Metformin protects the heart against chronic intermittent hypoxia through AMPK-dependent phosphorylation of HIF-1α

  1. Sophie Moulin
  2. Britanny Blachot-Minassian
  3. Anita Kneppers
  4. Amandine Thomas
  5. Stéphanie Paradis
  6. Laurent Bultot
  7. Claire Arnaud
  8. Jean-Louis Pépin
  9. Luc Bertrand
  10. Rémi Mounier
  11. Elise Belaidi

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

Chronic intermittent hypoxia (IH), a major feature of obstructive sleep apnea syndrome (OSA), is associated with a more severe myocardial infarction. In this study, we performed RNA sequencing of cardiac samples from mice exposed to IH, which reveals a specific transcriptomic signature of the disease, relative to mitochondrial remodeling and cell death. Corresponding to its activation under chronic IH, we stabilized the Hypoxia Inducible Factor-1α (HIF-1α) in cardiac cells in vitro, and observed its association with an increased autophagic flux. In accordance, IH induced autophagy and mitophagy that is decreased in HIF-1α +/_ mice compared to wild-type animals suggesting that HIF-1 plays a significant role in IH-induced mitochondrial remodeling. Next, we showed that the AMPK metabolic sensor, typically activated by mitochondrial stress, is inhibited after 3 weeks of IH in hearts. Therefore, we assessed the effect of metformin, an anti-diabetic drug and potent activator of AMPK, on myocardial response to ischemia-reperfusion (I/R) injury. Daily administration of metformin significantly decreases infarct size without any systemic beneficial effect on insulin-resistance under IH conditions. The cardioprotective effect of metformin is lost in AMPKα2 knock-out mice demonstrating that AMPKα2 isoform promotes metformin-induced cardioprotection in mice exposed to IH. Mechanistically, we found that metformin inhibits IH-induced mitophagy in myocardium and decreases HIF-1α nuclear expression in mice subjected to IH. In vitro demonstrated that metformin induces HIF-1α phosphorylation, decreases its nuclear localization and subsequently HIF-1 transcriptional activity. Collectively, these results identify the AMPKα2 metabolic sensor as a novel modulator of HIF-1 activity. Our data suggest that metformin could be considered as a cardioprotective drug in OSA patients independently of their metabolic status.

Article activity feed

  1. Review Commons

    Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

    Learn more at Review Commons

    Reply to the reviewers

    Manuscript number: RC-2024-02352

    Corresponding author(s): Elise, Belaidi

    1. General Statements

    We would like to thank the reviewers for their constructive suggestions and comments. We hope that the “point by point answer” and the revision plan proposed below will convince the reviewers and the editor.

    We thank the reviewer 1 for his/her comments. We would like to specify that the involvement of HIF-1 in IH-induced mitochondrial remodeling has indeed been initiated by RNA-seq analysis and confirmed in a cell-based model as well as in wild-type and HIF-1a+/- heterozygous mice subjected to intermittent hypoxia (IH). In vivo, we originally demonstrated that Metformin reversed IH-induced increase in myocardial infarct size through AMPKa2 and, we proposed that metformin could modify HIF-1 activity. Then, we validated our hypothesis in an in vitro model allowing to demonstrate that Metformin, by increasing HIF-1a phosphorylation decreases its activity. We acknowledge that we used several models and this is the reason why we detailed as much as possible the Materials and Methods section including all models, experimental sets designed and methods details. We hope that the point by point response that we made for the reviewer 1 will increase the clarity of our work and we hope that the new results provided will strengthen the evidences concerning the mechanisms by which metformin can inhibit and modulate the deleterious impact of HIF-1 on IH-induced an increase in myocardial infarct size.

    We thank the reviewer 2 for his/her conclusion highlighting that “our work opens new avenues for exploring the potential effects of metformin as a modulatory of HIF-1𝛂 activity in obstructive sleep apnea syndrome”. We hope that the clarifications and/or justifications brought will convince him/her.
    We thank the reviewer 3 for having underlined that “metformin induces HIF-1α phosphorylation, decreases its nuclear localization and subsequently HIF-1 transcriptional activity are very much interesting” and for having highlighting that “our study is convincing”. We hope that the justification and the corrections brought in the point by point answer will convince him/her.

    Alltogether, as underlined by the 3 reviewers, our study is very interesting for translational science in the fields of cardiovascular, respiratory and sleep medicine. We hope that the point by point answer and the revision plan proposed will allow the publication of our article in EMBO Molecular Medicine.

    2. Description of the planned revisions

    Please find below the revision that we plan to address to answer to the questions of the reviewer 1.

    Figure 1 - it would be helpful to list all of the DEGs (what genes are changed?). Including the expression of HIF-1α and PHD isoforms would be informative. If there is a robust HIF-1α signal, changes in the expression of HIF and PHD isoforms would be anticipated. Fig 1F - with regards to glycolysis and hypoxia pathway analysis, most of the DEGs are not canonical HIF-1α/hypoxia targets.

