PHD1 controls muscle mTORC1 in a hydroxylation-independent manner by stabilizing leucyl tRNA synthetase

mTORC1 is an important regulator of muscle mass but how it is modulated by oxygen and nutrients is not completely understood. We show that loss of the prolyl hydroxylase domain isoform 1 oxygen sensor in mice (PHD1KO) reduces muscle mass. PHD1KO muscles show impaired mTORC1 activation in response to leucine whereas mTORC1 activation by growth factors or eccentric contractions was preserved. The ability of PHD1 to promote mTORC1 activity is independent of its hydroxylation activity but is caused by decreased protein content of the leucyl tRNA synthetase (LRS) leucine sensor. Mechanistically, PHD1 interacts with and stabilizes LRS. This interaction is promoted during oxygen and amino acid depletion and protects LRS from degradation. Finally, elderly subjects have lower PHD1 levels and LRS activity in muscle from aged versus young human subjects. In conclusion, PHD1 ensures an optimal mTORC1 response to leucine after episodes of metabolic scarcity.

Specific points Figure 1. The authors should provide some discussion as to why the reduction in lean mass only contributed to the decrease in body mass in males only. To this point the authors should include the percent fat mass for the PHD KOs males and female animals. Lastly, the authors should specify in the figure legends whether males or females are being used for the in vivo assessment for the purpose of clarity. Figure 3. Did the authors observe any changes in muscle mass in the inducible muscle-specific PHD1 knockout animals? While the effect on S6K phosphorylation and other mechanistic components appears to be conserved in the mKOs, it would be good to understand whether the reduction in muscle mass also occurs.
--Reviewer #3 (Remarks to the Author): In this manuscript, D'Hulst et al evaluate the consequences of PHD1 deficiency on skeletal muscle mass and performance. Through the use of several novel mouse models, the investigators highlight a non-canonical role for PHD1 (independent of its hydroxylase function) in the regulation of mTORC1 activation by leucine. The authors perform both in vivo and in vitro experiments that clearly show that loss of PHD1 prevents leucine activation of mTORC1. The cell autonomous effects are confirmed in an inducible skeletal muscle specific knockout of PHD1. Mechanistically, the authors suggest leucyl tRNA synthetase as a potential link between PHD1 and mTORC1. Although interesting, there are several limitations and concerns that should be addressed that would further support the main conclusions of this study.
Major concerns: -Most of the focus is on LRS acting as a leucine sensor for mTORC1. This is based on the observation that in PHD1KO muscle there is a reduction in LRS protein content, accelerated degradation. Moreover, overexpression of LRS in culture increases mTORC1 activity. In addition, the authors note an unremarkable change in the other "putative" leucine sensor, Sestrin2. However, Sestrin2 expression is lowest in muscle and recent reports indicate Sestrin1 acts as a leucine sensor for mTORC1 in muscle (PMID: 30835510). The authors only look at Sestrin 2 (Figure 4B) which nevertheless appears to have increased Sestrin2 content in PHD1KO samples. A more detailed analysis of all Sestrins and their potential functional role as a link between PHD1 and mTORC1 must be explored further.
- Figure 5G, there is a substantial disconnect between the western blot and quantitation regarding the role of LRS OE and mTORC1 activity. For instance, starved levels of pS6K1 in LRS OE/PHD1KO appear much enhanced compared to wild-type and PHD1KO EV yet the bar graph indicates no increase? -In the skeletal muscle specific knockout, muscle weights and leucine-stimulated protein synthesis (via puromycin) should be measured and reported to confirm cell autonomous effect - Figure 3C, given the mild effect on pS6K compared to whole body knockout, more detailed analysis of mTORC1 signaling (4EBP1, pS6) should be performed -The study in human tissue as is does not add anything. Rather, it suggest that a reduction in PHD1 is not correlated with a decrease in LRS in human tissue. Moreover, recent reports suggest mTORC1 activity is elevated in aged muscle in rats (Joseph, Giselle et al BioRxiv 2019). Do the authors see a similar effect on mTORC1 activity in their samples?
Minor concerns: - Figure 4F, the ICC staining is not very convincing and should be improved -PHD1 protein levels in knockout should be reported

Reviewer #1:
In this study, D'Ulst and colleagues report that PHD1 controls muscular mass by regulating the stability of the leucine sensor LRS. This novel PHD1 function is independent of the enzymatic properties of PHD1 as a proline hydroxylase. The authors use an elegant combination of in vivo and in vitro approaches to test their working hypothesis. The study is novel, highly significant, and with translational implications. The quality of the data is excellent. The authors' conclusions are supported by the data as shown.
