Survival, tumor development and motor behavioral deficits of mice with individual and combined ablation of Palb2 and Atg7

As heterozygous disruption of Becn1 delays Palb2-associated mammary tumor development [27], we sought to further examine the role of autophagy in mammary tumor development following PALB2 loss. Palb2^f/f;Wap-Cre (Palb2-CKO) mice were crossed with Atg7-floxed mice to generate mice with single and combined knockout of the two genes in the mammary gland. Palb2-CKO females showed a median overall survival (T₅₀) of 649 days (Fig 1A), and most of these mice developed spontaneous tumors (Fig 1B). Notably, only 6/19 (32%) of Palb2-CKO mice developed mammary tumors (Fig 1C), likely due to a mostly C57BL/6 genetic background, which is known to be more resistant to mammary tumorigenesis. Atg7^f/f;Wap-Cre (Atg7-CKO) mice had slightly shorter overall survival (T₅₀ = 604 days) (Fig 1A); 1 of the 22 mice (4.5%) developed mammary tumor and 9 (45.5%) developed other tumors (Fig 1C).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Survival curves and motor deficits of Palb2 and Atg7 deletion mice.

(A-C) Overall survival (A), tumor-free survival (B) and observed causes of death (C) of Wap-Cre driven CKO mice. Black ticks in B indicate censoring events (mice that died without tumor). (D-G) Footprint test of the CKO mice at 6 weeks of age. Representative footprints and variables measured are shown in D, and results of stride, base and overlap measurements are shown in E, F and G, respectively. n = 4–7 mice per genotype. (H and I) Beam balance test of the CKO mice. Beam balance scores are shown in H and the percentage of mice that passed the beam in I. n = 4–7 mice per genotype. (J) Neurodegeneration-associated death in the CKO cohorts. (K-L) Survival curves (K) and stride length (L) of mice with Ubc-Cre-ERT2 driven, whole body knockout (WBKO) of Palb2 and/or Atg7. n = 8–13 mice per genotype.

https://doi.org/10.1371/journal.pgen.1010138.g001

Unexpectedly, Wap-Cre driven Atg7-CKO mice showed progressive motor deficits, manifesting primarily as gait ataxia suggestive of cerebellar involvement. Note that although the Wap-cre used here was designed to express Cre specifically in the mammary gland, Cre activity has been reported in the brain [32]. Gait analysis at 6 weeks of age revealed an ataxic walking pattern of Atg7-CKO mice (Fig 1D). Compared with Wap-Cre (control) mice, mean stride length of Atg7-CKO mice was significantly decreased (Fig 1E), and both mean base width and overlap were significantly increased (Fig 1F and 1G). Beam balance test also showed gradual impairment of motor coordination in Atg7-CKO mice (Fig 1H and 1I). Additionally, these mice showed limb clasping reflexes when suspended by the tails (S1 Fig). These motor behavioral phenotypes are consistent with previous reports that loss of Atg7 in the brain leads to neurodegeneration [17,33,34]. The mice were able to survive with motor deficits until approximately 400 days of age and then started to die. Eventually, 12/22 (54.5%) of the mice succumbed to neurodegeneration-associated death (Fig 1J), defined as death with severe motor deficits and without tumor or other overt pathological conditions seen upon dissection.

Remarkably, Palb2^f/f;Atg7^f/f;Wap-Cre (Palb2;Atg7-CKO) mice exhibited greatly reduced overall survival (T₅₀ = 280 days) (Fig 1A), even though their tumor development was also greatly reduced (Fig 1B). These double CKO mice showed more severe motor deficits, especially in terms of stride length and balance (Fig 1E, 1H and 1I), and died from motor deficits much earlier than Atg7-CKO mice (Fig 1J). Similar to Atg7-CKO mice, these double CKO mice also showed limb clasping reflexes when suspended by the tails (S1 Fig). Only 1 out of 25 (4%) of these mice developed cancer, and all the rest died from severe motor deficits. The more severe motor deficits of the double CKO mice and the striking acceleration of death associated with neurodegeneration in these mice suggest that the two genes may synergistically suppress neurodegeneration. Although Palb2-CKO mice did not show any overt behavioral abnormalities, footprint tests showed increased hind base width and overlap (Fig 1F and 1G), suggestive of mild motor dysfunction.

We also crossed Atg7- and Palb2-floxed mice with mice carrying a Ubc-Cre-ERT2 allele, which allow for tamoxifen (TAM)-induced whole-body knockout (WBKO) of floxed genes [35]. One week after (5 daily) TAM injections, efficient deletions of both Palb2 and Atg7 were observed in brain tissues (S2A Fig). The resulting Palb2-WBKO (Palb2^Δ/Δ) mice appeared normal and survived for at least a year. In contrast, a small number of Atg7^Δ/Δ mice died from infections within the first week, the rest soon manifested motor deficits that worsened progressively over time, and the mice died within 120 days (Fig 1K), as we have shown recently [17,36]. Notably, Palb2^Δ/Δ;Atg7^Δ/Δ mice had even shorter survival compared with Atg7^Δ/Δ mice (T₅₀ = 16 vs 30 days, p<0.05). Gait analysis at 4 weeks post TAM treatment showed that both Palb2^Δ/Δ mice and Atg7^Δ/Δ mice had significantly shortened hind stride length, and the length was even shorter in Palb2^Δ/Δ;Atg7^Δ/Δ mice (Fig 1L). These double knockout mice also died primarily from severe motor deficits, except those that died shortly after injection due to infections. Thus, loss of PALB2 in the brain leads to mild motor impairment but greatly exacerbates the neurological phenotype induced by Atg7 deficiency.

Loss of PALB2 exacerbates neurodegeneration induced by ATG7 loss

To better understand the cause of the neurological phenotype in our models, we assessed the “leaky” expression of Cre in multiple tissues. Indeed, deletion of Palb2 was observed in the brains of Palb2-CKO and Palb2;Atg7-CKO mice (S2B and S2C Fig). Additionally, we analyzed the protein level of p62/SQSTM1, an autophagy adaptor protein that accumulates upon autophagy inhibition [37]. Western blotting showed strong p62 accumulation in midbrain, cerebellum and cerebral cortex in Atg7-CKO and Palb2;Atg7-CKO mice (S2D Fig). Immunofluorescence (IF) further revealed aberrant accumulation of p62 throughout the brain, with intense staining signals in midbrains and pons, and scattered signals in cerebral cortex and cerebellum (Figs 2A and 2B, and S3A and S3B). Significantly stronger p62 staining was found at 10 weeks than 6 weeks, and double CKO mice showed more p62 accumulation in midbrain and pons than did in Atg7-CKO mice at both ages. Further increased p62 accumulation was detected by immunohistochemistry (IHC) at 40 weeks (S4A Fig), indicative of a progressive accumulation of protein aggregates.

