Hypoxic regulation of mitochondrial metabolism and mitophagy in nucleus pulposus cells is dependent on HIF-1α -BNIP3 axis

Mice

All animal experiments were performed under protocols approved by the IACUC of Maine Medical Center Research Institute and Thomas Jefferson University. MitoQC mice with constitutive knockin of mCherry-GFP-mtFIS1^101−152 on C57BL/6 background developed earlier were maintained as reported.⁽²⁷⁾ Foxa2^Cre;Hif1α^f/f mutant mice have been generated and genotyped as previously described.⁽²⁸⁾

Fluorescence microscopy and immunohistochemistry

MitoQC mice spines were fixed in 4% PFA for 8 h at 4°C, decalcified in 20% EDTA and treated with 30% (wt/vol) sucrose before embedding in OCT compound. 6–8 μm thick coronal sections were cut on a cryostat (CM3050S; Leica Biosystems), washed with PBS and mounted with ProLong Gold Antifade Mountant with DAPI. Images were taken on a Nikon A1R confocal microscope using 100x objective (CFI Plan Apo Lambda 100x/1.45 oil). For quantification of mitophagy we measured mCherry-positive structures (pixel²/area) using ImageJ software (http://rsb.info.nih.gov/ij/). E15.5 HIF-1α null or control embryos were fixed in 4% PFA and then embedded in paraffin and 6–8 μm coronal sections were generated. Sections were deparaffinized and citrate buffer pH 6 (Vector Labora-tories, H-3301) was used for antigen retrieval. Sections were incubated for 1 h in the appropriate blocking solution (either 5% Normal Goat Serum in PBS-T (0.4% Triton X- 100) or reagent from M.O.M.™Immunodetection Kit; Vector Laboratories, BMK-2202) and incubated with anti-Complex III (Abcam, ab14745; 1:100), anti-LC3 antibody (Novus Biologicals, NB100–2220; 1:200) or anti-NIX (Abcam, ab109414; 1:100) in appropriate blocking buffer at 4°C overnight. Sections were mounted with ProLong Gold Antifade Mountant with DAPI and colocalization of Complex III with LC3 was quantified as area (pixel²/cell) using the colocalization Plugin of ImageJ software (http://rsb.info.nih.gov/ij/).

Isolation of NP cells, treatments and hypoxic culture

Rat NP cells were isolated and maintained in Dulbecco’s Modification of Eagle’s Medium (DMEM) and 10% FBS supplemented with antibiotics.⁽²⁹⁾ To investigate the effects of hypoxia on mitochondria cells were cultured in hypoxia workstation (Invivo2 400; Baker Ruskinn, UK) with a mixture of 1% O₂, 5% CO₂, and 94% N₂. To investigate effects of HIF-1α, lentivirally transduced cells were cultured in a hypoxia for 24 h. To induce mitophagy, cells were treated with deferiprone (1 mM; Sigma) for 24 h. To inhibit autophagic flux, cells were treated with bafilomycin A1 (50 nM; Tocris) for 24 h in normoxia or 1 h prior to hypoxic culture. Cells were treated with leptomycin B (10 nM; Sigma) 1 h prior to 24 h of hypoxic culture to block nuclear export.

Lentiviral particle production and viral transduction

LV-shHIF-1α (clone #232222) and pLKO.1 shCtrl and psPAX2 (# 12260); and pMD2G (# 12259) (Addgene) were used. HEK 293T cells were plated in 10-cm plates (5 × 10⁶ cells/plate) in DMEM with 10% heat-inactivated FBS one day before transfection. Cells were transfected with ShCtrl or ShHIF-1α plasmids along with psPAX2 and pMD2.G and lentiviral particles were harvested at 48 to 60 h post-transfection. Cells were transduced with viral particles with 8 mg/ml polybrene and harvested for protein extraction 4 days after transduction.⁽³⁰⁾

