The class 3 PI3K coordinates autophagy and mitochondrial lipid catabolism by controlling nuclear receptor PPARα

The class 3 phosphoinositide 3-kinase (PI3K) is required for lysosomal degradation by autophagy and vesicular trafficking, assuring nutrient availability. Mitochondrial lipid catabolism is another energy source. Autophagy and mitochondrial metabolism are transcriptionally controlled by nutrient sensing nuclear receptors. However, the class 3 PI3K contribution to this regulation is unknown. We show that liver-specific inactivation of Vps15, the essential regulatory subunit of the class 3 PI3K, elicits mitochondrial depletion and failure to oxidize fatty acids. Mechanistically, transcriptional activity of Peroxisome Proliferator Activated Receptor alpha (PPARα), a nuclear receptor orchestrating lipid catabolism, is blunted in Vps15-deficient livers. We find PPARα repressors Histone Deacetylase 3 (Hdac3) and Nuclear receptor co-repressor 1 (NCoR1) accumulated in Vps15-deficient livers due to defective autophagy. Activation of PPARα or inhibition of Hdac3 restored mitochondrial biogenesis and lipid oxidation in Vps15-deficient hepatocytes. These findings reveal roles for the class 3 PI3K and autophagy in transcriptional coordination of mitochondrial metabolism.


Point-by point response
We thank the reviewers for their helpful comments and we appreciate they found specific interests in our work. In this revised version, we provide requested additional controls as well as a substantial amount of new functional data addressing the mechanistic links between the class 3 PI3K and lysosomal degradation of PPARα repressors. Briefly, we address three main points: 1) We now provide a large amount of new functional data on regulation of NCoR1 and Hdac3 by class 3 PI3K. In addition to Hdac3, in this revised version of the manuscript, we show that Hdac1 is also accumulated in liver of Vps15 mutant. Importantly, we now show that NCoR1 and Hdac3 are degraded by autophagy in vivo in liver. Significantly, we provide mechanistic insights into the selective autophagy of NCoR1 and Hdac3 stimulated by the class 3 PI3K. We now show that NCoR1 and Hdac3 interact with Atg8-like proteins GABARAP and LC3. Furthermore, we find that this interaction is modulated by Vps15 overexpression and by pharmacologic inhibition of class 3 PI3K. Altogether, these mechanistic analyses considerably reinforce our initial conclusion on the implication of class 3 PI3K in the control of PPARα transcriptional responses in liver.
2) We further strengthened our initial findings on the mitochondrial dysfunction in Vps15-null models. Thanks to reviewer guidance, we now show that mitochondrial respiration and mitochondrial DNA content are significantly decreased in Vps15-null liver. In addition, we have demonstrated that the fasting-induced increase in mitochondrial mass is Vps15 and PPARα dependent. We believe that these novel findings further support our important discovery that PPARα transcriptionally controls mitochondrial biogenesis in liver.
3) We complemented the analyses in the Vps15 mouse model and in cells with all requested controls. Essentially, in this revised manuscript, we have used genetic tools to downregulate Hdac3 and Vps34, the catalytic subunit of the class 3 PI3K, to demonstrate that those are sufficient to significantly restore or reproduce the mitochondrial phenotype of Vps15null models, respectively. Furthermore, as guided by the reviewer suggestions, we performed in vivo analyses of the adipose tissue and biochemical analyses of plasma in Vps15 hepatic mutant and now report, for the first time in this model, a complex phenotype of increased lipolysis and defective whole-body lipid metabolism.
There is a large amount of new data that are included in the 7 main figures, and 7 supplementary figures of this revised manuscript which support the novel function of class 3 PI3K upstream of PPARα and lipid degradation in liver. Below is our detailed response to the reviewer's comments. The modifications to text of the manuscript are highlighted in red.
Editorial Note: Parts of this Peer Review File have been redacted as indicated to maintain the confidentiality of unpublished data.

Reviewer 1:
We would like to thank the reviewer for highlighting the novelty and importance of our study towards a better understanding the mechanisms of metabolic adaptation in physiological fasting and the novel role of class 3 PI3K in control of PPARα transcriptional activity.