We also stained for tyrosine hydroxylase (TH), a marker of dopaminergic neurons in the substantia nigra that are degenerated in Parkinson’s disease. No significant difference in TH staining was observed between control and any of the CKO mice at either 6 or 10 weeks of age (Fig 2C), indicating that the dopaminergic neurons were likely intact by 10 weeks of age, consistent with a previous report (Friedman et al., 2012). However, p62 accumulation was observed in TH+ neurons in both Atg7-CKO and double CKO mice at 6 weeks, and a further accumulation was seen at 10 weeks. Moreover, 10 weeks old double CKO mice had more TH+;p62+ neurons than Atg7-CKO mice, again suggesting that loss of PALB2 exacerbated p62 accumulation induced by loss of ATG7/autophagy (Fig 2C and 2D). Moreover, we observed significantly reduced TH+ neurons in the substantia nigra of double CKO mice than in Atg7-CKO mice at 40 weeks (S4B Fig), suggesting that Atg7 loss induced degeneration of dopaminergic neurons were also accelerated by Palb2 loss.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Neuronal degeneration in Palb2 and Atg7 CKO mice.

(A-D) Representative immunofluorescence staining of p62/SQSTM1 and TH and quantification of their staining intensity in the whole midsagittal section (A and B) or substantia nigra (C and D) of the brain at 6 and 10 weeks of age. Nuclei were stained with DAPI. Images are representatives of two independent experiments. Scale bar = 1 mm in A and 20 μm in B as indicated. n = 3 mice per genotype. (E and F) Representative IHC staining of cerebellum with Purkinje cell marker Calbindin (E) and quantification of Purkinje cells (F) at 2, 4, 6, 8 and 10 weeks of age. n = 3 mice per genotype. Scale bar = 20 μm. (G and H) Representative TUNEL images (G) and quantification of positive cells (H) in the brains of CKO mice at 4 and 6 weeks of age. Scale bar = 100 μm. n = 3 mice per genotype.

https://doi.org/10.1371/journal.pgen.1010138.g002

It is well known that Atg7 deletion in the brain leads to loss of Purkinje cells in the cerebellum [24,34], which are the primary controller of motor coordination and balance, and the motor deficits observed in the present study are consistent with degeneration of these neurons. Indeed, IHC for Calbindin, a marker for Purkinje cells, demonstrated progressive loss of these cells in Atg7-CKO mice between 2 and 10 weeks of age (Fig 2E and 2F). Palb2-CKO mice showed only a modest reduction in the number of these cells at 2 weeks but no further reduction later, while the double CKO mice showed more pronounced and significantly accelerated Purkinje cell loss compared with Atg7-CKO mice (Fig 2F). Apoptotic cells were detected in both Purkinje cell layer and granule cell layer of both Atg7-CKO and Palb2;Atg7-CKO mice, with the double CKO mice showing more apoptotic Purkinje cells than Atg7-CKO mice at 4 weeks of age (Fig 2G). At 6 weeks, while Atg7-CKO mice still had some apoptotic Purkinje cells, these cells were hardly detectable in Palb2;Atg7-CKO mice. At this age, the double CKO mice also had significantly more apoptotic cells in the granule cell layer than did Atg7-CKO mice (Fig 2G and 2H). Collectively, the above observations indicate that Wap-Cre driven deletion of Atg7 induces neurodegeneration throughout the brain, with different neurons being affected differently in terms of timing and severity, and loss of Palb2 exacerbates and accelerates the neurodegeneration induced by Atg7 deletion. Since Purkinje cell is instantly recognizable and its loss is the most obvious histopathological change which also happens early and correlates with motor deficits caused by Atg7 loss, we mainly focus on these cells in the following studies.

DNA damage and oxidative stress in Palb2 and Atg7 deleted brains

To begin to understand the cause of the neurodegenerative phenotype of our mice and considering the known roles of PALB2 and autophagy in DNA repair and redox homeostasis, we assessed the levels of γH2AX, a marker for DSBs, 8-oxo-deoxyguanidine (8-oxo-dG), a marker for oxidative DNA damage, and 4-HNE, a product of lipid peroxidation, in cerebellum and midbrain of Palb2 and Atg7 CKO mice by IHC.

In cerebellum of both control and the CKO mice, positive γH2AX signals were only observed in Purkinje cells (Figs 3A and S5A and S5). Unlike the typical DNA damage-induced γH2AX nuclear foci, γH2AX in Purkinje cells was observed mostly in a single large, round structure in the nucleus, whose nature remains to be determined. At 6 weeks of age, about 16.6% of Purkinje cells in Wap-Cre control mice and 64.2% of Purkinje cells in Palb2-CKO mice showed positive γH2AX staining in the nucleus (Fig 3A and 3B). In Atg7-CKO mice and double CKO mice, most of the Purkinje cells had been lost, and the staining in remaining Purkinje cells cannot be interpreted with confidence due to cell shrinkage and possible ongoing apoptosis. In the midbrain, positive γH2AX signals were only observed in neurons and not in glial cells. Very few neurons in the control mice showed clear γH2AX foci and the signals were also weak, whereas most of neurons in Palb2-CKO, Atg7-CKO and Palb2;Atg7 double CKO mice were positive and the signals were also much stronger (Fig 3A). Notably, Palb2 single CKO neurons showed significantly more (but smaller) foci per cell, whereas Atg7 single CKO neurons showed 2–3 large foci per cell. Compared to Palb2 single CKO neurons, Palb2;Atg7 double CKO neurons showed stronger staining signal and larger foci although fewer in number; compared to Atg7 single CKO neurons, double CKO neurons showed both stronger staining signal and also more foci (Fig 3A and 3B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Increased DNA damage and oxidative stress and mitochondrial markers in Purkinje cells of Palb2 and Atg7 deletion mice.