Immunocytochemistry

NP cells were plated on glass coverslips and treated with 100 nM MitoTracker™ Red CMXRos (Thermo Fisher, M7512) for 30 min after completion of the experimental treatments. Cells were then fixed with 4% PFA or ice cold methanol for 15 min and permeabilized with 0.1% Triton X-100 for 10 min. and blocked with 1% BSA for 1 h. Cells were incubated with anti-BNIP3 (3769S), anti-BNIP3L (12396S), anti-LC3 (12741) (Cell Signaling), anti-TOMM20 (ab186734), anti-LAMP1 (ab24170), anti-PINK1 (ab23707) (Abcam), anti-PARKIN (sc-32282) (Santa Cruz), anti-DRP1 (611113), anti-OPA1 (612607) (BD Biosciences), anti-BECLIN1 (NB500–249) (Novus Biologicals) in blocking buffer at 1:100–200 at 4°C overnight. After washing, cells were incubated with Alexa Flour-488 or Alexa Flour 647 and mounted with ProLong Gold Antifade Mountant with DAPI. Negative controls were used to confirm the specificity of staining (Suppl. Fig. 4). Cells were visualized using a Nikon A1R confocal microscope using 60x objective (CFI plan Apo Lambda 60x/1.40 oil). Positive staining for markers and co-localization of LC3 and LAMP1 with mitochondria was measured as area (pixel²/cell) using the colocalization Plugin of ImageJ software (http://rsb.info.nih.gov/ij/).

Analysis of mitochondrial morphology

Mitochondrial number, branching and morphology were quantified in ImageJ using methods reported by Valente et al. and Picard et al (Suppl. Fig. 6).^(31,32) The aspect ratio (AR) represents the “length to width ratio” and form factor (FF), the complexity and branching aspect of mitochondria.

ATP and NAD:NADH measurements

NP cells were plated in white-walled tissue culture plates at 10,000 cells/well and placed in normoxia or hypoxia for 24 h. To examine the role of HIF-1α, cells were transduced with ShCtrl or ShHIF-1α lentivirus and cultured in hypoxia for 24 h. ATP and NAD/NADH measurements were performed using the ATP bioluminescence assay kit (Abcam, ab113849) and NAD/NADH-Glo™ detection reagent (Promega, G9071) using a luminometer (Tecan Infinite M200Pro) according to the manufacturer’s protocol. Measurements were normalized to total protein.

Mitochondria isolation, Protein extraction and Western blotting

Mitochondria were isolated according to manufacturer’s instruction (Thermo Fisher Scientific, 89874). Mitochondria were lysed and 20 μg of protein was electroblotted to PVDF membranes (EMD Millipore). The membranes were blocked and incubated overnight at 4°C with antibodies against ETC complexes; complex I (EMD Millipore, ABN302), complex II (anti-SDHA, Abcam ab14715), complex III (anti-UQCRC2, Abcam ab14745), complex IVi1 (NB110–39115), and complex IVi2 (H00084701-M01) from Novus. The membranes were incubated with IRdye conjugated secondary antibodies for 1 h and scanned with an Odyssey infrared scanner. Densitometric analysis was performed using ImageJ software. For ShCtrl and ShHIF-1α blotted membranes were incubated with anti-HIF-1α (1:500; R&D Systems), anti-BNIP3, anti-BNIP3L, anti-BECLIN1 (3495), anti-GAPDH (5174) (Cell Signaling), anti-LC3, anti-TOMM20 (Thermo Fisher, PA5–52843), anti-LAMP1, anti-DRP1, anti-OPA1, anti-PINK1, anti-PARKIN, or anti-β-tubulin (1:5000, DSHB, IA, USA). Immunolabeling was detected using ECL reagent (LAS4000; GE Life Sciences).

Seahorse XF analyzer

Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured as reported. ⁽⁵⁾ Briefly NP cells were plated (15,000 cells/well) in a Seahorse XF 24 V7 microplate (Seahorse Bioscience, 100777–004) and incubated in hypoxia for 24 h. On the day of the analysis, microplate was incubated for 1 h at 37°C, no CO₂ and in normoxia before OCR and ECAR were measured (Seahorse XF24, Agilent Technologies). All measurements were normalized to total protein. Mitochondrial OCR was calculated by subtracting the final OCR value (pmol O2/min) after Antimycin A treatment from the average of the three OCR values before Antimycin A treatment.

