Loss of PGC-1α in RPE induces mesenchymal transition and promotes retinal degeneration

Sustained loss of PGC-1α in RPE cells triggers mitochondrial/autophagic dysfunction and oxidative damage resulting in epithelial dedifferentiation and mesenchymal transition. RPE dysfunction caused by deletion of the PGC-1 coactivators in vivo causes retinal degeneration.

1. The authors show decreased expression of autophagy genes upon PGC1a knockdown and increased expression with PGC1a over-expression. They also show defective LC3 processing in PGC1a knockdown cells. However to confirm that toggling PGC1a is actually inhibiting autophagy as opposed to increasing the rate of autophagic flux, the authors need to repeat experiments shown in figure 2E and 2F in the presence or absence of bafilomycin A1 or chloroquine to determine whether LC3B-II now accumulates or not. 2. Work in figure 3 overstates its conclusions. While knocking down PGC1a appears to cause mitochondrial dysfunction and defective autophagy, and also loss of epithelial phenotype, the authors cannot say that the loss of epithelial phenotype is due to mitochondrial dysfunction and/or defective autophagy. This is merely correlative. 3. It is not at all clear from figure 7A, how many cells are present and this figure needs to be analyzed quantitatively. Nor is it clear why the format of analyzing expression in figure 7B is not used for the other markers also (COXIV, Twist, Vimentin etc). This is an interesting and provocative set of preliminary observations but currently much of what is presented is descriptive. Identifying a novel role for PGC1a in autophagy and/or EMT would be particularly significant. However, at the moment these findings are not sufficiently fleshed out and the connection between defective mitochondria/autophagy and EMT is not at all clear. The author needs to determine whether key autophagy genes shown to be down-regulated in the RPE cells are direct PGC1a targets -perform some ChIP to assess whether PGC1a binds the promoters of these genes. They also need to explain why mitochondria are dysfunctional when PGC1a is downregulated -due to defective mitophagy or is it due to reduced biogenesis? How does the mitochondrial defect relate to the impact on EMT with loss of epithelial phenotype? Overall more work is needed to "drill down" on each of these findings to make the work sufficiently compelling for publication.
Reviewer #2 (Comments to the Authors (Required)): Mariana Aparecida Brunini Rosales et al show that PGC-1alfa depletion leads to impaired autophagy and epithelial-mesenchymal transition in RPE cells resembling age-related macular degeneration (AMD). This is excellent manuscript. I liked very much to read it. Hypothesis is very relevant. Study is very well planned. Data is strong and convincing. Manuscript is very well written.
Discoveries may really open new ways to find better treatments to prevent or treat AMD. I have only minor comments to improve manuscript.
1. In Fig. 6A-D indicate organelle or tissue changes by arrow.
2. Autophagosomes and autolysosomes were not calculated to confirm autophagy flux by TEM, but this can be replaced by citing and discussing the recent publication: Loss of NRF-2 and PGC-1α genes leads to retinal pigment epithelium damage resembling dry age-related macular degeneration. Felszeghy S, Viiri J, Paterno JJ, Hyttinen JMT, Koskela A, Chen M, Leinonen H, Tanila H, Kivinen N, Koistinen A, Toropainen E, Amadio M, Smedowski A, Reinisalo M, Winiarczyk M, Mackiewicz J, Mutikainen M, Ruotsalainen AK, Kettunen M, Jokivarsi K, Sinha D, Kinnunen K, Petrovski G, Blasiak J, Bjørkøy G, Koskelainen A, Skottman H, Urtti A, Salminen A, Kannan R, Ferrington DA, Xu H, Levonen AL, Tavi P, Kauppinen A, Kaarniranta K. Redox Biol. 2019;20:1-12. doi: 10.1016/j.redox.2018 Reviewer #3 (Comments to the Authors (Required)): The authors of the manuscript titled 'Loss of PGC-1α in RPE induces mesenchymal transition and promotes retinal degeneration' have undertaken a thoroughly planned, detailed investigation on the consequences of PGC-1α functional loss in retinal pigment cells. In 2016 they had published an article describing the role of PGC1a with respect to influencing oxidative metabolism and antioxidant capacity in ARPE19 cells. In the current study the group analysed loss of PGC-1α in RPE and genetically modified mice (C57Bl/6;PGC-1α(fl/fl),PGC-1β(fl/fl) conditionally knocking both forms out in RPE via CRE activity. In the current manuscript, the authors convincingly show mitochondrial dysfunction and oxidative damage after repressing PGC-1α expression. Knock-down/silencing in RPE cells resulted in epithelial dedifferentiation and mesenchymal transition. Moving from isolated RPE cells to adult mice, they can show, that conditional knockout of PGC-1 coactivators resulted in a rapid RPE dysfunction and transdifferentiation in vivo associated with severe photoreceptor degeneration.
The study builds on thoroughly generated, reproduced data. Data are presented in a comprehensive and clear way. References are provided in a complete and meaningful way. Workplan and the resulting data are conceivable and convincing. Each figure is designed such, that it complies with the experimental plan of the study. The results are well discussed in the context of the current level of understanding.
Taken together, this is a convincing manuscript that merits publication. . However, I would encourage the authors to consider the following suggestions concerning further improvement of the manuscript: As human RPE cells as well mice are used, the coverage between mouse sequence and human sequence of PCG1a (line 495) in the overexpression experiment would be of interest.
EMT can be associated with reduced proliferation (i.e. some forms of cancer). I wonder whether cell numbers change after PCG1a silencing, I would have expected no differences in this case, because the energy metabolism switch and autophagy already require great effort for RPE cells. This is addressed in fig 3I: cell numbers increased. However, this is no further investigated or discussed.  Fig 1E) CSA activity increases at day 14, then decreases at 21 days for shPGC1A. Please discuss! Fig 1H) Total ROS activity appears at a maximum at day 7 as comparing to all other timepoints. What's the reason for that? (a point for discussion) Fig 1I) The authors point to downregulated genes, how about upregulation of HMOX1 and TXN2 genes at day 7 and day 14?  Figure 2B) It appears more suitable to adjust the order of the relative expression of genes within the graphs to the text as; LAMP1, MAP1LC3B, WIPI, ATG4D, ATG9B Figure 2E) Why are their two different western blotting bands for AAS and untreated conditions? To show two different repeats? Figure 2H) The label for " Untreated versus AAS" similar to graphs Figure   For the discussion part: -Discussion on the age dependency of degenerating mouse retina is given for the AAV-cre mice. Discussion on the age dependency of results of RPE silenced for PGC-1α, however, is missing. Based on the in vitro data using 7, 14 and 21 days, this appears attractive.
