Axl‐inhibitor bemcentinib alleviates mitochondrial dysfunction in the unilateral ureter obstruction murine model

Abstract Renal fibrosis is a progressive histological manifestation leading to chronic kidney disease (CKD) and associated with mitochondrial dysfunction. In previous work, we showed that Bemcentinib, an Axl receptor tyrosine kinase inhibitor, reduced fibrosis development. In this study, to investigate its effects on mitochondrial dysfunction in renal fibrosis, we analysed genome‐wide transcriptomics data from a unilateral ureter obstruction (UUO) murine model in the presence or absence of bemcentinib (n = 6 per group) and SHAM‐operated (n = 4) mice. Kidney ligation resulted in dysregulation of mitochondria‐related pathways, with a significant reduction in the expression of oxidative phosphorylation (OXPHOS), fatty acid oxidation (FAO), citric acid cycle (TCA), response to reactive oxygen species and amino acid metabolism‐related genes. Bemcentinib treatment increased the expression of these genes. In contrast, AKT/PI3K signalling pathway genes were up‐regulated upon UUO, but bemcentinib largely inhibited their expression. At the functional level, ligation reduced mitochondrial biomass, which was increased upon bemcentinib treatment. Serum metabolomics analysis also showed a normalizing amino acid profile in UUO, compared with SHAM‐operated mice following bemcentinib treatment. Our data suggest that mitochondria and mitochondria‐related pathways are dramatically affected by UUO surgery and treatment with Axl‐inhibitor bemcentinib partially reverses these effects.

However, although innovative potential therapeutic approaches for CKD have been proposed, over the last 20 years, no new drug has been approved to specifically prevent CKD or to improve kidney function. 3,4 Furthermore, since available information guiding kidney patients' care is limited, new approaches are necessary. 5 An important barrier to the development of new therapeutic approaches is represented by the limited understanding of molecular mechanisms underlying CKD, and the lack of therapeutic targets.
As highly metabolically active organ, kidney function is tightly dependent on mitochondria performance. Mitochondria are not only the cell powerhouse but also coordinate cellular adaptation to stressors, and regulate cell death, oxidative stress and cellular metabolism. 6,7 Therefore, mitochondrial damage and dysfunction have been associated to the pathophysiology of a broad spectrum of renal diseases, including kidney fibrosis development, [8][9][10]12 and diabetic kidney disease. 13 Axl, a member of the TAM family of receptor tyrosine kinases, is widely expressed in normal cells and tissues. Because of its function as a regulator of different physiological processes including cell survival, proliferation, migration and differentiation, Axl has been proposed as a promising treatment target for different malignancies. 14,15 A previous study from our group demonstrated the effectiveness of the Axl inhibitor bemcentinib in alleviating fibrosis development in a murine unilateral ureteral obstruction (UUO) model. 16 This model is widely used to elucidate the pathogenesis of obstructive nephropathy and the mechanisms responsible for progressive renal fibrosis, and as a model to investigate fibrosis attenuating treatments. 17 Furthermore, since it reflects the progression of acute kidney injury (AKI) to CDK, UUO provides an important model to study mitochondrial dysfunction in kidney diseases. 18 Based on this background, in this study, we addressed the effects of-Axl-inhibitor bemcentinib on mitochondrial dysfunction induced by UUO, by investigating renal cell transcriptome and amino acid metabolism.

| Animals and sample collection
Animal handling and sample collection were previously described in full detail. 16 Briefly, eight-to nine-week-old male C57Bl/6JOlaHSD mice were acquired from Envigo (Horst, the Netherlands) and kept and managed in the local animal facility at the Department of Biomedicine, University of Bergen, Norway. All surgeries were performed under general anaesthesia with isoflurane gas. Left ureter was identified through a subcostal incision and ligated with a silk ligature. Animals were divided into three groups: UUO model treated with bemcentinib diluted in vehicle (0.5% hydroxypropylmethylcellulose in 0.1% tween 80) (n = 6) or only with vehicle (n = 6) and SHAM-operated (n = 4).
The drug was administered twice daily by oral gavage at a dose of 50 mg/kg (10 ml/kg), from one day before surgery to 14 days post-surgery. Mice were sacrificed fifteen days post-surgery and blood was collected by retro-orbital method or cardiac puncture.
Kidneys were also harvested, cut into transverse slices, fixed in formaldehyde and embedded in paraffin according to standard procedures. RNA was extracted from frozen murine kidney poles and sequenced on the Illumina HiSeq4000 platform. The mRNA sequencing data were processed in the RStudio environment, where reads were aligned and counted. Log 2 CPM data were obtained and fold changes were calculated.

