De novo serine biosynthesis is protective in mitochondrial disease

  1. Christopher B Jackson
  2. Anastasiia Marmyleva
  3. Ryan Awadhpersad
  4. Geoffray Monteuuis
  5. Takayuki Mito
  6. Nicola Zamboni
  7. Takashi Tatsuta
  8. Amy E. Vincent
  9. Liya Wang
  10. Thomas Langer
  11. Christopher J Carroll
  12. Anu Suomalainen

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

Importance of serine as a metabolic regulator is well known in tumors and raising attention also in degenerative diseases. Recent data indicate that de novo serine biosynthesis is an integral component of metabolic response to mitochondrial disease, but the roles of the response have remained unknown. Here, we report that glucose-driven de novo serine biosynthesis maintains metabolic homeostasis in energetic stress. Pharmacological inhibition of the rate-limiting enzyme, phosphoglycerate dehydrogenase (PHGDH), aggravated mitochondrial muscle disease, suppressed oxidative phosphorylation and mitochondrial translation, altered whole-cell lipid profiles and enhanced mitochondrial integrated stress response (ISR mt ), in vivo, in skeletal muscle and in cultured cells. Our evidence indicates that de novo serine biosynthesis is essential to maintain mitochondrial respiration, redox balance, and cellular lipid homeostasis in skeletal muscle with mitochondrial dysfunction. Our evidence implies that interventions activating de novo serine synthesis may protect against mitochondrial failure in the skeletal muscle.

Bullet points

  • Serine becomes an essential amino acid in mitochondrial translation defects

  • Blocking de novo serine biosynthesis promotes progression of mitochondrial disease

  • De novo serine biosynthesis maintains phospholipid homeostasis upon mitochondrial insult

  • Serine biosynthesis sustains redox-balance and mitochondrial translation in disease

Article activity feed

  1. Review Commons

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

    Learn more at Review Commons

    Reply to the reviewers

    General Statements:

    We want to thank the reviewers for their time in assessing our study, their positive feedback and their constructive comments that we address here. We have carefully assessed each reviewers’ comments and address them point-by-point below. We appreciate that the reviewers consider our study to be relevant and advance the field in better understanding mitochondrial stress responses.

    There is a consensus as to why we have inhibited rather than depleted PHGDH via genetic means. PHGDH knockout is developmentally lethal at embryonal day 13.5, and therefore whole-body knockout crossings were not an option (Yoshida et al. JBC 2004). Furthermore, we and colleagues have shown that mitochondrial disease induces a complex systemic response including e.g., metabokines FGF21 and GDF15, secreted by the affected muscle and heart and modifying metabolism in the whole organism, indicating that tissue-specific knockouts are not optimal to resolve mechanisms in pathophysiology. Further, potential serine supplementation from other tissues could mask the effects on the muscle. Next, the slowly progressive phenotype of the mitochondrial myopathy mouse, fully manifesting close to 2 years of age, make double-transgenic strategies resource-intensive. Therefore, a PHGDH inhibitor, the specificity of which we carefully explore, was the approach of choice.

    Serine rich diet: We did consider a serine-rich diet. However, our evidence indicates that exogenous serine is not sufficient to rescue the phenotype in cultured cells. This suggests that the enzymatic intracellular localization of de novo serine synthesis is of crucial importance.

    Off-target effects: We agree and are aware that using a small-molecule molecule for pharmacological inhibition bears risks of off-target effect, as Arlt et al. 2021 reported for NCT-503 neuroblastoma cells lines. In agreement with the reviewers about the relevance of our results, the observation that we do not observe any meaningful alterations in WT muscle, even at the advanced age where metabolic muscle fitness is already challenged, whilst the myopathy phenotype worsens in myopathic mice support an on-target effect of NCT-503.

    Therefore, we had carefully considered the suggested experiments during the initial study and decided that PHGDH inhibitor was the best choice to answer our questions about importance of de novo serine biosynthesis.

    Reviewer #1 (Evidence, reproducibility and clarity):

    In this manuscript the authors describe interesting results regarding amino acid metabolism under conditions of mitochondrial stress. The authors used a selective PHGDH inhibitor (compound NCT503), and document that de novo serine synthesis is essential to sustain phospholipid biosynthesis, redox homeostasis, mitochondrial function, and mitochondrial protein synthesis in Deletor mice and in cell culture under mitochondrial stress. Interestingly, serine supplementation does not rescue those metabolic alterations, indicating a specific mitochondrial stress-dependent mechanism of serine utilization by the cells.
    In addition, the authors attempted to explore the role of serine in phospholipid biosynthesis under in vivo and ex vivo mitochondrial stress, and this is the weakest part of the manuscript. The specific reduction of mitochondrial PE using the PHGDH is interesting, and it can provide some clues into how mitochondrial membranes are synthetized under mitochondrial stress, but additional experimental approaches should be incorporated to generate a more robust body of evidence.

    In summary, the manuscript should be improved by using additional experimental approaches. In addition, genetic intervention through PHGDH gain or loss of function can help to elucidate the molecular mechanisms.

    We thank the reviewer for this proposition. As explained above in detail, knockout of PHGDH in mice is lethal in embryogenesis. Secondly, the cell culture studies show that in a mitochondrial disease background, cell-intrinsic de novo serine biosynthesis becomes essential. Therefore, the phenotype of a conditional knockout in skeletal muscle is likely to be drastic and does not reflect the situation of an adult-onset disease. Thirdly, if the gain- and loss of mutant double transgenics would live, the Deletor mice are an actual slowly progressing disease model, which manifests the phenotype at around two years. Therefore, this experiment takes three years, is not feasible and was considered to be of little informativeness. We carefully explored the specificity of PHGDH inhibition by NCT503 and found it to serve best the questions asked.

    The authors could also incorporate metabolomic studies using C13 serine or C13 glucose supplemented culture medium to differentiate carbons that come from extracellular serine or from de novo synthesis.