-Functional data (ERG) could be discussed deeper in the discussion part? How do functional and histological data relate? -With respect to Iba-1 as a marker revealing an immune response, is there a pattern? -Does photoreceptor morphology change upon PGC-1a silencing? -PGC-1α has a role in serving as a switch between mitochondrial biogenesis and oxidative damage by controlling the mitochondrial levels of ROS (Kaarniranta et al., 2018). What do the authors conclude with respect of a ptoential switch mechanism and effect on the mesenchymal transition? -Line 278 The authors state: "Our longitudinal analysis in vitro indicates that PGC-1α loss alters primarily 279 RPE metabolic and autophagic functions followed by EMT induction." How do these patterns as a consequence of loss of PGC-1α in retinal pigment epithelial (RPE) cells trigger mitochondrial/autophagic dysfunction and oxidative damage mechanistically? -Line 285 Mitochondrial oxidative damage and/or decreased ATP levels can directly impair lysosomal activity (Demers-Lamarche et al., 2016) and defective lysosomal degradation can reciprocally cause the accumulation of defective mitochondria (Osellame et al., 2013) Did the authors check ATP level of the cells after silencing PGC-1α? -Line 295 The rise in total ROS appeared to precede the detection of higher mitochondrial superoxide production which could be explained by an increase in mitochondrial calcium influx following mitochondrial stress (Ahmad et al., 2013) or reflect enhanced mitochondrial production of hydrogen peroxide and hydroxyl radical, both detectable with the probe used. Did the authors check for mitochondrial calcium influx and/or hydrogen peroxide and hydroxyl radical or is it an assumption? -Line 301 PGC-1α is well known as a major transcriptional inducer of mitochondrial biogenesis. However, PGC-1α silencing in RPE did not alter the expression of the mitochondrial replicationrelated genes TFAM and POLG expression or grossly reduced mitochondrial content. If mitochondrial protein composition is not severly affected, how can loss of mitochondrial dysfunction be mechanistically explained? -Line 324 ...autophagic flux and promoting the accumulation of defective organelles. Importantly our investigation of the faulty autophagic step in PGC-1α-deficient RPE cells point to the inability of AMPK to be activated upon starvation suggesting the existence of a positive feedback mechanism between AMPK and PGC-1α regulating RPE metabolism and autophagic functions. Is there any reference supporting this conclusion?
1st Authors' Response to Reviewers April 8, 2019 Reviewer #1: "The authors show that markedly reduced PGC1a expression in RPE cells causes mitochondrial dysfunction and defective autophagy that is associated with loss of epithelial phenotype in vitro. They follow up these findings in vivo using AAV delivered Cre to delete PGC1a and PGC1b in the RPE of mice that results in disorganization of the RPE and loss of apparently functional mitochondria (based on EM). Similar to results obtained in vitro, this was associated with loss of epithelial phenotype and in vivo resulted in retinal degeneration." We thank the reviewer for his/her thorough reading of our manuscript and detailed comments on how it could be improved.
Concern 1: "The authors show decreased expression of autophagy genes upon PGC1a knockdown and increased expression with PGC1a over-expression. They also show defective LC3 processing in PGC1a knockdown cells. However to confirm that toggling PGC1a is actually inhibiting autophagy as opposed to increasing the rate of autophagic flux, the authors need to repeat experiments shown in figure 2E and 2F in the presence or absence of bafilomycin A1 or chloroquine to determine whether LC3B-II now accumulates or not. " Answer: We agree with the reviewer's comment. As suggested we have repeated the starvation experiments in presence of the lysosomal inhibitor chloroquine and monitored the autophagic flux by quantification of LC3-II levels. As expected, we found that LC3-II levels increased in shCtrl cells following both serum starvation and chloroquine treatments, indicating enhancement of the autophagic flux (Mizushima et al., 2010). In contrast, LC3-II levels were unchanged in PGC-1α silenced cells following both serum starvation and/or chloroquine treatment (New Figure S2B). Thus, our results suggest that silencing PGC-1α impairs RPE's autophagic flux and further highlight the importance of PGC-1α in promoting autophagy. Concern 2: "Work in figure 3 overstates its conclusions. While knocking down PGC1a appears to cause mitochondrial dysfunction and defective autophagy, and also loss of epithelial phenotype, the authors cannot say that the loss of epithelial phenotype is due to mitochondrial dysfunction and/or defective autophagy. This is merely correlative." Answer: We thank the reviewer for his/her input. We agree that our data does not provide definitive evidence that the Oxphos and autophagic dysfunctions following PGC-1α loss are directly causative of the RPE dedifferentiation and EMT observed, as these biological processes are tightly linked and interdependent. Teasing out the hierarchic organization between metabolic, autophagic and phenotypic pathways is thereby highly challenging. Further investigation of a direct contribution of mitochondrial dysfunction and/or glycolytic switch to EMT of RPE cells (independently of PGC-1α) will be particularly valuable. However our central premise was to define the role of PGC-1α on RPE metabolic and functional maturation. Our conclusions are further supported by our longitudinal analysis indicative of a rapid metabolic dysfunction (by day 7) preceding detectable induction of the mesenchymal transcription factors Zeb1, Twist1 and repression of p53 at day 21. As suggested we have adjusted the conclusions drawn from Figure 3 as "Taken together, these data indicate that PGC-1α is required to maintain RPE epithelial phenotype and that sustained PGC-1α loss triggers EMT in RPE." (Lines 216-217 page 10). figure 7A, how many cells are present and this figure needs to be analyzed quantitatively. Nor is it clear why the format of analyzing expression in figure 7B is not used for the other markers also (COXIV, Twist, Vimentin etc)." Figure 7 are representative of selected regions associated with overt retinal degeneration and RPE phenotypic changes. In theses locations we observed increased EMT-markers in highly amorphic RPE. As described in Figure 6, RPE degeneration caused by PGC-1s deletion is regional (Fig 6A) and the phenotypic changes ranged from minor to severe (Fig 6B-E). In order to provide additional information on the global alteration in mitochondrial and EMT-marker expression we stained and quantified these proteins on RPE flat-mounted preparations as recommended. The new figure S5 shows that while the mitochondrial protein CoxIV was uniformly repressed in RPE, EMT markers were only found increased in a regional subset of amorphic RPE cells.

