Genetic or Pharmacologic Nrf2 Activation Increases Proteinuria in Chronic Kidney Disease in Mice

Abstract

The nuclear factor erythroid 2 related factor 2 (Nrf2) pathway upregulates key cellular defenses. Clinical trials are utilizing pharmacologic Nrf2 inducers such as bardoxolone methyl to treat chronic kidney disease, but Nrf2 activation has been linked to a paradoxical increase in proteinuria. To understand this effect, we examined genetically engineered mice with elevated Nrf2 signaling due to reduced expression of the Nrf2 inhibitor, Kelch-like ECH-associated protein-1 (Keap1). These Keap1^FA/FA mice lacked baseline proteinuria but exhibited increased proteinuria in experimental models evoked by adriamycin, angiotensin II, or protein overload. After injury, Keap1^FA/FA mice had increased glomerulosclerosis, nephrin disruption and shedding, podocyte injury, foot process effacement, and interstitial fibrosis. Keap1^FA/FA mice also had higher daytime blood pressures and lower heart rates measured by radiotelemetry. Conversely, Nrf2 knockout mice were protected from proteinuria. We also examined the pharmacologic Nrf2 inducer CDDO-Im. Compared to angiotensin II alone, the combination of angiotensin II and CDDO-Im significantly increased proteinuria, a phenomenon not observed in Nrf2 knockout mice. This effect was not accompanied by additional increases in blood pressure. Finally, Nrf2 was found to be upregulated in the glomeruli of patients with focal segmental glomerulosclerosis, diabetic nephropathy, fibrillary glomerulonephritis, and membranous nephropathy. Thus, our studies demonstrate that Nrf2 induction in mice may exacerbate proteinuria in chronic kidney disease.

Graphical Abstract

Introduction

Proteinuria is a major risk factor for the progression of chronic kidney disease (CKD). The abnormal urinary leak of serum proteins, including albumin, is generally due to glomerular injury or dysfunction and is a biomarker of many kidney diseases.¹ Current treatment of proteinuric individuals targets reduction of systemic and intraglomerular pressures through renin-angiotensin-aldosterone system (RAAS) blockade.^{2, 3} However, incomplete protection and adverse side effects have led to continuing testing of novel agents.

Nuclear factor erythroid 2 related factor 2 (Nrf2) is a transcription factor that regulates a multitude of target genes with roles in antioxidant and anti-inflammatory actions, electrophile detoxification, and general cytoprotection. Kelch-like ECH-associated protein-1 (Keap1) is an adaptor protein for the Cullin-3 ubiquitin ligase and a key cytoplasmic repressor of Nrf2. Keap1 interaction with Nrf2 leads to Nrf2 polyubiquitination and subsequent proteasomal degradation. In the presence of oxidative stress or electrophiles, key “sensor” sulfhydryl groups on Keap1 are modified, disrupting the degradation process and enhancing Nrf2 nuclear translocation and target gene transcription.⁴ Downstream targets include NAD(P)H: quinone oxidoreductase 1 (Nqo1) and the glutathione-S-transferases (GSTs).^{5, 6} Keap1 alkylating agents such as bardoxolone methyl (CDDO-Me), 1-[2-cyano-3-,12-dioxooleana-1,9(11)-dien-28-oyl]imidazole (CDDO-Im), and sulforaphane protect against acute kidney injury,^7-9 while Nrf2 null mice are sensitized to injury.^{10, 11}

Bardoxolone methyl is a synthetic triterpenoid that not only upregulates Nrf2 activity but is also anti-inflammatory, perhaps through additional actions on nuclear factor-κB (NF-κB)¹². This drug was used to treat diabetic nephropathy in the phase 2 Bardoxolone Methyl Treatment: Renal Function in CKD/Type 2 Diabetes (BEAM) and phase 3 Bardoxolone Methyl Evaluation in Patients with Chronic Kidney Disease and Type 2 Diabetes Mellitus: The Occurrence of Renal Events (BEACON) trials. While BEAM showed a promising increase in estimated glomerular filtration rate (eGFR), BEACON was terminated early due to serious adverse events.^{13, 14}

Both trials reported that CDDO-Me increased proteinuria. A post-hoc analysis showed an association between albuminuria and increases in eGFR.¹⁵ An explanation for an increase in glomerular filtration rate (GFR) may be increased glomerular surface area¹⁶ but concerns persist for increased renal plasma flow or intraglomerular pressure, which may be deleterious in chronic disease.^{17, 18}.

Bardoxolone methyl is currently in clinical trials for Alport’s Syndrome (CARDINAL, NCT03019185), IgA nephropathy, type 1 diabetic nephropathy, focal segmental glomerulosclerosis, and autosomal dominant polycystic kidney disease (PHOENIX, NCT03366337). An additional phase 3 study in patients with diabetic kidney disease is also being performed with a primary composite outcome as the time to onset of a 30% decline in eGFR or end-stage renal disease¹⁹ (AYAME, NCT03550443). Supporters advocate for continued testing with careful monitoring and patient selection,²⁰ but long-term deleterious effects remain a concern.¹⁷ In this context, additional preclinical studies would be invaluable.