    We thank reviewer #1 for her/his comments. The exogenous serine supplementation shows minor rescue of the phenotype in cultured cells but decreases somewhat the mitochondrial integrated stress response markers in serine-containing medium, suggesting some amelioration of the PHGDH inhibition.

    We have now incorporated new data of D3-serine flux in cells to the manuscript (p.14, FigS3G, p.33). These data indicate that some serine uptake occurs by the cells and that is not significantly affected by mitochondrial translation inhibition (actinonin), even if the combination of PHGDH plus actinonin is not viable.

    Manuscript Figure S3G. D3-serine uptake in C2C12 cells with and without actinonin. Normal non-labelled serine was used as a negative control for D3-serine contamination. D3-serine amounts were measured in culture media and in cells; biological replicates (n=3). In media, fold change (FC) was calculated relative to the sample with normal serine media without actinonin. In cells, FC was determined against cells in D3-serine without actinonin. No traces of D3 label were detected in normal serine samples. ACT = actinonin.

    Additional questions: Is PHDGH subcellular localization modulated under mitochondrial stress conditions?

    This is an interesting question. The insufficient resolution in muscle sections did not give a conclusive answer. We have, however, performed new immunofluorescence experiments in cells to investigate this potential mechanism and used a newly generated PHGDHKO cell line as a negative control (Rev Fig.1). The conclusions have been added to the manuscript (page 14, FigS4E,F, p.34). In essence, no altered localization of PHGDH in cells is observed – the localization appears to be cytoplasmic. The increasing PHGDH amount in increasing actinonin concentration is apparent, however.

    Revision Figure 1. Immunofluorescent staining of HEK293 and C2C12 cells. WT and PHGDHKO HEK293 cells were used as controls for the PHGDH antibody. (Hoechst staining of nucleus, blue; anti-PHGDH, green; mitochondrial outer membrane protein TOM20, red. EtOH = ethanol; ACT = actinonin; INH = PHGDH inhibitor.

    Is the extracellular serine uptake impaired under mitochondrial stress?

    As explained above, we have now incorporated data of D3-labeled serine uptake in cells to the manuscript, quantifying the uptake from the media and the levels in the cells. Mitochondrial translation inhibition by actinonin does not affect uptake, indicating that the uptake per se is not impaired. However, the de novo synthesis being essential in metabolic stress strongly points to the intracellular site to be important.

    While the extracellular serine can only little compensate for de novo synthesis, our new data, qPCR analysis on muscle of all mouse groups to quantify serine tRNA levels (MTTS1) indicates the levels to be increased, suggesting insufficient charging of the tRNA (MTTL1 as negative control; Rev Fig. 2). In addition, the presumed mitochondrial serine transporter SFXN1 is upregulated on protein level (p.11, Fig S2H, p.32). These data suggest that cell-intrinsic serine synthesis supports mitochondrial translation in metabolic stress situations.

    Revision Figure 2. RNA expression of mitochondrial tRNAs for leucine (MTTL1) and serine (MTTS1) in skeletal muscles of wild type (WT) and Deletor (DEL) mice; RT-qPCR. WT VEH n=6, DEL VEH n=5, WT INH n=7, DEL INH n=8. VEH = vehicle; INH = PHGDH inhibitor; FC = fold change.

    ATF4 is a transcription factor that regulates amino acid transport. In this connection, is ATF4-dependent serine transporters modulated under mitochondrial stress or under PHGDH treatment?

    As explained above, SFXN1 transporter was upregulated in protein level. In postmitotic skeletal muscle, the responses are under ATF5, as shown in Forsström, Jackson et al. Cell Metab 2019.

    Regarding the alteration in redox homeostasis, mitochondrial function and mitochondrial protein synthesis. Is the alteration of phospholipid synthesis upstream of all of those mitochondrial alterations?

    The authors are unclear of what the reviewer means by “all those mitochondrial alterations”. The disease in the Deletor mice is caused by dominant patient-homologous mutation of mitochondrial DNA helicase Twinkle, and in the cell culture model we block mitochondrial protein synthesis with actinonin. Therefore, the primary defect in both mice and cells is intramitochondrial, in mtDNA replication and protein synthesis, and the stress responses are a consequence of the primary mitochondrial dysfunction. The consequent secondary stagewise mitochondrial integrated stress response, ISRmt, affects both mitochondria and rest of the cell, with effects in metabolism and nuclear genome transcription. We have pioneered the discovery of ISRmt in mice and humans with mitochondrial defects/diseases in several studies (e.g., Cell Metab 2016, 2017, 2019, 2020). This response includes a remarkable remodeling of one-carbon metabolism, with major changes in methyl cycle, transsulfuration and nucleotide synthesis of the whole cell. PS synthesis is dependent of methyl groups deriving from the one-carbon cycle -driven methyl cycle. Therefore, the original mitochondrial replisome dysfunction causes a stagewise, progressive disease process, which is upstream of all the other responses. Phospholipid synthesis alteration, however, has high potential to modify mitochondrial membranes that can aggravate disease during its progression.

    The temporal metabolomics of cultured cells did show PE accumulation at later time points than on mitochondrial translation or other crucial cellular metabolites, suggesting its alterations to be a consequence rather than upstream. Indeed, further studies are needed to dig deeper into the dynamics of phospholipid synthesis in mitochondrial dysfunction.

    Additionally, in my opinion the results should be reorganized, because in the current format the manuscript is fragmented, and several panels lose the symmetry.

    The authors are unclear what the reviewer means by the panels losing symmetry. Without suggestions it is hard to make changes, but we have carefully reviewed the clarity of the presentation.

    Major concerns.

    Figure 1

    The authors should incorporate the Deletor mice amino acid levels in muscle. How does the PHGDH inhibitor treatment modulate the other non-essential amino acids?