    Figure 1 aimed at better understanding and manipulate the well-recognized involvement of HIF-1 in response to our specific IH stimulus (Semenza, Physiology 2009; Belaidi, Pharmacol & Ther. 2016). __The results provided by the RNA-seq analysis shows that IH induces cardiac oxidative and metabolic stress which are inter-related with HIF-1 activation. __We did not claim that these genes are HIF-1 targets genes. The RNA seq analysis did not allow to reveal HIF-1a and PHD1-3 transcript as the most dysregulated genes of the panel. In case of publication, bulk data and DEGS will be provided in an online file. We agree with the reviewer that the list of the 40 up and down-regulated genes would be very informative and would increase the value of the paper. Thus, we plan to add the name of the 40 up and down-regulated genes on Figure 1B.

    Figure 5G-I, show cytoplasmic HIF1a as well as nuclear.

    Alternatively, why not use IHC for subcellular localization?

    We think that the comments of the reviewer 1 concern Fig.4G-I and not 5G-I. In this figure, we showed that IH increases nuclear HIF-1____a____ expression compared to N condition and that this IH-effect is abolished in mice treated with Metformin, suggesting that, __upon IH, Metformin impacts HIF-1__a nuclear content and subsequently, its activity. The nuclear localization of HIF-1a is the most relevant mean to indicate its activation. We agree with the reviewer that IHC also allows for the indication of the nuclear localization of HIF-1a. Indeed, we previously performed IHC on nuclear HIF-1a localization and demonstrated that IH increased HIF-1a nuclear localization by IHC that was corroborated by Western-blot (Moulin S, TACD, 2020). Western-blot and IHC are both semi-quantitative techniques with different process of analyses. In this study, we choose Western-blot because we have the material to perform this technique and because IHC is associated with an analysis process (size of a slice, areas to analyze, colorimetry…) that is more complex than the analysis process of Western-blot (densitometry solely).

    While the nuclear localization of HIF-1a is the most relevant mean to indicate its activation; it could be interesting to see that HIF-1a cytosolic content was neither modify by IH nor by Metformin. This would also corroborate the results of the RNA-seq that did not demonstrate any difference in DEGs of HIF-1a or of other members of the HIF family. This would also confirm that Metformin plays a major role on HIF-1 activaty regulation (and not transcription) in the context of IH.

    Thus, we plan to perform a Western-blot of HIF-1a on cytosolic extracts of hearts from mice exposed to N or IH and treated or not with Metformin. These extracts are already available and Western-blot would be performed and replicated in 3 weeks. We could also provide a Western-blot in order to show the purity of our extraction protocol (nucleus vs cytosol).

    Figure 5F, it would be important to show the levels of expression of HIF1a in these experiments. Are there positive and negative controls that the authors could use for HIF21a activity in this experiment?

    In our manuscript, we aimed at demonstrating that Metformin decreases HIF-1 activity in a context of strong HIF-1____a____expression and/or stabilizion those mimics what happens after chronic IH in mice (Belaidi E, Int J Cardiol 2016, Moulin S, Ther Adv Chronic Dis 2020) and in apneic patients (Moulin S, Can J Cardiol 2020). Thus, we used a transfection allowing to overexpress HIF-1a that is one of the best means to increase HIF-1 activity. In the Figure 1 below, HA-HIF-1α-WT Addgene AmpR and 5 HRE GFP AmpR plasmids co-transfection induced a decrease in H9c2 viability and an increase in GFP-positive cells that were not observed in H9c2 transfected with pcDNA 3.1 HA-C AmpR (negative control). __This validates our *in vitro *model as a good positive control to mimic IH consequences. __ However, we agree with the reviewer that we could add a supplemental figure or a panel demonstrating that our transfection induced an increase in HIF-1a expression. Thus, we will perform a Western-blot targeting HIF-1a on H9c2 transfected with the control plasmid (pcDNA 3.1 HA-C AmpR) or the plasmid allowing the overexpression of HIF-1a (HA-HIF-1α-WT AmpR). This work would be performed in 2 months.

    Moreover, we already improved the lisibility of the Figure 5F to clarify the experimental conditions (table inserted under the graphic); we also completed the Materials and Methods section to specify the plasmid used (modifications are in red in the manuscript).

    Figure to see on the downloaded file.

    Figure 1 : GFP fluorescence in H9c2 cells transfected with pcDNA 3.1 HA-C AmpR (control condition) or HA-HIF-1α-WT AmpR (positive control, overexpression of HIF-1a) and 5 HRE-GFP AmpR plamids and treated with CoCl2 (1mM, 2h); magnification x100.

    This paragraph concerns only the point 4 of the fifth question.

    In these experiments, as well as subsequent studies, it would be very informative to use a specific AMPK activator e.g. MK-8772, to compare with metformin. It is well known that metformin has a number of other targets in addition to AMPK.

    We agree with the reviewer that metformin has pleiotropic effect. Very interestingly, we demonstrated that the reduced-infarct size is not related to the metabolic systemic effect of metformin since it failed to improve the IH-induced insulin resistance while it improves the answer to insulin in normoxic mice (supplemental Figure S3B). This demonstrates that in our model, the cardioprotective effects of Metformin are independent of a potential systemic effect. Then, we demonstrated that metformin protects the heart against ischemia-reperfusion through AMPK____a____2 activation by using AMPK____a____2KO exposed to IH in which Metformin failed to decrease infarct size (Fig.4N). MK-8772 is not widely used in vivo models. Moreover a recent study indicates that chronic treatment with MK-8772 (14 days 1 month in mice and rats, respectively) induces cardiac hypertrophy characterized by an increase in heart weight (Myers R, Science 2017). In vivo experiments with MK-8772 would be not clinically relevant as the use of metformin that is already used in clinic. However, in order to improve the mechanistic investigation concerning the role of AMPKa2 activation on inhibiting HIF-1 activity, we propose to perform the in vitroexperiments performed in Figure 5 with a specific allosteric small-molecule activator of AMPKa2 such as 991.