In as much as drugs can have off-target effects, genetic manipulations provide a method to study the Keap1/Nrf2 pathway with greater precision. Since Keap1 knockout mice die soon after birth,²¹ we utilized Keap1 knockdown or hypomorphic mice (recently denoted as Keap1^FA/FA to differentiate from other floxed strains) that have an 80% reduction in Keap1 expression.⁴ These mice are viable without developmental or growth defects and have constitutively enhanced Nrf2 activity.²² Keap1^FA/FA mice are protected from CKD progression after ischemic acute kidney injury as well as unilateral ureteral obstruction,^{23, 24} but proteinuric diseases have not been studied comprehensively.

We show that Keap1^FA/FA mice are sensitized markedly to glomerular injury and proteinuria, characterized by podocyte depletion and an increase in foot process effacement and interstitial fibrosis. We also find that the bardoxolone analogue, CDDO-Im, similarly worsened proteinuria when administered to mice. These data suggest that Keap1/Nrf2-targeted therapies could potentiate glomerular injury in the setting of proteinuric nephropathy.

Results

Keap1 hypomorphs are not proteinuric at baseline

We measured albuminuria in adult wild type (WT) and Keap1^FA/FA mice in the absence of experimental injury. Both groups had low urine albumin-to-creatinine ratios (UACR) (Figure 1a). Keap1^FA/FA mice exhibited normal glomerular histology and nephrin abundance and distribution (Figure 1b). The number of Wilms Tumor 1-positive (WT1+) podocytes were also similar between groups (Figure 1c). The expression of nephrin was also unchanged (Figure 1d). Therefore, Nrf2 activation did not affect baseline glomerular function or histology.

Figure 1.

Keap1^FA/FA mice exhibit no proteinuria or glomerular injury at baseline. (a) Urinary albumin-to-creatinine ratios (UACR) reveal no significant difference between overall levels of proteinuria in WT versus Keap1^FA/FA. (b) Glomerular histology as well as numbers of WT1+ podocytes and nephrin staining was similar between WT and Keap1^FA/FA mice. (c) Quantification of the number of WT1+ cells shows no differences between groups. (d) Nephrin mRNA levels were also not significantly different (t-test).

Keap1 hypomorphs are sensitized to proteinuric injury

We exposed Keap1^FA/FA mice to three models of experimental proteinuria. To overcome the proteinuria resistance of the C57BL/6 background, we performed a unilateral nephrectomy in all mice prior to injury.²⁵ Adriamycin (doxorubicin) causes proteinuria and focal segmental glomerulosclerosis.²⁶ An intravenous dose of 18 mg/kg adriamycin induced progressive proteinuria (Figure 2a and b). Keap1^FA/FA mice had higher UACR than WT starting at day 14 and achieving statistical significance at 21 and 28 days.

Figure 2.

Keap1^FA/FA mice are sensitized to different proteinuria models. (a, b) WT and Keap1^FA/FA mice (aged 21 weeks) were exposed to a single dose of adriamycin (ADR, 18mg/kg) and UACR measured at the indicated timepoints. (c, d) An osmotic minipump releasing angiotensin II at a dose of 1.5 mg/kg/day was implanted subcutaneously in mice (aged 13 weeks) and UACR examined. (e, f) Bovine serum albumin was repeatedly injected i.p. at the indicated doses to induce protein overload, with UACR measured on day 11. Mice were 16 weeks old. * P < 0.05 using multiple t-tests with Holm-Sidak correction for multiple comparisons (b, d) or student’s t-test (f).

In separate experiments, Keap1^FA/FA and WT mice were exposed to continuous angiotensin II (AngII) infusion. We used a 1.5 mg/kg/day dose that induces hypertension, glomerular injury, and proteinuria.²⁷ This led to significant UACR increases in Keap1^FA/FA compared to WT mice (Figure 2c and d). Finally, we tested a protein overflow model using serial injections of bovine serum albumin (BSA)²⁸ and found that Keap1^FA/FA again had increased proteinuria (Figure 2e and f). By utilizing three experimental models of proteinuria, each one with unique pathogenesis, we demonstrate that reduced Keap1 levels generally worsen proteinuria.

Keap1 hypomorphic mice are more susceptible to glomerular injury

After 28 days, adriamycin-treated mice had more glomerulosclerosis than vehicle controls, with Keap1^FA/FA exhibiting more extensive injury than WT mice (Figure 3a). These mice exhibited more nephrin disruptions and fewer WT1+ nuclei per glomerulus (Figure 3a and b). Keap1^FA/FA also had lower WT1 and nephrin (Figure 3c and d) and higher urinary nephrin levels (e.g. nephrinuria) (Figure 3e), a biomarker of podocyte injury.²⁹ The proteinuria was not due to altered tubular protein handling, since megalin expression was not different between Keap1^FA/FA and WT mice (Figure 3f-g). The majority of the protein detected in the urine was identified as albumin by gel electrophoresis (Figure 3h). We also confirmed these findings in the AngII model. Keap1^FA/FA mice had increased glomerulosclerosis and nephrin disruption, fewer WT1+ cells per glomerulus, and increased interstitial fibrosis (Supplemental Figure 1).

Figure 3.