    We have quantified all 11 non-essential amino acids from Deletor muscle by targeted metabolomics in combination or individually and added the new graphs (p.10, Figure S2D,E, p.32).

    Manuscript Figure S2D. Total non-essential amino acid quantification (NEAAs) from targeted metabolomics of skeletal muscles of WT and DEL mice. WT VEH n=6, DEL VEH n=5, WT INH n=7, DEL INH n=8. VEH = vehicle; INH = PHGDH inhibitor. Significance levels: n.s. = p > 0.05; * p ≤ 0.05; ** p ≤ 0.01; *** p ≤0.001.

    Manuscript Figure S2E. Individual NEAAs from targeted metabolomics of skeletal muscles of WT and DEL mice. WT VEH n=6, DEL VEH n=5, WT INH n=7, DEL INH n=8. VEH = vehicle; INH = PHGDH inhibitor. Significance levels: n.s. = p > 0.05; * p ≤ 0.05; ** p ≤ 0.01; *** p ≤0.001.2. The authors should incorporate the data of WT + vehicle and wt + NCT503 mice in figures 1C, 1D, 1F and 1G, in order to compare properly the effects of PHGDH inhibitorThe authors should incorporate the data of WT + vehicle and wt + NCT503 mice in figures 1C, 1D, 1F and 1G, in order to compare properly the effects of PHGDH inhibitor.The authors should incorporate the data of WT + vehicle and wt + NCT503 mice in figures 1C, 1D, 1F and 1G, in order to compare properly the effects of PHGDH inhibitor.

    The authors should incorporate the data of WT + vehicle and wt + NCT503 mice in figures 1C, 1D, 1F and 1G, in order to compare properly the effects of PHGDH inhibitor.

    We agree with the reviewer`s notion to depict all groups. No COX- fibres exist in our vehicle or inhibitor-treated WT mice and we have added this data into the manuscript in p.31, commented on in p.6, Figure S1B, also below).

    Manuscript Figure S1B. Histochemical analysis of combined cytochrome-c-oxidase (COX) and succinate dehydrogenase (SDH) activity in muscles of treated WT mice. (Brown fibres indicate high COX activity, translucent – low COX activity). Lower panel shows immunohistochemical detection of the mTORC1 downstream target - phosphorylated ribosomal S6 (p-S6). INH = PHDGH inhibitor.

    Figure 1H. Are there any effect of PHGDH in the mtDNA of WT mice? The authors might incorporate this information, to show properly that mitochondrial stress led to dependence of serine to sustain muscle homeostasis.

    In addition to the mtDNA deletion analysis, we have assessed mtDNA copy number via qPCR (p.27, Figure S1D). No mtDNA deletions exist in WTs and their mtDNA copy number raised slightly. These new data are included in p. 6 and Figure S1D, p.31.

    Manuscript Figure S1D. Mitochondrial DNA copy number analysis in muscles of WT and DEL mice. Measured with qPCR (n=5-8/group). VEH = vehicle; INH = PHGDH inhibitor. Significance levels: n.s. = p > 0.05; * p ≤ 0.05; ** p ≤ 0.01; *** p ≤0.001.

    Figure 2.
    Interestingly the authors observe a decrease in total mitochondrial lipids content, and an increase in mitochondrial PE content in Deletor mice compared to WT mice. These results suggest an alteration in the phospholipids flux between mitochondria and endoplasmic reticulum in this model of mitochondrial disease. Moreover, PHGDH treatment appears to be able to rescue this alteration. Some question related this issue: What is the expression of genes involved in the balance PC, PE, PS? Are the PSS1, PSS2, PSD and PEMT expression altered?

    Previous Deletor muscle studies cohorts showed non-altered PEMT in diseased mice of similar age nor was PEMT and PSD significantly altered RNAseq data from patients in contrast to PHGDH (Forsstrom et al., 2019; Rev Fig.3). 20-months old Deletor do not show any alteration in PEMT (qPCR quantification below) when analysed from total muscle. Whether these are changed in single fibers (mosaic manifestation of the disease) we cannot exclude.

    Revision Figure 3. Left: RNA expression of PEMT in skeletal muscles of 20-months-old WT and DEL mice. Measured with qPCR; n=9. Right: Gene expression of PEMT, PSD, and PHGDH enzymes in muscles of mitochondrial myopathy patients. Measured with RNA-Seq; n=8 in control; n=4 in patients. Significance levels: n.s. = p > 0.05; * p ≤ 0.05; ** p ≤ 0.01; *** p ≤0.001.

    Regarding the phospholipid synthesis. Is mitochondria or endoplasmic reticulum ultrastructure altered in Deletor mice muscle? The authors should explain the possible mechanisms.

    We have extensively characterized the pathology in Deletor muscle in publications previously (Tyynismaa et al. PNAS 2005; Tyynismaa Hum Mol Genet 2010 for morphology, especially, but also as a part of the physiology studies Nikkanen et al. Cell Metab 2016; Khan et al. Cell Metab 2017; Forsström, Jackson et al. Cell Metab 2019 and Mito et al. Cell Metab 2022). The morphological changes in mitochondria are quite extensive, and these are comparable to those found in patients with similar mutations: enlarged mitochondria with distorted and few cristae, various inclusions. We presume the reviewer means sarcoplasmic reticulum in skeletal muscle. The extent of the muscle pathology in vivo suggests that ER structure is changed as everything else is: the diseased muscle fibers are full of abnormal mitochondria in the Deletors, with only few traces of myofibers or fibrils. These fibers become more prevalent after PHGDH inhibition. Furthermore, the mice – WT and Deletors – show age-induced prevalence of non-pathology-related accumulation of sacroplasmic aggregates (so called tubular aggregates) as a known feature of C57Bl6 mice.