| Total gene expression analysis
An over-representation analysis (ORA) was performed on the differential expression dataset using the ClusterProfiler package and the Reactome pathway repository, 19 based on significantly expressed genes (FC ± 1.15, q < 0.05). A principal component analysis (PCA) was performed on the normalized log2 CPM expression data to evaluate the variability in the transcriptomics dataset. A loading plot from the first two (n = 2) principal components (PCs) was generated to elucidate which genes contributed most to the variance observed in PC 1 and PC2 and was visualized using the ggplot package. 20 Differential expression for genes involved in selected pathways was visualized using Gene Ontology (GO) 21 and the ggplot2 package. All analyses were performed using the R programming language. Sequencing data were published in a previous study from our group 16

| Mitochondrial gene expression analysis
Normalized log2 CPM expression data were filtered based on the public gene database Mouse MitoCarta 2.0 (www.broad insti tute. org/pubs/MitoC arta), and PCA was performed on selected genes with statistically significant differential expression values (FC ± 1.5, q-value < 0.05) to evaluate data variability. Loadings from the first n (n = 2) principal components (PCs) were extracted and variance was visualized using the ggplot package. 20 Genes in each PC were ranked based on their loadings, and the 40 genes with the highest loading score in each PC were selected for hierarchical clustering using Euclidean distance and Ward2-linkage, and visualized with ComplexHeatmap package. 22

| Mitochondrial DNA extraction and qPCR quantitation
Murine kidney tissue from FFPE blocks was cut into three 10

| Immunohistochemistry
Three-micron-thick formalin-fixed paraffin-embedded sections from ligated and non-ligated murine kidneys were deparaffinated in xylene and rehydrated in descending concentrations of ethanol. Epitope retrieval was performed in target retrieval buffer (pH6, Dako) using a microwave oven, and endogenous peroxidase activity was quenched, by 10 minutes incubation with peroxidaseblocking solution (Dako). Unspecific binding sites blocking was achieved by incubating sections with 10% normal goat serum in PBS (Dako) for 30 minutes. Sections were then incubated for 60 minutes with rabbit polyclonal anti-TOMM20 (catalogue number ab186735, Abcam) or anti-SDHB primary antibodies (catalogue number HPA002868, Sigma-Aldrich) at a 1:500 dilution in antibody diluent with background reducing agent (Dako). Primary antibodies were labelled using polymer-Horseradish peroxidase-conjugated anti-rabbit immunoglobulins (Envision+ ® system, Dako). Signal was visualized using 3,3′-diaminobenzidine (DAB, Dako) and sections were counterstained using haematoxylin (Dako), dehydrated and cover-slipped using a non-aqueous mounting medium. All reagents and kits were used according to manufacturer´s instructions.
All immunohistochemical reactions were performed on the autoimmunostainer intelpthFLX (BioCare) at room temperature. Digital

| Western Blot
Protein extraction from mouse kidney tissues was achieved by using RIPA buffer (Sigma-Aldrich, catalogue no. R0278) with the addition of complete protease inhibitor (Roche, catalogue no. 4693116001) and phosphatase inhibitor cocktail (Sigma-Aldrich, catalogue no. P5726).
Protein concentration was determined using Pierce BCA Protein Assay Kit (Thermo Scientific, catalogue no. 23225). Proteins were separated in Bolt 4-12% Bis-Tris Plus electrophoresis gels and transferred to nitrocellulose membranes using iBlot 2 System. Membranes were blocked with 5% BSA in PBS containing 0.1% Tween-20 and then incubated overnight with rabbit polyclonal Anti-TOMM20 antibody (Abcam, ab186735) 1:2000 dilution and mouse-monoclonal Anti-SDHB antibody (Abcam, ab14714) 4µg/mL. SeeBlue Plus2 Prestained Protein Standard (Invitrogen, LC5925) was used to visualize protein molecular weight. The blots were washed three times with a wash buffer (PBS, 0.1% Tween-20) and then incubated for 1 hour either with goat anti-rabbit (Abcam, ab205718) or goat antimouse secondary (ab205719) HRP-linked antibodies. The blots were washed again and developed using Pierce ECL Plus Western blotting substrate (Thermo Fisher). Chemo-luminescence signals were assessed using ChemiDoc Imaging System (Bio-rad). Densitometry analysis was performed using the ImageJ software.

| Serum amino acids profile
Blood samples were collected in Microvette 500Z-gel tubes (Sarsted, Germany, catalogue no. 201344) and centrifuged at 10000 g for 10 minutes to separate plasma, which was stored at −80℃. Briefly, Metabolites with more than 10% missing values were removed.
Remaining metabolites were imputed using the minimum method.
For multivariate analysis, data were autoscaled, and PCA and loadings were visualized using the ggplot package. 20 Fold changes were calculated and a Welch test was applied to the log-transformed dataset to calculate statistical differences between sample groups for each metabolite.

