Deletion of the Mitochondrial Complex-IV Cofactor Heme A:Farnesyltransferase Causes Focal Segmental Glomerulosclerosis and Interferon Response

Abstract

Mutations in mitochondrial DNA as well as nuclear-encoded mitochondrial proteins have been reported to cause tubulointerstitial kidney diseases and focal segmental glomerulosclerosis (FSGS). Recently, genes and pathways affecting mitochondrial turnover and permeability have been implicated in adult-onset FSGS. Furthermore, dysfunctioning mitochondria may be capable of engaging intracellular innate immune-sensing pathways. To determine the impact of mitochondrial dysfunction in FSGS and secondary innate immune responses, we generated Cre/loxP transgenic mice to generate a loss-of-function deletion mutation of the complex IV assembly cofactor heme A:farnesyltransferase (COX10) restricted to cells of the developing nephrons. These mice develop severe, early-onset FSGS with innate immune activation and die prematurely with kidney failure. Mutant kidneys showed loss of glomerular and tubular epithelial function, epithelial apoptosis, and, in addition, a marked interferon response. In vitro modeling of Cox10 deletion in primary kidney epithelium compromises oxygen consumption, ATP generation, and induces oxidative stress. In addition, loss of Cox10 triggers a selective interferon response, which may be caused by the leak of mitochondrial DNA into the cytosol activating the intracellular DNA sensor, stimulator of interferon genes. This new animal model provides a mechanism to study mitochondrial dysfunction in vivo and demonstrates a direct link between mitochondrial dysfunction and intracellular innate immune response.

Focal segmental glomerulosclerosis (FSGS) is a pathologic diagnosis made in the setting of proteinuric kidney disease. It is characterized by initial disease of podocytes of the glomerular tuft, frequently associated with tubulointerstitial disease, which is most often proportionate to the severity of glomerular disease. Several mutations in causal genes have been identified in familial FSGS, including those coding for structural proteins of the slit diaphragm of the podocyte.¹ Nevertheless, most patients do not have a recognized mutation that causes disease, and, in some circumstances, FSGS may be an adaptive manifestation to hyperfiltration. During the past 15 years, case reports have suggested mutations of mitochondrial DNA (mtDNA) may cause several familial and sporadic cases of FSGS, as well as tubular diseases, such as Fanconi syndrome and tubulointerstitial nephritis. Rarely, nuclear genes coding for mitochondrial proteins have been implicated, but these reports have often described kidney disease in the setting of systemic disorders, pointing to widespread mitochondrial disease.2, 3, 4 In addition, xenobiotics to treat viral infections and malignancies have been proposed to directly damage mitochondria, causing tubulointerstitial disease. Recently, polymorphisms in an innate immune protein, apolipoprotein L1 (APOL1), have been found to be associated with high risk of developing FSGS in patients of African ancestry, and potentially playing a causative role in a much larger fraction of patients who present with proteinuric progressive kidney disease.⁵ Biochemical and cell-based studies suggest intracellular, not circulating, APOL1 may directly disrupt the mitochondrial transmembrane potential in podocytes, to cause podocyte dysfunction.6, 7 In addition, disruption or dysfunction of mitochondrial turnover by loss of mitochondrial quality control (mitophagy) has been demonstrated to directly cause FSGS in patients and animal models.8, 9 Such studies strongly implicate mitochondrial dysfunction in the pathogenesis of proteinuric kidney diseases.10, 11 Moreover, single-nucleotide polymorphisms in the regulatory sequences of the energy sensor and mitochondrial regulator, AMP-activated protein kinase (PPKAG2), and other mitochondrial metabolism genes, such as carbamoyl phosphate synthase I (CPS1) and glucokinase regulator (GCKR), have been implicated in hypertensive and diabetic kidney diseases with proteinuria from highly powered genome-wide association studies.¹²

Several mechanisms by which mitochondrial dysfunction can result in FSGS have been reported.13, 14, 15, 16, 17, 18 Mitochondrial disease causes increased intracellular reactive oxygen species generation. In certain circumstances, oxidative damage to local organelles, including endoplasmic reticulum and DNA, can occur, although the absolute importance of mitochondrial reactive oxygen species to disease remains unclear. Impaired clearance of defective mitochondrial components, known as mitophagy, also leads to enlarged mitochondria generating reduced ATP and enhanced reactive oxygen species. A reduction in oxidative phosphorylation can compromise cellular function, particularly in cells that rely heavily on mitochondria for ATP generation, including podocytes and epithelium of the kidney.

Mitochondria have many functions in addition to metabolism and oxidative phosphorylation. They are critical in detoxification, iron-sulfur cluster biogenesis, intracellular calcium regulation, and central regulation of cell death/survival pathways, including caspase-mediated apoptosis. In addition, mitochondria have been shown to be important sensors and structural regulators of intracellular innate immune responses to pathogens.19, 20, 21, 22, 23, 24 Some of the intracellular sensors used to detect intracellular pathogens include the retinoic acid–inducible gene-I–like proteins, stimulator of interferon genes (STING), mitochondrial antiviral signaling protein, cyclic GMP-AMP synthase, and interferon-γ–inducible protein 16, which sense viral RNA and DNA.19, 20, 25, 26, 27, 28, 29 It has been proposed that the leakage of mtDNA from dysfunctioning or disrupted mitochondria could additionally directly activate STING, the DNA sensor, providing a direct potential link between mitochondrial disease and inflammatory response.20, 30 However, abundant extracellular mtDNA can be found in tissue injury but does not activate toll-like receptors that sense extracellular foreign DNA, except in rare circumstances in which the DNA is reported to be highly oxidized.³¹ Therefore, the role of mtDNA leakage in innate immune responses and its contribution to pathogenesis in the setting of noninfectious kidney disease are unexplored.

In the following studies, we validated the hypothesis that mutation of a single gene that contributes to complex IV (COX-IV) stability in the electron transport chain is sufficient to cause the pathologic manifestations of FSGS and determined in a controlled manner whether mutation-driven mitochondrial dysfunction directly activates sensors of viral DNA sufficient to trigger an interferon response. To this end, nephron-specific complex IV assembly cofactor heme A:farnesyltransferase (Cox10) conditional knockout mice (Cox10^fl/fl)³² expressing Cre under the control of the Six2 promotor (Six2-Cre⁺;Cox10^fl/fl:Cox10^Δ/Δ) were generated. The nephron is the structural and functional unit involved in all key processes, including filtration, reabsorption, secretion, and excretion. In bigenic, homozygous, floxed mice, Cox10 was disrupted in Six2-expressing nephron progenitors and their progeny, which form the mesenchyme-derived kidney epithelium, including podocytes, parietal epithelial cells, proximal tubules, loops of Henle, and distal tubules.33, 34 Cox10 encodes a heme A/farnesyl transferase,³⁵ which is required for the correct assembly of cytochrome c oxidase (COX-IV; alias the complex IV of the electron transport chain), a critical complex in mitochondrial respiratory ATP production.³⁶ Genetic deletion of Cox10 causes COX-IV deficiency, in turn leading to mitochondrial dysfunction.32, 37, 38

Materials and Methods

Mice

Mice harboring loxP-flanked Cox10 alleles³² were crossed to a mouse strain expressing Cre under the Six2 promotor to generate nephron-specific Cox10 conditional knockout mice (Six2-Cre⁺;Cox10^fl/fl:Cox10^Δ/Δ) and the littermate controls carrying either unexcised floxed or Cox10 wild-type (WT) gene on both alleles (Cox10^+/+). To test prenatal lethality of Cox10^Δ/Δ mice, Six2-Cre^−/−;Cox10^fl/fl mice were bred with Six2-Cre^+/−;Cox10^fl/+. Studies were performed under approved Institutional Animal Care and Use Committee protocols (0489-2013 and 755) at Biogen Inc. (Cambridge, MA) and the University of Washington (Seattle, WA; protocol 4244).³⁹