    New Figure S1 C, p.31 and below (Rev Fig.4): Ultrastructure of Deletor muscle shows enlarged mitochondria both inside the fibers (Fig S1C left panel, arrows) and subsarcolemmally, further aggravated by PHGDH inhibition. Tubular aggregates, as shown in the inhibitor treated Deletor in the Figure S1C seen in the light-microscopic image on the right (arrows), are part of the mouse substrain characteristics, not disease associated. These changes are also, as white aggregates inside the fibers in all genotypes and treatment groups We now added the light and electron microscopic image analysis to Figure S1 C. We now added the electron microscopic images to the manuscript as S1C (commented on p.6), indicating the aggravated phenotype of PHGDH inhibitor.

    Revision Figure 4. Left: Panel from the Manuscript Fig. S1C. Representative images of transmission electron microscopy (TEM) in skeletal muscles of WT and DEL mice. Black arrows – enlarged mitochondria. Right: Representative images from light microscopy (LM) of muscles of the same groups as in Left. Black arrows – tubular aggregates. VEH = vehicle; INH = PHGDH inhibitor.

    We have also performed analysis of mitochondrial ultrastructure in cells (see below) showing specific lipid accumulation. As mentioned above the alterations in PE levels are detected metabolically at later time points and after substantial loss of mitochondrial translation.

    We added this new data as Figure S3E, p.33, commented on p.6.

    Manuscript Figure S3E. Representative images of TEM analysis of 12h-treated C2C12 cells. The cells were incubated in media with or without serine and treated with either actinonin, PHGDH inhibitor or a combination of both. White arrow – myelinosome-like membranous lipid aggregates; M = mitochondria. INH = PHGDH inhibitor; ACT = actinonin; Ser = serine.

    Supplementary figure 3D
    Based on the metabolomic studies, the authors propose a time-dependent decrease in PSAT1 and phosphoserine in cells under mitochondrial stress (Figure S3D). To elucidate the direct role of PHGDH, the authors should analyze the phosphoserine and different phospholipid (described in figure 2E) in presence of PHGDH inhibitor. This will help to understand the link between the endogenous serine synthesis and mitochondrial PE accumulation.

    The temporal metabolomics show a time-dependent decrease of the metabolite phosphoserine (not PSAT1). Here, we show the PE and PC dynamics at 6 and 24 hours. These data have now been added to the manuscript, figure S3I, p.33, commented on page 14.

    Manuscript Figure S3I. Normalised PC and PE pools in C2C12 cells measured with untargeted metabolomics after treatments with actinonin, PHGDH inhibitor, or a combination of two for 6 and 24 hours. n=6-11/group; ACT = actinonin; INH = PHGDH inhibitor.

    Figure 3H shows a decrease in phosphoserine in the presence of PHGDH inhibitor but this figure is asymmetric compared to figure 3I. Can the authors use another experimental approach to detect the specific mitochondria phospholipid levels (used in Figure 2 for instance)?

    Figure 3H shows a decrease in phosphoserine in the presence of mitochondrial dysfunction induced through actinonin alone. Figure 3I shows a further decrease in phosphoserine when actinonin-treated cells are treated with the PHGDH inhibitor (phosphoserine is marked now also in our genetic model of mtDNA depletion, Figure 3I). The data are from untargeted metabolomics, parallel analysed samples, and therefore are comparable for the levels.

    We have now grouped the time-wise dynamics of the annotated phospholipids from the untargeted metabolomics decreasing at a late stage of the temporal treatment, suggesting that the consequences on of mitochondrial dysfunction by translation inhibition clearly precede phospholipid alterations in cells. (Figure below, also added to the manuscript as Figure S3D, p.33, and commented in the text in p. 13)

    Manuscript Figure S3D. Heatmap of selected top significantly altered metabolites in primary human myoblasts temporally treated with actinonin; untargeted metabolomics (n=5-6, run in technical duplicates). ACT = actinonin.

    The authors should incorporate mitochondrial PE analysis in figure 3 to link the cellular studies described in this figure with the studies done in Deletor mice muscle.

    Figure 3 I-K. The authors suggest an alteration in glutathione redox state and a further increase in mitochondrial superoxide production in cells treated with the PHGDH inhibition under mitochondrial stress. What are the total glutathione levels under these conditions? Could GSH regeneration improve the mitochondrial function and mitochondrial protein synthesis? Is extracellular serine able to rescue the reduced glutathione levels?

    We have previously shown in vivo (Nikkanen et al. Cell Metab 2016), in Deletors, that in the skeletal muscle and the heart glucose carbons show flux via serine to glutathione, without changes in steady-state levels of glutathione, indicating high usage. In our targeted metabolomics assay from muscle, the steady-state glutathione amount was below the limit of quantitation. The levels of glutathione oxidized vs reduced are dynamic, however. Our untargeted set shows a trend towards an increase. The GSSG/GSH data (Rev Fig.5 Right) was included already in the original manuscript, as Figure S2C, and the time-wise dynamics is here shown for the reviewer (Rev Fig.5 Left).

    Revision Figure 5. Left: Ratio between oxidized (GSSG) and reduced (GSH) glutathione temporally measured in primary human myoblasts with untargeted metabolomics; n=10-16/group. Right: Ratio between glutathione forms in muscles of WT and DEL mice. Measured by targeted metabolomics; WT VEH n=6, DEL VEH n=5, WT INH n=7, DEL INH n=8; VEH = vehicle; INH = inhibitor. Significance levels: n.s. = p > 0.05; * p ≤ 0.05; ** p ≤ 0.01; *** p ≤0.001.

    Minor concerns. Figure 3B,C does not show the statistical analysis so please incorporate this information.
    Include quantification of figure 3 E and supplementary figure S3E.

    We have added this information to both figures.

    Figure S3E. Improve flow cytometry histogram, the cell population data values cannot be observed.

    We have improved the resolution.

    Some methods and primers included in the material and methods section are not used in the manuscript.

    We have carefully edited the materials and methods sections for any unnecessary data.