Keap1^FA/FA mice had increased glomerular injury in response to adriamycin. Keap1^FA/FA mice were treated with a single dose of adriamycin (ADR) and sacrificed at 28 days. (a) Glomerulosclerosis was identified in all ADR-treated animals via trichrome stain. The Keap1^FA/FA mice had worsened glomerular injury. The Keap1^FA/FA also had reduced numbers of WT1+ podocytes (green) and greater nephrin (red) disruption (arrowheads) after ADR. (b) Quantification of WT1+ cells per glomerulus. Levels of WT1 (c) and nephrin (d) mRNA were also reduced in ADR-treated Keap1^FA/FA. (e) Nephrinuria was increased in Keap1^FA/FA mice. (f, g) Megalin levels were similar between WT and Keap1^FA/FA mice. (h) Gel electrophoresis for urine samples showing albumin as the most abundant urinary protein. CTL, nonproteinuric control for reference. * P < 0.05, between groups as indicated, t-test or one-way ANOVA, as appropriate.

Using electron microscopy, we found that podocyte foot processes (FP) were normal at baseline in both the Keap1^FA/FA and WT groups (Figure 4a and b, e and f). Adriamycin induced FP effacement in WT mice (Figure 4c and g, arrowheads), but Keap1^FA/FA mice had significantly more effacement (Figure 4d and h, arrowheads). In some cases this encompassed large areas (Figure 4i and j, arrowheads). These findings show increased podocyte injury in Keap1^FA/FA mice following adriamycin treatment.

Figure 4.

Keap1^FA/FA mice have increased foot process effacement in response to adriamycin. Transmission electron microscopy (TEM) was performed in wild type (WT) vehicle (a, e), Keap1^FA/FA vehicle (b, f), WT adriamycin (ADR) (c, g), and Keap1^FA/FA ADR (d, h) mice. Both groups of vehicle-treated animals had normal foot processes. WT ADR had areas of focal foot process (FP) effacement (arrowheads), but Keap1^FA/FA had much larger areas of FP effacement (arrowheads). (i,j) An example of a severely effaced foot process from a Keap1^FA/FA mouse treated with ADR. These severe findings were more common in the mutant mice. (k) Quantification of average FP lengths were performed. NS, not significant. Bar equals 500 nm. A 2-way ANOVA indicated a significant main effect for treatment (P<0.0001) and for genotype (P=0.0278) but no difference in the interaction (P=0.3263). Posthoc testing via Holm-Sidak test indicated a significant difference between WT ADR and Keap1^FA/FA ADR.

Increased renal fibrosis in Keap1 hypomorphic mice after proteinuric injury

To examine whether the enhanced proteinuria is associated with permanent scarring, we examined interstitial fibrosis 28 days after adriamycin exposure. Keap1^FA/FA mice exhibited greater histologic fibrosis (Figure 5a and b) and expressed more fibronectin (Figure 5c-e) and type 1 and 3 collagens (Figure 5f and g). These data suggest that the induced proteinuria in Keap1^FA/FA mice is associated with permanent kidney injury.

Figure 5.

Keap1^FA/FA mice have increased interstitial fibrosis compared to wild type mice. (a) Renal fibrosis was evaluated with Masson’s trichrome (blue staining, arrowheads) and picrosirius red (pink staining, arrowheads) staining. Collagen fibers are birefringent after picrosirius red staining when viewed with polarized light. Keap1^FA/FA had greater fibrosis compared to wild type (WT) mice. (b) Semiquantitative scoring of fibrosis reveals greater collagen deposition in Keap1^FA/FA ADR compared to WT ADR mice. Fibronectin protein (c, d) and mRNA (e), and Col1a1 and Col3a1 mRNA (f, g) are also increased in Keap1^FA/FA ADR mice, consistent with renal fibrosis. Bar equals 50 μm. * P < 0.05, t-test.

Keap1^FA/FA mice exhibit higher blood pressures

Blood pressure (BP) is correlated with degree of proteinuria and BP reduction is recommended in management of proteinuria.¹ To examine whether BP plays a role in our proteinuria models, we utilized radiotelemetry in the hypertension-inducing AngII model. At baseline, Keap1^FA/FA versus WT mice had an elevated daytime systolic (SBP, 122 ± 1 vs 112 ± 2 mmHg), diastolic (DBP, 88 ± 1 vs 82 ± 2), and mean arterial BP (MAP, 102 ± 1 vs 95 ± 2) (Figure 6a-c). Nocturnal BP did not differ between groups. AngII significantly increased SBP, DBP, and MAP in both WT and Keap1^FA/FA mice, but BP did not differ between groups at days 5-7 and 12-14. Although WT mice exhibited a clear diurnal rhythm at days 19-21 and 26-28, Keap1^FA/FA did not. Daytime BP of Keap1^FA/FA mice was significantly higher than daytime BP of WT mice. Both strains exhibited a diurnal heart rate pattern (Figure 6d) but nighttime heart rate of Keap1^FA/FA mice was significantly lower than WT mice at baseline and throughout the AngII infusion. A table of numerical results is provided (Supplemental Table 1). While these overall BP differences were small, they are significant and may provide at least a partial explanation for proteinuria differences.

Figure 6.