    We plan to:

    -Expose H9c2 to CoCl2 and treat them with 991 in order to measure HIF-1a phosphorylation.

    -Transfect H9c2 with our plasmids HA-HIF-1α-WT AmpR and 5 HRE GFP AmpR and treat them or not with 991 in order to measure HIF-1 activity (GFP fluorescence). These experiments would be performed in 2 months.

    __Please find below the revision that we plan to address to answer to the second question of the reviewer 2. __

    1. The WB images were cut and pasted. Please add the original images

    We acknowledge the reviewer's comment and will address it by submitting a supplementary file containing the uncropped immunoblot images. Since this file already exists, our plan is to standardize it by providing, for each slide (immunoblot), all relevant information pertaining to our experiments, including groups, molecular weight markers, cutting, membrane stripping, and other pertinent details.

    3. Description of the revisions that have already been incorporated in the transferred manuscript

    __Please find below the answers or the revisions that have already been incorporated in response to the comments of the reviewer 1. Please note that we provided new results and new figures at the discretion of the reviewer, but we are ready to insert them as figures or supplemental figures in a new revised manuscript if the reviewers and the editor think that it would improve our message. __

    Was mitochondrial content in the hearts after IH experiment measured e.g. mtDNA measurements? IH results in mitochondrial dysfunction/reduced mitochondrial content. It would have been good to show mitochondrial dysfunction by doing basic functional experiments (e.g. TMRM/MitoROS imaging etc.) by isolating cardiomyocytes from the N and IH experiments.

    We thank the reviewer for these questions about the mitochondrial function and content. The impact of IH on mitochondrial function has already been demonstrated in heart (Moulin S, Antioxydants 2022, Wei Q, Am J Physiol 2012). __Indeed, we previously showed that mitochondria isolated from hearts of mice exposed to IH had a decrease in maximal respiration in complex I and II that was not observed in HIF-1____a____+/- ____mice (Moulin S, Antioxidants 2022), indicating that HIF-1 is responsible for IH-induced mitochondrial dysfunction. __

    Figure 4 shows that Metformin abolished IH-induced mitochondrial remodeling similar to what we observed in HIF-1a+/-(Figure 2). This means that treating with Metformin or partially deleting the gene encoding for HIF-a induce the same impact on IH. Then, we demonstrated that Metformin can control HIF-1 activation and we concluded that metformin could be cardioprotective through inhibiting HIF-1 activation and subsequent mitochondrial stress and remodeling. In this study, we focused on the effects of Metformin on HIF-1 and we did not aim at directly test the effect of metformin on mitochondrial function. Actually, metformin exhibits biphasic effects on bioenergetics of cardiac tissue depending on the modality of administration (i. e. single injection, time of administration during an ischemia-reperfusion procedure); the dose administered and the tissue studied (i. e. hiPSC-CMs, isolated mitochondria…) (Emelyanova N, Transl Res 2021). But, we collected some data that we would like to submit at the discretion of the reviewer. Using oximetry, we measured maximal respiration in complex 1 and 2 on isolated mitochondria from hearts of mice exposed to N, IH and treated or not with metformin during the exposure__. While we observed that IH decreases maximal respiration in complex 1 and 2, we did not find any effect of metformin on mitochondrial respiration alteration induced by IH __(Figure 2A, B). Using spectrofluorometry, we measured the mitochondrial membrane potential using TMRM; we did not find any modification of membrane potential in IH or Metformin-treated mice (Figure 2C). Because we previously did not observe any impact of IH on mtDNA/gDNA ratio (Figure 2D), we did not test metformin on this parameter.

    To conclude, we think that these results are not directly in the scope of our work but if the reviewer thinks that they deserve to be discussed, we could add them in a supplementary figure.

    Please, see the figure on the dowloaded file

    Figure 2 : Mice were exposed to 21 days of Normoxia (N) or Intermittent Hypoxia (IH) (1-min cycle of FiO2 5%-21%) and treated with vehicle (Vh, CmCNa 0.01%, 0,1ml.10g-1) or Metformin (Met, 300mg.kg-1.d-1). (A, B) __Mitochondrial function __was measured by oximetry with sequential addition of substrate (state 2), ADP (200mM, state 3, maximal respiration) and oligomycin (12.5 mM, state 4) ; quantification of O2 consumption for NADH-linked mitochondrial respiration (complex I-glutamate-malate, GM, 20mM) (A), and for FADH2-linked mitochondrial respiration (complex II-succinate, S, 5mM, in presence of complex I inhibition by rotenone, 6.25mM) (B) (n=8). __(C) Mitochondrial membrane potential __measured by spectrofluorometry after Tetramethylrhodamine Methyl Ester (TMRM, 0.2mM) in presence of GM (basal condition), maximal respiration (ADP) and uncoupling condition Carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone, FCCP, 3mM); fluorescence intensity is expressed relative to fuorescence at baseline (before GM) (n=8). (D) Mitochondrial content assessed by the expression of mitochondrial DNA (mtDNA, COX1) relative to genomic DNA (gDNA, ApoB) measured after PCR (n=6); *p

    Fig 2A-D - CoCl2 is not a good model to mimic hypoxia due its effect on disrupting iron homeostasis in cells, which can mean that some of the effects are due to changes in iron levels and not HIF stabilisation.