    Reviewer #1 (Significance):

    In this manuscript the authors describe interesting results regarding amino acid metabolism under conditions of mitochondrial stress. The authors used a selective PHGDH inhibitor (compound NCT503), and document that de novo serine synthesis is essential to sustain phospholipid biosynthesis, redox homeostasis, mitochondrial function, and mitochondrial protein synthesis in Deletor mice and in cell culture under mitochondrial stress. Interestingly, serine supplementation does not rescue those metabolic alterations, indicating a specific mitochondrial stress-dependent mechanism of serine utilization by the cells.
    This data will be relevant to better understand the connection between alterations in mitochondrial function and amino acid metabolism in cells, and in organisms.

    We thank the reviewer #1 for the constructive comments and acknowledgement of interesting results and significance.

    Reviewer #2 (Evidence, reproducibility and clarity):

    Experiments reported in this manuscript indicate that NCT-530- an inhibitor of the de novo serine biosynthetic enzyme PHGDH- worsens mitochondrial pathology.
    Systemic administration of NCT-503 decreased serine levels only in Deletor mice- ubiquitously expressing a homologous dominant patient mutation in mitochondrial twinkle helicase.
    NCT-503 treatment induced a further increase of the mitochondrial integrated stress response in Deletor mice. Moreover, the metabolic profile and lipid balance was further modified by NCT-503 administration to Deletor mice. The relevance of NCT-503 treatment was finally evaluated in cellular systems exposed with different treatments inducing mitochondrial insults.

    COMMENTS:

    NCT-530 specificity should be confirmed by reducing PHGDH levels by either siRNA or even better by CRISPR-Cas9-mediated gene deletion.

    Firstly, we find clear results in Deletor mice but no pathology in WT mice treated with NCT-530. We interpret the reviewer to mean our cellular model, because in vivo siRNA or CRISPR-Cas9 approach is not feasible in mice that manifest the disease at 2 years of age. The compound is inhibiting the activity of PHGDH. The specificity has been described in the paper of Pacoult et al. (2016) previously, and as we find decreased serine and response of increased PHGDH transcript, the results are well consistent to what has previously been described. However, to respond to this reviewer****, we have now added new data on a PHGDHKO in HEK293 cells (Figure S4E,F,p34 and below, commented on page 14).

    Manuscript Figure S4F. Quantification of population doublings of WT and PHGDHKO cells treated with actinonin, PHDGH inhibitor or a combination of both. Immunofluorescence images are presented in Rev Fig 1. The data is presented as average values of three independent experiments; VEH = vehicle; INH = inhibitor; ACT = actinonin; KO = knockout; EtOH = ethanol. Significance levels: n.s. = p > 0.05; * p ≤ 0.05; ** p ≤ 0.01; *** p ≤0.001; **** p ≤0.0001.

    If the serine biosynthetic pathway is causally relevant to worsen the mitochondrial pathology, supplementing Deletor mice with a serine-rich diet or intraperitoneal injection of serine is expected to improve pathology.

    This is an exciting suggestion which we did consider. However, based on our cell experiments, we show that in the context of mitochondrial disease, the de novo serine synthesis becomes essential for cellular viability, and this cannot be compensated by extracellular serine supplementation alone. This suggests that the intracellular localization of serine synthesis is essentially important, which is an interesting finding and warrants further investigation, not in the scope of this paper.

    Reviewer #2 (Significance):

    Limited significance in the field.

    Reviewer #3 (Evidence, reproducibility and clarity):

    This study concludes that de novo serine biosynthesis fueled by PHGDH activity is an adaptive mechanism counteracting defects in mitochondrial function observed in mitochondrial myopathies, while being dispensable for mitochondrial function of healthy muscle. By using a mouse model of mitochondrial myopathies, as well as cells treated with different mitochondrial poisons, authors show that de novo serine biosynthesis is specifically needed to preserve phospholipid synthesis, some degree of mitochondrial function, and mitigate mitochondrial ROS production. These conclusions are drawn mostly from the data obtained from experiments treating cells and mutant mice with NCT-503, an inhibitor of PHGDH. The main limitation is that the approach does not allow us to discern whether the phenotypes observed are explained by on-target and muscle autonomous actions of NCT-503. The list of major concerns are as follows:

    We thank reviewer #3 for her/his comments.

    1. The authors do not cite the study by Vandekeere et al. Cell Metabolism 2018 that define the KO of PHGDH in endothelial cells. This study demonstrates that serine derived from PHGDH is required to synthesize heme to preserve mitochondrial function in endothelial cells. In addition, they demonstrate that the absence of PHGDH increases mitochondrial ROS and decreases electron transport chain function. These previous studies can indicate that the worsening of the muscle phenotype in the mice treated with NCT-503 might be driven by its actions in endothelial cells, and not in muscle. In addition, it raises the question on why NCT-503 has no effect on muscle of 24-month-old mice, which have a decline in metabolic fitness.

    This is an interesting comment, and we indeed have now included the missing reference. Firstly, we find it interesting as well that NCT-503 has no significant health-related effect in WT background in the tissues that we analysed but shows harmful effects upon mitochondrial dysfunction. Our aged WT mice, even at the age of 24 months, show no signs of respiratory chain deficiency in skeletal muscle and this is consistent with numerous other studies on old mice. Therefore, the metabolic fitness decline that the reviewer mentions does not make PHGDH induction essential in normal aging.