Renal Histopathology

Kidneys were resected after systemic perfusion with ice-cold phosphate-buffered saline (PBS). Paraffin-embedded sections of kidney fixed with neutral-buffered formalin were used for periodic acid-Schiff, methenamine silver, and picrosirius red staining, as described previously.⁹

Histologic Analysis

Glomerular pathology was evaluated by an observer-blinded pathologist (Y.W.) assessing all glomeruli of whole cortex of a sagittal section of the whole kidney from each mouse (×40 magnification) in periodic acid-Schiff– or methenamine silver–stained kidney sections, as described in several previous publications.9, 40, 41 Between 42 and 71 glomeruli were scored sequentially per animal. Each glomerular section was quantified for the percentage of sclerotic glomeruli (capillary loop obliteration and replacement with sclerotic matrix). Up to 25% of tuft involved scored 1, up to 50% of tuft scored 2, up to 75% of tuft scored 3, and up to 100% of tuft scored 4. Visceral epithelial cell hyperplasia with collapsing-type lesions (pseudocrescents) was defined as more than two layers of epithelial cells with the parietal basement membrane. Tubular pathology was evaluated by assessing 10 randomly selected microscopic fields (×10 magnification) of the outer medulla per animal in periodic acid-Schiff–stained kidney sections and determining the numbers of dilated, atrophic tubules and tubules with casts.

Electron Microscopy

The protocol for tissue preparation and staining for transmission electron microscopy has been described previously.⁴² Grids were scanned using a JEM-1380 electron microscope (Jeol, Peabody, MA).

Renal Function

Serum and urinary creatinine were measured using Creatinine Liquid Reagents Assay (Diazyme, San Diego, CA) or a urinary creatinine detection kit (Thermo Fisher Scientific, Carlsbad, CA). Blood urea nitrogen levels were detected by blood urea nitrogen (Pointe Scientific, Canton, MI). Urinary albumin concentration was measured using an Albuwell M kit (Exocell, Philadelphia, PA).

Immunostaining of Kidney Sections

Kidneys were fixed with periodate-lysine-paraformaldehyde fixative. After immersion in PBS, including 18% sucrose at 4°C overnight, they were embedded and frozen in optimal cutting temperature compound (Sakura Finetek, Torrance, CA). Cryostat-cut mouse kidney sections (5 μm thick) were stained for the following: fibrosis, using Cy3-conjugated anti–α-smooth muscle actin antibody (Ab;; 1:200; clone 1A4; Sigma-Aldrich, St. Louis, MO); tubular injury, using polyclonal rabbit anti-human kidney injury molecule-1 Ab (1:200; catalog ab47635; Abcam, Cambridge, MA); apoptosis, using rabbit anti-human cleaved caspase-3 Ab (1:50; clone 5A1E; Cell Signaling Technology, Danvers, MA); proximal tubules, using fluorescein isothiocyanate–conjugated lotus tetragonolobus lectin (LTL; 1:200; catalog FL-1321; Vector Laboratories, Burlingame, CA); and macrophages, using rat anti-mouse F4/80 Ab (1:400; clone BM-8; Thermo Fisher Scientific). This was followed by Cy3-conjugated donkey anti-rabbit IgG Ab, Alexa Fluor 488 donkey anti-rabbit IgG antibody, and Cy3-conjugated donkey anti-rat IgG Ab (1:400; Jackson ImmunoResearch, West Grove, PA). All sections were mounted with mounting medium containing 4′,6-diamidino-2-phenylindole (DAPI; catalog P36931; Thermo Fisher Scientific). Images were taken using an LSM 710 laser-scanning confocal microscope (Zeiss, Thornwood, NY). The magnitude of staining was analyzed in 10 microscopic fields per animal using Adobe Photoshop CC 2017 (Adobe Systems, San Jose, CA).

SDS-PAGE and Western Blot Analysis

SDS-PAGE and Western blot analysis were performed as described previously.9, 43 To detect total and phosphorylated STING, protein electrophoresis was performed using 7.5% SuperSep Phos-tag gel (Wako, Richmond, VA). Primary antibodies used were as follows: polyclonal rabbit anti-cytochrome c oxidase subunit I (COX1; 1:1000; catalog PA5-26688; Thermo Fisher Scientific), rabbit anti-COX10 (1:1000; catalog MBS2526567; MyBiosource Inc., San Diego, CA), monoclonal rabbit anti-interferon regulatory factor 1 (IRF1; 1:1000; clone D5E4; Cell Signaling Technology), rabbit anti–2′-5′-oligoadenylate synthase 1 (OAS1; 1:1000; clone D1W3A; Cell Signaling Technology), rabbit anti-STAT3 (1:1000; clone 79D7; Cell Signaling Technology), rabbit anti-STING (1:1000; clone D1V5L; Cell Signaling Technology), rabbit anti–TRAF family member-associated NFkappaB activator (TANK)-binding kinase-1 [TBK1; 1:2000; clone EPR2867(2)-19; Abcam], rabbit anti–phosphorylated TBK1 (1:250; clone D52C2; Cell Signaling Technology), and mouse anti–glyceraldehyde-3-phosphate dehydrogenase Ab (1:2000; clone GT239; GeneTex, Irvine, CA).

Gene Ontology and Pathway Enrichment Analysis

Total RNA was extracted from kidney biopsy specimens from Cox10^Δ/Δ and Cox10^+/+ mice using RNeasy Plus Mini Kit (Qiagen, Hilden, Germany). Purity was determined by measuring the 260/280 and 260/239 ratio. Gene expression profiling was performed using next-generation sequencing on the Illumina HiSeq 2500 platform producing 50-bp paired-end reads. Total reads were mapped using the aligner STAR.⁴⁴ Genes were quantified with RSEM,⁴⁵ and differential expression was determined by DESeq2 using Array Studio 10 (OmicSoft, Cary, NC).⁴⁶ Significantly differentially expressed genes were selected according to the following criteria: false discovery rate–corrected P < 0.001 and fold change >1.4-fold. Analysis was subjected to the enrichment test for gene ontology and pathway analysis through the use of Ingenuity Pathway Analysis version 01-10⁴⁷ (Qiagen) or DAVID version 6.8 (http://david.ncifcrf.gov).48, 49 Differentially expressed genes were subjected to functional protein association analysis using STRING database version 10.5 (http://www.string-db.org, last accessed March 16, 2018, registration required). Differentially expressed genes were interferon-stimulated genes (ISGs) and were identified using the online database INTERFEROME version 2.01 (http://www.interferome.org/interferome/home.jspx, last accessed March 19, 2018). Because the RNA was directly sequenced, PCR confirmation of transcriptional changes was not considered necessary. All RNA sequencing data and expression values are available at the Gene Expression Omnibus repository at the National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/geo; accession number GSE117681).

Isolation and Culture of Primary Kidney Tubular Epithelial Cells

Kidneys from WT mice at 2 to 3 months of age were harvested, decapsulated, diced, and then incubated at 37°C for 60 minutes with liberase thermolysin low (0.2 mg/mL; Roche, Indianapolis, IN) in the presence of DNase I (100 U/mL) in serum-free Dulbecco's modified Eagle's medium/F12 medium. After filtration (40 μm), proximal tubular epithelial cells (PTECs) were enriched using Dynabeads biotin-binder (Thermo Fisher Scientific) in combination with biotinylated LTL (Vector Laboratories). Unbound cells (including uncaptured PTECs) were further purified for TECs by immunomagnetic bead purification using magnetic-activated cell sorting anti-mouse CD326 [epithelial cell adhesion molecule (EpCAM)⁺] microbeads (Miltenyi Biotec, Auburn, CA). PTECs (LTL⁺) and TECs (EpCAM⁺) were cultured in renal epithelial cell growth basal medium renal growth basal medium (Lonza, Walkersville, MD) supplemented with renal epithelial cell growth medium SingleQuot Kit Supplements & Growth Factors (Lonza).