    __The capacity of CoCl2 to chelate iron is the main property of CoCl2 that we used in order to stabilize HIF-1____a____. Actually, prolyl-4-hydroxylases need Fe2+ to hydroxylate HIF-1____a____ and induce its degradation. __Then, intermittent hypoxia (IH) is characterized by very rapid changes in PO2. This stimulus was designed to reproduce sleep apnea syndrome and its associated disorders (i. e. insulin-resistance, hypertension, increase in myocardial infarct size). This model was firstly developed and validated in rodents (Dematteis M, ILARJ 2008, Belaidi E, Eur Resp Rev 2022, Harki , Eur Resp J 2022). Compelling evidence indicate that the involvement of HIF-1 in IH-deleterious consequences is related to the repetitive phases of oxygenation and especially to IH-induced oxidative stress (Semenza Physiology 2009, Belaidi Pharmacol. & Ther 2016). In order to increase the level of mechanistic insights on HIF-1, we next attempted to optimize in vitro models. A device was developed by Minoves et al. (Minoves M, Am J Physiol, 2017) to expose endothelial and cancer cells to IH. However, as illustrated below, this device does not mimic efficient rapid hypoxia-reoxygenation cycles able to induce cardiac cell death (Figure 3A). However, CoCl2 decreases H9C2 viability by 60% (Figure 3B) that is associated with a sustained stabilization of HIF-1____a (Figure 3C,D). Thus, we choose this *in vitro *model as it replicates cardiac cell death and HIF-1____a____overexpression or stabilization which we similarly observe in our in vivo model and in apneic patients (Moulin S, Can J Cardiol 2020).

    Please see the figure on the dowloaded file

    Figure 3 : (A-B) Cell viability measured by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-2H- tetrazolium bromide (MTT) of H9c2 cells exposed to 6 hours (h) of repetitive cycles of Intermittent hypoxia (2 minutes (min) PO2 16% - 2 min PO2 2%) (n=3) (A) or treated with CoCl2 (1mM, 2h) (n=6) (B). __(C-D), __quantification of total ____HIF-1____𝛂 expression relative to tubulin (C) and representative image of Wetsern-blot (D) (n=2-3); *p****p*

    Fig. 2K, change in BNIP3 expression is modest, but change in Parkin is very dramatic. BNIP3 is a HIF-1α target but Parkin is not, so it is plausible that mitophagy could be occurring through a HIF-1α independent mechanism.

    Fig. 2K is a representative panel of 2-3 independent experiments. The quantification reported in Figures 2I and 2J demonstrated a significant decrease in BNIP3 and Parkin expressions in HIF-1a heterozygous mice exposed to Intermittent Hypoxia (IH) compared to HIF-1a+/+ mice exposed to IH. While we acknowledge that only BNIP3 is a direct target of HIF-1, the role of HIF-1 in IH-induced auto/mitophagy is demonstrated by our experiments performed in HIF-1____a____heterozygous mice. This shows an important role for HIF-1 without excluding any impact of HIF-1a independent mechanisms.

    What are we meant to be looking at in Fig 2L?

    Figure 2L aims at illustrating mitochondrial remodeling under IH. Stars indicate that mitochondria have abnormal fate in IH conditions and arrows point autophagosomal membrane and formation. This figure was magnified to be clearer (please see new Figure 2L).

    Figure 3B-C, the reduction in pT172 on AMPK is modest. It would be good to include pACC as a downstream target for AMPK.

    As recommended by the reviewer, we inserted in the manuscript the quantification of 79Ser-P-ACC/ACC western-blot as well as a representative image of the Western-blot (See new Figure 3). We also modified the legend of the figure. 79Ser-P-ACC is an important target of AMPK; however, in our experimental conditions, its phosphorylation is not associated to the decrease in AMPK phosphorylation. This could be explained by many points. First, Metformin was administered every day and hearts were harvested 24 hours later after the last administration. Most studies demonstrating a modification of AMPK and ACC phosphorylation are experiments performed in vitro or directly (less than 1 hour) after a single dose of Metformin administration. In the context of myocardial ischemia-reperfusion, Yin et al. showed an increase in P-AMPK/AMPK directly after Metformin treatment without showing any data on P-ACC/ACC (Yin M; Am J Physiol 2011); similar data were published in models of chronic cardiac diseases (Soraya H, Eur J Pharmacol, Gundewar S, Circ Res 2009). Second, in line with the previous explanation, the lack of effect of metformin on P-AMPK and/or P-ACC in rodent models could be explained by its rapid distribution (Sheleme T Clin Pharmacokinetics of Metformin 2021) and its short half-life that is around 3.7 hours in mice (Junien N, Arch Int Pharmacodyn Ther 1979).