No significant weaknesses could be identified.
We would like to thank the reviewer for this positive evaluation of our work.
Minor point: Figure 6: It would be helpful if the authors could perform immunohistochemistry for PHD1 on human muscular tissue isolated from either young or old donors to provide information about the spatial localization of the signal.
To address the reviewer's comment, we have performed immunofluorescent stainings to get further insight into the spatial localization of PHD1. To do so, we used cryosections from young and old human skeletal muscle. In the original manuscript, we showed by western blot that PHD1 protein levels are lower in muscle samples from old subjects (see Figure 6). To validate our data, we have performed immunofluorescent staining for PHD1 in muscle cryosections form the same subjects. We could confirm the lower overall intensity of PHD1 in old muscles using signal intensity quantifications of stained muscles (see figure 6B). Also, PHD1 has been shown to be localized mainly in the cell nucleus (Metzen et al., 2003;Ortmann et al., 2016;Steinhoff et al., 2009) but several groups have also reported cytosolic localization (Bur et al., 2018;Couvelard et al., 2008;Moser et al., 2013;Soilleux et al., 2005;Zhang et al., 2015) and cytosolic functions (Moser et al., 2013) of PHD1. As highlighted in Figure 6B, we found clear cytosolic localization of PHD1 in muscle fibers of adult human subjects. Nuclear staining was detected, but to a lower extent. Immunofluorescent staining for PHD1 in cultured myoblasts a confirmed both nuclear and cytosolic localization (See figure RL 1). Of note, we could confirm the specificity of our antibody to detect PHD1 by the observation that there is a weaker signal in PHD1 KO myotubes compared to the signal in WT myotubes/myoblasts (See figure RL1). Altogether, our data shows that in human skeletal muscle and mouse primary myoblasts/myotubes, PHD1 is located in the cytosol as well as (to a lower extent) the nucleus.
We have included information about spatial localization of PHD1 in the text (p14, line 304-308): 'Lower PHD1 content was confirmed using immunofluorescent stainings ( Figure 6B, and 6C). Importantly, in other cell types PHD1 has been shown to be localized mainly in the nucleus 88-90 ,but several groups have also reported cytosolic localization 91-95 and function of PHD1 93 . We saw clear cytosolic localization of PHD1 in muscle fibers whereas nuclear staining was detected, but to a much lower extent.'  We would like to thank the reviewer for the positive evaluation of our work.
Specific points We agree with the reviewer's comment and have now included the percentage fat mass (both males and females) in the manuscript ( Figure 1C and S1C) and accordingly adapted the results section in the manuscript (p6, line 94-98). As reported before (Thomas et al., 2016), genetic loss of Phd1 leads to an increase in both subcutaneous as well as visceral fat mass. In our mice, the increase in FM was higher in female mice (more than doubling of percent fat mass) when compared to male mice (50% increase in percent fat mass). Unfortunately, only male mice were included in the aforementioned study by Thomas et al. In our female mice, the higher increase in fat mass compensated for the lower muscle weights in the PHD1 KO animals, overall resulting in the absence of a significant reduction in body weight. It is difficult to speculate about the underlying cause of these gender specific differences in adipose tissue mass upon loss of Phd1. It is known that female mice have different body temperature, different heart rate, different oxygen consumption rate, different adipose tissue biology and can respond differently to metabolic stress (Chang et al., 2018;Landsberg et al., 2009;Yang et al., 2007).
Following the reviewer's last comment, we have now adapted all figure legends and included a statement about whether males/females (or mixed) were used. We would like to mention that we observed reduced leucine mediated activation of mTORC1 in both males as well as females in both PHD1 KO and PHD1 mKO mice. While the data included in the original manuscript was obtained in females only, we now also provide experimental data showing blunted increase in mTORC1 activation in PHD1 KO muscles upon leucine administration in males as well. Therefore, we can conclude that the ability of PHD1 to control leucine mediated mTORC1 activity is independent of gender. We have included those data in the supplements (figure S2, panel A) and added a statement in the text (line 134-136): 'Inhibition of mTORC1 activation upon deletion of Phd1 in response to leucine stimulation was observed in both female ( Figure 2A) as well as male mice ( Figure S2A), so both genders were used for subsequent experiments.' Figure 3. Did the authors observe any changes in muscle mass in the inducible muscle-specific PHD1 knockout animals? While the effect on S6K phosphorylation and other mechanistic components appears to be conserved in the mKOs, it would be good to understand whether the reduction in muscle mass also occurs.