Keap1^FA/FA exhibit elevated blood pressure (BP) and suppressed heart rate. Systemic BP was measured in wild-type (WT) (open circles) and Keap1^FA/FA (closed squares) mice using telemetry. Data was separated into daytime and nighttime cycles and averaged from continuous 6-hour measurement windows over several days. The response to AngII (1.5mg/kg/day) is shown. Systolic (a), diastolic (b), and mean arterial (c) BP are shown and reveal significant differences at baseline and at days 19-21 and 26-28 only during the day. (d) Heart rates also showed a nocturnal suppression in the Keap1^FA/FA regardless of AngII exposure. N=7-10 per group at each timepoint. # P < 0.05, student’s t-test on baseline data only. * P < 0.05, multiple t-tests with multiple comparison correction using Holm-Sidak method. Daytime and nighttime measures were analyzed separately for statistical purposes. Mice were 10-11 weeks of age. SBP, systolic blood pressure; DBP, diastolic blood pressure; MAP, mean arterial blood pressure.

Keap1^FA/FA mice do not exhibit elevated GFR

To determine if proteinuria in Keap1^FA/FA mice resulted from increased glomerular filtration, GFR was measured by the transcutaneous recording of FITC-sinistrin elimination. There were no significant differences between Keap1^FA/FA and wild type mice at baseline, and our AngII injury model reduced GFR equivalently in animals of both genotypes (Supplemental Figure 2a). Serum creatinine was also not significantly different between Keap1^FA/FA and wild type mice in the AngII model (Supplemental Figure 2b). These data suggest that the increased proteinuria in the Keap1^FA/FA mice occurs independently of changes in GFR.

Pharmacologic Nrf2 upregulation increases proteinuria

We examined whether pharmacologic Nrf2 activation would increase proteinuria in a manner similar to genetic manipulation. Since bardoxolone methyl is reportedly harmful to rodents due to accumulation of toxic metabolites,^30-32 we used the bardoxolone analogue, CDDO-Im. This drug, at an intense dose/route combination of 10 μmol/kg, i.p., leads to strong induction of Nrf2 signaling in whole kidney as determined by Nqo1 expression (Supplemental Figure 3) but did not result in proteinuria (data not shown) in the absence of other injury. We went on to compare glomerular expression of Nrf2 targets and found that CDDO-Im induced greater Nrf2 activation when compared to Keap1^FA/FA mice or untreated control WT mice (Figure 7a, b).

Figure 7.

Both CDDO-Im and Keap1 hypomorphism lead to strong induction of Nrf2 downstream targets in glomeruli. Glomeruli were isolated from WT mice, Keap1^FA/FA mice, or mice treated with 7 days of CDDO-Im (aged 9 weeks). mRNA levels of Nqo1 (a) and Gstm1 (b) were compared between groups. * P < 0.05, one-way ANOVA.

We administered CDDO-Im to mice concomitantly receiving AngII. To minimize bias,³³ we exposed all mice to AngII and divided them into two groups with equivalent proteinuria based on urine from day 7. We then randomized the groups to receive either CDDO-Im or vehicle injections from days 9 through 16 (Figure 8a). WT mice treated with CDDO-Im had significantly increased proteinuria compared to vehicle-treated animals at day 16 (Figure 8b). Twelve days after withdrawal of CDDO-Im (day 28), proteinuria declined, leading to no residual difference between groups. We repeated the experiment with additional timepoints and found that proteinuria began to increase relative to vehicle-treated controls within 3 days of CDDO-Im exposure (day 12) and appeared to normalize by day 24 (Figure 8c). We evaluated glomerular expression of nephrin and podocin and found suppression of nephrin at the day 16 timepoint (time of maximal proteinuria) and enhanced nephrin and podocin at day 24 (as proteinuria was improving) (Figure 8d and e). CDDO-Im treatment was accompanied by high glomerular expression of Nqo1 and Gstm1, indicating successful engagement of Nrf2 signaling (Figure 8f and g). At day 16, we observed a reduction in WT1+ cells and nephrin protein (Figure 8h-j) and an increase in nephrinuria (Figure 8k). Podocyte FP effacement was also increased (Figure 8l).

Figure 8.

CDDO-Im enhances AngII-induced proteinuria. (a) Wild type C57BL/6 mice were all treated with unilateral nephrectomy and AngII (1.5 mg/kg/day, s.c. minipump) as shown. On day 9, mice were divided into two groups exhibiting equivalent proteinuria and randomized to receive vehicle or CDDO-Im (10 μmol/kg, daily i.p.) for 7 days. (b) Proteinuria dramatically increased during exposure to CDDO-Im (boxed area) but this regressed after drug withdrawal. (c) The experiment was repeated with additional timepoints to show the gradual increase and regression of proteinuria in response to CDDO-Im. (d-g) The mRNA expression of nephrin, podocin, Nqo1, and Gstm1 were measured in isolated glomeruli from sham, day 16 AngII, day 16 AngII + CDDO-Im, and day 24 (one week after CDDO-Im withdrawal) groups. (h) Day 16 CDDO-Im group exhibits fewer WT1+ cells and more extensive nephrin disruption (white arrowheads). (i, j) Quantification of WT1+ cells and nephrin disruption. (k) Nephrinuria was significantly increased in CDDO-Im-treated animals. (l) TEM reveals greater foot process effacement in the CDDO-Im-treated mice. Mice were 8-10 weeks old. * P < 0.05, one-way ANOVA or t-test, as indicated. ** P < 0.001, multiple t-tests with Holm-Sidak correction.