Cox10 Silencing

PTECs and TECs were transfected with 100 nmol/L of siRNA for Cox10 (Silencer Select siRNA; catalog number 4390771; identifier s88774) and control siRNA (Silencer Select Negative Control#2; catalog number 4390846) preincubated with Lipofectamine RNAiMAX in OptiMEM (all from Thermo Fisher Scientific) for 5 minutes, according to the manufacturer's protocol. At days 2 and 5 of transfection, transfected cells were analyzed for Cox10 knockdown and transcriptional changes.

Quantitative PCR

Total RNA was extracted using the RNeasy Plus Mini Kit (Qiagen). Purity was determined by measuring the 260/280 and 260/230 ratios. cDNA was synthesized using random primers and reverse transcriptase (iScript cDNA Synthesis Kit; Bio-Rad, Hercules, CA). TaqMan real-time PCR was performed using the QuantStudio 12K Flex Real-Time PCR System in a duplex reaction with the gene of interest (FAM probe) and Gapdh as the housekeeping gene (VIC probe). The mRNA expression was calculated using the ΔΔC_T method and expressed as a fold difference relative to the untreated group. All TaqMan probes used in this study are listed in Table 1.

Table 1.

TaqMan Primers Used in the Study

Gene	Assay identifier
Casp3	Mm01195085_m1
Casp6	Mm01321726_g1
Casp7	Mm01259818_m1
Casp8	Mm01255716_m1
Cox10 exon 3–4	Mm01332417_m1
Cox10 exon 6–7	Mm00617695_m1
Cxcl10	Mm00445234_m1
Ddx58	Mm01216853_m1
Gapdh∗	Mm99999915_g1
Ifih1	Mm00459183_m1
Ifit1	Mm00515153_m1
Ifit3	Mm01704846_s1
Ifna2	Mm00833961_s1
Ifna4	Mm00833969_s1
Ifnb1	Mm00439552_s1
Il1b	Mm00434228_m1
Irf7	Mm00516793_g1
Isg15	Mm01705338_s1
Nlrx1	Mm00617978_m1
Sdhd	Mm00546511_m1
Sod1	Mm01344233_g1
Sod2	Mm01313000_m1
Stat1	Mm00439531_m1
Tnf	Mm00443258_m1
Usp18	Mm01188805_m1

^∗Housekeeping gene.

Measurement of Mitochondrial Function

The oxygen consumption rate was measured using a Seahorse XFe96 extracellular flux analyzer (Agilent Technologies, Santa Clara, CA), according to the manufacturer's instructions. Total cellular ATP production by cultured TECs was measured using the CellTiter-Glo Luminescent Cell Viability Assay kit (Promega, Madison, WI), according to the manufacturer's instructions. Cell viability and mitochondrial membrane potential were measured using the HCS Mitochondrial Health Kit (catalog number H10295; Thermo Fisher Scientific), according to the manufacturer's instructions. Mitochondrial membrane potential was normalized by mitochondrial mass by coincubating TECs with 75 nmol/L MitoTracker Green FM (catalog number M7514; Thermo Fisher Scientific) and determining the ratio of mitochondrial membrane potential (MMP)/MitoTracker Green FM fluorescence intensity.

Immunofluorescence Microscopy for mtDNA

Before staining, cultured cells were washed in PBS, fixed with 4% paraformaldehyde for 20 minutes at room temperature, permeabilized with 0.1% Triton X-100 in PBS for 5 minutes, and blocked with 10% fetal bovine serum in PBS for 30 minutes. Then, cells were costained with polyclonal rabbit anti-human superoxide dismutase 2 Ab (1:50; catalog number ab13353; Abcam) for mitochondria and mouse anti-DNA Ab (1:50; clone AC-30-19; EMD Millipore, Burlington, MA), followed by Alexa Fluor 488–conjugated goat anti-rabbit IgG Ab and Alexa Fluor 594 goat anti-mouse IgG antibody (1:750; all from Thermo Fisher Scientific). Cells were washed 3× in PBS containing 5 μg/mL Hoechst 33342 (Thermo Fisher Scientific) for 5 minutes and imaged on an Opera Phenix High-Content Screening System with a 63× water-immersed objective (PerkinElmer, Waltham, MA).

Detection of Cytosolic mtDNA

Cytosolic and whole-cell DNAs were extracted as described previously.20, 50 Briefly, cultured TECs (8 × 10⁶) were divided into two equal aliquots. Whole-cell DNA was extracted from one aliquot using DNeasy Blood & Tissue Kit (Qiagen) after a 10-minute incubation with proteinase K (500 μL; 100 mmol/L tris-HCl, pH 8.5, 5 mmol/L EDTA, pH 8.0, 0.2% SDS, 200 mmol/L NaCl, and 100 μg/mL proteinase K). Cytosolic mtDNA was purified from the second aliquot using QIAquick Nucleotide Removal Columns (Qiagen) after digestion with digitonin [500 μL; 150 mmol/L NaCl, 50 mmol/L HEPES, pH 7.4, and 20 μg/mL digitonin (EMD Millipore)] and subsequent centrifugations: 3× at 980 × g for 5 minutes and 1× at 17,000 × g for 25 minutes. Quantitative PCR was performed on both whole-cell and cytosolic fractions using nuclear DNA primers (Tert) and mtDNA primers (Dloop1 to Dloop3), and the C_T values for mtDNA-encoded genes obtained from whole-cell extracts served as normalization controls for the mtDNA values obtained from the cytosolic fractions. Primers used were as follows: Tert, 5′-CTAGCTCATGTGTCAAGACCCTCTT-3′ (forward) and 5′-GCCAGCACGTTTCTCTCGTT-3′ (reverse); mtDNA Dloop1, 5′-AATCTACCATCCTCCGTGAAACC-3′ (forward) and 5′-TCAGTTTAGCTACCCCCAAGTTTAA-3′ (reverse); mtDNA Dloop2, 5′-CCCTTCCCCATTTGGTCT-3′ (forward) and 5′-TGGTTTCACGGAGGATGG-3′ (reverse); and mtDNA Dloop3, 5′-TCCTCCGTGAAACCAACAA-3′ (forward) and 5′-AGCGAGAAGAGGGGCATT-3′ (reverse).

mtDNA Preparation and Transfection into TECs

The 2.6- and 3.6-kb fragments of mtDNA were amplified from whole-cell DNA extracted from untreated TECs in the presence or absence of 8-oxo-dGTP (TriLink Biotechnologies, San Diego, CA), as described previously.⁵¹ Thereby, we used following primers: 2.6 kb, 5′-AATCTACCATCCTCCGTGAAACC-3′ (forward) and 5′-ATACCGCGGCCGTTAAA-3′ (reverse); and 3.6 kb, 5′-AACGGATCCACAGCCGTA-3′ (forward) and 5′-TGTTGGGTTGTTTGATCCTG-3′ (reverse). Without further purification, amplified mtDNA fragments (4 μg/mL) were transfected into TECs using Lipofectamine RNAiMAX Transfection Reagent (Thermo Fisher Scientific).

Statistical Analysis

Data represent the means ± SEM. Differences between groups were assessed by the U-test. P < 0.05 was considered statistically significant.

