Dapagli ozin alleviates diabetic kidney disease via HIF‐1α/HO1 mediated ferroptosis

AIMS
Diabetic kidney disease (DKD) is the leading cause of end-stage kidney disease. Sodium-glucose cotransporter 2 inhibitors (SGLT2i) showed excellent renoprotective effects; however, the underlying mechanism remains not fully understood. Previous studies have revealed the importance of ferroptosis, which is closely related to oxidative stress, in the progression of DKD. In the current study, we hypothesized that SGLT2i could relieve ferroptosis and thereby alleviate renal injury in DKD due to their anti-oxidative stress effects.


RESULTS
Typical changes of ferroptosis including massive lipid peroxidation, compromised antioxidant capability, and iron overload were found in db/db mice and high glucose/high fat (HG/HF)-treated HK-2 cells. Furthermore, increased expression of hypoxia inducible factor 1α (HIF1α) and heme oxygenase 1 (HO1) was observed in db/db mice and HG/HF-treated HK-2 cells as well. Dapagliflozin treatment significantly ameliorated the ferroptosis-related changes via attenuating over-activation of the HIF1α/HO1 axis in vivo and in vitro. Besides, downregulation of the HIF1α/HO1 axis alleviated ferroptosis, while overexpression of HIF1α and HO1 aggravated ferroptosis induced by HG/HF in HK-2 cells.


INNOVATION AND CONCLUSION
This study revealed that SGLT2i played a renoprotective role in DKD, at least in part, through alleviating HIF1α/HO1 mediated ferroptosis.


Introduction
Diabetic kidney disease (DKD) develops in approximately 40% of type 2 diabetes mellitus (T2DM) patients and has become the leading cause of end-stage kidney disease (ESKD) 1, 2, 3 . Despite multifactorial risk management hyperglycemia and hypertension, the residual risk of DKD progression remains high, indicating an unmet medical need. Sodium-glucose cotransporter 2 inhibitors (SGLT2i) inhibit the coupled reabsorption of sodium and glucose from the proximal tubule, reducing glucose reentry from tubular uid into the bloodstream. Many recent clinical trials showed the renoprotectionve effects of SGLT2i in patients with T2DM 4, 5, 6, 7 , but the underlying mechanisms were not fully clear.
Multiple factors, including hemodynamic dysfunction, oxidative stress, hypoxia, and in ammation contribute to the development and progression of DKD 8 . Ferroptosis is a form of regulated cell death culminating with the accumulation of redox-active iron and iron-dependent lipid peroxidation 9 . Oxidative stress is crucial in ferroptosis for reactive oxygen species (ROS) generation and polyunsaturated fatty acids (PUFAs) peroxidation. Glutathione peroxidase 4 (GPX4), solute carrier family 7 member 11 (SLC7A11), and nuclear factor erythroid 2-related factor 2 (NRF2) function as negative regulators of ferroptosis by relieving ROS accumulation through antioxidant capacity 10,11,12 , while transferrin receptor (TFRC), NADPH oxidase, and p53 act as positive regulators of ferroptosis by stimulating iron uptake and/or promoting ROS production 13,14 . Previous studies suggested that ferroptosis is involved in the pathogenesis of DKD by bioinformatics analysis as well as in vivo and in vitro experiments 15,16,17 . Antioxidative stress is a prominent pathway related to the renoprotective effects of SGLT2i 18, 19,20 . Therefore, it is reasonable to speculate that SGLT2i play a renoprotective role via inhibiting ferroptosis.
Hypoxia-inducible factor-1α (HIF-1α), which binds to hypoxia response elements (HREs; 5'-RCGTG-3'), regulates a variety of genes including iron homeostasis associated genes such as heme oxygenase-1 (HO1, catalyzing the rate-limiting heme oxidation to biliverdin, carbon monoxide, and free ferrous iron 21,22 ) and TFRC 23,24 , and thus makes it a key regulator of ferroptosis. Although Feng X's study found that ferroptosis could aggravate albuminuria, damage renal tubules, and enhance renal brosis through HIF-1α/HO-1 pathway in diabetic mice 25 , several lines of evidence showed the survival-promoting effects of HIF-1α and HO1 on inhibiting ferroptosis in other models 26,27,28,29,30,31 . Moreover, Jiang N's study revealed that HIF-1α exerted a protective effect against tubular injury by improving mitochondrial quality in DKD through HO1 upregulation 32 . Given the controversial role of HIF-1α and HO1 in ferroptosis and DKD, it is of interest to elucidate the renoprotective effects of SGLT2i in DKD via regulating ferroptosis by HIF-1α/HO1 axis since SGLT2i can attenuate hypoxia as well as HIF-1α accumulation in vivo and in vitro 33, 34 .