Metabolomics and flux analysis

Cells were transduced with ShCtrl and ShHIF-1α viral particles as described above. For metabolomics, on the 3^rd day media was changed to DMEM without pyruvate, 10% dialyzed FBS (Sigma F0392) and then cultured in hypoxia for 24 h. On harvest, cells were rinsed and collected in ice-cold 150 mM ammonium acetate and centrifuged for 3 min at 1100 g. Cell pellet was snap frozen in liquid nitrogen and stored in −80°C until use. Pellets were extracted with of 80% methanol solution (containing 2.5μM U¹³C succinate, and 5μM U¹³C citrate as internal standards) with 3 freeze-thaw cycles. After centrifugation, the supernatant was transferred to GC sampling vial. The samples were dried under gentle nitrogen flow, and a two-step derivatization with methoximation and silylation according to our previous procedure.⁽³³⁾ The derivatized metabolites were separated on a 30-meter DB-5MS column analyzed with gas chromatography time-of-flight mass spectrometry (GC-TOFMS premier, Waters, USA). The data was analyzed with Genedata Expressionist (Genedata, Basel, Switzerland) software. The output data was normalized to internal standards and the sum of all the metabolites within each sample. When flux was assessed in the [1,2-¹³C]-glucose labeling experiment, labeled glucose was added to be 50% of the total DMEM glucose concentration, and an ethyl chloroformate (ECF) derivatization method was used instead of silylation. A volume of 700 μl of cell culture media was used for ECF derivatization that followed our previously published procedure with minor modifications. ⁽³⁴⁾ The derivatized metabolites were extracted with 300 μl of chloroform, and injected 1 μl into the GC-MS system in splitless mode (GC-MS, Agilent, USA). The data was analyzed with MassHunter Quantitative Analysis software (Agilent, USA). M1 of glutamate around m/z 128 reflects the presence of ¹³C at the 4^th carbon position and reflects pyruvate entering the acetyl units originating from glucose to pyruvate entering through pyruvate dehydrogenase. M2 glutamate m/z 128 reflects the presence of ¹³C at the 2^nd and 3^rd carbon, and reflects pyruvate entering the oxaloacetate pool through pyruvate carboxylase.⁽³⁵⁾ On the other hand, M2 glutamate m/z 202 reflects the presence of both ¹³C at the 2^nd and 3^rd carbon positions and the presence of both ¹³C at the 4^th and 5^th carbon positions, summing as the overall TCA cycle flux. The two ¹³C atoms (M2) come from either pyruvate entering via pyruvate carboxylase (carbons 2 and 3) or via pyruvate dehydrogenase (carbons 4 and 5). The fraction of glucose converted to pentose phosphate in the pentose cycle is defined as: Pentose cycle = (M1/M2)/(3 + M1/M2), where the M2 isotope yield in lactate is the product of glycolysis from [1,2] ¹³C glucose, and the recycled (from glucose through the Pentose cycle) isotope yield of lactate as M1 lactate.⁽³⁶⁾ Lactate isotopomers were also derived from the ECF derivatization method, M0 lactate has m/z 145, M1 lactate m/z 146 and M2 lactate m/z 147.

Statistical analysis

Statistical analysis was performed using Prism7 (GraphPad, La Jolla, CA, USA). For both in vitro and in vivo quantitative parametric data are represented as mean ± SD or box and whisker plot with median and interquartile range. Data distribution was checked with Shapiro-Wilk normality test, and Mann-Whitney test was used for non-normally distributed data. Differences between two groups were assessed by t-test and one-way ANOVA was used for comparisons between more than two groups. All in vitro experiments were performed in triplicate at minimum.