| Statistics
Data are presented in dot plots (median/interquartile ranges) for the number of samples. Mann-Whitney U test was used to assess statistical significance. Data were analysed and figures produced by Graphpad Prism 8. P-values lower than .05 were considered significant.

| Genome-wide transcriptome analysis of UUOmurine model treated with Bemcentinib
To investigate the effect of bemcentinib in the UUO model, we ana-  (Table 1).
To obtain a general overview, we performed a PCA, which revealed that the majority of the variance (66.5%) in the dataset could be attributed to ligation (PC1), whereas bemcentinib treatment ex- In sharp contrast, AKT/PI3K signalling pathway genes were generally up-regulated upon ligation, but bemcentinib largely reverted this transcription pattern (Figure 2A).
Since Axl is a tyrosine kinase signal transductor, we analysed differential expression of genes from MAPK-related signalling cascades and pathways acquired from the KEGG database. This analysis revealed that a majority of genes in these pathways were also up-regulated upon ligation, and bemcentinib treatment resulted in a considerable reversal of their transcription pattern ( Figure 2B).

| Effect of ligation and bemcentinib treatment in mitochondrial-related gene expression
To further investigate how ligation and treatment with bemcentinib affected transcription of mitochondrial-related genes, we performed several multivariate analyses on a data subset filtered through MitoCarta v2 public database.
A PCA on significant features revealed that the majority of variance (90.7%) in the dataset may be attributed to ligation, as shown in principal component 1 (PC1). The effect of bemcentinib, as explained by PC2, consisted in a 3.0% difference between the two treatment groups ( Figure 3A).
We also performed hierarchical clustering on the top 50 genes with the highest loading in both PC1 and PC2 to investigate which genes most contributed to the variance seen in the PCA, and how ligation and treatment affected their expression. Top-ranked genes in PC1 mostly showed a clear pattern of down-regulation upon ligation, whereas bemcentinib treatment appeared to mildly reverse these effects ( Figure 3B). However, in PC2, we observed a more diverse transcription pattern, identifying four distinct gene clusters where bemcentinib partially reversed the effects of ligation ( Figure 3C).

| Determination of mitochondrial biomass and dysfunction
Three approaches, two based on proteins and the other on DNA quantification, were used to determine if changes in mitochondrial gene expression could be associated with mitochondrial biomass alterations and dysfunction.
At the protein level, we determined two mitochondrial proteins: SDHB, located on the inner membrane of the mitochondria and participating in Citric Acid Cycle and electron respiratory chain, and TOMM20, a translocase located in the mitochondrial outer membrane by immunohistochemistry and western blot. Anti-SDHB staining in ligated vehicle-treated kidneys displayed a clearly weaker signal than in non-ligated kidneys, and the same pattern was observed for anti-TOMM20 ( Figure 4A, B). Most importantly, bemcentinib treatment partially reverted these effects. Comparative positive pixels (%) quantification ( Figure 4C, D) confirmed this data. However, the effect of Bemcentinib was not significant when measured by western blot ( Figure 4F), probably because of the lower method sensitivity and because total protein was extracted including fibrotic tissue, that was removed from the immunohistochemistry quantification.
Notably, with both techniques, SDHB/TOMM20 ratio was significantly lower in ligated kidneys and bemcentinib treatment also reverted this effect ( Figure 4E and G).
At the genomic level, we quantified mitochondrial gene MT-ND1,  0.360 ± 0.126; P = .045). There was no significant difference between non-ligated kidneys with or without bemcentinib treatment ( Figure 4H).

| Metabolomic profiles in bemcentinib and vehicle-treated kidneys
Considering the key role of mitochondria in amino acid metabolism, to complement our transcriptomics study, we performed a serum