We thank the reviewer for this comment. To address the reviewer's comment we have included below ( Figure RL2) the weight for TA and GAS muscle form WT and PHD1 mKO mice. We did not observe any decrease in muscle weight (TA or GAS) at this time point in the inducible muscle-specific PHD1 knockout mice (see Figure RL2). However, we would like to mention that, since (i) gene recombination was only induced less than two weeks before the experiments were performed and (ii) loss of Phd1 only reduces the leucine-dependent increase in mTORC1 and not contraction/insulin response, we did not expect changes in muscle mass at this time point. In this line, a recent report using muscle-specific Raptor deficient mice (where mTOR -downstream signaling is severely blunted in response to any stimulus) showed no decrease in muscle mass 21 days after recombination (You et al., 2018).
Nonetheless, to address whether inducible muscle deletion of PHD1 affected muscle mass, we tried to mimic the full KO condition by inducing muscle deletion at early postnatal stages. To do so, we injected HSA-Cre-ER T2 -/x Phd1 fl/fl (WT) and HSA-Cre-ER T2 +/x Phd1 fl/fl mice (PHD1 mKO ) females with tamoxifen (5 x 0.1 mg per mouse at P13-14). Since it is known that there is still significant contribution of muscle stem cells to muscle growth during and maintenance up till 14-16 postnatal weeks (Bruusgaard et al., 2006;Murach et al., 2017;Pallafacchina et al., 2013;Pawlikowski et al., 2015;White et al., 2010), we applied additional Tamoxifen injections every two weeks (1 x 1 mg per mouse, see scheme in figure  RL3. Subsequently, mice were sacrificed at identical age when compared to the PHD1 KO experiments (16 weeks). We found that TA and GAS weight were lower in PHD1 mKO compared to WT (11.3% and 14.3% respectively), but due to lower variation, only TA reached significance (p = 0.03 for TA and p = 0.2 in GAS -See figure RL3). Even though these data underscore the role of PHD1 in muscle growth, we prefer not to include them in the manuscript. The main reason for this is that the HSA-Cre-ER T2 +/inducible mouse model is not commonly used to follow up the effect of gene deletions during growth. Moreover, the repeated use of Tamoxifen might interfere with hormonal household of the mice, which prevents us from making sound and convinced statements about our observations.   Figure 4B) which nevertheless appears to have increased Sestrin2 content in PHD1KO samples. A more detailed analysis of all Sestrins and their potential functional role as a link between PHD1 and mTORC1 must be explored further.
We thank the reviewer for this valuable concern. As requested by the reviewer, we provide a more indepth analysis of the other SESTRINs, their role in leucine mediated activation of mTORC1 and their potential interaction with PHD1. First, we measured mRNA as well as protein expression levels of SESN 1-3 (mRNA) in PHD1 KO and PHD1 mKO muscle (used for in vivo experiments) as well as in myotubes derived from primary myoblasts that were isolated from WT and PHD1 KO mice. loss of PHD1 in vivo did not affect mRNA nor protein levels of SESN1-3. We do acknowledge that there was a slight increase in the (protein) levels of SESN1 in m. tibialis anterior of PHD1 mKO , however this failed to reach statistical significance, and we also did not observe any difference in the PHD1 KO . In addition, we did not find any difference in SESN1 protein content in primary myotubes, arguing against an active regulation of SESNs expression by loss of Phd1. We have included the protein expression levels data of SESN1 and 2 in the main manuscript ( Figure  Notwithstanding, to further investigate whether SESNs are involved in PHD1 mediated leucine-induced mTORC1 activation, we silenced both SESN1 and 2 in primary myoblasts derived from PHD1 KO muscle, leading to a ~70% knock down efficiency (see Figure S4E-F). Subsequently, we generated myotubes from these cells and repeated our leucine stimulation experiments (identical conditions as the ones described in the original manuscript). Interestingly, silencing of SESN1 or SESN2 in PHD1 KO myotubes failed to rescue leucine mediated mTORC1 activation ( Figure S4G-H).
Overall, since loss of Phd1 does not affect SESNs protein levels and silencing SENS1 or SESN2 in PHD1 KO myotubes doesn't rescue leucine-dependent mTORC1 activation to the WT levels, our data do not support a role for SESNs in the PHD1-mTORC1 signaling pathway. We would however like to highlight (as we also did in the original version of the manuscript) that this data does not indicate that SESNs are not involved in leucine mediated mTORC1 activation in muscle. Off note, we tested several short hairpin sequences and found that knock down of SESN1 induced a significant compensatory upregulation of other SESNs, which underscores the crucial role of these genes in myoblasts/myotubes (See figure S4E-F). We included the data on SESN1 and 2 knock-down in figure S4 and edited the revised manuscript (Line 231-234):

'To further investigate the role of LRS and SESNs, we evaluated whether SESN knock-down or LRS overexpression could rescue leucine-dependent mTORC1 activation in Phd1 deficient myotubes.