    To conclude, since we performed all our analysis 24h after the last treatment and exposure to hypoxia, we argue that the slight but significative decrease in AMPK phosphorylation that we observed in our study highlight a robust impact of chronic IH. However, this would be elegant to confirm this result by measuring AMPK through its phosphorylation capacity (Cool B, Cell Metab 2006, Ducommun S, Am J Physiol 2014). We already sent hearts from mice exposed to Normoxia or Intermittent Hypoxia to Luc Bertrand’s lab (IREC, Belgium) where they used to perform this assay.

    Fig. E-G, show data for mice treated with vehicle.

    In Figure 4 I-J­­, we demonstrated that __Metformin significantly decreases infract size in IH condition only and this validates our main hypothesis regarding the specific beneficial effect of this drug in the context of chronic IH. __In order to show that the cardioprotective effect of Metformin is relative to AMPKa2 activation, we first showed that 79Ser-PACC/ACC, one of the main downstream targets of AMPKa2 was increased (Fig. E-G). We did not find it necessary to does not exhibit cardioprotective effects. However, as shown in Figure 3 below, Metformin also increases 79Ser-PACC/ACC in Normoxic mice validating the treatment. Thus, in normoxic conditions, AMPK____a____2 activation does not exert any cardioprotective effect. We acknowledge that this reinforces our result about the specificity of AMPK____a____2 activation by Metformin under chronic IH condition. We could add this Figure in supplemental results.

    Please, see the figure on the dowloaded file

    __Figure 4 __: AMPK activation in Normoxic mice treated with vehicle (Vh, CmCNa 0.01%, 0,1ml.10g-1) or __ __metformin (Met, 300mg.kg-1.d-1 : __ __172Thr-P-AMPK/AMPK (A) and 79Ser-P-ACC/ACC (B) ratio and representative image of Western-blot (C) (n=3-6); *p

    Fig. 3K - what cre is used for the a2 KO mice?

    As written in the Materials and methods section, AMPKa2KO mice are not inducible Knock-out mice. Constitutive AMPKα2 knockout mice were kindly generated by Benoit Viollet (Viollet B, JCI, 2003).

    Include normoxia data for the a2 KO mice studies.

    The question of the reviewer concerning the cardioprotective effects of metformin is interesting but is not aligned with the objectives of the study. Indeed, we did not treat normoxic mice with Met for several reasons. First, the objective of the study was to find a cardioprotective strategy against IH-induced an increase in infarct size. Second, Fig. 3I shows that Met significantly reduced infarct size upon IH only; this suggests that AMPK____a____2 activation is specifically involved in IH-induced increase in infarct size but not in reducing infarct size in normoxic mice. Moreover, the beneficial impact of metformin in standard models of myocardial ischemia-reperfusion is controversial and has been extensively discussed (Foretz et al. Cell Metab. 2014). Overall, using AMPKa2 mice was legitimated in the context of IH only. We validated the involvement of AMPKa2 in the cardioprotective effect of metformin especially in IH conditions.

    Figure 5G, what is the rationale for switching to CoCl2 in the mice to prove metformin reduced HIF-1α expression? Why not use reduced O2 tension in mice.

    We respectfully disagree with the reviewer since mice were exposed to N and IH and treated or not with Metformin to demonstrate that this drug abolished IH-induced increase in HIF-1a nuclear expression (Figure 4 H, I). The same model was used in Figure 3 to demonstrate the impact of Metformin on infarct size. Fig. 5G was conducted to demonstrate the potential link between AMPKa2 and HIF-1a phosphorylation in basal conditions of AMPKa2 content or in absence of AMPKa2 (AMPKa2-/- mice). The single presence of AMPK____a____2 demonstrates an increase in HIF-1____a____ phosphorylation if its stabilization is increased by CoCl2; this was not observed in AMPK____a____2-/- mice highlighting that AMPK____a____2 plays an important role in HIF-1____a phosphorylation.

    Please find below the answer to the first question asked by the reviewer 2.

    1. Why did authors choose the IH protocol illustrated in Fig. S1A

    The choice of the hypoxic stimulus was based on literature and mainly on our recognized expertise in preclinical studies aiming at better understanding obstructive sleep apnea syndrome (OSA); a chronic pathology associated with several comorbidities such as diabetes, hypertension… We are conscious that the hypoxic stimulus used in this study is very severe, with a nadir arterial oxygen saturation (SaO2) around 60%. __However, this experimental design is required to induce detrimental cardiovascular effects __in the absence of any confounding factors (i.e., obesity) or genetic susceptibility for complications (i. e. genetic susceptibility to hypertension) (Dematteis M, ILARJ 2008). Especially in the context of myocardial infarction, exposing rodents to 14 to 21 days of IH at 5% and subjected them to a myocardial ischemia-reperfusion protocol allows us to reproduce the increase in infarct size in rats (Belaidi E, J. Am. Coll. Cardiol., 2009; Bourdier G, Am J Physiol, 2016) and in mice (Belaidi E, Int J Cardiol 2016; Moulin S, Can J Cardiol 2020) similar to what has been observed in apneic patients (Buchner S, EHJ 2014). Moreover, we recently conducted a meta-analysis based on 23 preclinical studies aiming at investigating the impact of the IH pattern (duration, FiO2, repetition of cycles…) on infarct size and cardiomyocyte death (Belaidi E, Eur. Resp. J 2022). We showed that IH significantly increases infarct size when IH is applied several days (especially 14 to 21 days) and when FiO2 is around 5%; whereas IH decreases infarct size when it is applied a single day at a FiO2 at 10%. This meta-analysis provided the confirmation that we need to apply a chronic and severe stimulus to reproduce an increase in infarct size that is observed in apneic patients which are exposed every day, during several days to a decrease in SaO2. If the reviewers and the editor consider that this point should be discussed in the discussion section, we will be happy to include it.

