.M. is a recipient of a fellowship from the Consortium de Néphrologie de l’Université de Montréal (2018) and from the American Society of Nephrology Ben J. Lipps Research Fellowship program (2019–2020).

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

Author Contributions. S.Z., C.-S.L., K.N.M., A.G., X.Z., I.C., J-F.C., J.E., and J-B.L. contributed to the in vivo and in vitro experiments and data collection. K.N.M., J.G.F., and J.R.I. contributed to the discussion and reviewed/edited the manuscript. S.L.Z. contributed to the data research and discussion of the manuscript. S.L.Z. and J.S.D.C. contributed as principal investigators to the study conception and design. J.S.D.C. drafted the manuscript. All authors approved the final version for publication. S.L.Z. and J.S.D.C. are guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in poster format at the Annual Meeting of the American Society of Nephrology, Washington, DC, 5–10 November 2019, and the Annual Meeting of the American Society of Nephrology, Denver, CO, 22–25 October 2020.