Response:
We agree with the reviewer that special precautions have to be taken when experimenting with virally expressed recombinant Cre protein. Indeed, we have made those controls on wild-type mice before proceeding with experiments on cells from Vps15 floxed mice and when expressing Cre in vivo in livers of Vps15 floxed mice. We express Cre protein using non-replicative adenoviral vectors (Ad5-Cre-GFP) in the lowest dose in which efficient deletion of Vps15 is achieved without any noticeable effect on cell metabolic activity and viability. Importantly, we always use a control infection with the adenoviral vectors expressing GFP protein (Cre protein that we use is a GFP-fusion). As an example, we present in this revised manuscript that Cre expression in primary hepatocytes isolated from wild type mice does not affect mitochondrial respiration assessed by Seahorse assay or the expression of PPARα targets. These new data are presented in revised manuscript as new Supplementary   Fig. 2e, 2f.
Notably, our work is entirely focused on liver and we have used all generally accepted controls to assure adequate interpretation of the phenotypes (transduction with control viral vectors; Alb-Cre transgene always kept in Het state in mice; analyses of Alb-Cre negative and Alb-Cre positive littermates). We employed a combination of acute (shRNA, Ad5-Cre-GFP) and chronic (Alb-Cre) genetic depletion approaches and pharmacologic (Vps34 inhibitor) approaches. These approaches are complementary and the results support the hypothesis that functional class 3 PI3K signalling is required for transcriptional activity of PPARα.

Response:
We would like to thank the reviewer for these insightful suggestions that allowed us to reinforce our main conclusions on mitochondrial dysfunction in Vps15 mutant. As suggested by all reviewers, in this revised version of the manuscript, we provide new data on the biochemical profiling of the respiratory chain complex activities in liver extracts of control and Vps15 mutants. We revealed a significant inhibition of the total respiratory chain activity due to decreased activity of all five complexes in liver tissue of Vps15 mutant (new Fig. 1d, new Supplementary Fig. 2c). Notably, this defective respiration in Vps15 mutants was also accompanied by significant increase in lactate dehydrogenase activity in livers of Vps15 mutant (new Supplementary Fig. 2d).
Also, as suggested by the reviewer, we now show that respiration is not affected in wild type hepatocytes transduced with adenoviral vectors expressing Cre recombinase (new Supplementary Fig. 2e). In addition, we now demonstrate that not only inactivation of Vps15 but also knockdown of Vps34, a lipid kinase subunit in class 3 PI3K complex, inhibit mitochondrial respiration (new Supplementary Fig. 6a). As suggested by reviewer, we have also improved the presentation of Seahorse measurements and now show, for each experiment, the graphs of spare respiratory capacity and ATP production with the statistical analyses conducted (new Fig. 1e, 5a, 6g and Supplementary Fig. 6a). Finally, we also performed staining with the potential sensitive dye MitoTracker Red in primary hepatocytes isolated from wild-type and Vps15-LKO mice (Figure for Reviewer 1a). This staining revealed weak signal which is consistent with a significant mitochondrial depletion in Vps15-null hepatocytes. To this end, as suggested by the reviewer, we also measured mtDNA content in liver tissue of Vps15 hepatic mutant. We now show that, consistent with decreased mitochondrial mass, mtDNA is depleted in liver of Vps15 mutant (new Supplementary Fig.   2b). In addition, in this revised manuscript we show that fasting-induced increase in mtDNA content in liver tissue of control mice requires expression of Vps15 (new Fig. 7c).
Furthermore, we now show that this response is also dependent on PPARα, as it is fully abrogated in fasted hepatic mutant of PPARα (new Fig. 7d).
Moreover, as advised by the reviewer, we complement the immunofluorescent analyses of the mitochondrial network in Vps15 depleted hepatocytes by performing additional staining with Tom40, an integral protein of mitochondrial membrane (new Fig. 1c). These immunofluorescent analyses show that Tom40-positive mitochondrial compartment is disturbed (less mitochondria and fragmented network) in Vps15-null hepatocytes. This is consistent with our new data on the activated fission in livers of Vps15-LKO mice (new Supplementary Fig. 7a) Fig. 1a, new Supplementary Data 2). Importantly, by performing transcript and protein expression analyses in liver tissue of Vps15 mutants, we have confirmed that the signature of oxidative stress response is highly upregulated upon Vps15 inactivation (new Supplementary Fig. 1). It is evidenced by significant overexpression of GST, Nqo1, Gpx7 and Gstm2 both at protein and transcript level (new Supplementary Fig. 1c, 1d). This is fully consistent with increased protein expression of Nrf2 transcription factor that we have observed in livers of Vps15 hepatic mutants (new Supplementary Fig. 1b). In line with these findings in livers of Vps15-LKO mice, we also conducted additional metabolomic analyses in primary hepatocytes which revealed that GSH:GSSG ratio is highly downregulated in primary hepatocytes depleted of Vps15 further advocating on-going oxidative stress in Vps15-null cells (Figure for Reviewer 1b). Finally, analyses of protein carbonylation in livers of Vps15-LKO mice suggested pattern of increased protein oxidation in liver tissue of Vps15 mutant compared to wild-type mice (Figure for Reviewer 1c). In sum, we believe that these extensive new analyses, combined to the previous transcriptomic, metabolomic and ultrastructural findings, firmly support our conclusion on the mitochondrial dysfunction upon Vps15 inactivation.