| D ISCUSS I ON
In a previous study from our group, we found that tyrosine kinase receptor AXL is involved in the progression of renal fibrosis in a UUO murine model and that AXL inhibitor bemcentinib attenuates disease development. 16 Mitochondrial dysfunction and oxidative stress have been demonstrated to play a role in the pathogenesis of renal fibrosis in both, humans and animal models. 24,25 In this study, we investigated the effects of Axl-inhibitor bemcentinib on mitochondrial dysfunction associated with renal fibrosis.
The main function of mitochondria is to produce energy through respiration and molecular catalysis, thereby playing a central role in cell metabolism. Targeting defective mitochondria-related pathways has also been proposed as a potential treatment of a variety of diseases, including different types of fibrosis, 26,27 diabetic kidney disease and CKD. 28 Here, using RNA sequencing data from a murine UUO model, 29,30 we identified a pattern of dysregulated genes suggesting an impaired mitochondrial bioenergetics in ligated kidneys following 14 days post-surgery, compared to non-ligated kidneys.
In general, mitochondria-related genes were significantly downregulated in ligated compared to non-ligated kidneys. However, this pattern was at least partially reverted in bemcentinib-, compared to vehicle-treated animals, thus indicating that bemcentinib has a potential beneficial effect on mitochondrial dysfunction occurring during renal fibrosis. Interestingly, unlike in ligated kidneys where bemcentinib had a clear effect, in non-ligated kidneys, bemcentinib had little to no effect at the transcriptome level.
Mitochondrial homeostasis is tightly regulated and the disruption of the dynamic processes of mitochondrial biogenesis, fission/ fusion and mitophagy impacts on renal injury and recovery. 31 To compare mitochondrial biomass and dysfunction in our model, we used two methods. Firstly, we quantified mtDNA. Moreover, since mitochondria are dynamic organelles with a variable number of circular mtDNAs, we also analysed them at the protein level using two mitochondrial biomarkers, TOMM20 and SDHB. TOMM20 is a translocase located in the mitochondrial outer membrane that has an essential role in the specificity of mitochondrial protein import, 32 whereas SDHB is part of the complex II of the respiratory chain located on the inner membrane of the mitochondrion linking citric acid cycle and oxidative phosphorylation, two critically important pathways in energy conversion. By using either approach, we found that mitochondria biomass was significantly decreased in ligated compared to non-ligated kidneys, but bemcentinib significantly increased mitochondria biomass compared to vehicletreated animals.
A common by-product of mitochondrial OXPHOS is represented by ROS, which are efficiently removed by different scavenging systems such as glutathione oxidation-reduction cycle and superoxide dismutases (SODs). Increased ROS induces oxidative stress, 33 associated with CKD and fibrosis progression. 11, 34 We observed a F I G U R E 4 Visualization and quantification of TOMM20 and SDHB staining. Immunohistochemical analysis of sections from ligated and non-ligated kidneys with or without bemcentinib treatment from male C57BI/6 mice after 14 d of ureteral obstruction for (A) SDHB and (B) TOMM20 protein expression. Quantitative analysis of positive pixels (%) with fibrotic tissue subtracted from quantification is provided for (C) TOMM20, (D) SDHB, (E) SDHB/TOMM20 ratio (paired samples). F, Western blot analysis of SDHB and TOMM20. G, Western blot protein quantification SDHB/TOMM20 ratio (paired samples). H, mtDNA quantification. All data were analysed by Mann-Whitney U test down-regulation of antioxidant genes upon ligation, but a positive effect of bemcentinib, that increased the expression of these genes.
Mitochondria play key roles in amino-acid metabolism and are involved in both catabolic and anabolic processes. 7 In particular, renal mitochondria critically contribute to nitrogen and amino-acid homeostasis, and ammonia disposal, by renal deamidation of glutamine to glutamate (GLU). 35  were decreased by bemcentinib treatment. In contrast, we did not observe any differences in levels of urea between the bemcentiniband vehicle-treated mice, suggesting an adaptation of hepatic production to compensate for the reduced renal capacity to dispose of urea through GLU-GLN metabolism.
It should be noted that although it is widely used as a renal fibrosis model, UUO requires aggressive surgery that leads to a rapid interstitial inflammation at difference with the slow progression observed in humans. 17,45 Nevertheless, this model has proven useful both for the identification of biomarkers and for the development of new treatments. [46][47][48] Furthermore, our study documents that bemcentinib was able to improve mitochondrial function at the molecular level despite the intrusiveness of the experimental method.

F I G U R E 5
Multivariate analysis of serum amino acids and related metabolites. A, PCA shows groups of samples clustering according to ligation and treatment. The variance seen in PC2 reflects ligation and how treatment with bemcentinib appears to reverse the effects of ligation whereas PC1 represents differences between samples. B, Loading plot from PC2 shows how serum levels of modified amino acids separate different treatment groups whereas PC1 mainly consists of essential amino acids. C, Fold changes and statistics of serum metabolites In conclusion, our data indicate that mitochondria and mitochondrial-related pathways are dramatically affected by UUO surgery. Bemcentinib partially reverses the effects of UUO, without affecting non-ligated kidneys or SHAM-operated mice, and thereby qualifies as a promising treatment in kidney disease.

ACK N OWLED G EM ENTS
We are very grateful to Giulio Spagnoli, University of Basel, Switzerland, for having provided reviewing and language editing services. We are thankful to DA Sandnes for the help with IMH. This project was supported by open project funding from the Western Norway Regional Health Authority (HelseVest F-11546, # 912233) and by BerGenBio ASA. BerGenBio ASA also provided the bemcentinib for this study.

CO N FLI C T O F I NTE R E S T
This study was partly supported by BerGenBio ASA, which has also provided the bemcentinib. JBL declares ownership in BerGenBio ASA. GG is employed by BerGenBio ASA.

DATA AVA I L A B I L I T Y S TAT E M E N T
Transcriptomics data analysed in this study were a re-analysis of existing data, which are available at the repository Gene