Results

Cox10^Δ/Δ Mice Develop a Severe Form of FSGS

To disrupt the Cox10 gene in the nephron only, Cox10^fl/fl mice were crossed with a well-validated bacterial artificial chromosome transgenic mouse that expresses Cre under the control of the Six2 promoter to generate nephron-specific Cox10 knockout mice (Cox10^Δ/Δ). Six2 encodes a transcription factor exclusively expressed in cap mesenchyme during nephrogenesis.³³ Six2⁺ cap mesenchyme cells are progenitors that produce all of the epithelial cells of the nephron, except the collecting duct. Thereby, exon 6 of Cox10 encoding part of the enzymatic active site of COX10 was deleted in nephron progenitors and their progeny (podocytes, parietal epithelial cells, proximal tubules, loops of Henle, and distal tubules),9, 33 resulting in the out-of-frame truncation and loss of COX10 protein in the glomerular and tubular epithelium, except the collecting ducts.³² Cohorts of mice were bred, and kidney pathology in Cox10^Δ/Δ and WT littermates (Cox10^+/+) was evaluated at 3 months of age. Successful mutation of the Cox10 gene was confirmed by quantification of COX10 protein. COX10 protein was markedly decreased in the whole kidney of Cox10^Δ/Δ compared with Cox10^+/+ kidneys. Correspondingly, Cox10^Δ/Δ kidneys showed decreased levels of the main subunit of cytochrome c oxidase, COX-IV (named COX1), indicating that Cox10^Δ/Δ kidneys harbor less fully assembled COX-IV because of conditional gene targeting (Supplemental Figure S1A). Cox10^Δ/Δ kidneys have typical features of FSGS (Figure 1A). These features include a segmental pattern of scarring, showing obliterated capillary loops, hyaline deposition, and adhesion of tuft to the Bowman capsule; glomeruli with perihilar predominance of sclerosis or hyalinosis; segmental sclerosis at the tubular pole of the glomerulus; and glomeruli with tuft collapse associated with glomerular epithelial cell proliferation (Figure 1, A and B). These are similar to the features of FSGS in the nephron-specific autophagy-deficient mouse model reported previously.⁹

Figure 1.

Cox10 deficiency in the nephron causes severe forms of focal segmental glomerulosclerosis. A and B: Representative photomicrographs taken from periodic acid-Schiff (PAS)–stained kidney cortical sections of Cox10^Δ/Δ (Six2-Cre⁺;Cox10^fl/fl; knockout phenotype) and Cox10^+/+ (Six2-Cre⁻;Cox10^fl/fl; wild-type phenotype) mice (A) and methenamine silver (silver)–stained kidney cortical sections of Cox10^Δ/Δ mice at 3 months of age (B). C: Evaluation of FSGS in Cox10^Δ/Δ and Cox10^+/+ kidneys using an index system and based on the percentage of sclerotic glomeruli and glomeruli with crescent formation. Statistics analyzed using the U-test. D: Representative electron micrographs taken from glomeruli of Cox10^Δ/Δ and Cox10^+/+ kidneys. Arrows indicate podocyte foot processes. Data are expressed as means ± SEM (C). n = 4 to 9 (C). ^∗P ≤ 0.05, ^∗∗P ≤ 0.01. Scale bars: 50 μm (A and B); 2 μm (D).

Cox10^Δ/Δ mice showed an advanced stage of glomerulosclerosis at 3 months of age, where almost all glomeruli had segmental scars. Segmental sclerosis was present in 18.0% ± 4.6%, and global glomerulosclerosis was present in 48.3% ± 8.7%. In addition, glomerular epithelial proliferation (pseudocrescents), similar to the collapsing form of FSGS, was found in 45.3% ± 8.5% (Figure 1C). Electron microscopy of glomeruli from Cox10^Δ/Δ mice at 2.5 months of age showed typical features of a podocytopathy with extensive effacement of podocyte foot processes along the glomerular basement membrane and reduction of primary and secondary podocyte processes. In addition, mitochondria of remaining podocytes in sclerotic glomeruli of Cox10^Δ/Δ mice showed morphologic changes characterized by increased diameter, decreased length, and partial loss of cristae (Figure 1D).

Cox10^Δ/Δ Mice Are Susceptible to Prenatal and Postnatal Lethality with Renal Failure

While breeding mice, it was noticed that Cox10^Δ/Δ mice have a shorter lifespan than WT littermates and are born at submendelian ratios. Hence, survival was first examined in 55 Cox10^Δ/Δ mice, and the median survival time of Cox10^Δ/Δ mice was found to be 38 days; all Cox10^Δ/Δ mice that were analyzed died before 20 weeks of age (Figure 2A). In addition, Six2-Cre^−/−;Cox10^fl/fl mice were crossed to Six2-Cre^+/−;Cox10^fl/+ mice to verify that the disruption of Cox10 in the nephron results in prenatal lethality. Mendelian ratios would predict equal proportions for all possible genotypes (Six2-Cre^+/−;Cox10^fl/+, Six2-Cre^+/−;Cox10^fl/fl, Six2-Cre^−/−;Cox10^fl/+, and Six2-Cre^−/−;Cox10^fl/fl). However, the number of Cox10^Δ/Δ mice was only 4 of 37 pups obtained from five litters with this breeding strategy, which was a submendelian ratio (Figure 2B), suggesting that a substantial proportion of Cox10^Δ/Δ mice die in utero. These results predict that disease occurs in the early development of the mouse and should be detectable early after birth. Body weights of male Cox10^Δ/Δ mice surviving beyond the first 3 months showed a 25% reduction at 2 months and a 37% reduction at 3 months compared with WT littermates (Figure 2B). The same tendency was observed in female groups (Supplemental Figure S1B). At 2 to 3 months, kidney weight and volume were markedly decreased in Cox10^Δ/Δ mice compared with age-matched controls (Figure 2C and Supplemental Figure S1C). Consistent with the severity of glomerulosclerosis, Cox10^Δ/Δ mice showed impaired kidney function with albuminuria, as evaluated by the levels of serum creatinine, blood urea nitrogen, urinary albumin, and urinary albumin/creatinine ratio compared with age- and sex-matched mice (Figure 2D and Supplemental Figure S1D).

Figure 2.

Cox10 deficiency in the nephron produces prenatal and postnatal lethality with renal failure. A: Left panel: Kaplan-Meier curve showing the survival of Cox10^Δ/Δ compared with Cox10^+/+ mice. Right panel: Breeding scheme to test Cox10 deficiency–associated embryogenic lethality and a pie chart showing genotype distribution of the offspring. B: Body weights of male Cox10^Δ/Δ compared with Cox10^+/+ mice at 2 and 3 months of age. C: Representative photomicrographs and graphs showing kidney weights of male Cox10^Δ/Δ compared with Cox10^+/+ mice at the time of natural death and euthanasia. D: Assessment of kidney function of Cox10^Δ/Δ and Cox10^+/+ mice by serum creatinine, blood urea nitrogen (BUN), and urinary albumin levels. Statistics analyzed using the U-test. Data are expressed as means ± SEM (B–D). n = 55 (A, right panel); n = 4 mice aged 2 months (B); n = 7 to 11 mice aged 3 months (B); n = 8 (C); n = 11 (D). ^∗P ≤ 0.05, ^∗∗∗∗P ≤ 0.0001. Scale bar = 1 cm (C).

Cox10^Δ/Δ Mice Develop Tubulointerstitial Damage and Fibrosis

Human FSGS is frequently associated with severe tubulointerstitial disease.⁹ Although the precise drivers of nonglomerular disease in FSGS are not fully elucidated, prevailing models suggest that this is predominantly secondary to podocyte disease.9, 52 Nevertheless, the causes of podocyte disease may also simultaneously cause tubular disease, as could be the case when mitochondrial function is disrupted. In Cox10^Δ/Δ mice, in which COX10 is deleted in podocytes as well as tubules, severe tubular pathology was seen at 2.5 to 3 months of age. A considerable number of renal tubules in the outer medulla of Cox10^Δ/Δ kidney were atrophic (30.0 ± 2.0 versus 5.1 ± 0.6 tubules per microscopic field in Cox10^+/+ kidney) and dilated (7.9 ± 1.4 versus 1.3 ± 0.2 tubules per microscopic field in Cox10^+/+ kidney). A total of 5.3% ± 0.4% of renal tubules per microscopic field (versus 0.1% ± 0.0% tubules in Cox10^+/+ kidney) showed casts in Cox10^Δ/Δ kidney (Figure 3A). Interstitial fibrosis is a hallmark of tubulointerstitial disease⁵³ and represents a common pathway leading to progression of chronic kidney disease.⁵⁴ Therefore, interstitial fibrosis was studied in kidneys of Cox10^Δ/Δ mice at 2.5 to 3 months of age by two different methods, picrosirius red staining and immunofluorescence staining of α-smooth muscle actin, a marker for activated fibroblasts. Large areas in the cortex and outer medulla of Cox10^Δ/Δ kidney were found to be fibrotic. Fibrotic lesions were especially prominent in the periglomerular region and cortical interstitium (Figure 3, B and C). For evaluation of tubular injury, expression of the tubular injury marker Kim-1 was studied.⁵⁵ Kim-1 is up-regulated in proximal tubules after tubular injury and fibrosis.⁵⁶ To colocalize Kim-1 to proximal tubules, fluorescently tagged LTL was applied to detect proximal tubules. In Cox10^Δ/Δ kidneys at 2.5 to 3 months of age, most LTL⁺ proximal tubules coexpressed Kim-1 (Figure 3D). In addition, increased expression of cleaved caspase-3, an indicator of cellular apoptosis, was seen in renal tubules of Cox10^Δ/Δ kidneys (Figure 3E). Next, the kidneys from Cox10^Δ/Δ and Cox10^+/+ mice were examined for the magnitude of macrophage infiltration. Macrophages are found in normal kidney and in increased numbers in kidney injury. When inflammation occurs, the pool of tissue-resident and monocyte-derived macrophages is expanded, being provided with growth factors required for activation, survival, and proliferation.57, 58 In Cox10^Δ/Δ kidneys, increased numbers of macrophages were present in patches, indicative of ongoing or nonresolving inflammation (Figure 3F).