NP cells evidence increase in age-dependent basal mitophagy in vivo

A limited attempts to visualize mitochondrial morphology in intervertebral disc cells in vivo suggested that NP cells possess a few mitochondrial that are under-developed. Since previously employed EM-based methods are prone to fixation artifacts, we used confocal microscopy and transgenic mitoQC mice wherein outer mitochondrial membrane protein fission1 is genetically tagged with a pH-sensitive fluorescent reporter mCherry-EGFP to delineate mitochondrial morphology and mitophagy in-vivo (Fig.1A, B). The NP cells showed a well-developed mitochondrial network in early post natal as well as skeletally mature mice (Fig.1C, D, Suppl. Fig. 1 movie). Noteworthy, when damaged mitochondria experience a loss in membrane potential, they undergo fission when separated from the mitochondrial network and ultimately they are eliminated through mitophagy. Accordingly, in mitoQC mice mitochondria undergoing degradation display the mCherry signal as EGFP undergoes acidic quenching following fusion with lysosomes (Fig.1A). Given that an accumulation of dysfunctional mitochondria can lead to pathophysiological conditions,⁽²⁵⁾ mitophagy is essential to ensure recycling of old and damaged mitochondria. Interestingly the NP cell of lumbar discs in skeletally mature mice (4-month onwards) showed elevated mitophagy compared to the discs from 2–4 weeks old animals (Fig. 1C, E). Likewise, in the caudal discs there was increased mitophagy in 4 and 14 month old animals, compared to 2–4 weeks animals (Fig. 1D, F). Additionally, the overall mitochondrial number was lower in lumbar but not caudal disc between early post-natal and skeletally mature animals (Fig. 1G, H). Taken together, these results suggest that in physiologically hypoxic disc, NP cells contain a healthy mitochondrial network and evidence mitochondrial quality control, which serves to eliminate damaged and/or old mitochondria.

Hypoxia-induces mitophagy in NP cells and silencing of HIF-1α does not affect overall incidence of mitophagy

Since NP tissue is avascular and hypoxic and as our previous studies have shown an inductive effect of hypoxia on macroautophagy, we determined if hypoxia induces mitophagy? NP cells were cultured under hypoxia and mitochondrial co-localization with LC3 phagosomes as well as LAMP1 positive lysosomes was assessed. As controls, cells were treated with bafilomycin A1, to block lysosomal degradation of mitochondria, or deferiprone (DFP), an iron chelator that depolarizes mitochondria and induces mitophagy. In agreement with our previous work,⁽³⁰⁾ we observed an increase in LC3-positive autophagosomes and LC3 II levels in hypoxic NP cells as well as those treated with DFP (Fig. 2A–D). Bafilomycin A1 treatment also resulted in accumulation of autophagosomes when compared to non-treated controls, indicating an active basal autophagic flux.⁽³⁰⁾ We further quantified “mitophagosome” by colocalizing LC3-positive puncta and mitochondria. We found a significant increase in mitophagosome formation in the hypoxia, BafA1- and DFP-treated NP cells when compared to normoxic controls (Fig. 2A, C). To analyze the final step of mitophagy, we assessed the expression and localization of lysosome associated membrane protein 1 (LAMP1). The number of green LAMP1-positive puncta, representing lysosomes, and LAMP1 levels were significantly higher under hypoxia and DFP-treatment, interestingly, however, the mitolysosome percentage decreased in hypoxia as well as in DFP-treated cells suggesting an increased rate of degradation (Fig. 2E–H). Furthermore, the levels of mitochondrial outer membrane protein, TOMM20 decreased under hypoxia and DFP treatment-indicating active mitophagy (Fig. 2I–K). We then investigated contribution of HIF-1α to NP cell mitophagy.⁽³⁷⁾ HIF-1α knockdown was confirmed by Western blot in NP cells cultured under hypoxia (Fig. 3A). Imaging and Western blot analysis showed that HIF-1α silenced cells under hypoxia showed no significant changes in co-localization of LC3 or LAMP1 with mitochondria (Fig. 3B–E) and their levels (Fig. 3F, G). Western blot analysis revealed an increase in the TOMM20 (Fig. 3H–J) possibly suggesting increased mitochondrial mass. Together, these results suggested that HIF-1α silencing did not decrease the percentage of mitophagosomes or mitolysosomes in hypoxia.