To examine the causes of increased proteinuria due to CDDO-Im, we examined mouse hemodynamics, tubular megalin levels, and glomerular filtration rate (GFR). Mice were treated as in Figure 8a. AngII increased BP in both groups but CDDO-Im treatment did not further increase BP compared to vehicle (Figure 9a-c). Interestingly we observed that CDDO-Im suppressed nocturnal heart rate and SBP (Figure 9d). Bardoxolone methyl decreased megalin in primates,³¹ but we did not find this to be the case in mice administered CDDO-Im (Figure 9e and f). Consistent with this outcome, gel electrophoresis of proteins in the urine samples confirmed the abundance of albumin, but not other urinary proteins suggestive of proximal tubular dysfunction, in response to CDDO-Im treatment in the AngII model. Urinary retinol binding protein 4 (RBP4), a sensitive marker for reduced proximal tubule protein reabsorption,³⁴ was not increased after CDDO-Im exposure (Supplemental Figure 4). Since elevated GFR has been posited to increase proteinuria with bardoxolone methyl,¹⁷ we measured GFR in mice treated with 7 days of CDDO-Im or vehicle and did not find any differences (Figure 9g). Overall, these results indicate that pharmacologic Nrf2 enhancement worsens proteinuric and glomerular injury independent of blood pressure, megalin expression, and GFR.

Figure 9.

CDDO-Im proteinuria is not explained by BP, GFR, or changes in megalin expression. (a-d) Wild type mice (aged 9 weeks) were exposed to AngII and CDDO-Im as shown in Figure 8a. BP and HR were measured with radiotelemetry. Day 14-16 measurements represent the timepoint at which mice were exposed to vehicle or CDDO-Im. (e, f) Megalin levels were unchanged in either group. (g) Naïve wild type mice were treated with vehicle or CDDO-Im alone for 7 days prior to measurement of GFR, which was not significantly different between groups. N=3 for AngII alone, N=7 for AngII + CDDO-Im up to day 16, and N=4 for AngII + CDDO-Im for days 22-25. * P < 0.05, multiple t-tests with Holm-Sidak correction. Daytime and nighttime measures were analyzed separately for statistical purposes.

Nrf2 contributes to proteinuric injury

We examined proteinuria in WT mice and Nrf2 knockout mice (Nrf2^−/−). Nrf2^−/− mice were protected from AngII-induced proteinuric CKD (Figure 10a). In addition, Nrf2^−/− mice were insensitive to the effects of CDDO-Im (Figure 10b), indicating dependence on Nrf2 for the drug action.

Figure 10.

Nrf2^−/− mice are protected from AngII and CDDO-Im effects on proteinuria. (a) WT and Nrf2^−/− mice were treated with AngII and UACR determined as shown in Figure 2c. *** P < 0.0005, multiple t tests with Holm-Sidak correction. (b) WT and Nrf2^−/− mice were treated as in Figure 7a with AngII and CDDO-Im. UACR was checked at day 16. While WT mice increased UACR in response to CDDO-Im, Nrf2^−/− mice did not. Mice were 10-13 weeks old. A two-way ANOVA was performed and revealed a significant main effect for treatment (P=0.0227) and genotype (P=0.0338) as well as the interaction between treatment and genotype, (P=0.0370). * P < 0.05, ** P < 0.01, two-way ANOVA with Holm Sidak multiple comparisons.

Nrf2 activity is increased in human proteinuric CKD

We examined the expression of Nrf2 in a small number of human kidney disease cases. Compared to normal human kidney tissue, we found that there was generalized upregulation of glomerular Nrf2 in focal segmental glomerulosclerosis (FSGS), diabetic nephropathy, fibrillary glomerulonephritis, and membranous nephropathy (Figure 11). In addition, nuclear translocation of Nrf2 was increased in these cases and colocalized with WT1, demonstrating that Nrf2 activity was increased in podocytes (Figure 11, insets). In contrast, we found that Nrf2 levels and nuclear translocation were not increased in amyloidosis (Figure 11m-o). To verify the robustness of our findings, we repeated this analysis in an independent set of samples and again found increased glomerular Nrf2 levels and podocyte nuclear translocation in patients with proteinuric CKD (Supplemental Figure 5).

Figure 11.

Nrf2 is upregulated in proteinuric CKD. Human kidney tissue was assessed for Nrf2 expression. In comparison to normal control kidneys (a-c), glomeruli from patients with FSGS (d-f), diabetic nephropathy (g-i), fibrillary glomerulonephritis (j-l), and membranous nephropathy (m-o) exhibited higher levels of Nrf2 (red) in the cytoplasm and in the nucleus, where it often colocalized with WT1 (blue, purple overlap) indicating podocyte expression. In the case of amyloidosis (p-r), there was no increase in Nrf2 or nuclear translocation. Hematoxylin and eosin (H&E) stains from these patients are shown for reference. Low power and magnified views from representative photos are shown. Scale bars are 50 μm as indicated.