Response:
We thank the reviewer for raising several important points on the complex metabolic phenotype of liver-specific mutant of Vps15.This led us to further highlight the importance of hepatic class 3 PI3K signalling for whole-body metabolic homeostasis. Notably, our findings in liver-specific Vps15 knockout mice are consistent with the metabolic phenotype of a wholebody and liver specific PPARα knockout mice (Nemazanyy, et al. 2015), (Leone, et al. 1999), (Montagner, et al. 2016). The PPARα mutants are viable, however, when challenged with starvation, are hypoketogenic and hypoglycaemic. We agree with the reviewer that most likely the metabolic plasticity of these mutants is limited under starvation stress and, if fasting were prolonged, would be detrimental to mutant viability. We have challenged both PPARα and Vps15 mutants by fasting for 24 hours, the longest duration allowed by our ethics permit, and observed a significant decrease in the metabolic responses such as ketone body production.
Moreover, the liver is not the only site of ketogenesis and gluconeogenesis, as the kidneys and intestines are also reported to contribute (Takagi, et al. 2016), (Owen, et al. 1969), (Zhang, et al. 2011), (Bekesi and Williamson 1990), (Penhoat, et al. 2014). Their contribution in fasting induced ketogenesis in hepatic mutant of Vps15 is out of the scope of this work. As suggested by reviewer, we have also measured the lactate levels in fasted and fed mice (new Supplementary Fig. 3f). These analyses revealed that plasmatic lactate is non-significantly decreased (p value 0.059) in fed Vps15 mutants with fasting decreased circulating lactate to a comparable level in control and Vps15 mutants. These findings suggest that lactate is not an alternative energetic source in Vps15 mutants after prolonged fast. Our future work will determine the metabolic processes that are induced in the liver tissue of Vps15 hepatic mutant to compensate the lack of PPARα transcriptional activity.

Another metabolic paradox is related to lipid metabolism. During fasting there is an important lipolysis that induces a liver overload causing liver steatosis. Liver steatosis should be enhanced in case of mitochondrial beta oxidation inhibition due to CPT1 reduction. Instead authors found a decrease of FFA or triglyceride in liver of VPS15 KO mice. The only explanation is that mitochondrial are uncoupled and even in face of a reduced CPT1/CPT2 expression are still able to greatly use acyl-CoA for beta-oxidation. Authors must check the blood levels of FFA in fasting and the expression of UCP proteins in liver. Finally, quantification of WAT in VPS15 KO mice in this experimental
setting must be also performed.

Response:
The reviewer is absolutely right that fasting induces the lipolysis in adipose tissue and provokes an influx of lipids in liver used as a substrate for the ketogenesis and β-oxidation. To assess the changes in adipose tissue in response to fasting, we have performed the benchmark measurements of adipose tissue content using dual energy X-ray absorptiometry (DEXA) scan in control and Vps15 hepatic mutants in fed and fasted states. Three major conclusions could be drawn from these new data presented as new Supplementary Fig. 3gj. First, hepatic mutants of Vps15 present significantly lower adiposity both in fed and fasted state (new Supplementary Fig. 3g). Second, this decreased adiposity is paralleled by significantly increased lean mass in Vps15 liver-specific mutant mice (new Supplementary   Fig. 3h). Finally, despite significantly decreased adiposity of Vps15 liver mutants, the fasting induced loss of fat in these mice was higher as compared to wild type mice (new Supplementary Fig. 3j). The body fluid content was unmodified in fed/fasted wild type and Vps15 mutant mice (new Supplementary Fig. 3i). In sum, these findings suggest that hepatic class 3 PI3K signalling has a whole-body effect on adipose tissue maintenance in response to fasting and, potentially, acts as a brake of the fasting induced lipolysis. The mechanistic studies in this direction are out of scope of current manuscript and will be pursued in future.
Following the suggestion of all three reviewers, we have also performed extensive biochemical studies in plasma of Vps15 hepatic mutants in fed and fasted state. These new data are presented as new Supplementary Fig. 3f. These analyses have revealed that, as expected, fasting in control mice resulted in increased plasmatic levels of glycerol and free fatty acids (new Supplementary Fig. 3f). Moreover, fasting led to a decrease in plasmatic TG levels consistent with lipid uptake by liver, as also witnessed by hepatic steatosis in wild type mice (new Supplementary Fig. 3b-3d, 3f). At the same time, in Vps15 hepatic mutants, the levels of free fatty acids and TG were significantly increased in fed state and TG levels were unmodified by 24 hour fasting (new Supplementary Fig. 3f). Notably, consistent with changes in fat mass measured by DEXA scan, the fasting of Vps15 hepatic mutants resulted in further increase in plasmatic levels of free fatty acids (new Supplementary Fig. 3f). It was also paralleled by significantly lower levels of circulating glycerol both in fed and fasted state (new Supplementary Fig. 3f). In sum, these findings demonstrated that plasmatic lipid homeostasis is defective in Vps15 hepatic mutants. These observations advocate that lipid uptake is controlled by class 3 PI3K in hepatocytes.
Although we cannot exclude that the uncoupling in mitochondria takes place in Vps15-null hepatocytes, we believe extensive data presented in this revised manuscript strongly advocates a positive role of Vps15 in control of PPARα transcription for mitochondrial biogenesis and fatty acid degradation. To this end, an increased ratio of acyl-carnitine comparable in liver tissue of fed mice, they are significantly decreased in livers of fasted Vps15 mutant compared to fasted control mice. Altogether, mitochondrial mass reduction, inhibition of the respiratory chain activity, decreased expression of fatty acid trafficking proteins, significant accumulation of long chain fatty acid acyl-carnitines together with lower levels of acetyl-CoA and free carnitine support our initial hypothesis on defective fatty acid degradation in Vps15-null hepatocytes.