(A-D) Representative IHC images and quantification of DNA DSB marker γH2AX (A and B) and oxidative DNA damage marker 8-oxo-dG (C and D) in cerebellum/Purkinje cells (upper images) and midbrain neurons (lower images) of the Wap-Cre driven CKO mice at 6 weeks. n = 3 mice per genotype. (E and F) Representative IHC images and quantification of lipid peroxidation marker 4-HNE in the cerebellum of the CKO mice at 6 weeks of age. Scale bar = 50 μM. n = 3 mice per genotype. (G) GSH:GSSG ratios in the cerebellum and midbrain of Ubc-Cre-ERT2 driven WBKO mice. GSH and GSSG were measured by LC–MS/MS at 4 weeks after tamoxifen injection. n = 6–9 for cerebellum and n = 9–15 for midbrain. (H-K) Representative IHC images and quantification of MT-CO1 (H and I) and SOD2 (J and K) in cerebellum/Purkinje cells (upper images) and midbrain neurons (lower images) of CKO mice at 8 weeks. Red dots indicate Purkinje cells. Scale bar = 50 μm. n = 3 mice per genotype.

https://doi.org/10.1371/journal.pgen.1010138.g003

γH2AX was also analyzed in the brains of the afore-described WBKO (Δ/Δ) mice. As expected, increased γH2AX signal was found in Purkinje cells of Palb2^Δ/Δ mice at 4 weeks after tamoxifen injection (S6A Fig). At this time, Purkinje cells had been mostly lost in Atg7^Δ/Δ and Palb2^Δ/Δ;Atg7^Δ/Δ mice; however, increased γH2AX was detected in a subset of granule cells in these mice, suggesting that ATG7 plays a role, directly or indirectly, in DNA damage repair, consistent with our recent findings [17]. Moreover, we subjected the mice to 10 Gy of whole-body gamma radiation. As expected, Palb2^Δ/Δ mice showed substantially reduced survival compared with either control or Atg7^Δ/Δ mice (S6B Fig). Notably, however, Palb2^Δ/Δ;Atg7^Δ/Δ mice exhibited similar survival to that of Palb2^Δ/Δ mice post radiation, suggesting that the DNA repair deficit of Palb2^Δ/Δ;Atg7^Δ/Δ mice stems primarily from PALB2 deficiency.

In the cerebellum, 8-oxo-dG staining signals were observed in both Purkinje and granular cells, but any clear difference between control and CKO mice was mainly found in Purkinje cells (S5C and S5D Fig). At 2 weeks of age, very few Purkinje cells in the control mice showed clear positive 8-oxo-dG staining and the signals were also weak, whereas more cells in Palb2-CKO and Atg7-CKO mice were positive and the signals were slightly stronger. Notably, Palb2;Atg7-CKO mice had even more 8-oxo-dG positive Purkinje cells at this time point, and the signals were also substantially stronger. At 4 weeks, the staining in Palb2-CKO mice was much stronger than that at 2 weeks and also much stronger than that in either control or Atg7-CKO mice, although Atg7-CKO mice showed slightly higher staining intensity than the control. At 6 weeks of age, about 22% of Purkinje cells in the control mice showed clear positive 8-oxo-dG staining and the signals were also weak, whereas 54% cells in Palb2-CKO mice were positive and the signals were much stronger (Fig 3C and 3D). In midbrain of 6 weeks mice, since most of neurons showed positive 8-oxo-dG staining, we only measured staining intensity. Compared with control mice, much stronger 8-oxo-dG staining was found in both Palb2-CKO mice and double CKO mice, and Atg7-CKO mice showed moderately higher staining intensity (Fig 3C and 3D). These results suggest that PALB2 is critical for preventing DNA oxidation in both Purkinje cells and midbrain neurons, whereas ATG7 plays a less prominent role. On the other hand, 4-HNE, a lipid peroxidation by-product, was evidently more intense in Atg7-CKO and Palb2;Atg7-CKO brains (Fig 3E and 3F). Thus, ATG7 and autophagy play a more important role in suppressing lipid peroxidation than does PALB2.

Finally, we also determined the ratio of reduced to oxidized glutathione (GSH:GSSG). Reduced glutathione (GSH) is known to play a pivotal role in the cellular oxidant defense system and is indispensable for preventing lipid peroxidation by free radicals and oxygen [38], and decreased GSH:GSSG ratio is a key indicator of decreased ability to reduce cytotoxic lipid peroxides [39]. Cerebellum and midbrain were collected from the WBKO mice and tissue metabolites were measured by liquid chromatography-mass spectrometry (LC-MS). As shown in Fig 3G, in the brain of the WBKO mice, GSH:GSSG ratio was modestly (though not statistically significantly) decreased in the cerebellum of both Palb2^Δ/Δ and Atg7^Δ/Δ mice, and the ratio was further and statistically significantly decreased in the cerebellum of Palb2^Δ/Δ;Atg7^Δ/Δ mice. In the midbrain, the same trends were observed, and all the differences were larger and statistically significant. These results further support the increase in lipid peroxidation in the cerebellum and midbrain induced by combined deletion of Palb2 and Atg7. Taken together, our results reveal both similar and distinct impacts of PALB2 and ATG7 loss on DNA damage and oxidative stress in the brain.

Requirement of PALB2 for mitochondrial homeostasis in the brain

Given the key role of autophagy in mitochondrial quality control and mitochondrial dysfunction in neurodegenerative diseases, we analyzed two mitochondrial proteins by IHC. First, we examined Cytochrome c oxidase subunit I (MT-CO1), a subunit of respiratory Complex IV, which is located within the mitochondrial inner membrane and is the final enzyme of the electron transport chain of mitochondrial oxidative phosphorylation [40]. At 2 weeks of age, both Palb2-CKO and Atg7-CKO mice appeared to have slightly higher levels of MT-CO1 than control; at 4 weeks, MT-CO1 signals in Palb2-CKO Purkinje cells were much stronger than control, while signals in Atg7-CKO and Palb2;Atg7 double CKO cells were hard to interpret due to cell shrinkage (S5E and S5F Fig). At 8 weeks of age, both Purkinje cells and midbrain neurons had much higher staining intensity of MT-CO1 in Palb2-CKO mice than in control; in Atg7-CKO and Palb2;Atg7 double CKO midbrain neurons, MT-CO1 was moderately higher than that in control (Fig 3H and 3I). Additionally, increased MT-CO1 was also found in Purkinje cells of Palb2^Δ/Δ mice (S6A Fig). Next, we stained for SOD2 (MnSOD), a superoxide dismutase that is localized in the mitochondrial intermembrane space and converts superoxide anion resulting from leaked electrons from the electron transport chain to hydrogen peroxide (H₂O₂) [41]. At 8 weeks, both Purkinje cells and midbrain neurons in Palb2-CKO mice showed much stronger SOD2 staining than controls, signals in Atg7-CKO midbrain neurons were even stronger than in Palb2-CKO mice, while neurons in the double CKO midbrain showed strongest staining (Fig 3J and 3K).