Figure 3.

Cox10 deficiency in the nephron leads to tubulointerstitial damage. A: Representative photomicrographs taken from periodic acid-Schiff (PAS)–stained kidney sections of Cox10^Δ/Δ and Cox10^+/+ mice at 3 months of age. Using PAS-stained kidney sections, tubular damage was evaluated on the basis of the percentage of dilated, atrophic tubules (T); tubules were in Cox10^Δ/Δ compared with Cox10^+/+ kidney. B–F: Evaluation of fibrosis (B), activated fibroblasts (C), proximal tubular injury (D), apoptosis (E), and macrophage infiltration in kidneys (F) from Cox10^Δ/Δ and Cox10^+/+ mice at 3 months of age using immunostaining. All statistics analyzed using the U-test. Data are expressed as means ± SEM (A–F). n = 4 to 9 (A and B); n = 3 (C–F). ^∗P ≤ 0.05, ^∗∗P ≤ 0.01. Scale bar = 100 μm (A). Original magnification, ×20 (B–F). KIM-1, kidney injury molecule-1; LTL, lotus tetragonolobus lectin; α-SMA, α-smooth muscle actin.

Cox10^Δ/Δ Kidneys Up-Regulate Nephrotoxicity- and Interferon-Stimulated Genes

To identify genes differentially regulated in Cox10^Δ/Δ versus Cox10^+/+ kidneys and investigate molecular pathways being activated by mitochondrial dysfunction, global transcriptional profiling analysis of sequences mapped to the mouse transcriptome was performed after RNA sequencing of whole-kidney tissue from Cox10^Δ/Δ and Cox10^+/+ mice at 2.5 to 3 months of age. As expected, principal component analysis of the data set separated the samples into two groups: Cox10^Δ/Δ and Cox10^+/+ (Supplemental Figure S1E). A total of 780 genes were found to be significantly up-regulated, and 385 genes were significantly down-regulated, in Cox10^Δ/Δ kidneys (Figure 4A and Supplemental Table S1). The differentially expressed genes showed changes consistent with tubulointerstitial disease, with up-regulation of innate immunity, complement, phagocytosis, and extracellular matrix receptor interaction pathways (Figure 4B). In addition, the genes were enriched for viral signaling pathways, including human T-lymphotropic virus and herpes simplex virus (Figure 4B). Strikingly, there was marked down-regulation of lipid and fatty acid metabolism consistent with a loss of mitochondrial function, as detected by loss of metabolic pathways, butanoate metabolism, and peroxisome and CoA biosynthesis (Figure 4B and Supplemental Figure S2). Unexpectedly, there was also marked loss of multiple transcripts encoded from mtDNA (Figure 4C), in keeping with a high degree of loss of mitochondrial function. In addition, Ingenuity Pathway Analysis showed strong enrichment for genes involved in kidney failure, consistent with the histologic and clinical parameters described in Figures 1, 2, and 3 and summarized in Figure 4D. Because there was strong enrichment for viral-induced signaling pathways, and a notable increase in certain interferon response genes, the gene set was studied, and strong enrichment for activation of ISGs was found. Further analysis of all of the differentially expressed genes indicated that >30% were ISGs, and a secondary analysis of these ISGs indicated that most (80%) were type I interferon response genes (Figure 4, E and F, and Supplemental Table S2). The up-regulation of the ISGs, OAS1, STING (encoded by Tmem173), and the interferon-inducing and interferon-induced transcription factors, IRF1 and STAT3, was validated in Cox10^Δ/Δ versus Cox10^+/+ kidneys at protein levels using Western blot analysis (Figure 4G). This strong type I interferon signature was unexpected but is consistent with mitochondrial dysfunction being causally linked to type I interferon pathway activation, similar to the effects of viral infection and as seen in certain autoimmune diseases, such as systemic lupus erythematosis.⁵⁹ Cell death–associated genes were also detected (Supplemental Figure S1F). Because cell death and release of damage-associated molecular patterns can cause inflammation, the impact of mitochondrial dysfunction on the innate immune response was studied using in vitro approaches.

Figure 4.

Cox10 deficiency in the nephron up-regulates nephrotoxicity genes and interferon-stimulated genes (ISGs). Comparison of gene expression profile in Cox10^Δ/Δ versus Cox10^+/+ kidneys using RNA sequencing. A: A volcano plot visualizing differentially expressed genes in Cox10^Δ/Δ versus Cox10^+/+ kidneys with fold changes significantly (P < 0.001) >1.4. Significance versus fold change is plotted on the y and x axes, respectively. B: Pathway enrichment analysis using DAVID. C: A heat map featuring differently expressed mtDNA-encoded genes. D: Clustering and annotation of differentially expressed genes associated with nephrotoxicity using Ingenuity Pathway Analysis (IPA) Software version 01-10. E: Identification of ISGs among differentially expressed genes using INTERFEROME version 2.01. F: Expression and clustering of all differentially expressed ISGs graphically presented as a heat map. G: Western blot analysis for selected ISGs in Cox10^Δ/Δ versus Cox10^+/+ kidneys. n = 4 (A–F). DEG, differentially expressed genes; ECM, extracellular matrix; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; HTLV-1, human T-lymphotropic virus; IRF, interferon regulatory factor; KEGG, Kyoto Encyclopedia of Genes and Genomes; OAS, 2'-5'-oligoadenylate synthetase 1; STING, stimulator of interferon genes; TNF, tumor necrosis factor.

Cox10 Ablation Triggers the Release of mtDNA into the Cytosol and Activates the Intracellular DNA Sensor STING