Response:
We thank the reviewer for raising this point that improved the illustration of our findings.
We have now included the images of gross liver morphology and HE stained liver sections of fed and fasted WT and AlbCre + ;Vps15 f/f mice (new Supplementary Fig. 3b, 3c). We have already published the images of gross liver morphology and HE stained liver sections of Vps15 f/f mice transduced with Ad5-Cre-GFP vectors as well as AlbCre + ;Vps15 f/f mice (Nemazanyy et al., Nat Comm. 2015). Those are also presented on Figure for Reviewer 3.
We have added this information when describing the model and mentioned that, in these models, Vps15-depletion concurs with autophagy block. In addition, we present new data clearly showing that while fasting induces a significant decrease in liver size in control mice, the Vps15 hepatic mutants are resistant to this physiological response (new Supplementary   Fig. 3e).

Response:
We thank reviewer for raising this important point that helped us to understand better the regulation of PPARα downstream of Vps15. The lack of activating ligand and the presence of co-repressors activate proteosomal degradation of PPARα. Notably, we have demonstrated that fenofibrate treatment significantly rescued nuclear levels of PPARα (new Supplementary   Fig. 4b). We now show that PPARα transcript (new Supplementary Fig. 3a and new Supplementary Fig. 4h) and protein ( Fig. 2f and new Supplementary Fig. 4g) are downregulated in livers of Vps15 hepatic mutants. As for protein turnover, we discovered that PPARα protein ubiquitination is increased in livers of Vps15 hepatic mutants (Figure for Reviewer 4). We used in this experiment Tandem Ubiquitin Binding Entity beads that bind with equal affinity K63 and K48 ubiquitinated proteins (Tube2). As presented on Figure for Reviewer 4a, despite profound PPARα protein depletion in the livers of Vps15 mutant mice, similar levels of PPARα protein are pulled down on Tube2 beads from total protein extracts of control and Vps15-null liver. These observations are paralleled by our findings of increased   protein expression of Huwe1 protein, an E3 ubiquitin ligase that was recently shown to control polyubiquitination of PPARα in hepatocytes (Figure for Reviewer 4b) (Zhao, et al. 2018). We have also demonstrated that Huwe1 is activated in livers of Vps15 mutants, as evidenced by its increased binding to Tube 1 beads which have higher affinity to K63 ubiquitinated proteins, a common posttranslational modification for active E3 ligases (Figure for Reviewer 4c).
Finally, we show that PPARα protein turnover is modulated by Vps15 expression. To this end, as shown in the new Supplementary Fig. 5b, Vps15 overexpression in HEK293T cells increased basal levels of recombinant PPARα protein and slowed its degradation, as monitored upon cyclocheximide addition. These are in line with our new findings presented in new Fig. 4d and 4e, that Vps15 overexpression decreases complex of PPARα with Hdac3. In sum, these new findings are concordant with the positive role of PPARα in control of its own transcription and, also, they are in agreement with known mechanism of PPARα proteosomal degradation. We believe that, although these new data are interesting, they are out of the main focus of the current report and, if reviewer agrees, they will not be included in the manuscript to avoid distraction. Fig 2d)? Is PPARa protein affected by lysosomal inhibition? In case that lysosomes control the NCoR1 and HDAC3 degradation how con fenofibrate reduce their levels (Fig  3c)  PPARα Huwe1

Point 7. It is not clear how the changes of NCoR1 HDAC3 proteins are due. The fig 3e suggest that are consequent to autophagy but more data are required. Is it mediated by p62 or do they contain LIR domains and are directly delivered to autophagosomes? How they can get out of the nucleus and incorporated into the autophagosomes (Suppl
Vps15