Regulation of mitochondrial gene expression by PALB2 in human medulloblastoma cells

To confirm the role of PALB2 in regulating mitochondria, we conducted RNA-seq analysis following siRNA-mediated knockdown of PALB2. Six different PALB2 siRNAs (along with five controls) were individually transfected into DAOY medulloblastoma cells, which have a cerebellar origin [42], to deplete the protein (Fig 4A). A total of 540 genes were significantly up- or down-regulated in PALB2 knockdown cells (S1 File). Top 50 upregulated and top 50 downregulated differentially expressed genes (DEGs) affected by PALB2 knockdown are shown in Fig 4B. Gene set enrichment analysis (GSEA) revealed highly enriched DEGs in oxidative phosphorylation (Fig 4C and 4D), mitochondrial gene expression (Fig 4E and 4F), mitochondrial translation, respiratory chain complex assembly, mitochondrial electron transport chain, ATP synthesis coupled electron transport and mitochondrial complex I (NADH:ubiquinone oxidoreductase) (S1 File), again implicating PALB2 in mitochondrial regulation. Quantitative reverse transcription PCR (qRT-PCR) further validated the RNA-seq results and confirmed the upregulation of MT-CO1, MT-CO2 and COX4i1, which encodes Cytochrome c oxidase subunit 4 isoform 1, another subunit of respiratory Complex IV, in PALB2 knockdown cells. Furthermore, the expression of PGC1A, encoding peroxisome proliferator-activated receptor gamma coactivator 1-alpha, a transcriptional coactivator and central inducer of mitochondrial biogenesis [43,44], was also increased in PALB2 knockdown cells, supporting a role for PALB2 in mitochondrial biogenesis. Next, we analyzed mitochondria using MitoTracker Green and MitoTracker Red, which measure mitochondrial mass and mitochondrial membrane potential, respectively. Both fluorescence microscopic visualization and flow cytometry analysis showed significantly increased mitochondrial mass and membrane potential in PALB2 knockdown cells (Fig 4H–4J).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Impact of PLAB2 loss on mitochondrial gene expression and function in human medulloblastoma cells.

(A) Western blot analysis of PALB2 levels in DAOY cells treated with transfection reagent alone, 4 different control siRNAs and 6 different PALB2 siRNAs. β-ACTIN was used as loading control. (B) Heat map showing top 50 upregulated and top 50 downregulated genes from two-way hierarchical clustering analysis of the 11 samples (5 control and 6 siPALB2). (C and D) GSEA enrichment plot (C) and heat map (D) of the expression of genes in the “oxidative phosphorylation” gene set in control and PALB2 knockdown cells. (E and F) GSEA enrichment plot (E) and heat map (F) of the expression of genes in the “mitochondrial gene expression” gene set in control and PALB2 knockdown cells. (G) qRT–PCR analysis of the expression of PGC1α, MT-CO1, MT-CO2 and COX4i1 in control and PALB2 knockdown cells. Results are averages of the 5 control and 6 PALB2 knockdown cells from three independent experiments, each performed in two technical replicates. (H) Representative fluorescence images of MitoTracker Green and MitoTracker Red from control and PALB2 knockdown cells. Scale bar = 10 μm. (I and J) Representative flow cytometry overlay plot (I) and quantified fluorescence intensity (J) of MitoTraker Green and MitoTracker Red signals in the above control and PALB2 knockdown cells. n = 3 independent experiments.

https://doi.org/10.1371/journal.pgen.1010138.g004

Mitochondrial abnormality, oxidative stress, and cell death upon combined loss of PALB2 and ATG7 in human medulloblastoma cells

To further assess the role of PALB2 in mitochondrial biogenesis/function and its functional relationship with ATG7 in human cells, we used CRISPR/Cas9 to knock out PALB2 and ATG7 in the DAOY medulloblastoma cells. Clones were screened using western blotting (S7A Fig). PALB2;ATG7 double knockout (DKO) clones were then generated by further knocking out ATG7 in a sequence validated and functionally characterized PALB2-KO clone. Consistent with the essential role of ATG7 in LC3 lipidation and autophagosome formation [45], ATG7-KO and PALB2;ATG7-DKO cells showed reduced LC3-I to LC3-II conversion and increased p62 levels (Fig 5A). MT-CO1, MT-CO2 and SOD2 were all increased in PALB2-KO, ATG7-KO and PALB2;ATG7-DKO cells, implying increased mitochondrial mass upon loss of either PALB2 or ATG7. The growth rate of ATG7-KO cells was slightly slower than that of control cells, while PALB2-KO cells grew significantly slower, and a further decrease in growth rate was seen for PALB2;ATG7-DKO cells (Fig 5B). Both PALB2-KO and ATG7-KO cells showed moderate boundary shrinkage, and the DKO cells showed overt boundary shrinkage along with soma cavitation (Fig 5C). Spontaneous cell death, including both apoptosis and necrosis, was increased in PALB2-KO and ATG7-KO cells and further increased in the DKO cells (Fig 5D and 5E).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Mitochondrial abnormality, oxidative stress, and apoptosis in PALB2 and ATG7 KO human medulloblastoma cells.

(A) Western blot analysis of whole cell homogenates of WT and KO cells. All 9 cell lines were from single clones. The WT cell lines were false positive clones from the PALB2 CRISPR/Cas9 KO procedure. (B) Doubling times of WT and KO cells of the 9 cell lines in A. n = 3 independent experiments. Cells of the same genotype were grouped together. (C) Representative bright field microscopy images of the cells. Scale bar = 10 μm. (D and E) Representative flow cytometry plots (D) and quantification (E) of FITC Annexin V analysis of the cells. n = 3 independent experiments, each performed in technical duplicates. Cells of the same genotype were grouped together in E. (F and G) Mitochondrial analyses of WT and KO cells with MitoTracker-Green and MitoTracker-Red. Representative images of stained cells are shown in F. Flow cytometry-based quantification of the signals in all cell clones are shown in G. Avg. R/G, average ratios of Red vs Green signals (in percentage). Scale bar = 10 μm. (H) Oxygen consumption rate (OCR) in WT and KO cells as measured by the Seahorse assay. n = 5 technical replicates. (I) Bar graph of mitochondrial functional parameters as measured in panel H. (J and K) Levels of reactive oxygen species in WT and KO cells as measured by DCF (J) and MitoSOX (K) staining followed by flow cytometry-based detection and quantification.

https://doi.org/10.1371/journal.pgen.1010138.g005

We also analyzed mitochondria using MitoTracker Green and MitoTracker Red in all the positive clones at early passage and quantified the signals by flow cytometry (Figs 5F and 5G, and S7A and S7B). Consistent with previous reports [36,46,47], ATG7-KO cells showed an increase in total mitochondrial mass. Notably, an even larger increase in mitochondrial mass was observed in PALB2-KO cells. Both PALB2 and ATG7 single KO cells showed modestly increased membrane potential compared with WT cells, whereas the DKO cells showed no change. The ratio of MitoTracker Red to Green signals (R/G) of PALB2-KO, ATG7-KO and the DKO cells was 85.09%, 92.82% and 88.76%, respectively, of wt level, indicative of impaired mitochondrial quality in all 3 cell types (Fig 5F and 5G). The increased mitochondrial mass in PALB2-KO cells corroborates with our RNA-seq results (Fig 4) and in vivo findings (Fig 3) and lends further support to the notion that loss of PALB2 leads to increased mitochondrial biogenesis, while the decreased R/G ratio in the cells suggests lower overall mitochondria quality or a defect in the clearance of damaged mitochondria. To evaluate if mitochondrial respiration was affected by PALB2 and/or ATG7 inactivation, oxygen consumption rate (OCR) was measured using the Seahorse assay. Basal mitochondrial respiration, ATP-linked respiration and maximal respiration were moderately decreased in both PALB2-KO and ATG7-KO cells, but dramatically decreased in PALB2;ATG7-DKO cells (Fig 5H and 5I). These data suggest that a strong defect in mitochondrial respiration may be a major contributor to the spontaneous death of DKO cells.