Patients and samples
Twenty patients with T2DM and biopsy-proven DKD, diagnosed from January 2017 to December 2017 in Peking University First Hospital 35 , were enrolled in this study. T2DM was de ned according to the criteria proposed by the American Diabetes Association in 2017 36 . None of the patients had coexisting nondiabetes-related renal disease. DKD was de ned as previously described 35 . Healthy control kidney samples were obtained from healthy kidney poles of individuals (n = 10) receiving tumor nephrectomies without diabetes or other kidney diseases. All healthy control kidney samples were con rmed by pathological examinations including immuno uorescence, light microscopy, and electron microscopy. Clinical data of the patients at the time of renal biopsy was systematically recorded. Biopsies were scored independently by two experienced pathologists respectively. Interstitial brosis and tubular atrophy (IFTA) scores were assessed semi-quantitatively based on the proportion of the tubulointerstitial compartment affected (0, none; 1, < 25%; 2, 25-50%; 3, > 50%) 35 . The investigation was conducted according to the Declaration of Helsinki and was approved by the Ethics Committee of Peking University First Hospital (2017 − 1280). Written informed consent was obtained from each participant at renal biopsy.
Urinary HO1 levels were measured at week0 and 13 (ab205524; Abcam, Cambridge, MA, USA). Kidney tissues were collected before the time of euthanization and preserved for examination. All animal experiments were approved by the Laboratory Animal Ethics Committee of Peking University First Hospital (J202138).

Renal histology
Staining of kidney sections was performed using periodic acid Schiff (PAS; K1433; BioVision, CA, USA). Glomerular area was measured by tracing around the perimeter of the glomerular tuft. The mesangial matrix expansion area was assessed from the images of glomeruli and presented as a proportion of PASstained per glomerular cross-sectional area. The tubulointerstitial injury index was determined by assessing the extent and severity of tubular dilation, atrophy, and loss of tubular cells. Twenty images of a kidney section (magni cation ×400) were scored as follows: 0 for no injury, 1 for < 25%, 2 for 25-50%, 3 for 50-75%, and 4 for > 75% tubulointerstitial injury 35 . Quantitation analyses were performed on Image-Pro Plus software V.6.0 (Media Cybernetics, Bethesda, MD).

Transmission electron microscopy
The renal cortical tissues of mice were xed in 3% glutaraldehyde and further sample handling was performed by the Laboratory of Electron Microscopy, Peking University First Hospital. Imaging was performed by a Hitachi HT7800 transmission electron microscope (Hitachi, Japan). The measurement of glomerular basement membrane (GBM) thickness and mitochondria analysis of renal tubular epithelial cells was processed using Image-pro plus.

Quanti cation of lipid peroxidation
HK-2 cells were incubated with 2.5 µM C11-BODIPY 581/591 (D3861; Thermo sher Scienti c, MA, USA) for 45 min at 37℃. Then cells were collected and washed once with PBS followed by FACS. Fluorescence intensity was analyzed using FlowJo_V7 software.

Assessment of mitochondrial activity
Mitochondrial activity in HK-2 cells was assessed (C1035; Beyotime, Beijing, China) via FACS analysis or confocal laser scanning microscope. The uorescence intensity of FACS was analyzed using FlowJo_V7 software.