Figure for Reviewer 4. Ubiquitination of PPARα is induced in livers of Vps15-LKO mice. a Endogenous
PPARα and Huwe1 proteins were pulled-down on Tube2 agarose from 1mg of total liver extract of six week old random-fed Vps15 f/f and AlbCre + ;Vps15 f/f . The precipitation with control agarose beads served as a control of non-specific binding. Immunoblot analyses with indicated antibodies revealed similar pattern of binding of analysed proteins precipitated from liver tissue extracts of wild-type and Vps15-LKO mice. b Immunoblot analysis of total protein liver extracts of random-fed six week old Vps15 f/f and AlbCre + ;Vps15 f/f using indicated antibodies. Immunoblot with GAPDH antibody served as a loading control. c Endogenous Huwe1 protein was pulled-down on Tube1 agarose from 1mg of total liver extract of six week old randomfed Vps15 f/f and AlbCre + ;Vps15 f/f . The precipitation with control agarose beads served as a control of nonspecific binding. Immunoblot analyses with anti-Huwe1 antibody revealed increased binding of Huwe1 (K63-Ubiquitination) in liver tissue extracts of Vps15-LKO mice.

Response:
We completely agree with the reviewer that additional mechanistic insights would further deepen our understanding of the role of Vps15 and class 3 PI3K in control of NCoR1 and Hdac3 proteins autophagic degradation. A similar point was also raised in the comments of two other reviewers. Following the insightful comments of all reviewers, we approached this question by further studying the in vivo autophagic degradation and the mechanisms of selective autophagy of these transcriptional repressors. First, we show that Hdac3 and NCoR1 proteins are degraded in lysosomes upon starvation-induced autophagy. It is evidenced by their accumulation in livers of fasted wild type mice treated with Leupeptin inhibitor (lysosomal protease inhibitor) (new Fig. 3e, new Supplementary Fig. 4e). Notably, this degradation is abrogated in Vps15-null livers, consistent with defective autophagic flux in mutants (new Fig.   3e, new Supplementary Fig. 4e). This is consistent with our findings of Hdac3 and NCoR1 accumulation in primary hepatocytes treated with Bafilomycin A1 (new Fig. 3f). In addition, in this revised manuscript, we show that endogenous Hdac3 and NCoR1 proteins co-localize with Lamp2 protein in primary hepatocytes (Fig. 4a) further supporting our conclusion on their lysosomal degradation.
Second, guided by the reviewer, we tested if Hdac3 and NCoR1 proteins could be targets of selective autophagy. In this revised manuscript, we now show that, Hdac3 and NCoR1 proteins interact with GABARAP and LC3B proteins (new Fig. 4b).
Notably, the interaction of NCoR1 was observed predominantly with GABARAP while Hdac3 was pulled down with GABARAP and to lesser extent with LC3B, suggesting distinct mechanisms for the induction of their selective autophagy. Our further studies have demonstrated that interaction of Hdac3 and NCoR1 with GABARAP protein was increased by Vps15 overexpression (new  repressors in livers of Vps15-null mice (new Supplementary Fig. 4c). The quantification of these immunoblots showed no significant changes in total levels of repressors in livers of Vps15-LKO mice treated with fenofibrate.

Response:
We agree with the reviewer that the genetic rescue experiment would provide an additional elegant confirmation of Hdac3 repressive role in PPARα function in Vps15-null models. In this revised manuscript, we include additional data on knockdown of Hdac3 using specific shRNA in Vps15-depleted hepatocytes. First, we show that knockdown of Hdac3 in hepatocytes prepared from AlbCre + ;Vps15 f/f mice rescues transcript levels of PPARα targets such as Aox and Cpt1 (new Supplementary Fig. 6b, 6c). Second, we show that defective mitochondrial respiration in Vps15-and Vps34-depleted primary hepatocytes could be significantly increased by acute co-depletion of Hdac3 (new Fig. 5a and new Supplementary   Fig. 6a). Altogether, these novel observations in vitro and our findings using VPA in vivo support our conclusions on the inhibitory role of Hdac3 in PPARα driven transcription. Fig 3c is poor and does  not support anything). It is also unclear how VPS15 would enhance PPARa activity (Fig  5d). Is it because NCoR1/HDAC3 are decreased? In this case they must be studied and shown.