Mitochondria are both a major producer and a primary target of ROS [48]. Consistent with our in vivo observations, cellular ROS level measured by DCF, which mainly detects H₂O₂, was significantly higher in PALB2-KO cells and even higher in PALB2;ATG7-DKO cells; however, no change in DCF readout was observed in ATG7-KO cells (Fig 5J). In contrast, mitochondrial superoxide, as measured by MitoSOX, was increased in ATG7-KO cells but not in PALB2-KO or PALB2;ATG7-DKO cells (Fig 5K). The increased superoxide level in ATG7-KO cells is consistent with recent reports [49,50], and the lack of any change in DCF readings suggests that the overproduced superoxide in these cells was not converted to H₂O₂ in a timely fashion, or that any increase in H₂O₂ had been promptly disposed of or reacted with polyunsaturated fatty acids leading to lipid peroxidation. On the other hand, the increase in DCF signals in PALB2-KO cells without any concomitant increase in superoxide suggests that the higher ROS level may not originate from mitochondria, or the overproduced superoxide has been converted to H₂O₂ by SOD2. These data are largely consistent with our findings from the mouse brain and support the notion that a combination of abnormal mitochondria function and increased oxidative stress leads to apoptosis in Purkinje cells and certain other neurons, hence the motor deficits.

Loss of BRCA2 delays the onset of neurodegeneration induced by ATG7 deficiency

To assess whether an HR deficiency and increased DSB formation were a significant cause of the more severe neurodegeneration in Palb2;Atg7-CKO mice than in Atg7-CKO mice, we analyzed Brca2^f/f;Atg7^f/f;Wap-cre (Brca2;Atg7-CKO) mice. Interestingly, unlike Palb2;Atg7-CKO mice, Brca2;Atg7-CKO mice had similar overall survival and moderately reduced neurodegeneration-associated death compared with Atg7-CKO mice (Fig 6A–6C). Also unlike Palb2, co-deletion of Brca2 with Atg7 did not exacerbate the neurological phenotype of Atg7-CKO mice; instead, it significantly moderated Purkinje cell loss caused by Atg7 deficiency from 2 to 10 weeks of age (Fig 6D). Both gait analysis and beam balance test also revealed a rescue of motor coordination and balance defects by co-deletion of Brca2 with Atg7 (Fig 6E and 6F).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Comparative analyses of the impacts of Brca2 and Palb2 deletions in wt and Atg7-CKO mice.

(A-C) Overall survival (A), neurodegeneration-associated death (B) and cause of death (C) of the CKO mice. Atg7^f/f;Wap-Cre, n = 22; Palb2^f/f; Atg7^f/f;Wap-Cre, n = 25; Brca2^f/f; Atg7^f/f;Wap-Cre, n = 24. (D) Purkinje cell numbers in CKO mice at 2 to 10 weeks of age. n = 3 mice per genotype per time point. (E) Footprint measurements of 6 weeks old CKO mice. n = 5–7 mice per genotype. (F) Beam balance scores of the CKO mice at different ages. (G) Representative images of γH2AX, 8-oxo-dG and MT-CO1 IHC in the cerebellum of 6 weeks old CKO mice. Red dots indicate Purkinje cells. Scale bar = 50 μm. (H-J) Quantification of γH2AX positivity (H), 8-oxo-dG positivity (I) and MT-CO1 staining intensity (J) in Purkinje cells (n = 3 mice per genotype).

https://doi.org/10.1371/journal.pgen.1010138.g006

To understand why ablation of Brca2 produced a different effect from that of Palb2 ablation, we compared γH2AX, 8-oxo-dG and MT-CO1 in the brains of the mice at 6 weeks of age. As shown in Fig 6G–6J, γH2AX positivity in Purkinje cells of Brca2-CKO mice was slightly lower than that in Palb2-CKO mice but still much higher than that in control mice; in contrast, their 8-oxo-dG and MT-CO1 staining signals were substantially weaker than those in Palb2-deleted Purkinje cells. The modestly higher γH2AX positivity in Palb2-deleted Purkinje cells compared with Brca2-deleted cells may be attributed to higher oxidative stress, which can cause DSBs. Compared with the remaining Purkinje cells in Atg7-CKO mice, the same cells in Brca2;Atg7-CKO mice showed similar γH2AX positivity. The staining patterns of both MT-CO1 and 8-oxo-dG in Purkinje cells of Brca2;Atg7-CKO mice were also similar to that in Atg7-CKO mice. Therefore, loss of DNA repair function of PALB2 is unlikely to be a significant cause of the more severe neurological phenotype of Palb2;Atg7-CKO mice, and that increased oxidative stress and/or mitochondrial dysfunction may underlie this exacerbated phenotype.

Loss of p53 delays Purkinje cell death and prolongs the survival of Palb2;Atg7-CKO mice

Our recent study showed that ATG7 limits p53 activation and p53-induced neurodegeneration [17]. To determine whether p53 played a role in the shortened survival of Palb2;Atg7-CKO mice, we set up cohorts of these mice either without or with floxed Trp53. Compared with Palb2;Atg7-CKO mice, overall survival of Palb2;Atg7;Trp53-CKO mice was significantly prolonged (T₅₀ = 448 vs 280 days, p = 0.0165) (Fig 7A), and neurodegeneration-associated death was greatly reduced and delayed (Fig 7B). Moreover, inactivation of Trp53 combined with prolonged survival allowed for efficient tumor development (Fig 7C). While 95% of Palb2;Atg7-CKO mice died from neurodegeneration, 45% of Palb2;Atg7;Trp53-CKO mice developed mammary tumors and 21% developed other tumors, with neurodegeneration causing the death of only 30% of the mice.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Role of p53 in Purkinje cell death and the survival of Palb2;Atg7-CKO mice.