Answer: Sections shown in
Concern 4: "The author needs to determine whether key autophagy genes shown to be downregulated in the RPE cells are direct PGC1a targets -perform some ChIP to assess whether PGC1a binds the promoters of these genes." Answer: As a transcriptional co-factor, PGC-1α does not directly bind to DNA sequences but forms a complex with many transcription factors (TFs) in order to co-activate a large and intricate transcriptional network. Genome-wide analysis of the binding sites for PGC-1α-TFs complexes has identified numerous transcription factor targets including CEBPB, ERRα, GABP (Charos et al., 2012;Chang et al., 2018) known to transcriptionally promote autophagy-and lysosome-related genes (Guo et al., 2013;Kim et al., 2018;Zhu et al., 2014). Our findings, based on gain and loss experiments and showing that PGC-1α transcriptionally regulates multiple autophagy-associated genes in RPE cells, are consistent with previous works in other cells and tissues (Vainshtein et al., 2015;Tsunemi et al., 2012). Concern 5: "They also need to explain why mitochondria are dysfunctional when PGC1a is down-regulated -due to defective mitophagy or is it due to reduced biogenesis? How does the mitochondrial defect relate to the impact on EMT with loss of epithelial phenotype? Overall more work is needed to "drill down" on each of these findings to make the work sufficiently compelling for publication."