Hypoxia and HIF-1α modulates mitochondrial morphology, composition, and energy metabolism

Primary NP cells were cultured under hypoxia or treated with DFP for 24 hours and stained with MitoTracker red to visualize mitochondrial morphology. Under normoxia, mitochondria were tubular and networked, whereas cells cultured in hypoxia and treated with DFP showed an increased incidence of fragmented mitochondria (Fig. 4A). Accordingly, mitochondrial mass (Fig. 4B), aspect ratio and form factor reflecting their length and branching (Fig. 4C) was decreased under hypoxia and DFP treatment. Interestingly, despite these morphological changes there was no change in ATP levels in hypoxia however DFP treated cells showed a significant decrease in ATP levels (Fig. 4D). Notably, there was a significant increase in NAD/NADH ratio in hypoxia; whereas DFP treatment had no effect (Fig. 4E) suggesting that glucose was being converted to lactate through lactate dehydrogenase activity. Notably, mitochondrial levels of ETC complex proteins were decreased in hypoxic cells (Fig. 4F, G). We then determined if hypoxia mediated changes in mitochondrial activity can be reversed by re-exposure to normoxia by measuring ECAR (Fig. 4H, I) and OCR (Fig. 4J, K). In comparison with other cell types, the basal metabolic rate of NP cells is relatively low.⁽³⁸⁾ Interestingly, short-term re-oxygenation resulted in decreased ECAR and increased OCR suggesting diminished H⁺/lactate export. Surprisingly, inhibiting the electron transport chain with Antimycin A showed that increased OCR was not due to elevation in mitochondrial OCR which on the contrary was decreased. These results suggested that short-term re-oxygenation was not sufficient to jump start mitochondrial metabolic function likely compromised by their fragmentation under hypoxia. Together, these results clearly showed that hypoxia alters mitochondrial morphology, composition and metabolism.

Since HIF-1α is a critical mediator of the hypoxia-dependent metabolic program,⁽²⁸⁾ we investigated if it plays a role in hypoxia-induced fragmentation of mitochondria in NP cells. The mitochondrial morphology after HIF-1α knockdown revealed that the hypoxia-induced fragmentation of mitochondria was reverted after HIF-1α silencing (Fig. 4L). Particularly, the mitochondria in HIF-1α silenced cells were tubular in morphology and more numerous. Moreover, similar to cells under normoxia, mitochondria in HIF-1α silenced cells under hypoxia showed increased branching, and length compared to cells transduced with ShCtrl (Fig 4M, N). Additionally, we determined if hypoxia dependent changes in mitochondrial number were related to mitochondrial biogenesis by measuring expression of PGC-1α, key regulator of this process. Our data showed that PGC-1α levels were insensitive to hypoxia (Suppl. Fig. 2). Silencing HIF-1α also resulted in decreased ATP levels (Fig. 4O), as well as a significantly decreased NAD/NADH ratio (Fig. 4P). This suggested that HIF-1α silenced cells lowered glycolytic activity and that either the NADH dehydrogenase and/or the regeneration of NAD⁺ by the conversion of pyruvate to lactate by LDH, and/or possibly the reductive amination of α-ketoglutarate to glutamate, are not keeping up with energy needs.