Silencing SESN1 nor SESN2 changed the responsiveness to leucine in myotubes derived from PHD1KO muscle ( Figure S4E-H).' And adapted the discussion section (line 387-391): We did not observe differences in SESN1-2 protein content upon loss of Phd1, nor did we observe a rescue of leucine mediated mTORC1 activation upon knockdown of these genes in PHD1KO cells. It is however important to mention that these data do not exclude a role for SESNs in regulating leucine mediated mTORC1 activation in muscle.'

there is a substantial disconnect between the western blot and quantitation regarding the role of LRS OE and mTORC1 activity. For instance, starved levels of pS6K1 in LRS OE/PHD1KO appear much enhanced compared to wild-type and PHD1KO EV yet the bar graph indicates no increase?
We agree with the reviewer that the representative image does not completely overlap with the quantitation, and apologize for this mistake. We have now added a western blotting image which fully represents the obtained data (See Figure 4E). We would like to emphasize that the quantification, and thus the interpretation of the data which we presented in the original manuscript are/were correct.
Of note, to improve the coherence of the revised manuscript, the data have now been moved to figure 4. In this section, both SENS knock-down experiments as well as LRS overexpression experiments are presented.

-In the skeletal muscle specific knockout, muscle weights and leucine-stimulated protein synthesis (via puromycin) should be measured and reported to confirm cell autonomous effect
We thank the reviewer for this comment. See also our reply to Reviewer#2. To address the reviewer's comment we have included below ( Figure RL2) the weight for TA and GAS muscle form WT and PHD1 mKO mice. We did not observe any decrease in muscle weight (TA or GAS) at this time point in the inducible muscle-specific PHD1 knockout mice (see Figure RL2). However, we would like to mention that, since (i) gene recombination was only induced less than two weeks before the experiments were performed and (ii) loss of Phd1 only reduces the leucine-dependent increase in mTORC1 and not contraction/insulin response, we did not expect major changes in muscle mass at this time point. In this line, a recent report using muscle-specific Raptor deficient mice (where mTOR -downstream signaling is severely blunted in response to any stimulus) showed no decrease in muscle mass 21 days after recombination (You et al., 2018).
Nonetheless, as requested by the reviewer, we did measure leucine-stimulated protein synthesis (via puromycin incorporation) in PHD1 mKO mice, and observed a 56% reduction in leucine induced protein synthesis in the PHD1 mKO compared to WT littermate controls. This data confirms the observations we made in the PHD1 KO animals and provide evidence for a cell autonomous effect of loss of Phd1. We adjusted the supplementary figure (S3C) and added this to the result section (line 184-185).
Edited in the revised manuscript: 'Moreover, leucine-induced protein synthesis assessed via puromycin incorporation was reduced as well (Figure S3C).' Nonetheless, to address whether inducible muscle deletion of PHD1 affected muscle mass, we tried to mimic the full KO condition by inducing muscle deletion at early postnatal stages. To do so, we injected HSA-Cre-ER T2 -/x Phd1 fl/fl (WT) and HSA-Cre-ER T2 +/x Phd1 fl/fl mice (PHD1 mKO ) females with tamoxifen (5 x 0.1 mg per mouse at P13-14). Since it is known that there is still significant contribution of muscle stem cells to muscle growth during and maintenance up till 14-16 postnatal weeks (Bruusgaard et al., 2006;Murach et al., 2017;Pallafacchina et al., 2013;Pawlikowski et al., 2015;White et al., 2010), we applied additional Tamoxifen injections every two weeks (1 x 1 mg per mouse, see scheme in figure  RL3. Subsequently, mice were sacrificed at identical age when compared to the PHD1 KO experiments (16 weeks). We found that TA and GAS weight were lower in PHD1 mKO compared to WT (11.3% and 14.3% respectively), but due to lower variation, only TA reached significance (p = 0.03 for TA and p = 0.2 in GAS -See figure RL3). Even though these data underscore the role of PHD1 in muscle growth, we prefer not to include them in the manuscript. The main reason for this is that the HSA-Cre-ER T2 +/inducible mouse model is not commonly used to follow up the effect of gene deletions during growth. Moreover, the repeated use of Tamoxifen might interfere with hormonal household of the mice, which prevents us from making sound and convinced statements about our observations.
- Figure 3C, given the mild effect on pS6K compared to whole body knockout, more detailed analysis of mTORC1 signaling (4EBP1, pS6) should be performed.