Discussion

Although Nrf2 protects against nonglomerular, nonproteinuric kidney diseases, including acute kidney injury and unilateral ureteral obstruction in animal models,^{7, 9, 10, 23, 24, 35} its role in glomerular disease is unclear. In human clinical trials, bardoxolone methyl use led to an increase in urinary albumin excretion in patients with diabetic nephropathy. Other side effects included BP elevation, hypomagnesemia, and muscle cramps.¹³ The BEACON trial was terminated early due to serious adverse events, primarily related to exacerbations of congestive heart failure.¹⁴ In animal studies, the bardoxolone derivative RTA 405 worsened proteinuria, glomerulosclerosis, and tubular injury in diabetic rats. When an impurity was found in the RTA 405 preparation, the use of another bardoxolone analogue, dh404, also led to a mild increase in proteinuria. Furthermore, dh404 increased renal inflammation and pseudotumor formation.³² In another study, Nrf2^−/− mice were protected from diabetic nephropathy, as evidenced by reduced albuminuria, SBP, glomerular area, and interstitial fibrosis.³⁶

These findings are consistent with our own results, which for the first time show that genetic Nrf2 enhancement (through partial loss of Keap1 expression) increases proteinuria when challenged with experimental injury models. Keap1^FA/FA mice exhibit increased podocyte injury in response to adriamycin which manifests as reduced nephrin and WT1 and increased FP effacement. Although Nrf2 can be antifibrotic,²⁴ Keap1^FA/FA mice exhibit increased fibrosis after adriamycin. This suggests either disease-specific actions or the possibility that heavy proteinuria causes permanent kidney scarring and overwhelms Nrf2-mediated protection. We also evaluated Nrf2^−/− mice and found that they were protected from AngII-induced proteinuria.

To increase the ability to translate our findings, we tested CDDO-Im, a synthetic triterpenoid and inducer of the Nrf2 pathway. CDDO-Im was used instead of bardoxolone methyl because of concerns that the latter leads to accumulation of toxic metabolites in rodents.^{30-32, 37} Similar to the enhancement of Nrf2 in Keap1^FA/FA mice, CDDO-Im treatment alone in wild-type mice did not induce proteinuria. However, in combination with AngII, CDDO-Im greatly enhanced proteinuria and was dependent on the presence of Nrf2. These results confirm involvement of the Nrf2 pathway on proteinuria and argue against this being an off-target effect of pharmacologic inducers.

BP has direct effects on proteinuria, and BP reduction is important in the care of proteinuric individuals. Radiotelemetry remains the gold standard for rodent BP measurement and provides accurate and continuous data compared to tail-cuff measures.³⁸ Using radiotelemetry we find that Keap1^FA/FA mice have higher daytime SBP, DBP, and MAP compared to WT mice. Since rodents sleep in the daytime, this could represent loss of blood pressure dipping during sleep, which is a preserved sympathetic response in rodents and humans.³⁹ Absence of dipping is a risk factor for adverse cardiovascular outcomes.⁴⁰ These data, coupled with the unexpected finding of reduced HR only at night (which was replicated by CDDO-Im), might suggest an effect of Keap1/Nrf2 on the sympathetic nervous system, cardiac conduction, or circadian rhythms and warrants further investigations beyond the scope of this study. It is likely that the BP increase in Keap1^FA/FA contributes at least partially to proteinuria. However, it is unlikely to be the entire explanation due to the small (7.3 mmHg) baseline difference in MAP and the temporal dissociation between the proteinuria and BP (proteinuria increases at day 14 while BP are similar at that timepoint). In addition, CDDO-Im increased proteinuria without increasing BP values beyond levels induced by AngII alone.

The mechanism of Nrf2 effects on proteinuria therefore requires further investigation. It has been hypothesized that Nrf2 increases intraglomerular pressure (i.e. glomerular hyperfiltration)¹⁷. Our studies did not identify any increase in GFR with CDDO-Im administration alone, or in Keap1^FA/FA mice (compared to wild type mice) either at baseline or after prolonged exposure to AngII. This may be related to our use of CDDO-Im rather than bardoxolone methyl, or to species-specific effects of Nrf2 activation. We do acknowledge that uninephrectomy, which was required to induce injury in the resistant C57BL/6 background, may have altered our results due to compensatory changes in the remaining kidney. Another limitation to our study is that we were not able to measure single nephron GFR, which could be elevated in kidney disease, particularly in the setting of disease-related reduction in nephron mass. A reduction in megalin has also been reported in cynomolgus monkeys treated with bardoxolone methyl.³¹, but we find that megalin expression is unchanged in both Keap1^FA/FA mice and with CDDO-Im administration.

Our data provide evidence that the Nrf2-enhanced proteinuria in our models occurs independently of hyperfiltration and proximal tubular dysfunction, namely through a toxic effect on podocytes. The reduction in podocyte number and nephrin expression, as well as increased FP effacement, may indicate a direct toxic effect on podocytes. This effect may be reversible to some extent, as proteinuria lessens and nephrin suppression reverts after CDDO-Im is withdrawn. Interestingly, we found that urinary nephrin shedding was dramatically induced by Nrf2 activity. Nephrinuria has been evaluated as a biomarker for various kidney diseases including diabetes and preeclampsia.^{29, 41-43} The loss of nephrin may be one mechanism by which Nrf2 enhances proteinuria, and the molecular pathways mediating this require further study. Collectively, our studies show that Nrf2 activation can cause specific podocyte defects that markedly worsen an existing proteinuric injury.