Quantitative real-time PCR (qRT-PCR)
Total RNA was extracted (DP441; Tiangen, Beijing, China) and reverse transcribed into cDNA (4374966; Applied Biosystems, MA, USA). The qRT-PCR analysis was carried out in an ABI Prism 7500 sequence detection system using SYBR Green Master Mix (A25742; Applied Biosystems). Primers are listed in Table  S1.

Bioinformatic analysis
The transcriptome sequencing and analysis were conducted by OE Biotech Co., Ltd. (Shanghai, China).
Raw data were processed using Trimmomatic 37 . After removing the low-quality reads, clean data were mapped to the reference genome using hisat2 38 . FPKM 39 value was calculated using cu inks 40 , and the read counts were obtained by htseq-count 41 . Differentially expressed genes (DEGs) were identi ed using DESeq 42 R package, and p < 0.05 and fold change > 2 or fold change < 0.5 was set as the threshold for signi cantly differential expression. KEGG 43 pathway enrichment analysis of DEGs was performed using the Cluster Pro ler R package.

Statistical analysis
Normally distributed data were presented as mean ± standard deviation (SD), and non-normally distributed data were presented as median and interquartile range (IQR). Groups were compared using unpaired two-tailed Student's t-tests or one-way analysis of variance (ANOVA) as appropriate. Pearson or Spearman correlation analysis was used to evaluate the association between immunohistochemistry staining intensity (IOD/area) and clinicopathological parameters as appropriate. A p value less than 0.05 was considered statistically signi cant (*p < 0.05, **p < 0.01, ***p < 0.001). All the results were plotted using GraphPad Prism 8 software.

Dapagli ozin alleviated renal ferroptosis in db/db mice
The level of MDA which increased in kidney cortex of vehicle-treated db/db mice was signi cantly alleviated in dapagli ozin-treated db/db mice, indicating the decreased lipid peroxidation upon dapagli ozin treatment (Fig. 1G). GPX4, which mainly expressed in tubulointerstitium, signi cantly increased in kidney cortex of dapagli ozin-treated db/db mice compared with vehicle-treated db/db mice (Fig. 1H-K).
Transmission electron microscopy revealed that typical changes of mitochondria in ferroptosis in kidney tubular epithelial cells including ruptured mitochondrial membrane and disappeared mitochondrial cristae 44 , were improved in dapagli ozin-treated db/db mice, as compared with vehicle-treated db/db mice (Fig. 1E). Consistently, the form factor and aspect ratio of mitochondria were improved by dapagli ozin treatment (Fig. 1Land M).
Kidney cortical tissues of mice were isolated for metabolomics analysis. 49 and 10 differential metabolites were found in db/db-vehicle vs. m/m group and db/db-dapagli ozin vs. db/db-vehicle group respectively ( Fig. 2A and B). Arachidonic acid signaling was upregulated in vehicle-treated db/db mice compared with m/m and dapagli ozin-treated db/db mice, and lipid peroxidation products derived from arachidonic acid have been claimed as the proximate executioners of ferroptosis 12, 45 ( Fig. 2C and D).
In RNA-Seq analysis, there were 2187 and 285 DEGs in db/db-vehicle vs. m/m group and db/dbdapagli ozin vs. db/db-vehicle group respectively (Fig. 2E). HO1, a key regulator of oxidative stress and strongly associated with ferroptosis, stood out among a group of DEGs in db/db-dapagli ozin vs. db/dbvehicle group (Fig. 2F). KEGG pathway enrichment analysis based on the RNA-Seq data showed that arachidonic acid (endocannabinoid) signaling was simultaneously enriched in db/db-vehicle vs. m/m group and db/db-dapagli ozin vs. db/db-vehicle group, which was consistent with the metabolomics analysis ( Fig. S1A and B).
We veri ed the RNA-Seq results on protein level and con rmed that dapagli ozin treatment led to a remarkable reduction of HIF-1α and HO1 expression, as compared with vehicle-treated db/db mice ( Fig. 3A-D). Meanwhile, the urinary HO1 level was also signi cantly reduced in dapagli ozin-treated db/db mice compared with vehicle-treated db/db mice (Fig. 3E). These results suggested that HIF-1α/HO1 axis may be associated with the ferroptosis inhibition effects of dapagli ozin in db/db mice.
Signi cantly reduced mitochondrial activity ( Fig. 4B and C) was found in HG/HF group compared with control group. Lipid peroxidation level (Fig. 4D) increased, while GPX4 expression decreased upon HG/HF treatment (Fig. 4E-G). These alterations were accompanied by increased expression of HIF-1α (in both whole cell and nucleus), HO1 and TFRC (Fig. 4E-G) as well as elevated iron levels ( Fig. 4H-J).
Under HG/HF condition, cells in dapagli ozin group have signi cant restoration of the cell viability (Fig. 4A), mitochondrial activity ( Fig. 4B and C), and GPX4 expression compared with vehicle group (HG/HF and solvent treated cells) (Fig. E-G). The lipid peroxidation level (Fig. 4D) and iron concentration (Fig. 4H-J) were signi cantly lower in dapagli ozin group than in vehicle group.