To study the impact of mitochondrial dysfunction on the cellular level, an in vitro approach was used. PTECs (LTL⁺) or TECs (EpCAM⁺) were isolated from kidneys of Cox10^Δ/Δ mice to obtain Cox10-deficient PTECs and TECs. However, they were not viable in vitro, emphasizing the profound impact of Cox10 deletion on epithelial cell health (Figure 5A). Instead, PTECs or TECs from wild-type mice were isolated and Cox10 expression was silenced in primary cultures of these cells by delivering exogenous siRNA to primary cell cultures (Figure 5A and Supplemental Figure S3A). TECs transfected with Cox10 siRNA (siCox10) reduced Cox10 transcripts by 50% to 75% compared with untreated cells, and they were reduced by approximately 50% compared with TECs transfected with control siRNA at days 2 and 6 of transfection, respectively (Figure 5B and Supplemental Figure S4, A and B). Cox10 silencing in PTECs was as efficient as in TECs (Cox10 mRNA reduction by 70% to 85%). Unexpectedly, Cox10 mRNA was also diminished in control PTECs but to a much lesser extent than in Cox10-silenced PTECs, suggesting PTECs are sensitive to siRNA transfection, causing transient down-regulation of mitochondrial biogenesis (Supplemental Figure S3B). More important, no differences in cell viability between Cox10-silenced and control TECs through 5 days were observed (Supplemental Figure S5A). Next, changes in the mitochondrial function in TECs after Cox10 silencing were studied (Figure 5C). The oxygen consumption rate of Cox10-silenced TECs was compared with control TECs before and after serial additions of oligomycin (an inhibitor of ATP synthase), carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (an uncoupling ionophore), and rotenone with antimycin A (blocking agents for the complexes I and III of the electron transport chain, respectively). Cox10-silenced TECs had reduced basal respiration, maximal respiratory capacity, and mitochondrial ATP production compared with controls (Figure 5C). In addition, reserve respiration capacity, which is calculated as the difference between maximal respiratory capacity and basal respiration, was reduced by 22.1% ± 15.1%. Taken together, these data showed that loss of COX10 affects mitochondrial oxidative phosphorylation within 5 days. The quantity of total ATP produced by Cox10-silenced TECs was also studied. Similarly, ATP production was decreased at day 5 after silencing (Supplemental Figure S5B). Next, MMP was assessed in Cox10-silenced TECs. MMP was modestly reduced by approximately 12% in Cox10-silenced versus control TECs at day 6 after transfection (Supplemental Figure S5C). This result is consistent with the effects of Cox10 silencing on mitochondrial respiration (Figure 5C), but the more modest reduction of MMP after Cox10 silencing may result from compensatory mechanisms exerted by complexes I and III, while complex IV is failing. In addition to the decline of ATP production, Cox10 silencing led to a subsequent reduction of mtDNA content. Previously, Diaz et al⁶⁰ reported that human fibroblasts displayed abnormal mitochondrial morphology with rounder and larger appearance after COX10 deletion. In line with those reports, changes were detected in mitochondrial morphology and location at day 5 after Cox10 silencing: the mitochondria appeared swollen and more concentrated within the perinuclear region (Figure 5D). Moreover, individual mitochondria lacking mtDNA could be detected only after Cox10 silencing. To determine whether the loss of mtDNA was attributable to release into the cytosol, extramitochondrial mtDNA was assessed in Cox10-silenced versus control TECs. Pure cytosolic extracts did not contain detectable nuclear DNA, as expected (Figure 5E and Supplemental Figure S6). Those cytosolic extracts were analyzed for mtDNA content. After silencing of Cox10, a fourfold to fivefold and a threefold to fourfold increase of specific mtDNA fragments from the Dloop regulatory region was detected at days 2 and 5 of transfection, respectively; this was consistent with modest liberation of mtDNA into the cytosol in response to loss of Cox10 (Figure 5F). Next, it was evaluated whether the presence of mtDNA in the cytosol could activate an important DNA sensor, STING, which is highly up-regulated in the Cox10^Δ/Δ kidneys. In the gene silencing experiments in vitro, both monomeric and dimeric forms of the active, phosphorylated forms of STING increased in response to transfection with siRNA, but the increase was more pronounced in response to Cox10 silencing and consequent mtDNA release (Figure 5G). TBK1, a recognized effector of the type I interferon signaling pathway and a substrate for STING, was also activated in the setting of Cox10 silencing (Figure 5H).

Figure 5.

Cox10 silencing in kidney tubular epithelial cells (TECs) results in mitochondrial dysfunction and release of mtDNA into the cytosol and activates stimulator of interferon genes (STING). A: Schematic representation: TECs isolated from Cox10^Δ/Δ mice are not viable in vitro. Thus, Cox10-targeted (siCox10) and control (siControl) siRNAs were introduced into cultured wild-type (WT) TECs by transfection (each at a final concentration of 100 nmol/L) to study the role of Cox10 in TECs. B: At day 6 of transfection, cells were harvested and analyzed for Cox10 mRNA expression. C: The oxygen consumption rate (OCR) was measured at baseline and in response to oligomycin (Oligo), carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP), and antimycin plus rotenone (Ant + Rot) using the Seahorse XFe extracellular flux analyzer. D: Confocal microscopy images were taken from TECs stained with anti-DNA (DNA) and anti–superoxide dismutase (Mito) at days 2 and 5 of transfection. Arrowheads indicate mitochondria not colocalizing with DNA staining. E: Nuclear-encoded Tert gene expression was quantitated via real-time quantitative PCR in cytosolic and whole-cell extracts of untreated TECs. F: Cytosolic mtDNA content in transfected TECs. G and H: Western blot analysis for phosphorylation of STING (G) and TANK-binding kinase-1 (TBK1; H) in transfected TECs. Dashed lines indicate untreated cells (B, C, and F). Data are expressed as means ± SEM (B, C, E, and F). n = 3 to 4 (B, C, E, and F). ^∗P ≤ 0.05 (U-test). Scale bars = 10 μm (D). Original magnification, ×63 (D). GAPDH, glyceraldehyde-3-phosphate dehydrogenase.

Cox10 Ablation Induces Up-Regulation of Oxidative Stress and Pro-Apoptotic Genes

Because Cox10 silencing causes mitochondrial dysfunction and triggers the release of mtDNA in vitro, it was next examined whether Cox10 ablation induces oxidative stress by detecting a compensatory expression of the cytosolic and mitochondrial superoxide dismutases Sod1 and Sod2. Both genes are up-regulated more than twofold in Cox10-silenced TECs at 6 days of transfection (Figure 6A). When functional enrichment analysis of differentially expressed genes associated with renal necrosis/cell death was performed in Cox10^Δ/Δ versus Cox10^+/+ kidneys, pathways involved in apoptosis of TECs, podocytes, and mesangial cells and in the necrosis of TECs were identified (Supplemental Figure S1E). Hence, it was examined whether mRNA transcripts of caspases are elevated in TECs after Cox10 silencing. Six days after transfection, silencing of Cox10 led to elevated expression of executioner caspases, Casp3, Casp6, and Casp7, as well as the initiator caspase, Casp8⁶¹ (Figure 6B). Neither caspases nor the oxidative stress indicators, Sod1 and Sod2, were elevated in Cox10-silenced TECs at 2 days of transfection (Supplemental Figure S4, C and D). However, both TECs and PTECs up-regulated Sod1, Sod2 (Supplemental Figure S3C), and Casp3, Casp6, and Casp8 at day 6 after Cox10 silencing. Unlike TECs, Cox10 silencing in PTECs caused an early increased expression of Sod1, Casp3, and Casp8 within 2 days but did not up-regulate Casp7 (Supplemental Figure S3, D and E).

Figure 6.

Cox10 silencing or introduction of mtDNA into the cytosol up-regulates apoptotic, proinflammatory, and interferon transcripts in tubular epithelial cells (TECs). Wild-type TECs were transfected with 100 nmol/L of siCox10 and control siRNA (siControl). A–C: At day 6 after silencing, cells were harvested and analyzed for mRNA expression of oxidative stress-related (A), apoptosis-related (B), and proinflammatory (C) genes. Dashed lines indicate untreated cells. D: Transcriptional expression of Ifna4 and Ifnb1 in TECs transfected with 2.6- or 3.6-kb fragments of synthesized oxidized mtDNA at 4 hours. Data are expressed as means ± SEM (A–D). n = 3 to 4 (A–D). ^∗P ≤ 0.05 (U-test). ND, not detected.