We agree with the reviewer and have performed a more detailed analysis of mTORC1 downstream signaling (including, p-RPS6 and p-4E-BP1). Our new data confirms and extends the data on p-S6K1 which was included in the initial manuscript. Data are now included in figure 3 (panel C) and will replace the data on mTOR ser2448 phosphorylation since there is debate about whether or not this phosphorylation site mirrors downstream mTORC1 signaling (Figueiredo et al., 2017).
-The study in human tissue as is does not add anything. Rather, it suggest that a reduction in PHD1 is not correlated with a decrease in LRS in human tissue. Moreover, recent reports suggest mTORC1 activity is elevated in aged muscle in rats (Joseph, Giselle et al BioRxiv 2019). Do the authors see a similar effect on mTORC1 activity in their samples?
We acknowledge the reviewer's comment, but respectfully disagree that the human data does not add anything to the manuscript. While due to high inter-subject variation and low subject number the reduction in LRS protein content did not reach statistical significance, we did find reduced LRS activity (assessed as RAGa Kleu 142 leucylation) in the human samples and this correlated with PHD1 protein content, r = 0.56; p = 0.02, see line 312). Also, we are aware of recent reports showing increased mTORC1 activity or protein synthesis in aged mouse muscle (Joseph et al., 2019;Miller et al., 2019), but this concerns basal (unfortunately, the exact dietary status of the mice was not reported) protein synthesis, and changes in protein synthesis upon leucine stimulation were not investigated. Concerning the reviewer's question about baseline mTORC1 activity, we kindly refer to the original publication describing the human study (Smeuninx et al., 2017): here, the authors did not observe changes in baseline mTORC1 activity. Nevertheless, we repeated the blot for p-S6K1 and p-RPS6 in the muscle lysates we received from Smeuninx et al. and found a very weak, to almost absent signal for both phosphorylated kinases (see RL 5) in young as well as aged subjects. This is not unexpected because subjects were fasted overnight before the muscle sample collection. Thus, at the specific dietary conditions when muscle biopsies were harvested, we did not see an increase in basal mTORC1 signaling in old human muscle. We would like to emphasize that the inclusion of the human dataset did not aim to address mechanisms underlying the regulation of basal mTORC1 signaling. Rather, they show that an impaired activation of muscle protein synthesis in response to leucine (a phenomenon observed in humans and described as anabolic resistance and well evidenced in literature -see for instance the references we included in the manuscript line 294) coincides with reduced PHD1 content and LRS activity. Even though we fully realize that these data do not imply causality, they are consistent with and extend our observations in mice.
Minor concerns: - Figure 4F, the ICC staining is not very convincing and should be improved We thank the reviewer's comment. We have now included an additional merged picture in Figure 4E to make the representative pictures clearer for the reader.

-PHD1 protein levels in knockout should be reported
We thank the reviewer for their comment. First, we already included in the original manuscript that we observed a 90% decrease in Phd1 mRNA in PHD1 mKO muscle when compared to WT (line 180). Given the contribution of other non-muscle cells to muscle samples, we were there confident to state that Phd1 was indeed knocked out, also in our muscle specific mouse model. This data is strengthened by our observations showing that PHD1 mKO muscle have an identical phenotype when compared to PHD1 KO muscle: 1) PHD1 mKO muscle show impaired leucine mediated mTORC1 activation; 2) our new data showing that PHD1 mKO muscle have lower protein synthesis upon leucine stimulation; 3) PHD1 mKO muscle retains its responsiveness to insulin; 4) PHD1 mKO muscle has lower LRS protein content.
Unfortunately, we have been facing major difficulties in detecting mouse PHD1 protein in tissue samples using western blotting while it easily detects human PHD1 -a problem that is very well known in the field. In fact, we tried to validate several commercially available antibodies on muscle samples from our whole body PHD1 KO animals (which cannot have any PHD1 protein since the DNA has been removed), but were not successful in generating high-quality western blotting images which would allow us to make a reliable statement about PHD1 protein levels (see figure RL6). Nevertheless, for  unknown reasons, we were able to able to pick up PHD1 in primary myotubes from WT and PHD1 KO mice: Figure 3E of the manuscript has a representative western blot for PHD1 protein levels, which clearly shows absent PHD1 protein levels in the KO cells and a recovery in the PHD1 overexpressing cells. In addition, we performed immunostainings in WT and PHD1 KO myoblasts, and could show a low signal in the Phd1 deficient cells (see our reply to reviewer #1, Figure RL1) compared to WT. Thus, because we are uncertain of the reliability of the PHD1 ab in mouse muscle lysates, we would like to not add data on PHD1 protein content in the main manuscript.