Inhibition of HIF-1α/HO1 axis rescued HG/HF induced ferroptosis in vitro
Under HG/HF condition, inhibition of HIF-1α or HO1 (by PX-478 and Znpp respectively) restored the cell viability impaired in vehicle group (Fig. 5C). PX-478 and Znpp also improved mitochondria activity ( Fig. 5D and E) and reduced lipid peroxidation level (Fig. 5J) compared with vehicle group. Upregulation of GPX4 (Fig. 5A, B and F) and relieved iron overload (Fig. 5G-I) were also observed in PX-478 and Znpp treated cells.
Interestingly, TFRC was upregulated in DMOG and Znpp treatment but not in PX-478 treatment whether with or without hemin (Fig. 5A, B and F), which indicated HO1 and TFRC were regulated by HIF-1α while GPX4 was regulated by HO1. Collectively, these ndings showed the inhibition of the HIF-1α/HO1 axis could suppress ferroptosis of HK-2 cells induced by HG/HF.

Ferroptosis was associated with the disease severity in DKD patients
We further analyzed the correlation between ferroptosis and disease severity in DKD patients (general data in Table S2). GPX4 level in kidney tubulointerstitium was signi cantly lower in DKD patients than in healthy controls ( Fig. 6A and B), while HO1 and TFRC levels were signi cantly higher in kidney tubulointerstitium of DKD patients than in healthy controls ( Fig. 6A and B). HO1 level positively correlated with serum creatinine (Scr) (r = 0.54, p < 0.05; Fig. 6C) and negatively correlated with estimated glomerular ltration rate (eGFR) (r =-0.59, p < 0.01; Fig. 6D), and GPX4 level negatively correlated with Scr (r =-0.63, p < 0.01; Fig. 6E) and positively correlated with eGFR (r = 0.69, p < 0.001; Fig. 6F). These data suggested that ferroptosis was associated the disease severity of DKD in patients.