    Please find below the answers or the revisions that have already been incorporated in response to the comments of the reviewer 3. They appear in red in the new manuscript except the modifications performed in the “references” section which appear in black.

    1 The authors used H9c2 rat cardiac cells in vitro experiments although they used mouse model in vivo experiments. Using mouse P19.CL6 cardiac cells instead of rat H9c2 cells may much clearer. Why the authors did not use P19.CL6 cells should be explained.

    We thank the reviewer for his/her suggestion. P19CL6 cell line has been isolated from pluripotent P19 embryonal carcinoma (EC) cells after long term culture under conditions for mesodermal differentiation (Habara-Ohkubo A, Cell Struc Funct 1996). Therefore, these cells are mostly used to ____study the differentiation of cardiac muscle. Indeed, they were recognized to avoid large variations in the differentiation rates which were extensively reported (Mueller I, J Biomed Biotechnol 2010). To our knowledges no ventricular non-beating mice cell line. In this study, we used H9c2 which are extensively used and recognized as a gold standard cellular model to study the biology of cardiomyocytes including mechanisms involved in cardiac ischemia-reperfusion injury (Paillard M, Circulation, 2013; Zhang G Circulation 2021__), cardiac hypertrophy__ (Zhang N, Cell Death Diff 2020; Hu H, Cardiovasc Res 2020), intra-organites calcium exchanges (Moulin S, Antioxydants, 2022, Paillard M, Circulation, 2013) as drug testing (Beshay NM, J Pharm and Tox Methods, 2007). Recently, H9c2 and P19.CL6 were exposed to intermittent hypoxia (70 cycles of FiO2 1% (5 min) - FiO2 21% (10%)) in order to “mimic OSA” and investigate the transcription level of a pool of genes. The authors show some similiraties and differences of mRNA expression between the two cell lines that, indeed, could be attributed to variations in the cell origin (Takasawa S, IJMS 2022). However, in this study, there are no experiment allowing to assess the state of cardiac cells (apoptosis, life, metabolism, remodeling) questioning the pathophysiologic transposability of the model. Moreover, the number of experiments conducted on H9C2 (pubmed references : 7000 vs 100 for P19.CL6) to understand the mechanism involved in acute and chronic cardiac pathologies makes our choice confident and relevant.

    2 The authors described, "The protocol was approved by the French minister (APAFIS#23725-2020012111137561.v2)." in Animals (page 16) without showing approval date, the authors should clearly show the approval date together with their approval numbers.

    We added the approvement date in the materials and methods section. It was approved on February 20, 2020.

    3 In Figure 2 L, scale bar(s) should be added because figures are magnified and/or reduced by printer.

    We agree with the reviewer that scale bars were not visible; we highlighted them.

    4 In Figure 3J and 3N, scale bar(s) should be added.

    Scale bars have been now added on figures 3J and 3N. All pictures were acquired at the maximal zoom of a camera placed at an equal distance from the slices. Then, analyses were performed, slice per slice, with Image J with the same zoom (x5 to get an image at 100%). In this context, scale was added based on a photo of slice taken close to a ruler.

    5 In introduction, "HIF-1" should be changed to "hypoxia-inducible factor 1 (HIF-1)".

    Thank you, we replaced HIF-1 by Hypoxia Inducible Factor-1 (HIF-1) in the introduction section.

    6 In Results, "Angpt1, Txnip, Nmrk2, Nuak1 or Pfkfb1" should be changed to "Angiopoietin 1 (Angpt1), Thioredoxin-interacting protein (Txnip), Nicotinamide riboside kinase 2 (Nmrk2), NUAK family SNF1-like kinase 1 (Nuak1) or 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 1 (Pfkb1)".

    We did the modification and let the abbreviation in italic since it concerns genes name.

    7 In Chronic intermittent hypoxia, "FiO2" should be changed to "FiO2".

    Thank you, we changed it in the materials and methods section.

    8 In Western-blot, "Bio-Rad, California, USA" should be changed to "Bio-Rad, Hercules, CA".

    Thank you, we did it.

    9 In Western-blot, what "tubulin" (α-tubulin or β-tubulin) should be clarified.

    We agree with the reviewer that this point should be specified; α-tubulin was stained, we added the “α”.

    As mentioned below, we have done all the modifications required by the reviewer in the references section.