Our results challenge the notion that Nrf2 enhancement is protective in glomerular disease. The reason for this may be multifactorial. The genetic and pharmacologic Nrf2 activation may be much higher than that achieved in other studies or in ongoing human trials, and Nrf2 could have markedly different effects depending on the degree and mode of activation. In our genetic models there is continuous sustained Nrf2 pathway activation throughout the lifetime of the mouse. Meanwhile, pharmacological interventions using CDDO-Im cause an intermittent activation of Nrf2. Dose may also play a role, as the bardoxolone derivative dh404 at low doses attenuated renal injuries, while higher doses worsened disease.^{44, 45} This suggests that Nrf2 activity levels may require monitoring when used to treat human disease.

Furthermore, the protection seen in prior studies was incomplete or confounded by other effects. For instance, Keap1^FA/FA mice were protected from histologic injury but not proteinuria in immunotoxin-induced glomerular injury.⁴⁶ Sulforaphane reduced proteinuria in diabetic mice, but the improvements did not persist after the drug was discontinued.⁴⁷ Nrf2-mediated protection in another diabetic model was confounded by a reduction in hyperglycemia.⁴⁸ A well-performed study targeting the Nrf2 inhibitor glycogen synthase kinase 3 beta (GSK3β) strongly implicated Nrf2⁴⁹ but did not rule out effects on other pathways.⁵⁰

A limitation of our studies is inability to isolate the effects of Nrf2 to a particular cell type. The enhanced Nrf2 signaling in Keap1^FA/FA mice is not limited to podocytes⁵¹ and we would not expect CDDO-Im administration to have cell specificity. To truly assess the effects of Nrf2 in podocytes, conditional cell-specific model mice would be needed. Furthermore, while we cannot rule out effects of Keap1 that are independent of Nrf2, we did find that Nrf2^−/− are protected from AngII-induced proteinuria and CDDO-Im exacerbation.

In accordance with our animal data, we were able to demonstrate that Nrf2 is upregulated in the glomeruli of patients with proteinuric kidney diseases including FSGS, diabetic nephropathy, membranous nephropathy, and fibrillary glomerulonephritis. Nrf2 also localized to the nuclei of podocytes, as shown with costaining for WT1. Amyloidosis was an exception, and this may be due to the deposition of fibrils in the glomeruli. These findings require further study.

In conclusion, we show that both genetic and pharmacologic Nrf2 induction in the setting of proteinuric disease challenge potentiates proteinuria and glomerular injury. Sustained dampening of Keap1 expression, as seen in the Keap1^FA/FA mice, leads to permanent injury with interstitial fibrosis following a “second hit” challenge. These results highlight a need for careful monitoring of renal effects in existing and future clinical trials utilizing inducers of the Nrf2 pathway.

Methods

Animal studies

All studies were performed to the ethical and scientific standards recommended by the Guide for the Care and Use of Laboratory Animals from the NIH. The animal protocol was approved by the Institutional Animal Care and Use Committee at the University of Pittsburgh. Keap1^FA/FA were originally generated and provided by Dr. Masayuki Yamamoto⁵² and crossed into and maintained in the congenic C57Bl/6J Tyr^c-2J (albino) background. Age matched C57BL/6J albino mice (Jackson Laboratory, cat#000058, Bar Harbor ME) were used as controls for the Keap1^FA/FA mice. Nrf2^−/− mice (Jackson cat#017009) were also used along with their control C57BL/6J mice (Jackson cat#000664).

We subjected mice to unilateral nephrectomy seven days prior to one of three proteinuria models. Model 1 utilized adriamycin (18mg/kg, Sigma, St. Louis MO) via tail vein injection. Model 2 utilized subcutaneous osmotic minipumps (Model 2004, Alzet, Cupertino CA) containing angiotensin II (1.5 mg/kg/day, Bachem, cat#4006473, Torrance CA) in 0.01M acetic acid as previously described.²⁷ Model 3 utilized escalating daily doses of endotoxin-free BSA (Sigma, cat#A9430): 2, 4, 6, 8, and 10mg/g followed by two rest days and then 10mg/g daily until sacrifice on day 11. ²⁸ In some experiments, mice were given CDDO-Im (10 μmol/kg daily i.p., Tocris/Bio-Techne, cat#473710, Minneapolis MN). For all experiments, we collected spot urine and kidney tissue.

Biochemical Measurements

Urine albumin excretion was determined with a mouse albumin ELISA kit (Bethyl Laboratories, Worthington TX) and values divided by urine creatinine determined by an enzymatic assay kit (Pointe Scientific, Canton MI) to obtain UACR. Serum creatinine was measured with the same enzymatic kit.

Histology

Kidneys were fixed in 10% buffered formalin and paraffin-embedded sections stained with Masson’s trichrome, Periodic Acid Schiff, and picrosirius red staining kits (Thermo). Fibrosis scoring was performed in a blinded fashion using at least 10 random fields from each kidney slice. A grid was overlaid on each image and boxes containing fibrotic tissue were counted. Human kidney biopsy slides were obtained from the pathology archive at the University of Pittsburgh Medical Center.