Response:
We apologize that we may have been unclear in the description of the figure in previous version. Indeed, the rationale for Vps15 expression was to increase autophagic flux. It is Immunoblot analysis of total protein liver extracts of ex vivo autophagic flux experiment in six-week old wild type mice using indicated antibodies. The livers were perfused with PBS solution, dissected and minced. The tissue was incubated during four hours at +37 o C in EBSS solution supplemented with Chloroquine at concentration 100μM or 500μM. Tissue was collected by brief centrifugation and washed once with PBS before proceeding with protein extraction. Immunoblot with anti-Lamp2 and anti-LC3 antibodies served as a control of lysosomal inhibition (accumulation of Lamp2 and LC3-II protein) and anti-GAPDH as a loading control. evidenced by significantly decreased p62 and LC3-II protein levels upon Vps15/PGC1α expression that we have now quantified (new Supplementary Fig. 7e). These are two commonly used read-outs of active autophagy. Autophagy activation is expected consequence of Vps15 expression. It was widely documented in cellular models upon expression of another regulatory subunits of the class 3 PI3K complex, e.g. Beclin protein overexpression (Liang, et al. 1999). Notably, PPARα transcriptional activity was increased in this experimental setting (Fig. 6d, 6e). As suggested by the reviewer, we analysed in the same experiments Hdac3 protein expression and now we show that its protein levels are also decreased upon Vps15/PGC1α expression (new Supplementary Fig. 7e). Following reviewer's suggestion, we have also performed the flux experiment upon Vps15/PGC1α expression. It shows that in this condition autophagic degradation of Hdac3 is induced (new Supplementary Fig. 7f). Altogether, these observations are consistent with the new data presented on new Fig. 4 showing that Vps15 expression promotes Hdac3 dissociation from PPARα and increases its binding to Atg8-like proteins.

Response:
We apologize for low resolution of electron microscopy figure in the merged PDF file of submitted manuscript. In this revised version of the manuscript, we have changed the captures of mitochondria in all conditions to higher resolution (new Fig. 6a). This figure is now much improved and clearly shows present cristae in hepatocytes of fenofibrate treated mice (new Fig. 6a). As suggested by the reviewer, in addition to mitochondria area, we have quantified the mitochondria number in all experimental conditions. Those quantifications revealed that the mitochondrial number is not modified between conditions (new Fig. 6a). It is also consistent with our novel data that demonstrates activated fission machinery in livers of Vps15 hepatic mutant (new Supplementary Fig. 7a). These additional analyses together with new data on quantification of mitochondrial mass, support our conclusions that while mitochondrial number is not modified in Vps15-null hepatocytes, the mitochondrial mass is significantly reduced upon Vps15 depletion.

Response:
We apologise for been unclear in the previous version of our manuscript. We have modified the introduction to better convey the message on growing recognition of requirement of functional mitochondria for autophagy. We cited works that demonstrated that mitochondrial respiration, phosphatidylethanolamine production and mitochondrial membrane are required for effective autophagic flux (Hailey, et al. 2010), (Rockenfeller, et al. 2015), (Thomas, et al. 2018). Furthermore, the revised version of the manuscript is now edited to highlight the liver tissue context of reported findings on transcriptional control of autophagy, which is also the focus of our manuscript. As suggested by the reviewer, in addition to PPARα/PGC1α dependent mitochondrial gene transcription and biogenesis, we have also mentioned the mechanism of mitochondrial fission as a possible contributor to the observed mitochondrial phenotype of hepatic Vps15 mutant. We thank the reviewer for these insightful comments.

Reviewer #2 (Remarks to the Author):
We would like to thank the reviewer for the interest in our study, for the appreciation of the major novelty regarding the role of class 3 PI3K in control of transcriptional activity of PPARα and for the constructive criticism that improved our work.

This manuscript reports the study that attempted to demonstrate a class 3 PI3Kmediated mechanism linking autophagy and mitochondrial function to the control of liver lipid metabolism. The presented results showed the transcriptional suppression of liver mitochondrial gene program in liver-specific Vps15 KO mice. Using fenofibrate, Bafilomycin A1 and valproic acid (VPA) treatments as well as PGC-1alpha expression, the data indicated that Vps15 might function to regulate PPARalpha-dependent mitochondrial metabolism through autophagic degradation of HDAC3/NCoR1. The authors conclude that "the class 3 PI3K acts upstream of nuclear receptors and exerts a broad transcriptional control in the liver to match autophagic activity with mitochondrial metabolism during fasting". Overall this is a potentially interesting study that may expand our molecular understanding of the mechanistic links between authophagy, mitochondrial function and lipid homeostasis. However, the current study is kind of descriptive at the gene expression levels; more mechanistic evidence is needed to support that Vsp15dependent autophagy mediated the degradation of HDAC3/NCoR1.
Specific points:

Response:
We thank reviewer for raising this important point that we addressed by performing ex vivo and in vivo flux experiments in liver of control and hepatic mutant of Vps15 (new Fig. 3e and new Supplementary Fig. 4e). We now show that endogenous NCoR1 and Hdac3 proteins are degraded in lysosomes in control liver unlike in autophagy deficient Vps15 mutant.
This degradation is also stimulated by 24-hour fasting (in vivo flux) (new Fig. 3e) or incubation of liver explants in EBSS media (ex vivo flux) (new Supplementary Fig. 4e). As suggested by reviewer, we also performed additional immunoblot analyses in total and nuclear liver extracts of fed and fasted for 24 hours control mice and hepatic mutant of Vps15. These new data show that fasting induced nuclear PPARα accumulation is inefficient in Vps15 mutants and is oppositely correlated with levels of repressor proteins NCoR1 and Hdac3 (new Supplementary Fig. 4g, 4i). Altogether, these new data further reinforce our initial findings on the lysosomal degradation of PPARα repressors.