Discussion
SGLT2i have emerged as a promising antidiabetic drug with renoprotection in T2DM patients 47 , and the role of ameliorating oxidative stress in SGLT2i mediated renal protection has aroused increasing attention recently 19 . In the current study, we found that dapagli ozin relieved ferroptosis in db/db mice and HK-2 cells, including lipid peroxidation products accumulation, weakened antioxidant capacity and iron overload, by downregulating the HIF-1α/HO1 axis. In addition, we found that HO1 and GPX4 in kidney tubulointerstitium signi cantly correlated with the disease severity in DKD patients.
In diabetic kidneys, enhanced glucose reabsorption by renal proximal tubular cells (RPTCs) (via SGLT2) increases oxygen consumption and predisposes the renal cortex to hypoxia 48 , and increased HIF-1α expression in RPTCs suggested that hypoxia of renal tubules was a hallmark of DKD 49,50 . Some studies showed the pathogenic role of HIF-1α in DKD 51,52,53 , while others found that modulation of metabolic disorders and in ammation in diabetic mice by hypoxia-inducible factor stabilizers (prolyl hydroxylase inhibitors) was protective 54,55 . Adaptive upregulation of HIF-1α under hypoxia could be protective to a certain degree, while overactivation of HIF-1α would lead to tissue damage. In addition, except HIF-1α prolyl hydroxylase inhibitors also stabilize HIF-2α 55 . HIF-2α played a protective role in DKD by inhibiting in ammation and brosis 56 , which should be taken into consideration of studies of prolyl hydroxylase inhibitor. SGLT2i improved hypoxia, which explained the prevention of HIF-1α accumulation by SGLT2i mechanistically.
Accumulation of HIF-1α in DKD induces the expression of HO1 and TFRC and regulates iron metabolism 57 . Although previous evidence indicated that HO1 had protective effects in DKD 58, 59, 60 , several studies found elevated HIF-1α and HO1 levels in kidneys of diabetic models 25,61 . HO1 help attenuate oxidative stress in the early stage, while continuous HO1 upregulation will trigger ferroptosis 62, 63 , which indicates that the effect of HO1 in DKD may depend on different pathological conditions. We found that upregulation of the HIF-1α/HO1 axis promoted ferroptosis in vivo and in vitro, and inhibition of the HIF-1α/HO1 axis alleviated ferroptosis. Moreover, upregulation of the HIF-1α/HO1 axis weakened the ferroptosis-relieving effect of dapagli ozin, which suggested alleviated iron overload by SGLT2i through the HIF-1α/HO1 axis could, at least to some extent, explain its renoprotective effects.
Cyst(e)ine/glutathione (GSH)/GPX4 signaling axis constitutes the predominant ferroptosis defense system 10,64,65 . However, we found dapagli ozin could rescue ferroptosis induced by RSL3 but not erastin (ferroptosis inducer by inhibiting SLC7A11 activity) in HK-2 cells (Fig. S2), which indicated that dapagli ozin inhibits ferroptosis by regulating GPX4 expression rather than SLC7A11. GPX4 utilizes GSH as its cofactor to prevent ferroptosis by eliminating membrane phospholipid hydroperoxides and maintaining the integrity of phospholipid bilayers 66 . GPX4 de ciency is regarded as one of the biomarkers of ferroptosis, and GPX4 depletion caused massive renal tubular epithelial cells to undergo ferroptosis 10,67,68 . Dapagli ozin could reverse the GPX4 reduction in vivo and in vitro, which suggested that the restored antioxidant capacity against lipid peroxidation caused by the iron overload of SGLT2i was also necessary for its renoprotective role.
There were many other pathways with signi cant changes in metabolomics and RNA-Seq ( Fig. 2C and D;  Fig. S1), so more detailed insights are needed to elaborate the effects of these pathways in the renoprotective effect of SGLT2i.
In conclusion, dapagli ozin alleviated DKD progression by mitigating ferroptosis through HIF-1α/HO1 axis (Fig. 7). The present study explained the renoprotective effect of SGLT2i from a new angle.   Ferroptosis played an important role in the renoprotective effects of dapagli ozin.
The metabolites difference between m/m mice and vehicle-treated db/db mice (A). The metabolites difference between vehicle-treated and dapagli ozin-treated db/db mice (B). The KEGG pathway clustering results of differential metabolites between m/m mice and vehicle-treated db/db mice (C) and vehicle-treated and dapagli ozin-treated db/db mice (D). The intersection of differentially expressed genes in m/m, vehicle-treated db/db and dapagli ozin-treated db/db mice (E). The differentially expressed genes between vehicle-treated and dapagli ozin-treated db/db mice (F). HO1, heme oxygenase-1; m/m, wild type mice; db/db-Veh, vehicle-treated db/db mice; db/db-Dapa, dapagli ozintreated db/db mice.