(A-C) Overall survival (A), neurodegeneration-associated death (B), and tumor free survival (C) of Palb2;Atg7-CKO and Palb2;Atg7;Trp53-CKO mice. Where applicable, black ticks indicate censoring events. Palb2;Atg7-CKO, n = 25; Palb2;Atg7;Trp53-CKO, n = 22. (D and E) Representative p53 IHC images (D) and quantification of p53 positive Purkinje cells (E) in the cerebellum of control and CKO mice at 2 and 4 weeks of age. Red dots indicate Purkinje cells. Scale bar = 50 μm. n = 3 mice per genotype per time point. (F) Representative H&E and IHC images of γH2AX, 8-oxo-dG and MT-CO1 in the cerebellum of the CKO mice at 6 weeks of age. Images are representatives from two independent experiments. Red dots indicate Purkinje cells. Scale bar = 50 μm. (G) Purkinje cell numbers in the CKO mice at 6 weeks of age, n = 2–3 mice per genotype. (H) Beam balance scores of the CKO mice at 11 weeks of age. n = 5–7 mice per genotype. (I and J) Quantification of γH2AX positive (I) and 8-oxo-dG positive (J) Purkinje cells at 6 weeks of age. n = 2–3 mice per genotype. (K) Number of apoptotic cells in the cerebellum of 6 weeks old control and CKO mice. n = 2–3 mice per genotype.

https://doi.org/10.1371/journal.pgen.1010138.g007

The above results suggest that p53 was activated upon loss of PALB2 and/or ATG7 in the brain which subsequently led to neuronal apoptosis. Indeed, IHC analysis revealed that at 2 weeks of age, p53 protein level was moderately increased in Atg7-CKO Purkinje cells in terms of both staining intensity and number of positive cells, and even higher p53 accumulation was seen in Purkinje cells of Palb2-CKO and the double CKO mice (Fig 7D and 7E). The overall situation was similar at 4 weeks, except that some or most Purkinje cells in Atg7-KO mice and Palb2;Atg7-CKO mice had been lost.

Co-deletion of Trp53 produced no significant effect on Purkinje cell numbers in control, Palb2-CKO or Atg7-CKO mice; however, the severe loss of Purkinje cells in Palb2;Atg7-CKO mice was significantly rescued (Fig 7F and 7G). Interestingly, when Trp53 was co-deleted, Purkinje cell number in Palb2;Atg7-CKO mice was restored to a level similar to that in Atg7-CKO mice (Fig 7G), suggesting that the severe neurodegenerative phenotype in Palb2;Atg7-CKO mice is likely an Atg7 null phenotype exacerbated by a further induction of p53 elicited by loss of PALB2. Consistent with the rescue of Purkinje cell number, motor function of Palb2;Atg7;Trp53-CKO mice was substantially preserved, with the mice achieving similar beam balance scores as Atg7-CKO mice at 11 weeks of age (Fig 7H).

Deletion of Trp53 alone (in Wap-Cre control mice) did not cause any significant effect on γH2AX staining in Purkinje cells but led to a substantial reduction in 8-oxo-dG positivity (Fig 7I and 7J), and the low basal level of apoptosis (detected in whole brain sagittal sections) was unaffected (Fig 7K). Co-ablation of Trp53 with Palb2 had no effect on any of these three parameters, whereas inactivation of Trp53 in Atg7-CKO and Palb2;Atg7-CKO mice led to a decrease in all three parameters (Fig 7I–7K); however, the results of γH2AX and 8-oxo-dG need to be interpreted with caution because most of the Purkinje cells were lost in mice with wt p53. Taken together, these results demonstrate an important role for p53 in inducing neuronal death upon combined loss of PALB2 and ATG7/autophagy.

Partial rescue of survival of Palb2^Δ/Δ;Atg7^Δ/Δ mice by an ROS scavenger

Collectively, results presented above suggest that excessively high oxidative stress, rather than increased DNA damage, in the brain, particularly Purkinje cells, underlies the severe neurodegenerative phenotype of Palb2;Atg7-CKO and Palb2^Δ/Δ;Atg7^Δ/Δ mice. To test this notion, we treated Palb2^Δ/Δ;Atg7^Δ/Δ mice with N-acetylcysteine (NAC), an ROS scavenger, starting from the first day of 4 daily TAM injections, and monitored their survival. Compared with untreated mice, NAC-treated animals showed similar survival during the first 35 days or so and then survived significantly longer (Fig 8A and 8B). The divergence suggests that there likely were two different causes of death in the mice and that NAC protects against late death or the second cause, which was neurodegeneration. The early death appeared to be caused by tamoxifen toxicity or Streptococcus infection as we showed in our previous study [51].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. Partial rescue of Palb2;Atg7-WBKO mouse survival by an ROS scavenger and a summary of key events occurring in the brain of the model mice in this study.

(A and B) Survival curves of untreated (n = 11) and NAC-treated (n = 9) Palb2;Atg7-WBKO mice. Panel A includes all cause deaths, while in panel B deaths that occurred within the first 35 days were treated as censoring events due to causes unrelated to neurodegeneration. (C) PCR detection of tamoxifen induced, Ubc-Cre-ERT2-mediated Palb2 or Atg7 deletion in the liver of the mice (note that the whole brain was fixed for all mice). (D) Representative IHC images of Calbindin, γH2AX and 8-oxo-dG in the cerebellum and midbrain of Palb2;Atg7-WBKO mice with or without 8 days of NAC treatment. Red dots indicate Purkinje cells. Scale bar = 200 μm or 50 μm as indicated. n = 3 mice per condition. (E) Quantification of Purkinje cell number, γH2AX positivity and 8-oxo-dG staining intensity in the cerebellum/Purkinje cells and midbrain neurons of Palb2;Atg7-WBKO mice with or without NAC treatment. n = 3 mice per condition. (F) A summary of the impact of PALB2 and ATG7 loss on DNA damage (DSBs), DNA oxidation, lipid peroxidation, mitochondria and p53 status. Colors from light to dark denote mild to strong effects. See texts for details.

https://doi.org/10.1371/journal.pgen.1010138.g008

To test the effect of NAC on DNA damage, oxidative stress, and neuronal survival, we separately treated another set of Palb2^Δ/Δ;Atg7^Δ/Δ mice with NAC for 8 days, also starting from the first day of 4 daily TAM injections, which generated efficient deletion of the genes (Fig 8C). IHC results showed that by the end of the 8 days, Purkinje cells in NAC untreated mice were almost all lost, and this loss was strongly rescued by NAC treatment (Fig 8D and 8E). As few Purkinje cells remained in the control mice, comparison of DNA damage and oxidative stress was not feasible in these cells. However, in midbrain neurons, it was evident that NAC treatment significantly reduced 8-oxo-dG but not γH2AX staining (Fig 8D and 8E). These results provide further support that excessive oxidative stress, rather than increased DNA damage, caused by co-deletion of Palb2 with Atg7 underlies the severe neurodegenerative phenotype of Palb2;Atg7-CKO and Palb2^Δ/Δ;Atg7^Δ/Δ mice.

Ethics statement

All animal works were approved by the Institutional Animal Care and Use Committee (IACUC) of Rutgers University.