Response:
Following guidance of all three reviewers, in this revised version of manuscript, we present a large amount of new mechanistic data (see point 7 to Reviewer 1. Altogether, these data strongly advocate lysosomal degradation of NCoR1 and Hdac3 through selective autophagy by binding to Atg8-like proteins (new Fig. 4).
Moreover, following reviewer's suggestion, we also performed additional immunofluorescent analyses in primary hepatocytes in which autophagic flux was blocked by Bafilomycin A1 treatment. In addition to the previously presented co-localization of NCoR1 and Lamp1, in this revised manuscript, we show that both NCoR1 and HDAC3 are targeted to Lamp2-positive lysosomal compartments (new Fig. 4a).

Response:
We thank the reviewer for asking this interesting question on the role of hepatic class 3 PI3K in whole-body lipid homeostasis. A similar question was raised by the Reviewer 1. In response to both comments we have performed the biochemical studies in plasma of Vps15 hepatic mutants in fed and fasted state (see point 6 to Reviewer 1). These new data are presented as new Supplementary Fig. 3f.

Response:
We thank the reviewer for raising this important point that led us to a deeper understanding of the role of Vps15 in mitochondrial maintenance. Since similar questions were also raised by the first reviewer, see our response to point 2 to Reviewer 1 on page 3. In sum, we have quantified mtDNA content in control and hepatic mutants of Vps15 and PPARα.
We now show, first, that mitochondrial mass is significantly reduced in Vps15 mutant and, second, that fasting induced increase of mtDNA in liver requires Vps15 and PPARα (new Fig.   7c, 7d). Our molecular analyses of mitochondrial fractions also show that pro-mitophagic Parkin pathway is activated in Vps15-null livers. These new data are presented as new Supplementary Fig. 7a and 7b.

Response:
We thank reviewer for this comment that allowed us to identify the differences in PPARα protein expression upon pharmacological rescue experiments. In agreement with our previously reported immunohistochemistry in liver tissue samples, PPARα protein levels were significantly upregulated in livers of fenofibrate treated hepatic Vps15 mutants (new Supplementary Fig. 4b). Unlike fenofibrate, treatment with VPA did not have the same normalizing effect on PPARα protein expression (new Supplementary Fig. 6d, 6f). These findings are in agreement with the literature as ligand binding and association with co-activator proteins such as PGC1α stabilize PPARα protein. Moreover, since PPARα is known to control its own transcription, its direct activation with a synthetic ligand such as fenofibrate is expected to promote its expression, unlike an indirect effect of VPA through inhibition of Hdac activity.
In VPA treated livers, Hdac activity is reduced, however the potent activation of PPARα, as observed in fenofibrate treated mice, is not achieved. Therefore, we believe the ligand availability (e.g. free fatty acids that we find depleted in livers of Vps15 mutants) could be one of the causes as to why VPA treatment is less effective as compared to fenofibrate in vivo.

Response:
We apologize for the lack of clarity of our previous version of the manuscript. The lower migrating protein band on immunoblot with anti-Vps15 antibody is a truncated non-functional form of Vps15 that is transcribed from the start codon in exon 4 upon effective Cre-mediated recombination in Vps15 gene locus. We have characterized this truncated protein in our previous report and demonstrated that it has no dominant negative effect (Nemazanyy, EMBO Mol.Med, 2013). We have included this information in the text of revised manuscript. We have also indicated the protein sizes on all immunoblots.

Response:
We thank the reviewer for this comment that improved the presentation of our findings.
We have now provided bioinformatic analyses on significantly upregulated genes (David pathway analysis). These new analyses are presented as new Supplementary Fig.1a and new Supplementary Data 2. The complete list of up-and down-regulated genes is also presented in Supplementary Data 1.

Response:
This important question on the mitochondrial respiration in livers of Vps15 mutant was raised by three reviewers. We addressed it, see detailed response to point 2 of Reviewer 1 Briefly, we measured activity of mitochondrial respiratory chain complexes in liver extracts of control and Vps15 liver-specific knockout mice. Those biochemical analyses showed decreased respiratory chain activity in livers of Vps15 mutants. These findings are presented as new Fig. 1d and new Supplementary Fig. 2c.

Reviewer #3 (Remarks to the Author):
We would like to thank the reviewer for the appreciation of our work and for noting its integrative aspect in use of different experimental approaches and the importance of our findings for the field. We are also grateful for the reviewer's comments that allowed us significantly improve our work.