Cox10 Ablation Induces Interferon Response

These results support a model in which deletion of COX10 in TECs causes mitochondrial dysfunction, leading to the release of mtDNA into the cytosol, with immunogenic and inflammatory properties, only subsequently initiating cell death pathways.20, 30, 51 This may provide an explanation for the prominent type I interferon signature in Cox10^Δ/Δ kidneys. To test this hypothesis, transcript levels of proinflammatory cytokines, including interferon genes, Ifna4 and Ifnb1, and other selected cytokines implicated in the pathogenesis of chronic kidney diseases, Il1b,⁶² Nlrx1,63, 64, 65 and Tnfa,66, 67 were assessed in Cox10-silenced PTECs and TECs. Compared with control-treated cells, those with Cox10 silencing had persistent elevation of interferons, particularly interferon β (Ifnb1), as well as modest elevations of Nlrx1 and Tnfa (Figure 6C). Treatment of epithelial cells with siRNA induces a sequence nonspecific immune response. siRNAs are double-stranded RNAs and may serve as low-level agonists for pattern recognition receptors. However, knowing that siRNA-mediated interferon response is transient and resolves, it was found that the transcriptional expression of Ifna4 and Ifnb1 in TECs increased markedly at day 2, but decreased to baseline levels by day 6 (Figure 6C and Supplemental Figure S4E).

To test directly whether cytosolic mtDNA can provoke an interferon response, 2.6- and 3.6-kb fragments of mtDNA were synthesized. In one set of experiments, oxo-GTP was incorporated instead of GTP to generate oxidized mtDNA. Transfection of mtDNA into TECs caused a marked increase in Ifnb1 transcripts and a smaller increase in Ifna4 transcripts. In the case of Ifnb1 responses, the oxidized forms of mtDNA had a more marked effect on Ifnb1 transcription (Figure 6D).

Next, the expression of genes that are activated in response to type I interferons (alias ISGs), previously identified as genes activated by mtDNA, was examined.²⁰ Consistent with elevated production of interferons in response to Cox10 silencing, increased expression of ISGs was observed, including Cxcl10, Ddx58, Ifih1, Ifit1, Ifit3, Irf7, Isg15, Stat1, and Usp18, in Cox10-silenced TECs at day 6 of transfection (Figure 7A). Accordingly, these ISGs were found to be up-regulated in Cox10^Δ/Δ versus Cox10^+/+ kidneys at 2.5 to 3 months of age, significantly or by trend (Figure 7B). Therefore, the in vitro results are consistent with a model in which Cox10 ablation leads to mitochondrial dysfunction, leading to interferon production and sensitization of kidney epithelial cells to increased interferon responses.

Figure 7.

Cox10 silencing elevates the expression levels of interferon-stimulated genes (ISGs) in tubular epithelial cells (TECs). A: Wild-type (WT) TECs transfected with 100 nmol/L of siCox10 and control siRNA (siControl). At day 6 of transfection, cells were harvested and analyzed for mRNA expression of differentially expressed ISGs. Dashed lines indicate untreated cells. B: Comparison of ISG expression in Cox10^Δ/Δ versus Cox10^+/+ kidneys using RNA sequencing. Select ISGs with fold changes >1.4 are highlighted in red in the volcano plot (left panel) and visualized in a heat map (right panel). Data are expressed as means ± SEM (A). n = 3 (A); n = 4 (B). ^∗P ≤ 0.05 (U-test).

Taken together, we have identified a novel mechanism, which may contribute to FSGS pathogenesis. The data suggest that release of mtDNA into the cytosol is a mechanism inducing innate immune activation in podocytes and epithelial cells upstream of cell death. Cell death may result in release of damage-associated molecular patterns, further activating innate immune responses. Therefore, mtDNA release is one possible mechanism driving pathogenesis in FSGS caused by mitochondrial disease.

Discussion

These studies unequivocally demonstrate that loss of COX10, a cofactor of COX-IV in the electron transport chain, is sufficient to cause severe FSGS and tubulointerstitial disease. RNA sequencing of whole kidneys identified a strong interferon response to loss of COX10. The data link COX10 deficiency to mitochondrial dysfunction, leading, in turn, to leakage of mtDNA into cellular cytosol, which directly activates an interferon response and initiates an inflammatory cascade. These cascades are mediated by intracellular DNA sensor proteins, including STING.

An increasing number of mutations in mtDNA or in nuclear genes encoding critical mitochondrial proteins, including COX-IV proteins, have been documented to be causative in forms of FSGS and tubulointerstitial diseases. Herein, the dysfunction of the mitochondrial electron transport in the kidney was modeled by deleting a COX-IV cofactor in the nephron. COX10 is thought to be required for the correct assembly of COX-IV, the terminal enzyme in the respiratory electron transport chain of mitochondria catalyzing the transfer of electrons from cytochrome c to oxygen.³⁶ For its catalytic function, COX-IV must dimerize and form a complex with prosthetic groups: heme (α and α₃), two copper centers (Cu_A and Cu_B), zinc, and magnesium.⁶⁸ COX10 is also required in the heme α biosynthesis.³⁵ In humans, germline COX10 loss-of-function mutations result in fatal outcomes, and they present with ataxia, muscle weakness, hypotonia, a pyramidal syndrome, and leukodystrophy accompanied with elevated cerebrospinal fluid and blood lactate levels.36, 69 They may also have anemia, sensorineural deafness and hypertrophic cardiomyopathy, Leigh syndrome,⁷⁰ and Leigh-like syndrome.⁷¹ Patients have been reported to present with a Fanconi syndrome because of a proximal tubulopathy.⁷² These results provide a possible explanation why patients experiencing mitochondrial cytopathies frequently manifest glomerular diseases and indicate that loss-of-function mutations in COX-IV subunits and auxiliary proteins cause FSGS. Moreover, these studies are consistent with a model in which multiple insults to mitochondrial health, by either genetic predisposition or environmental exposures, can trigger FSGS in patients.

A major consequence of loss of COX10 is the loss of ATP production by cells of the nephron that require high energy production to perform the physiological roles of the nephron. It is likely, therefore, that loss of cell function is a major consequence of COX-IV disruption and is in keeping with our reports of the consequences of mitochondrial dysfunction when mitochondrial turnover is disrupted.⁹ Such loss of function leading to cellular degeneration may be sufficient to cause all of the hallmarks of kidney disease demonstrated in these studies. However, as described above, mitochondria also sense and coordinate innate immune responses, and these findings suggest disrupting mitochondrial function may additionally lead to activation of the innate immune response.

Although loss of COX10 and consequent COX-IV dysfunction could be highly deleterious to cell health, previous studies using myofiber- and neuron-specific Cox10^Δ/Δ mice showed unexpected results.32, 73 Surprisingly, myofiber-specific Cox10^Δ/Δ mice were indistinguishable from controls from birth until early adulthood (<3 months of age), and muscles in adult Cox10^Δ/Δ mice (at 2.5 months of age) contracted at 80% to 100% of controls without any sign of oxidative damage or apoptosis. COX-IV activity in muscles of myofiber-specific Cox10^Δ/Δ mice was not completely abolished, maintaining 2% to 13% of normal activity, depending on age, which was indicative of bypass mechanisms of COX10 deletion.³² Similarly, oxidative stress in the central nervous system of neuron-specific Cox10^Δ/Δ mice was not evident in mice <4 months of age.⁷³ These findings imply that glycolysis and residual COX-IV activity can compensate for the deletion of Cox10 in these tissues, whereas such compensatory mechanisms are not sufficient in the nephron. Recent studies in human kidney cells indicate that, in disease states, oxidative phosphorylation is decreased. Glycolysis is highly up-regulated in kidney stromal cells, whereas epithelial cells fail to compensate by up-regulating glycolysis.⁷⁴ It remains elusive why glycolysis is not up-regulated in the nephron and needs further investigation beyond the scope of this study. Nevertheless, the kidney epithelium, including podocytes, appears to be particularly susceptible to loss of COX10 and COX-IV.