    10 In Ref. 8, "Antioxidants (Basel) 11 (2022)" should be changed to "Antioxidants (Basel) 11, 2326 (2022)".

    Done

    11 In Ref.10, "Pharmacol Ther (2016)" should be changed to "Pharmacol Ther 168, 1-11 (2016)".

    Done

    12 In Ref.13, "Antioxidants (Basel) 11 (2022)" should be changed to "Antioxidants (Basel) 11, 1462 (2022).

    Done

    13 In Ref. 21, "Diabetes (2017)" should be changed to "Diabetes 66, 2942-2951 (2017)".

    Done

    14 In Ref. 28, "Adv Biol (Weinh), e2300292 (2023)" should be changed to "Adv Biol (Weinh), 8, 2300292 (2023)".

    Done

    15 In Ref. 37, "J Am Heart Assoc 6 (2017)" should be changed to "J Am Heart Assoc 6, e006680 (2017)".

    Done

    16 In Ref. 41, "Eur Respir Rev 32 (2023)" should be changed to "Eur Respir Rev 32, 230083 (2023)".

    Done

    17 In Ref. 46, "Int J Mol Sci 22 (2020)" should be changed to "Int J Mol Sci 22, 268 (2021)".

    Done

    18 In Ref. 47, "Int J Mol Sci 21 (2020)" should be changed to "Int J Mol Sci 21, 2428 (2020)". Done

    4. Description of analyses that authors prefer not to carry out

    __Please find below the answers to the comment of the reviewer 3 that we cannot provide and that is not in the scope of the study. __

    Figure 5, multiple phosphorylation sites have been identified on HIF1a. What is the nature of the Thr/SerP-HIF1a antibody? It would be far more preferable (essential?) to identify the site(s) within HIF1a that are phosphorylated by AMPK.

    The antibody was provided by Cell Signalling, ref. 9631. Phospho-(Ser/Thr) Phe Antibody detects phospho-serine or threonine in the context of tyrosine, tryptophan or phenylalanine.

    The identification of the Phosphorylation sites will require a long-time consuming phosphoproteomic analysis and subsequent functional validation *in vivo *and in vitro (directed mutagenesis, knock-in mice, …) which are out of the scope of our paper.

  2. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #3

    Evidence, reproducibility and clarity

    Although the findings that metformin inhibits intermittent hypoxia (IH)-induced mitophagy in myocardium and decreases hypoxia-inducible factor 1-α (HIF-1α) nuclear expression in mice subjected to IH. In vitro demonstrated that metformin induces HIF-1α phosphorylation, decreases its nuclear localization and subsequently HIF-1 transcriptional activity are very much interesting, numbers of points need clarifying and certain statements require further justification. These are given below.

    Point

    1. The authors used H9c2 rat cardiac cells in vitro experiments although they used mouse model in vivo experiments. Using mouse P19.CL6 cardiac cells instead of rat H9c2 cells may much clearer. Why the authors did not use P19.CL6 cells should be explained.
    2. The authors described, "The protocol was approved by the French minister (APAFIS#23725-2020012111137561.v2)." in Animals (page 16) without showing approval date, the authors should clearly show the approval date together with their approval numbers.
    3. In Figure 2 L, scale bar(s) should be added because figures are magnified and/or reduced by printer.
    4. In Figure 3J and 3N, scale bar(s) should be added.
    5. In introduction, "HIF-1" should be changed to "hypoxia-inducible factor 1 (HIF-1)".
    6. In Results, "Angpt1, Txnip, Nmrk2, Nuak1 or Pfkfb1" should be changed to "Angiopoietin 1 (Angpt1), Thioredoxin-interacting protein (Txnip), Nicotinamide riboside kinase 2 (Nmrk2), NUAK family SNF1-like kinase 1 (Nuak1) or 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 1 (Pfkb1)".
    7. In Chronic intermittent hypoxia, "FiO2" should be changed to "FiO2".
    8. In Western-blot, "Bio-Rad, California, USA" should be changed to "Bio-Rad, Hercules, CA".
    9. In Western-blot, what "tubulin" (α-tubulin or β-tubulin) should be clarified.
    10. In Ref. 8, "Antioxidants (Basel) 11 (2022)" should be changed to "Antioxidants (Basel) 11, 2326 (2022)".
    11. In Ref.10, "Pharmacol Ther (2016)" should be changed to "Pharmacol Ther 168, 1-11 (2016)".
    12. In Ref.13, "Antioxidants (Basel) 11 (2022)" should be changed to "Antioxidants (Basel) 11, 1462 (2022).
    13. In Ref. 21, "Diabetes (2017)" should be changed to "Diabetes 66, 2942-2951 (2017)".
    14. In Ref. 28, "Adv Biol (Weinh), e2300292 (2023)" should be changed to "Adv Biol (Weinh), 8, 2300292 (2023)".
    15. In Ref. 37, "J Am Heart Assoc 6 (2017)" should be changed to "J Am Heart Assoc 6, e006680 (2017)".
    16. In Ref. 41, "Eur Respir Rev 32 (2023)" should be changed to "Eur Respir Rev 32, 230083 (2023)".
    17. In Ref. 46, "Int J Mol Sci 22 (2020)" should be changed to "Int J Mol Sci 22, 268 (2021)".
    18. In Ref. 47, "Int J Mol Sci 21 (2020)" should be changed to "Int J Mol Sci 21, 2428 (2020)".