    Secondly, we are aware of the Vandekeere et al. study on endothelial cells. Our mouse is a constitutive transgenic mouse, and therefore isolated changes in single tissues as in the endothelial KO mouse, because of metabolite signaling between cell types, is hard to compare with ours. In our mice, expressing a dominant patient mutation in the mitochondrial replicative helicase Twinkle, PHGDH induction in the skeletal muscle is a key component of the mitochondrial integrated stress response (ISRmt) as we described in e.g., in Tyynismaa et al. 2010 and Forsström, Jackson et al. Cell Metab 2019. Similar findings have been reported by Kuhl et al. ELife 2017 and other groups. We have also shown that PHGDH expression is dependent on prior induction of FGF21 in the skeletal muscle (Cell Metab 2019). We cannot fully exclude a contribution of endothelial cells in the Deletor phenotype, but the upregulation of PHGDH is robust exactly in the Deletor muscle, and endothelial cells show no phenotype in ultrastructural analysis of these mice. Therefore, our strong view is that in order to study the effect of PHGDH induction caused by primary mitochondrial disease in the skeletal muscle, the use of an inhibitor is the first choice.

    Hence, we find that the Vandekeere et al. study supports our findings in that increasing PHGDH could represent an adaptive response to support electron transport chain function and decreasing mitochondrial ROS in a tissue where PHGDH is induced.

    In this respect, we would also like to emphasize that a 1-month treatment of old mice, as in our study, differs from a life-long tissue-specific knockout, which is not a natural disease presentation nor an avenue to understand whether the opposite – increasing PHGDH activity – could represent a viable treatment option.

    1. No experiments with PHGDH knockdown are performed in vitro (or ideally in vivo using muscle electroporation if possible) to confirm the specificity of NCT-503. This is validation is key in muscle, as the on-target actions of NCT-503 were mostly shown in different cancers with low and high PGHDH expression (Pacold et al). Therefore, whether there are no off-target effects of NCT-503 in muscle is still unknown and should be defined.

    Firstly, we find effects in serine levels and also response to PHGDH expression, and effects on ISRmt. As argued, full-body genetic ablation is associated with lethality in early embryogenesis and is not an option. Muscle-specific PHGDH-KO was considered, but with a late-manifesting mouse is a several-year experiment and therefore was not considered to be in the scope of this article. In an addition to the arguments, we raised above, muscle electroporation or intramuscular AAV infection would represent the only option, but with limited spreading in adult aged mice (our own and colleagues’ experience), and in a mosaic disease would offer low information.

    We have however, now added data on a PHGDHKO HEK293 cell line with and without pharmacologically inducing mitochondrial dysfunction, p.14, Fig S4E,F.

    1. The data showing that exogenous serine supplementation cannot override the effects of NCT-503 treatment on mitochondria could also be compatible with an off-target effect of NCT-503 in models of mitochondrial dysfunction (see point 2). If the experiments suggested in point number 2 demonstrate on-target action of NCT-503, authors should then decrease the expression of the mitochondrial transporter of serine (SFXN1) to demonstrate that specific intracellular pools of serine are needed to mitigate mitochondrial dysfunction. If the authors' conclusion is conclusive, knock-down of SFXN1 should be highly toxic in in vitro and in vivo models of mitochondrial myopathies, while having barely any effect in controls (similarly to serine deprivation in vitro being almost harmless).

    Thank you for the comment. To our knowledge and based on our own experience elsewhere, the role of SFXN1 as the only serine transporter is still somewhat unclear (Kory et al, Science 2018; see also below). Therefore, the relevant metabolomic serine-dependent changes, PHGDH response and effects on the relevant stress response to our opinion are strong evidence of on-target-effects.

    1. The authors list SFXN1 in the antibodies used in the paper, but I could not find any data. Are the protein levels of SFXN1 changed in mice with mitochondrial myopathies? Do they strongly correlate with PHGDH expression?

    We actually had probed SFXN1 ab in the muscle but because of the extent of the data in the paper, had decided to omit the data – but accidentally left it in the methods. We do see an increase in the abundance of SFXN1 in treated Deletors. However, as also explained above, although SFXN1 has been described as a serine transporter, it is not fully characterized, also transports other amino acids, and has some redundancy with other SFXN family members requiring multiple KDs to get the required effect (Kory et al., Science 2018).

    These data are now included to Figure S2H (p.32 and below) and commented in page 11.

    Manuscript Figure S2H. Evaluation of SFXN1 protein expression in muscles of WT and DEL mice with Western Blotting. Left: image of the SFXN1 bands and the total protein lanes. Right: quantification of the band intensities. n=5/group; VEH = vehicle; INH = inhibitor. Significance levels: n.s. = p > 0.05; * p ≤ 0.05; ** p ≤ 0.01; *** p ≤0.001.

    1. Serine tracing experiments would be required to conclude that serine is needed for phospholipid synthesis. Otherwise, the defects observed can just be downstream of mitochondrial dysfunction and ISRmt activation.

    Thank you for the point. Our study aims to characterize the role of PHGDH in skeletal muscle, in a mitochondrial disease. In vivo tracing experiments would be out of the scope for this study. We have modified our conclusions in respect to this in page 11.

    1. Authors use isolated mitochondria from muscle to determine whether OPA1 processing is changed in mice with mitochondrial myopathy and treated with NCT-503. The blots show higher total OPA1 content per mitochondria in the group with the greatest mitochondrial dysfunction, depolarization, and ROS production (myopathy + NCT-503). These data strongly suggest that dysfunctional mitochondrial from dysfunctional fibers do not survive the isolation procedure, showing just OPA1 content of resilient mitochondria surviving isolation. Indeed, heme depletion (as expected from PHGDH inhibition Vadekeere et al. Cell Metabolism 2018) and OPA1 processing catalyzed by OMA1 activation by ROS and depolarization can both activate the mitochondria ISR, via HRI. Therefore, authors should analyze OPA1 processing in total lysates to include mitochondria from dysfunctional fibers. Maybe increased OMA1 activity as a result of increased mitochondrial ROS and depolarization could explain the exacerbation of the mitochondrial ISR induced by NCT-503 treatment.