A series of recent studies by other investigators implicate the release of mtDNA into the cytosol as a link between mitochondrial dysfunction and high interferon activity.20, 30, 75 mtDNA shares certain similarities with prokaryotic DNA and has been reported to interact with cyclic GMP-AMP synthase and thereby activate STING, which, in turn, activates a cascade of signaling intermediates, including TBK1 and IRF3. These reports, however, have focused on immune cells, including macrophages, dendritic cells, and T cells.27, 76, 77, 78, 79, 80, 81 The studies reported to date have also focused on cell death or transfection as methods to release mtDNA into the cytosol, and they also demonstrated an important role of mtDNA signaling inside of immune cells in enabling clearance of pathogens. One study generated a heterozygous mouse model of mitochondrial transcription factor A deficiency, which led to mitochondrial dysfunction in fibroblasts and was sufficient to elicit mtDNA-dependent activation of STING and an interferon response.²⁰ In the studies presented herein, a cofactor of COX-IV formation was deleted to elicit mitochondrial dysfunction in the epithelium. This was sufficient to drive mitochondrial dysfunction, intracellular mtDNA release, and activation of STING. In vivo as well as in vitro, this triggered a strong ISG response; however, there were several notable differences between reports of the consequences of mtDNA release in leukocytes and fibroblasts compared with our results in the epithelium. First, although specific activation of STING and TBK1 was detected, activation of IRF3 was not identified in in vitro studies; rather, IRF7 was strongly activated. IRF7 and IRF5 are important independent nuclear factors that result in interferon production.82, 83 Although further studies are required, it may be that in the epithelium STING mediates its effect predominantly by an IRF3-independent mechanism. Second, it was demonstrated that the epithelium elicits a strong interferon-β response rather than a strong interferon-α response. Such a bias toward interferon-β has been reported elsewhere and may have implications for the sensitivity of the kidney to the interferon response.⁸⁴

Although this current report provides some evidence for mtDNA release as a cause of kidney inflammation, further work is required to establish whether the mtDNA-STING pathway is important in human disease and whether mitochondrial dysfunction triggered by toxins or other disease states, such as diabetes or Alport syndrome, similarly causes mtDNA release in the setting of mitochondrial dysfunction. Furthermore, the significance of local interferon production in the kidney has been little studied, and the contribution to local injury and inflammation requires further study.⁸⁵ Nevertheless, in the setting of mouse models of kidney injury and glomerular inflammation, deficiency in the major interferon receptor, interferon-α/β receptor, has been reported to result in less severe kidney injury86, 87, 88, 89; deficiency of other sensors of intracytoplasmic nucleic acids, including mitochondrial antiviral signaling protein, has also been reported to contribute to disease severity in the kidney.⁹⁰

To study the impact of mitochondrial disruption on the cellular level, TECs were isolated from Cox10^Δ/Δ kidneys, but purified TECs from Cox10^Δ/Δ kidneys were not viable in cell culture. Thus, an alternative approach was used by silencing Cox10 mRNA expression in WT TECs. In Cox10-silenced cells, a decrease in mitochondrial function in basal conditions and at maximal respiration was detected, indicating a reduction in reserve capacity. MMP was modestly reduced, and no change was detected in viability. This finding was in line with the data from MitoTracker staining of Cox10-ablated fibroblasts.⁶⁰ One possible explanation why there was a reduction in oxygen consumption rate, whereas the MMP was relatively preserved, is that proton pumping by the complexes I, III, and IV (COX-IV) maintains the MMP. The complexes I and III may compensate during COX-IV failure. Indeed, F1F0 ATP synthase (complex V) inhibition has been shown to cause MMP hyperpolarization.⁹¹ Nevertheless, this observation suggests that mtDNA can leak into the cytosol in the absence of overall mitochondrial depolarization, although it does not exclude the possibility of localized MMP depolarization, as has been identified in other studies.⁹² It is worth noting that the knockdown protocol resulted in approximately 75% reduction in Cox10 mRNA levels; therefore, the complete effects of COX10 deficiency may not be manifest, whereas purification of epithelium from Cox10^Δ/Δ kidneys resulted in nonviable cell cultures.

The generated mouse model causes loss of function of COX10 in both podocytes and tubular cells of the nephron. Future work to further dissect the specific impact of COX10 mutation on podocytes in vivo using podocyte-restricted expression of Cre recombinase to mutate COX10 would add to our understanding of the kinetics of glomerular pathology progression in the absence of coexistent tubular disease.

In summary, deletion of the mitochondrial COX-IV cofactor heme A:farnesyltransferase in nephron epithelial cells causes severe FSGS, directly implicating mitochondrial dysfunction in the causal pathway of FSGS. Mitochondrial dysfunction resulting from COX-IV deficiency also drives an interferon response that may contribute to kidney pathology and organ failure. This study provides evidence to suggest that the interferon response in the kidney epithelium is caused by intracellular release of mtDNA from the dysfunctioning mitochondria.

Supplemental Data

Supplemental Figure S1.

Principal component analysis (PCA) and functional annotation analysis of cell death–associated genes. A: Western blot analysis for COX10 and COX1 protein expression in Cox10^Δ/Δ and Cox10^+/+ kidneys. B: Body weights of female Cox10^Δ/Δ compared with Cox10^+/+ mice at 2 and 3 months of age. C: Kidney weights of female Cox10^Δ/Δ compared with Cox10^+/+ mice at the time of natural death and euthanasia. D: Urinary albumin/creatinine ratio of mice at 2.5 to 3 months of age. E: PCA analysis of RNA sequencing data from kidneys of Cox10^Δ/Δ versus Cox10^+/+ mice. F: Clustering and annotation of differentially expressed genes associated with cell death using Ingenuity Pathway Analysis software version 01-10. n = 4 mice aged 2 months (B); n = 9 to 10 mice aged 3 months (B); n = 8 (C); n = 3 to 7 (D). ^∗P ≤ 0.05, ^∗∗P ≤ 0.01, and ^∗∗∗P ≤ 0.001. GAPDH, glyceraldehyde-3-phosphate dehydrogenase.

Supplemental Figure S2.

Functional protein association networks among differentially expressed genes in Cox10^Δ/Δ versus Cox10^+/+ kidneys. Differentially expressed genes in Cox10^Δ/Δ versus Cox10^+/+ kidneys were subjected to functional protein association network analysis using STRING database version 10.5.

Supplemental Figure S3.

Cox10 silencing up-regulates apoptotic and proinflammatory genes in proximal tubular epithelial cells (PTECs). A: Schematic representation: 100 nmol/L of siCox10 and control siRNA (siControl) was introduced into cultured wild-type (WT) PTECs by transfection. B–E: At days 2 and 6 of transfection, cells were harvested and analyzed for mRNA expression of Cox10 (B) and genes related to oxidative stress (C) and apoptosis (D) or proinflammatory genes (E). Dashed lines indicate untreated cells. Data are expressed as means ± SEM (B–E). n = 3 (B–E). ^∗P ≤ 0.05 (U-test). ND, not detected.

Supplemental Figure S4.

Oxidative stress and apoptosis-related genes are not up-regulated during the first 2 days of siCox10 transfection in tubular epithelial cells (TECs). A: Schematic representation: wild-type (WT) TECs were transfected with 100 nmol/L of siCox10 and control siRNA (siControl). B–E: At day 2 of transfection, cells were harvested and analyzed for mRNA expression of Cox10 (B) as well as genes related to oxidative stress (C) and apoptosis (D) or proinflammatory genes (E). Dashed lines indicate untreated cells. Data are expressed as means ± SEM (B–E). n = 3 to 4 (B–E). ^∗P ≤ 0.05 (U-test).

Supplemental Figure S5.

Cox10 silencing increases ATP production but does not influence cell viability. Cells were analyzed for cell viability (A), ATP production at day 5 of transfection (B), and mitochondrial membrane potential normalized by mitochondrial mass by costaining with MitoHealth and MitoTracker Green FM (C). Dashed lines indicate untreated cells. Data are expressed as means ± SEM (A–C). n = 4 (A–C). ^∗P ≤ 0.05 (U-test).

Supplemental Figure S6.

Nuclear-encoded Tert gene expression in cytosolic and whole-cell extracts of transfected tubular epithelial cells (TECs) and TEC transfection with mtDNA fragments. Nuclear-encoded Tert gene expression was quantitated via real-time quantitative PCR in cytosolic and whole-cell extracts of TECs transfected with siCox10 or control siRNA (siControl) at days 2 and 5. Data are expressed as means ± SEM. n = 3. ^∗P ≤ 0.05 (U-test).