    Significance

    the study is convincing

  3. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #2

    Evidence, reproducibility and clarity

    In the submitted paper by Moulin et al., the authors investigated the effect of metformin treatment (an oral anti-diabetic drug and AMPK activator) on cardiac response to ischemia-reperfusion in mice exposed to intermittent hypoxia (IH), a major component of obstructive sleep apnea (OSA). The data demonstrates that IH alters the expression profile of a subset of genes involved in myocardial mitochondrial dysfunction, such as Angpt1, Txnip, Nmrk2, Nuak1 or Pfkfb1, hallmark genes of hypoxia and glycolysis. In H9c2 cells treated with CoCl2 with or without chloroquine, the authors claim that IH induces mitophagy through HIF-1 activation, associated with an increase in autophagic flux. This effect is abolished in HIF-1𝛂+/- mice. Metformin treatment in mice reverses IH-increased in infarct size through AMPK activation, decreases IH-induced mitophagy and HIF-1𝛂 nuclear localization.

    The paper is interesting but needs some clarifications:

    1. Why did authors choose the IH protocol illustrated in Fig. S1A)
    2. The WB images were cut and pasted. Please add the original images

    Significance

    This work opens new avenues for exploring the potential effects of metformin as a modulatory of HIF-1𝛂 activity in obstructive sleep apnea syndrome.

  4. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #1

    Evidence, reproducibility and clarity

    Summary: The manuscript by Moulin and colleagues begins by exploring changes in gene expression in mice subjected to intermittent hypoxia (IH). Based on these results, the authors postulate a link with HIF1a, and so use a cell-based model to examine this link. Next, the authors turn to AMPK showing that IH decreases Thr172 phosphorylation on AMPK. In view of this, the authors use metformin as a way to activate AMPK during IH and monitor necrosis in hearts following ischemia-reperfusion. Metformin treatment reduces necrosis in mice subjected to IH. Mechanistically, the authors link decreased HIF1a activity following AMPK activation in response to metformin and suggest that this is due to phosphorylation of HIF1a by AMPK. Overall, this reviewer found some of these connections to be quite weak and in some cases the evidence supporting the claims was equivocal. In some instances, the lack of appropriate controls or the use on non-specific reagents, undermined the findings. Adding to these issues, there was a lack of experimental detail both in the Figure legends and in the M&M section. Below I have listed some of the major concerns that if addressed satisfactorily would greatly strengthen the manuscript.

    Major Comments

    Major Points

    1. Figure 1 - it would be helpful to list all of the DEGs (what genes are changed?). Including the expression of HIF-1α and PHD isoforms would be informative. If there is a robust HIF-1α signal, changes in the expression of HIF and PHD isoforms would be anticipated. Fig 1F - with regards to glycolysis and hypoxia pathway analysis, most of the DEGs are not canonical HIF-1α/hypoxia targets.
    2. Was mitochondrial content in the hearts after IH experiment measured e.g. mtDNA measurements? IH results in mitochondrial dysfunction/reduced mitochondrial content. It would have been good to show mitochondrial dysfunction by doing basic functional experiments (e.g. TMRM/MitoROS imaging etc.) by isolating cardiomyocytes from the N and IH experiments.
    3. Fig 2A-D - CoCl2 is not a good model to mimic hypoxia due its effect on disrupting iron homeostasis in cells, which can mean that some of the effects are due to changes in iron levels and not HIF stabilisation. Fig. 2K, change in BNIP3 expression is modest, but change in Parkin is very dramatic. BNIP3 is a HIF-1α target but Parkin is not, so it is plausible that mitophagy could be occurring through a HIF-1α independent mechanism. What are we meant to be looking at in Fig 2L?
    4. Figure 3B-C, the reduction in pT172 on AMPK is modest. It would be good to include pACC as a downstream target for AMPK. Fig. E-G, show data for mice treated with vehicle. Fig. 3K - what cre is used for the a2 KO mice? Include normoxia data for the a2 KO mice studies. In these experiments, as well as subsequent studies, it would be very informative to use a specific AMPK activator e.g. MK-8772, to compare with metformin. It is well known that metformin has a number of other targets in addition to AMPK.
    5. Figure 5G-I, show cytoplasmic HIF1a as well as nuclear. Alternatively, why not use IHC for subcellular localisation?
    6. Figure 5, multiple phosphorylation sites have been identified on HIF1a. What is the nature of the Thr/SerP-HIF1a antibody? It would be far more preferable (essential?) to identify the site(s) within HIF1a that are phosphorylated by AMPK.
    7. Figure 5F, it would be important to show the levels of expression of HIF1a in these experiments. Are there positive and negative controls that the authors could use for HIF21a activity in this experiment?
    8. Figure 5G, what is the rationale for switching to CoCl2 in the mice to prove metformin reduced HIF-1α expression? Why not use reduced O2 tension in mice.

    Minor Comments

    Nothing at this stage of the process. The authors need to focus on the major points to improve the quality of the manuscript.

    Significance

    Overall, the lack of robust mechanistic studies makes it difficult to fully interpret the impact of the study.

    This feels like a preliminary exploration of a potentially important biological system, but the manuscript seems incomplete and requires more attention to details.

    Upon revision, I think the study would be of interest to basic and clinical researchers working in the filed of cardiac metabolism and hypoxia.

    My expertise is in metabolism and cell signalling.

  5. Version published to 10.1101/2024.01.11.575152v1 on bioRxiv