    The notion that stress-induced OPA1 processing present in mitochondrial myopathy is a valid one, which we did ask in a previous study in the Deletor mouse (Forsstrom et al., Cell Metabolism 2019). Surprisingly, we did not detect any significant levels of OPA1 processing in the Deletors, in the total cell lysates in that paper, nor in a previous paper of mitochondrial myopathy treated with Atkins diet (Ahola et al. EMBO Mol Med, 2016). The differential centrifugation protocol is unlikely to induce selection bias at the high centrifugation forces used (10000xg). We have previously described the increase in stress-related proteins in mitochondria-enriched fractions (Forsstrom et al., Cell Metabolism, 2019) and Rev Fig.6, below from the current study.

    Revision Figure 6. Immunoblot analysis of mitochondria-enriched fractions from skeletal muscles of all treatment groups from this study. n=5/group; VEH = vehicle; INH = inhibitor.

    Reviewer #3 (Significance):

    This study is significant, as it aims to understand the pathways that mediate adaptation to mitochondrial myopathies and such knowledge is necessary to find novel therapeutic targets. It could be of high interest to basic researchers studying metabolism and mitochondrial regulation, as well as to clinicians treating mitochondrial diseases.

    We appreciate the reviewer noting its importance and emphasizing the significance of this pathway for its clinical relevance.

  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

    This study concludes that de novo serine biosynthesis fueled by PHGDH activity is an adaptive mechanism counteracting defects in mitochondrial function observed in mitochondrial myopathies, while being dispensable for mitochondrial function of healthy muscle. By using a mouse model of mitochondrial myopathies, as well as cells treated with different mitochondrial poisons, authors show that de novo serine biosynthesis is specifically needed to preserve phospholipid synthesis, some degree of mitochondrial function, and mitigate mitochondrial ROS production. These conclusions are drawn mostly from the data obtained from experiments treating cells and mutant mice with NCT-503, an inhibitor of PHGDH. The main limitation is that the approach does not allow us to discern whether the phenotypes observed are explained by on-target and muscle autonomous actions of NCT-503. The list of major concerns are as follows:

    1. The authors do not cite the study by Vadekeere et al. Cell Metabolism 2018 that define the KO of PHGDH in endothelial cells. This study demonstrates that serine derived from PHGDH is required to synthesize heme to preserve mitochondrial function in endothelial cells. In addition, they demonstrate that the absence of PHGDH increases mitochondrial ROS and decreases electron transport chain function. These previous studies can indicate that the worsening of the muscle phenotype in the mice treated with NCT-503 might be driven by its actions in endothelial cells, and not in muscle. In addition, it raises the question on why NCT-503 has no effect on muscle of 24-month-old mice, which have a decline in metabolic fitness.
    2. No experiments with PHGDH knockdown are performed in vitro (or ideally in vivo using muscle electroporation if possible) to confirm the specificity of NCT-503. This is validation is key in muscle, as the on-target actions of NCT-503 were mostly shown in different cancers with low and high PGHDH expression (Pacold et al). Therefore, whether there are no off-target effects of NCT-503 in muscle is still unknown and should be defined.
    3. The data showing that exogenous serine supplementation cannot override the effects of NCT-503 treatment on mitochondria could also be compatible with an off-target effect of NCT-503 in models of mitochondrial dysfunction (see point 2). If the experiments suggested in point number 2 demonstrate on-target action of NCT-502, authors should then decrease the expression of the mitochondrial transporter of serine (SFXN1) to demonstrate that specific intracellular pools of serine are needed to mitigate mitochondrial dysfunction. If the authors' conclusion is conclusion, knock-down of SFXN1 should be highly toxic in in vitro and in vivo models of mitochondrial myopathies, while having barely any effect in controls (similarly to serine deprivation in vitro being almost harmless).
    4. The authors list SFXN1 in the antibodies used in the paper, but I could not find any data. Are the protein levels of SFXN1 changed in mice with mitochondrial myopathies? Do they strongly correlate with PHGDH expression?
    5. Serine tracing experiments would be required to conclude that serine is needed for phospholipid synthesis. Otherwise, the defects observed can just be downstream of mitochondrial dysfunction and ISRmt activation.
    6. Authors use isolated mitochondria from muscle to determine whether OPA1 processing is changed in mice with mitochondrial myopathy and treated with NCT-503. The blots show higher total OPA1 content per mitochondria in the group with the greatest mitochondrial dysfunction, depolarization, and ROS production (myopathy + NCT-503). These data strongly suggest that dysfunctional mitochondrial from dysfunctional fibers do not survive the isolation procedure, showing just OPA1 content of resilient mitochondria surviving isolation. Indeed, heme depletion (as expected from PHGDH inhibition Vadekeere et al. Cell Metabolism 2018) and OPA1 processing catalyzed by OMA1 activation by ROS and depolarization can both activate the mitochondria ISR, via HRI. Therefore, authors should analyze OPA1 processing in total lysates to include mitochondria from dysfunctional fibers. Maybe increased OMA1 activity as a result of increased mitochondrial ROS and depolarization could explain the exacerbation of the mitochondrial ISR induced by NCT-503 treatment.

    Significance

    This study is significant, as it aims to understand the pathways that mediate adaptation to mitochondrial myopathies and such knowledge is necessary to find novel therapeutic targets. It could be of high interest to basic researchers studying metabolism and mitochondrial regulation, as well as to clinicians treating mitochondrial diseases.

  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

    Experiments reported in this manuscript indicate that NCT-530- an inhibitor of the de novo serine biosynthetic enzyme PHGDH- worsens mitochondrial pathology.

    Systemic administration of NCT-503 decreased serine levels only in Deletor mice- ubiquitously expressing a homologous dominant patient mutation in mitochondrial twinkle helicase.

    NCT-503 treatment induced a further increase of the mitochondrial integrated stress response in Deletor mice. Moreover, the metabolic profile and lipid balance was further modified by NCT-503 administration to Deletor mice. The relevance of NCT-503 treatment was finally evaluated in cellular systems exposed with different treatments inducing mitochondrial insults.