Mouse models

The Palb2^flox2-3;Wap-Cre (Palb2-CKO) mice were described previously [27]. They were crossed to strains carrying Atg7^flox [24], Brca2^flox11 [65], Trp53^flox2-10 [65], or Ubc-Cre-ERT2 (The Jackson Laboratory) alleles to generate all the genotypes in this study. Wap-Cre driven CKO females were all mated to go through two rounds of pregnancy and lactation to induce Cre expression and then monitored for survival and tumor development; males were used for studying neurodegeneration. To generate whole-body somatic Palb2 and/or Atg7 knockout in adult mice, 8–10 weeks old males with floxed alleles and Ubc-Cre-ERT2 were subjected to 5 consecutive daily intraperitoneal injections (unless otherwise specified) of tamoxifen (TAM, 200 μl of 10 mg/ml solution per mouse). A PCR based genotyping method was used to validate Palb2 deletion as described before [66], and primers used for validation of Atg7 gene deletion were as follows: forward 5’-aggcagggaggctaaatggt-3’, reverse 5’- gggcgccagttaagaacgat-3’. Diagnostic criteria for neurodegeneration-associated death are 1) a mouse has strong motor coordination and balance impairment; 2) loss of more than 15% body weight; and 3) no tumors found upon dissection.

Motor coordination assessments in mice

Motor coordination was assessed as previously described [67] with minor modifications. For the footprint test, forepaws and hindpaws of the mice were coated with red and green nontoxic paints, respectively. Mice were trained to walk along a 50-cm-long, 10-cm-wide, paper-covered runway (with 10-cm-high walls) into an enclosed box. All mice (5–7 mice per group) were given three runs per day at each time point (6, 9 and 12 weeks of age). A fresh sheet of white paper was placed on the floor of the runway for each run. Footprint patterns were assessed quantitatively by three measurements: stride length, hindbase width and front/hind footprint overlap. For WBKO mice, hind stride length was measured. Beam balance test was performed using a homebuilt experimental setup of a 50 cm-long, 12 mm-diameter round horizontal beam. At each time point (7, 9, 11, 13, 15, 17, 19 and 20 weeks of age), mice (5–7 per group) were first trained to walk on the beam. For data collection, three trials were performed per animal per day. The time the animal remained on the beam and the steps taken without falling were recorded. Both falling and hanging on the beam were counted as falls (a cushioned pad was used to prevent injury). Mice that successfully walked on the beam for 10 steps without falling were assigned 100 points. For mice that could not pass the “10 steps no fall rule”, we used a walk and stay combined rule in which their score depended on how many steps they could walk and how long they could stay at the spot where they eventually fell off. In this case, the number of steps taken and length of stay each accounted for half, or 50 points, of the full score, and each step walked or every 6 seconds of stay was given 5 points; these were then added to yield the score for these mice. The final daily beam walking score was the mean score of the three beam-walking trials.

Whole-body γ-radiation

Mice were retained in a Rodent RadDisk and irradiated using a Gammacell 40 Extractor (MDS Nordion) γ-irradiator at a dose rate of 91.6 cGy/min to a total of dose 10 Gy.

N-Acetyl-L-Cysteine (NAC) treatment

Eight to ten weeks old Ubc-Cre-ERT2 model mice were subjected to 4 daily injections of tamoxifen to delete genes of interest and then separated into 2 groups. One group was treated with sterile filtered water supplemented with 1% NAC (Sigma A9165) (pH ~7.4) and the other group was given control water. Water was changed every 2–3 days and the mice were monitored for survival. For histopathology assessment, six 10-week-old female Palb2^f/f;Atg7^f/f;Ubc-CreERT2 mice were subjected to 4-day schedule of TAM injection to delete Palb2 and Atg7. Starting at the first day of TAM administration, three mice were given NAC via drinking water (1 g NAC per 100 ml) and the other three given water without NAC. Mice were euthanized and tissues collected after 8 days of NAC treatment.

Tissue collection and immunostaining

At indicated time points, mice were anesthetized and perfused first with phosphate-buffered saline (PBS) and then with 4% paraformaldehyde (PFA). Brains were dissected, post fixed in 4% PFA for 24–72 h at 4°C, and then transferred to 70% ethanol prior to further processing. Paraffin-embedded block production and sectioning were conducted by The Histopathology Shared Resources of Rutgers Cancer Institute of New Jersey. IHC was performed on 5 μm sections as described before [68]. Paraffin sections were stained with antibodies against SQSTM1/p62 (1:1000 for IHC and 1:5000 for IF, Abcam, ab91526); Tyrosine Hydroxylase (EP1532Y, 1:300 for IHC and 1:2000 for IF, Abcam, ab137869); Calbindin (EP3478, 1:4000, Abcam, ab108404); γ-H2AX (1:300; Millipore, 05–636), 8-oxo-dG (1:2000; Trevigen); MT-CO1 (1:300, Abcam, ab14705), 4-HNE (EP1532Y, 1:500, Abcam, ab48506); p53 (CM5, 1:500; Leica Biosystems) and SOD2 (1:200, EMD Millipore, 06–984). Immunostained midsagittal sections were scanned using an Olympus VS120 instrument and images were generated using OlyVIA software for viewing.

Three animals per condition and two sections per animal were used for every quantification. Purkinje cell numbers were measured by total Calbindin positive Purkinje cells per section divided by the total length of Purkinje cell layer measured by OlyVIA. For γ-H2AX staining, in cerebellum Purkinje cells, ≥1 foci/cell were counted as positive; in midbrain neurons, foci number/cell were counted. For quantification of 8-oxo-dG positive ratios in Purkinje cells, both light and dark brown staining were counted as positive. The intensity of staining was recorded as 0, 1, 2, and 3 and the percentage of positive staining was recorded from 0 to 100%. The results of staining were scored using quick (Q) score, which was obtained by multiplying the percentage (P) by the intensity (I) (Q = P × I; maximum = 300). At least 100 Purkinje cells or all Purkinje cells in a brain section (when the total number in the section was less than 100) and at least 50 midbrain neurons/section were counted for each sample. For 4-HNE staining, the intensity of staining was recorded as 0, 1, 2, and 3 and the percentage of positive staining area was recorded from 0 to 100%. For p62 IF staining, images were converted to grayscale and measured by ImageJ on the midsagittal brain sections.