Hypoxia and HIF-1α are key regulators of NP cell glycolytic and mitochondrial metabolism

Considering that both oxygen tension and HIF-1α modulation significantly alter mitochondrial morphology and levels of metabolic enzymes, we assessed their specific impact on cell metabolism. Metabolic profiling was used in conjunction with ¹³C tracer analysis to examine inter-relationship between glycolysis and the TCA cycle regulated by HIF-1α. The metabolic profiles of NP cells in normoxia and hypoxia clustered separately along principal components (Fig. 5A). We delineated metabolites with significantly different concentrations using a FDR <5% and a significance analysis of microarrays (SAM) plot and heatmap representation (Fig. 5B, C). The same analysis was performed on cells treated with shCtrl and shHIF-1α cultured under hypoxia (Fig. 5D–F). Metabolic profiling revealed that in hypoxia, glycolytic metabolite levels were increased and, as expected, TCA cycle intermediates - citrate, and isocitrate were decreased (Fig. 5C, G). KEGG pathway enrichment analysis showed that the galactose metabolism pathway was enriched in the subset of metabolites with elevated concentrations in hypoxia, whereas the glyoxylate and dicarboxylate metabolism pathway was enriched in the metabolites that were reduced in hypoxia (Fig. 5H). On the other hand, the metabolic profile of HIF-1α silenced cells showed accumulation of glucose and glucose-6-phosphate paired with decreased DHAP and pyruvate levels, decreased levels of TCA cycle intermediates and no change in cytosolic lactate, which is quickly exported out of the cells through MCT4⁽⁸⁾ (Fig. 5F, I). Accordingly, KEGG pathway analysis of HIF-1α knocked down cells showed that altered metabolites were enriched in the TCA cycle along with alanine, aspartate, and glutamate metabolic pathways (Fig. 5J).

To gain further insights into glycolysis and TCA interactions the HIF-1α silenced cells were cultured for 24 hours under hypoxia, with the addition of 50% enrichment in glucose with [1,2] ¹³C-glucose. The framework for ¹³C tracer analysis using [1,2-¹³C]-glucose is shown in Fig. 6A. As shown in Fig. 6A, [1,2]-¹³C glucose yields [2,3]-¹³C lactate accumulating in media via glycolysis (M2 lactate), while M1 lactate is generated after transit of glucose through the oxidative limb of the pentose cycle. [2,3]-¹³C pyruvate entering the TCA cycle (rather than exiting via lactate dehydrogenase) has 2 options for entry: via pyruvate dehydrogenase (PDH) and via pyruvate decarboxylase (PC). The direction of entry can be monitored by observing the distribution of carbons in media Glutamate 128 (m/z 128 M0, 129 M1, 130 M2), which is the 3 carbon, C2-C4 fragment of glutamate. As shown in Fig 6A, [2,3]-¹³C pyruvate entry via PC results in labeling glutamate at the 2^nd and 3^rd glucose position, while [2,3]-¹³C pyruvate entry via PDH results in glutamate labeling at the 4^th and 5^th carbon position. M1 Glutamate 128 reflects PDH flux, as it can occur (mainly) as the fragment containing the ¹³C at position 4 of glutamate. M2 Glutamate 128 reflects PC flux as it can occur (mainly) for the glutamate fragment containing ¹³C at positions 2nd and 3rd of glutamate. Glutamate 202 on the other hand, is the sum of PC and PDH flux, as it can contain ¹³C at positions 2 and 3, reflecting PC flux, or ¹³C at positions 4 and 5, reflecting PDH flux.

Steady state was seen at 24 hours for flux analysis on the basis of the total TCA cycle flux (Glu 202) (Suppl. Fig. 5A) as well as pentose cycle flux (calculated from lactate 146 and 147) (Suppl. Fig 5B). Pentose cycle flux was approximately 20% greater for shHIF-1α compare to shCtrl at both 12 hours and 24 hours at 5 mM glucose, while for total TCA cycle flux, a 2 fold change between shHIF-1α and shCtrl was seen only at 24 hours at 5 mM (Suppl. Fig. 5A). At 12 hours, total TCA cycle flux was only 30% higher for shHIF-1α vs shCtrl (Suppl. Fig 5A). Based on these findings, tracer analysis experiments were carried out at 24 hour.