References
Answer: Though PGC-1α is known to promote mitochondrial biogenesis through NRF1dependent induction of Tfam, our data suggests that mitochondrial biogenesis is not significantly affected by PGC-1α loss of function in RPE. Indeed mitochondria, though abnormal, are still prominently observed in our in vivo and in vitro systems ( Figures 1C and 6E) and gene expression analysis shows no repression in Polg and Tfam ( Figure S1B) which point to defective mitophagy has the main cause of defective mitochondria. Evidence of deficient mitophagy include our report of impaired autophagic flux combined with hindered AMPK activation following PGC-1α silencing (Williams et al., 2017;Laker et al., 2017) which are consistent with our observed accumulation of abnormal mitochondria ( Figure 1C), increased mitochondrial superoxide production ( Figure 1G) and defective mitochondrial function ( Figure 1F). Definitive determination of mitophagic flux by use of specific tools such as MitoTimer or Mt-Keima is unfortunately not possible in our experimental system as our lentivirally-transduced cells constitutively express the fluorescent marker eGFP.
We agree with the reviewer that insights on the intersecting processes by which PGC-1, oxidative metabolism and EMT interact are of high interest although outside the scope of this manuscript. We are currently investigating the role of metabolic reprograming in promoting EMT in RPE. Once completed, this study should provide important information on the causal role of PGC-1s and RPE metabolic dysfunction in EMT.  Answer: As discussed in our manuscript, we used CSA to measured mitochondrial mass however it appears that CSA does not correlate with mitochondrial content in metabolically defective RPE cells. We believe that changes in CSA represent a compensatory mechanism to sustain citrate production and/or support cell proliferation. As noted by the reviewer, there is a slight reduction in CSA activity in shPGC1A cells from day 14 to day 21. Interestingly our new evaluation of cellular proliferation does indicate high proliferative activity at day 7 in shPGC1A cells but not at days 14 and 21 ( Figure S2C), which would support a role for citrate production for cell division (at the earliest time-point investigated). However we consider this interesting concept to be highly speculative at this stage. More definitive evidence on a role for citrate production in assisting RPE proliferative capacity would require significant investigations involving metabolomic profiling which is outside the scope of this manuscript. However, we did modify this point of discussion to integrate our new evaluation of cell proliferation (line 338).

References
Comment 4 " Figure 2B) It appears more suitable to adjust the order of the relative expression of genes within the graphs to the text as; LAMP1, MAP1LC3B, WIPI, ATG4D, ATG9B."

Answer:
The change has been made as suggested.
Comment 5: Figure 2E) Why are their two different western blotting bands for AAS and untreated conditions? To show two different repeats?
Answer: The reviewer is correct. As the results were quite unexpected we elected to show two sets of independent samples in the final figure and the quantification was performed on 4 independent samples. We are now complementing this figure with direct evaluation of autophagic flux further demonstrating impaired autophagy following PGC-1α silencing (see response to Reviewer #1, Concern 1). Figure 2H) The label for " Untreated versus AAS" similar to graphs Figure  Answer: Figure 2H shows increased APOE expression in PGC-1α silenced cells at day 14 of differentiation and cells were not amino acid starved. As explained in our results section (line 183): "As susceptibility to oxidative stress and impaired autophagy was noticed in primary RPE cells from AMD patients accompanied by upregulation of AMD-associated genes such as apolipoprotein E (APOE) expression (Golestaneh et al., 2017), we quantified the expression of APOE and found its protein level increased in shPGC-1α RPE cells at 14 days (Fig. 2H)." To avoid confusion we have added a title to this figure.

Comment 6:
Comment 7: Figure 3 E, F, G, H) -> For western blotting experiments GAPDH bands are not equal (lower in the treated panel). Did the authors check for equal protein concentration?
Answer: As described in our methods, all protein lysates were quantified using the bicinchoninic acid (BCA) method and the same protein content (30µg/sample) was loaded. We did noticed also that GAPDH was reduced in shPGC1A cells at 21 days which could be explained by the drastic phenotypic change between control and experimental cells and potential alteration of standard housekeeping proteins. However, since our findings show increased