Telemetry assessment of BP and heart rate

BP was measured using telemetry as described previously⁵³. Briefly, mice were anesthetized with 2% isoflurane (100% O₂). PA-C10 radiotelemetry units (Data Sciences, Inc., St. Paul MN) were inserted into the abdominal aorta via the femoral artery and transmitter body secured in the flank. Mice were singly housed and given 1 week to recover. Data was collected using Spike2 software and daytime (10AM-4PM) and nighttime (10PM-4AM) values averaged. Systolic and diastolic BP were analyzed on a beat-to-beat basis and mean BP calculated as diastolic plus 1/3 pulse pressure. Heart rate was derived from the pulsatile BP signal. After baseline measurements, AngII was infused as described above and hemodynamic measurements repeated. Data from indicated days were averaged and group means determined for reporting.

GFR Measurements

GFR was assessed via measurement of the extinction of FITC-sinistrin using a transdermal probe as previously described.⁵⁴ Briefly, mice were injected with 7.5mg/100g body weight FITC-sinistrin (Medibeacon GmbH, Mannheim, Germany) through the tail vein. A transdermal GFR monitor was affixed directly to shaved skin on the dorsum of the animal (Medibeacon GmbH) and levels of FITC-sinistrin measured over 90 minutes. A 3-compartment fit model was used to determine the half-life using Medibeacon software provided with the probe. GFR was calculated according to previously published methods.⁵⁵

Statistical Analysis

Data are presented as means ± standard error of the mean (SEM). Two group and three or greater group comparisons were made with student’s t-test or one-way ANOVA, respectively. Two-way ANOVA with Holm-Sidak multiple comparisons were utilized when appropriate as indicated. Measurements over a timecourse including proteinuria and hemodynamics over time were analyzed with multiple t-tests with Holm-Sidak corrections for multiple comparisons. For BP and HR, daytime values were analyzed together, as were nighttime values. Statistical analyses were performed with Graphpad Prism (version 8, San Diego CA).

Supplementary Material

1

Supplemental Figure S1. Keap1^FA/FA mice exposed to angiotensin II exhibit worse injury than wild type mice. (a, top panels) Trichrome staining of kidney tissue demonstrating greater glomerulosclerosis (yellow arrow) and interstitial fibrosis (black arrowheads) in Keap1^FA/FA mice. (a, bottom panels) Immunofluorescence staining for WT1 and nephrin showing fewer WT1+ nuclei and more nephrin disruption (white arrowhead) in the Keap1^FA/FA mice. (b) WT1+ cells are significantly reduced per glomerular section in Keap1^FA/FA mice. * P < 0.05, t-test.

Supplemental Figure S2. GFR is not elevated in Keap1^FA/FA mice. (a) GFR was measured in 10-12 week old wild type and Keap1^FA/FA mice at baseline and after 28 days of AngII injury. While injury reduced GFR, there was no significant difference between the two strains. (b) Serum creatinine was also similar between strains treated with AngII.

Supplemental Figure S3. Whole kidney Nrf2 pathway activation by CDDO-Im administration. Wild type mice were treated with vehicle, 3, 10, or 30 μmol/kg p.o. or 3 or 10 μmol/kg i.p. CDDO-Im as indicated (N=2 for each dose). Whole kidney mRNA levels of Nqo1 (a) and Gstp1 (b) were assessed 24 hours after injection. The 10 μmol/kg i.p. dose was chosen for proteinuria experiments.

Supplemental Figure S4. The composition of urine in mice treated with AngII and CDDO-Im. (a) Gel electrophoresis (1μg urine creatinine loaded per lane) reveals that albumin is the major protein increased by AngII + CDDO-Im exposure compared to AngII alone. (b) Immunoblot for RBP4 does not reveal any increases due to CDDO-Im exposure. These results suggest that the albuminuria is not due to defects in proximal tubular reabsorption of protein.

Supplemental Figure S5. Nrf2 is upregulated in proteinuric CKD. We repeated our analysis from Figure 11 independently in a separate set of human kidney biopsies from control kidneys (a-c) and from patients with FSGS (d-f), diabetic nephropathy (g-i), fibrillary glomerulonephritis (j-l), and membranous nephropathy (m-o). Representative images are shown. CKD patients exhibited higher levels of Nrf2 (red) in the cytoplasm and in the nucleus, where it colocalized with WT1 (blue, purple overlap) indicating podocyte expression. Hematoxylin and eosin (H&E) stains from these patients are shown for reference. Scale bars indicate 50 μm.

Supplemental Table S1. Hemodynamic measurements for WT and Keap1^FA/FA mice exposed to AngII infusions.

Supplemental Table S2. Primers used in quantitative real-time PCR.

Supplemental Table S3. Primary antibodies used in this study.

Translational Statement.

Ongoing clinical trials target activation of the nuclear factor erythroid 2 related factor 2 (Nrf2) transcription factor to treat chronic kidney disease. However, the Nrf2 inducer, bardoxolone methyl, increased proteinuria in diabetic nephropathy patients. Mice with genetic Nrf2 activation exhibited increased proteinuria and podocyte injury when exposed to kidney injury. Nrf2 knockout mice were protected from injury. High doses of a bardoxolone methyl analogue (CDDO-Im) also increased preexisting proteinuric injury in a Nrf2-dependent fashion. Therefore, in some situations, Nrf2 activation promotes podocyte injury and proteinuria in mice and suggests a need for careful monitoring in clinical trials.