Figure 6B and C show that the rate of lactate production was decreased with the HIF-1α knockdown, both at 5 mM and 10 mM glucose. Moreover the unlabeled (M0) media lactate (Fig. 6C, Suppl. Fig. 5C) and aspartate (Fig. 6D) showed a small decrease at 5 mM glucose, but it was more pronounced at 10 mM glucose. Media M0 glutamate was increased at 5 mM glucose by ~50% (Fig. 6E). Moreover, the decrease in media M2 lactate enrichment implied that flux was being re-directed into the TCA cycle, when the increased glutamate M0 was taken into account. The balance between flux entering the TCA cycle via pyruvate dehydrogenase (PDH) or pyruvate carboxylase (PC) with HIF-1α silencing was then assessed using glutamate ¹³C enrichments. The total PC+PDH flux was increased at both 5 mM and 10 mM glucose in HIF-1α knockdown cells (Fig. 6F). Moreover, the ratio of PDH to PC flux was also increased both at 5 mM and 10 mM glucose in HIF-1α knockdown cells (Fig. 6G). Thus, the flux through PDH alone was increased 2 fold at 5 mM glucose and nearly 3-fold at 10 mM glucose (Fig. 6H). Significant increase was also observed in the pentose cycle flux as measured by lactate (m/z 146 M1 and 147 M2) at both 5 and 10 mM glucose (Fig. 6I). These findings suggested that silencing of HIF-1α in NP cells resulted in decreased glycolysis and that it caused a dysregulation is glycolytic/mitochondrial TCA cycle flux interactions as well as increase in pentose cycle shunt activity. When these results are reconciled with metabolic profile of HIF-1α silenced cells it suggested that TCA flux was being directed to glutamate to try to normalize the NAD/NADH ratio to drive glycolysis and TCA cycle reactions (Fig. 6J).

Hypoxia and HIF-1α affect mitochondrial dynamics in NP cells

A few studies have shown a close connection between hypoxia and mitochondrial dynamics.^{(17, 39)} To investigate this dynamic process in NP cells, we examined the effects of hypoxia and HIF-1 loss-of-function on expression of proteins related to mitochondrial fission/fusion. NP cells culture in hypoxia increased the expression of fission protein, DRP1, without affecting levels of the fusion protein, OPA1 implying that DRP1 may contribute to hypoxia-dependent mitochondrial fragmentation (Fig. 7A–D). DFP, an inducer of mitophagy was used a positive control and showed changes in DRP1 and OPA1 levels similar to hypoxia (Fig. 7A–D). These results suggested that in hypoxic NP cells mitochondrial fission driven by elevated DRP1 levels may promote mitophagic signaling. We then investigated the contribution of HIF-1α in this process. HIF-1α silencing resulted in tubular mitochondria and a concomitant increase in levels of OPA1, without changes in DRP1 levels (Fig. 7E–H). These findings suggested that HIF-1α was responsible for mitochondrial fragmentation in hypoxic NP cells by modulating balance between levels of OPA1 and DRP1.

Hypoxia promotes mitochondrial translocation of BNIP3

Since hypoxia and HIF-1 affected mitochondrial fusion and fission, we assessed the effect of hypoxia on receptor-mediated mitophagy. We measured levels of BNIP3 and NIX in NP cells cultured under hypoxia (Fig. 8A–D). Interestingly, BNIP3 was mostly restricted to the nuclear compartment in normoxia (Fig. 8A). On the other hand, hypoxic treatment of NP cells resulted in a significant translocation of BNIP3 to mitochondria, similar to what we observed following DFP treatment (Fig. 8A,C). Hypoxia and DFP also induced overall levels of BNIP3 (Fig. 8D). However, NIX, another important mitophagy receptor protein did not show any hypoxia-dependent changes (Fig. 8B). Unlike BNIP3, NIX levels did not show significant changes under hypoxia and DFP treatment (Fig. 8D). Several studies have shown a cross talk between receptor and non-receptor mediated pathways to fine-tune mitophagy.⁽⁴⁰⁾ To understand the changes in non-receptor mediated pathways during NP cell mitophagy; we analyzed effect of hypoxia on PINK1 and PARK2 localization and expression levels. Upon depolarization of mitochondria, PINK1 localizes to mitochondria and recruits PARK2 to initiate classical non-receptor mediated mitophagy. We did observe increase in the levels of PINK1 with a trend of increasing levels PARK2 in hypoxic cells, a similar overall trend was observed in DFP treated cells (Fig. 8E–G).