    Comments:

    1. NCT-530 specificity should be confirmed by reducing PHGDH levels by either siRNA or even better by CRISPR-Cas9-mediated gene deletion.
    2. If the serine biosynthetic pathway is causally relevant to worsen the mitochondrial pathology, supplementing Deletor mice with a serine-rich diet or intraperitoneal injection of serine is expected to improve pathology.

    Significance

    Limited significance in the field.

  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

    In this manuscript the authors describe interesting results regarding amino acid metabolism under conditions of mitochondrial stress. The authors used a selective PHGDH inhibitor (compound NCT503), and document that de novo serine synthesis is essential to sustain phospholipid biosynthesis, redox homeostasis, mitochondrial function, and mitochondrial protein synthesis in Deletor mice and in cell culture under mitochondrial stress. Interestingly, serine supplementation does not rescue those metabolic alterations, indicating a specific mitochondrial stress-dependent mechanism of serine utilization by the cells.
    In addition, the authors attempted to explore the role of serine in phospholipid biosynthesis under in vivo and ex vivo mitochondrial stress, and this is the weakest part of the manuscript. The specific reduction of mitochondrial PE using the PHGDH is interesting, and it can provide some clues into how mitochondrial membranes are synthetized under mitochondrial stress, but additional experimental approaches should be incorporated to generate a more robust body of evidence.
    In summary, the manuscript should be improved by using additional experimental approaches. In addition, genetic intervention through PHGDH gain or loss of function can help to elucidate the molecular mechanisms. The authors could also incorporate metabolomic studies using C13 serine or C13 glucose supplemented culture medium in order to differentiate carbons that come from extracellular serine or from de novo synthesis.

    Additional questions:

    Is PHDG subcellular localization modulated under mitochondrial stress conditions?

    Is the extracellular serine uptake impaired under mitochondrial stress?

    ATF4 is a transcription factor that regulates amino acid transport. In this connection, is ATF4-dependent serine transporters modulated under mitochondrial stress or under PHGDH treatment?

    Regarding the alteration in redox homeostasis, mitochondrial function and mitochondrial protein synthesis. Is the alteration of phospholipid synthesis upstream of all of those mitochondrial alterations?

    Additionally, in my opinion the results should be reorganized, because in the current format the manuscript is fragmented, and several panels lose the symmetry.

    Major concerns.

    Figure 1

    1. The authors should incorporate the Deletor mice amino acid levels in muscle. How does the PHGDH inhibitor treatment modulate the other non-essential amino acids?
    2. The authors should incorporate the data of WT + vehicle and wt + NCT503 mice in figures 1C, 1D, 1F and 1G, in order to compare properly the effects of PHGDH inhibitor
    3. Figure 1D is mislabeled
    4. Figure 1H. Are there any effect of PHGDH in the mtDNA of WT mice?

    The author might incorporate this information, to show properly that mitochondrial stress led to dependence of serine to sustain muscle homeostasis

    Figure 2

    Interestingly the authors observe a decrease in total mitochondrial lipids content, and an increase in mitochondrial PE content in Deletor mice compared to WT mice. These results suggest an alteration in the phospholipids flux between mitochondria and endoplasmic reticulum in this model of mitochondrial disease. Moreover, PHGDH treatment appears to be able to rescue this alteration. Some question related this issue:
    What is the expression of genes involved in the balance PC, PE, PS?, Are the PSS1, PSS2, PSD and PEMT expression altered?
    Regarding the phospholipid synthesis. Is mitochondria or endoplasmic reticulum ultrastructure altered in Deletor mice muscle?
    The authors should explain the possible mechanisms.

    Figure 3

    Based on the metabolomic studies, the authors propose a time-dependent decrease in PSTA1 and phosphoserine in cells under mitochondrial stress (Figure 3D). To elucidated the direct role of PHGDH, the authors should analyze the phosphoserine and different phospholipid (described in figure 2E) in presence of PHGDH inhibitor. This will help to understand the link between the endogenous serine synthesis and mitochondrial PE accumulation.
    Figure 3H shows a decrease in phosphoserine in the presence of PHGDH inhibitor but this figure is asymmetric compared to figure 3I. Can the authors use another experimental approach to detect the specific mitochondria phospholipid levels (used in Figure 2 for instance)?
    The authors should incorporate mitochondrial PE analysis in figure 3 to link the cellular studies described in this figure with the studies done in Deletor mice muscle.
    Figure 3 I-K. The authors suggest an alteration in glutathione redox state and a further increase in mitochondrial superoxide production in cells treated with the PHGDH inhibition under mitochondrial stress. What are the total glutathione levels under these conditions? Could GSH regeneration improves the mitochondrial function and mitochondrial protein synthesis? Is extracellular serine able to rescue the reduced glutathione levels?

    Minor concerns.

    Figure 3B,C does not show the statistical analysis so please incorporate this information.

    Include quantification of figure 3 E and supplementary figure S3E.

    figure S3E. Improve flow cytometry histogram, the cell population data values cannot be observed.

    Some methods and primers included in the material and methods section are no used in the manuscript.

    Significance

    In this manuscript the authors describe interesting results regarding amino acid metabolism under conditions of mitochondrial stress. The authors used a selective PHGDH inhibitor (compound NCT503), and document that de novo serine synthesis is essential to sustain phospholipid biosynthesis, redox homeostasis, mitochondrial function, and mitochondrial protein synthesis in Deletor mice and in cell culture under mitochondrial stress. Interestingly, serine supplementation does not rescue those metabolic alterations, indicating a specific mitochondrial stress-dependent mechanism of serine utilization by the cells.

    This data will be relevant to better understand the connection between alterations in mitochondrial function and amino acid metabolism in cells, and in organisms.

  5. Version published to 10.1101/2023.03.23.533952v1 on bioRxiv