Cell culture, RNA interference, RNA-Seq analysis and qRT-PCR

DAOY cells were purchased from American Type Culture Collection (ATCC) and cultured in DMEM/F12 medium with 10% heat-inactivated fetal bovine serum (FBS) and 1X penicillin-streptomycin (Pen-Strep) in a humidified incubator with 5% CO₂. For PALB2 knockdown, cells were seeded at 1x10⁵ cells per well in 6 well plates and transfected with siRNAs using Lipofectamine RNAiMax (ThermoFisher, #13778150) following manufacturer’s instructions. The final concentration of siRNAs was 8 nM. The sequences of the sense strands of the PALB2 siRNAs are listed in S2 file. These siRNAs were custom synthesized by Sigma Genosys. Control AS siRNA was purchased from Qiagen (AllStars, #1027281), other nonspecific siRNAs were synthesized by Sigma Genosys. Three days after transfection, total RNA was extracted using RNeasy Plus Mini Kit (#74134, QIAGEN). Library preparation, RNA-seq and bioinformatics analysis was performed at Novogene (Beijing, China). Briefly, the sequencing reads were aligned to human genome (version Hg19) and the resulting binary alignment (BAM) files were used to calculate the gene counts that represent total number of sequencing reads aligned to a gene. DEGs between control and PALB2 siRNA-treated samples were selected based on fold change of ≥1.5 for upregulated and ≤0.67 for downregulated genes, Cutoff P value is 0.05. Top 50 upregulated DEGs and top 50 downregulated DEGs were depicted as heatmap, GSEA was performed using GSEA software (version 4.0.0; https://www.gsea-msigdb.org) [69].

For qRT-PCR, total RNA was extracted and used for poly-T-based reverse-transcription (SuperScript III First-Strand Synthesis System, 18080051, Invitrogen) according to manufacturer’s protocol. qRT-PCR was performed using an ABI-7300 sequence detection system using SYBR Green qPCR master mix (Applied Biosystems A25742). Each measurement was performed in duplicate and expression levels of ACTB were used for normalization. Only primer pairs resulting in a single peak in the melting curve analysis were used. Oligonucleotides used for qPCR are listed in S2 File.

CRISPR/Cas9 knock out of PALB2 and ATG7

pSpCas9-2A-GFP (PD1301) V2.0 constructs containing gRNAs targeting PALB2 and pSpCas9(BB)-2A-Puro (PX459) V2.0 constructs containing gRNAs targeting ATG7 were transfected into DAOY cells using X-tremeGENE 9 (Roche, Mannheim, Germany). Single cell clones were isolated by puromycin selection or fluorescence activated cell sorting (FACS). Clones were first assayed by western blotting for complete loss of protein expression, and genomic DNA of candidate clones were then sequenced to verify the disruption of the genes. The target sites and sequences of guide RNAs were listed in S2 File. Cell doubling time was calculated by the following formula: Where N(t) = the number of cells at time t; N0 = the number of cells at time 0; r = growth rate; t = time (usually in hours)

Western blotting

Cells were lysed with NETNG-400 (400mM NaCl, 1 mM EDTA, 20 mM Tris-HCl [pH7.5], 0.5% Nonidet P-40, and 10% glycerol) with Complete protease inhibitor cocktail (Roche). Tissues were homogenized in the same buffer with a TissueRuptor (Qiagen) followed by sonication. Samples were resolved on 4–12% or 4–20% Tris-glycine SDS-polyacrylamide gels and transferred to a nitrocellulose membrane (0.45 micron, Bio-Rad) overnight at 4°C. Resolved proteins were detected following standard procedures using the following antibodies: ATG7 (1:2000; Sigma A2856), LC3 (1:1500; Novus Biologicals NB600-1384), p62 (1:10000; Abcam, ab109012), PALB2 M11 (1:2000, homemade), SOD2 (1:2000, EMD Millipore, 06–984), MT-CO1 (1:2000, Abcam, ab14705), MT-CO2 (1:1000, Abcam, ab198286), GAPDH (1:1000; Santa Cruz Biotechnology sc-365062), and β-Actin (1:5000; Sigma A1978). Blots were developed using Immobilon Western Chemiluminescent HRP Substrate (EMD Millipore).

GSH/GSSG measurement

In brief, brain tissues were collected from the Palb2, Atg7, and Palb2;Atg7 WBKO mice at 4 weeks post tamoxifen treatment and snap frozen in liquid N2. Frozen tissue samples were weighed (~20–30 mg) and ground in liquid N2 using Cryomill (Retsch). The powdered samples were then mixed with methanol: acetonitrile: water (40:40:20) solution containing 0.5% formic acid, followed by 10 sec vortexing, 10 min incubation on ice and centrifugation at 16,000 x g at 4°C. The supernatants were collected, neutralized with 50 μL/mL of 15% ammonium bicarbonate, and then centrifuged at 16,000 x g for 10 min. The final supernatants were transferred to LC–MS autosampler vials and sent for LC-MS analysis which was done using Q Exactive Plus hybrid quadrupole orbitrap mass spectrometer (ThermoFisher). Raw data analysis was done using Maven v.707. 4 [70]. GSH and GSSG contents in the brain were measured as part of a comprehensive metabolomics analysis using liquid chromatography-mass spectrometry (LC-MS).

Terminal deoxynucleotidyltransferase dUTP nick end labeling (TUNEL) assay

Staining was performed on 5 μm sections using the DeadEnd Fluorometric TUNEL System (Promega) according to manufacturer’s instructions, and all apoptotic cells in the cerebellum of each midsagittal section were counted.

Measurements of ROS, mitochondrial mass, mitochondrial membrane potential

Cells were washed with PBS and then incubated at dark for 30 min in phenol red–free DMEM with 10% FBS and 100 μmol/L 2’,7’-dichlorofluorescein diacetate (DCF-DA; D6883, Sigma) or 2.5 μmol/L MitoSOX Red (M36008, Invitrogen) for ROS measurement [60], 200 nmol/L of MitoTracker Green FM (M7512, Invitrogen) for mitochondrial mass measurement, and 200 nmol/L of MitoTracker Red CMXRos (M7512, Invitrogen) for mitochondrial membrane potential measurement. After incubation, cells were trypsinized, spun down, and resuspended in PBS at a density of approximately 2 × 10⁵ cells per mL. Signals were analyzed by flow cytometry.

Cell death measurement

Three wells of the above cells were directly trypsinized and used to measure apoptosis and necrosis using the FITC Annexin V Apoptosis Detection Kit with PI (BioLegend, #640914) according to manufacturer’s instructions.

Assessment of oxygen consumption rate (OCR)

OCR of control and KO DAOY cells were measured using a Seahorse Biosciences extracellular flux analyzer (XF24). Cells were seeded at 2.0 × 10⁴ cells per well in the XF24 plates overnight prior to XF assay. Real-time OCR measurements were performed in DMEM/F12 with 2 mM glutamine, or 1 mM dimethylα-KG for 3 h, and measurements were taken every 15 min. Relative OCR (percentage) was normalized to the 0-min time point.

Statistical analyses

Comparisons of survival curves were made using the log-rank test, and unpaired two-sided Student’s t test was used for column analysis with GraphPad Prism 8.4.3. All error bars represent standard error of the mean (SEM). Significance is denoted as follows: ns, p≥0.05; *, p<0.05; **, p<0.01. ***, p<0.001.