HIF-1α is essential for mitochondrial translocation of BNIP3 and maintenance of PINK1 levels

In order to investigate if mitochondrial translocation of BNIP3 and NIX is controlled by HIF-1α, we assessed the status of these receptor-mediated pathway proteins after HIF-1α silencing (Fig. 9A–D). Knockdown of HIF-1α resulted in reduced mitochondrial and nuclear BNIP3 expression, and most of BNIP3 appeared to be sequestered in the nucleus, suggesting that HIF-1α is necessary for both the induction and mitochondrial translocation of BNIP3 (Fig. 9A, C, D). In contrast, NIX expression was increased with silencing of HIF-1α without any obvious changes in localization (Fig. 9B, D), suggesting NIX may be compensating for loss of mitochondrial BNIP3 availability. Likewise, PINK1 showed elevated expression after HIF-1α silencing however, PARK2, which is downstream of PINK1 did not alter expression (Fig. 9E–G).

Hypoxia induced mitophagy is BNIP3 dependent in NP cells

While hypoxia induces BNIP3⁽³⁰⁾ and HIF-1α is essential for its mitochondrial translocation, it is unclear whether this phenomenon is necessary for inducing mitophagy in NP cells. To address this issue, we blocked nuclear export of BNIP3 using leptomycin B, an inhibitor of the nuclear export receptor exportin 1 (CRM1), which causes nuclear accumulation of proteins. Immunofluorescence analysis showed that leptomycin B treatment for 24 h under hypoxia causes nuclear accumulation of BNIP3 and inhibited its mitochondrial translocation (Fig. 10A, B). Likewise, there was nuclear accumulation of Beclin1 without much effect on mitochondrial localization (Suppl. Fig. 3). Consequently, staining for LC3 in leptomycin B treated cells under hypoxia revealed a marked reduction in the number of mitophagosomes (Fig. 10D, E). These results are in agreement with previous studies that showed that cells treated with leptomycin B or expressing Beclin1 with deficiency in nuclear export resulted in nuclear accumulation, ultimately blocking its autophagic function.⁽⁴¹⁾ Accordingly, we observed inhibition of mitochondrial fragmentation and increased mitochondrial mass and networking as reflected by form factor and aspect ratio when BNIP3 and Beclin1 nuclear export was blocked (Fig. 10F–H). A similar increase in mitochondrial mass and dysfunctional mitochondria was reported in liver of BNIP3-deficient mice.⁽⁴²⁾

FoxA2^Cre-HIF-1α ^f/f mice show elevated NIX levels and maintenance of mitophagic flux

In order to further investigate whether loss of HIF-1α results in changes in mitophagy in vivo, we utilized NP-specific HIF-1α null mice by crossing HIF-1α floxed mice with mice expressing Cre recombinase under control of the notochord specific Foxa2 promoter/enhancer (Fig. 11A). ⁽²⁸⁾ We analyzed colocalization of LC3 with a mitochondrial marker, electron transport chain Complex III as a way to measure mitophagy. We observed a slight increase in the colocalization of LC3 with Complex III (Fig. 11B, D) with a concomitant increase in mitochondrial mass as reflected by % area stained for Complex III (Figure. 11C, E). Similarly, in agreement with our in vitro results, we observed a robust increase in NIX abundance in NP cells of HIF-1α mutant mice (Fig. 11F, G).

Together these results suggested that mitochondrial translocation of BNIP3 in NP cells is required for hypoxia-induced mitophagy and thus for mitochondrial quality control and that this process is HIF-1 dependent. On the other hand it is plausible that elevated levels of NIX, in the absence of HIF-1, compensate for loss of BNIP3 and continue mitophagic flux (Fig. 11G).