Response:
We thank reviewer for raising this interesting point that led us to strengthen the control of PPARα transcriptional activity by the autophagic degradation of its repressors. We have addressed this question by using in selected assays the pharmacologic inhibitor of Vps34 lipid kinase, PIK-III, and by knocking down Vps34 using specific shRNA. In both approaches, we have observed that targeting Vps34 expression and activity was sufficient to inhibit the mitochondrial respiration and recruitment of NCoR1 and Hdac3 to Atg8-like proteins. These new data are presented as new Fig. 4b and new Supplementary Fig. 6a. In addition, our comparative analyses of publicly available microarray from autophagy deficient Atg7 liver knockout mice show significant overlap in transcriptional responses with PPARα and Vps15 hepatic mutants (Figure for Reviewer 6). Although we cannot exclude that the Vps15 protein might have a role beyond the complex with lipid kinase Vps34, our analyses point to the mechanism of selective autophagy of transcriptional repressors dependent on lipid kinase activity of Vps34/Vps15 complex. (Fig.1c) (Fig.1d).

Response:
We thank reviewer for raising these important questions that allowed us better understand the impact of class 3 PI3K inactivation on mitochondrial function. Since similar questions were also raised by the other reviewers, see our point 2 to the Reviewer 1 Briefly, we have performed additional immunofluorescent analyses with Tom40 as a mitochondrial membrane protein (new Fig. 1c), we have quantified the mitochondrial number on the electron microscopy captures (new Fig. 6a), we have measured the mitochondrial mass (mtDNA) (new Fig. 7c, 7d, new Supplementary Fig. 2b) and mitochondrial respiration activity (new Fig. 1d, new Supplementary Fig. 2c) in liver tissue of Vps15-null mice. These new results together with the findings already presented in the previous version allow us to conclude that class 3 PI3K acts upstream of mitochondrial biogenesis and mitochondrial maintenance in hepatocytes. Fig.2F). If the authors thinks there may be residual levels that are sufficient for recruiting the corepressor to PPARalpha target genes, this should be confirmed by chromatin immunoprecipitation. An alternative possibility is that accumulation of NCoR and HDAC3 affects transcription through other nuclear receptors.

Response:
We apologize for being unclear on this important point. Our data suggest that although protein levels of PPARα is drastically decreased, it is still expressed at low levels in Vps15null livers. This is evidenced by the new immunoblot analyses presented in new Supplementary Fig. 4b, 4g, 6d and 6f and could be also seen on immunohistological analyses presented in the previous version (Fig. 3c). This conclusion is fully consistent with the observations that PPARα transcriptional activity could be partially restored either by fenofibrate or by VPA treatment. Although we cannot exclude that other transcription factors might contribute to the phenotype of hepatic mutants of Vps15, the rescue that was achieved with fenofibrate, a selective ligand of PPARα, and lack of responses in PPARα hepatic mutant suggest that the mitochondrial dysfunction in Vps15-null hepatocytes is largely PPARαdependent. We have modified the text of the revised manuscript to convey this point on residual PPARα expression.

Response:
We apologize that we may have been unclear in our description. In this experiment, we asked whether co-expression of Vps15 would potentiate a co-activating effect of PGC1α, which indeed was the case (Fig. 6d and 6e). We have now provided the evidence that expression of Vps15 in these conditions promoted autophagic flux and degradation of Hdac3 protein (new Supplementary Fig. 7f). In this revised manuscript, we also show that overexpression of Vps15 is sufficient to promote an association of Hdac3 and NCoR1 corepressors with Atg8-like proteins (new Fig. 4c). However, we cannot exclude and will test in future that Vps15 protein might have a wider and positive role in the control of PGC1α stability and transcription co-activating function. Finally, following the reviewer suggestion we also measured in this experimental setting other PPARα target involved in fatty acid degradation conclusion is further reinforced by our novel findings of interaction between NCoR1 and Hdac3 and Atg8-like proteins (new Fig. 4b, 4c).

Response:
We apologize for being unclear in the previous version of the manuscript. We have now modified the text to highlight that other Hdacs might contribute to the phenotype of Vps15 hepatic mutants. To this end, we now show that levels of Hdac1 are also increased in livers of Vps15 mutants (new Supplementary Fig. 6d). Furthermore, in this revised version of the manuscript, we show that acute knockdown of Hdac3 partially restores mitochondrial activity and expression of PPARα targets in Vps15-null hepatocytes (new Fig.5a and new Supplementary Fig. 6a). These new data together with evidence already presented in the previous version advocate the role of the repressors in PPARα inhibition in liver of Vps15 mutant mice.

Response:
We thank reviewer for this insightful suggestion that has improved the presentation of our data. As suggested by the reviewer, we have converted the colour code in the heatmap of microarray (new Fig. 2d and 7a).