MARBL: A Live-Cell Method for Profiling Bioenergetic Heterogeneity by Noncanonical Methionine Labeling of the Cell Surface Proteome

Reviewer #1 (Public review):

The idea behind this paper is to have an alternate, reliable and quantitative approach to assess cell-cell metabolic heterogeneity. This study tries to achieve that using translationally-coupled energetic responses to metabolic stress. This is interesting because, in general, most quantitative measurements of metabolic outputs are 'bulk' and average for many cells. To overcome this, many recent studies use some read-outs of translation (presuming that translation is the single major energy sink in cells - however, this is objectively correct only in rapidly proliferating cells). That said, the authors take an interesting approach - to use two clickable methionine analogs, and assess baseline vs metabolically coupled translation within the same cell.

The highlight is the methodology development where two distinct, clickable CMAs are used (to replace methionine in proteins). The labelling and approach are clever, and can be useful if carefully used. But there is going to be a challenge in using this, since the depletion of methionine itself (required for labelling), and a bias towards incorporation in some proteins (because these reagents are not highly permeable) will make it challenging to obtain precise metabolic state information - which otherwise can be obtained directly and far more precisely using a combination of other methods (ATP/flux measurement, respiratory capacity, translation rates etc). I therefore only make broader comments in my review below - to help structure this study better, clearly identify key limitations (and there are several that are clearly seen), and better clarify what MARBL may be useful for.

(1) These reagents used for MARBL are not highly cell permeable/transported into cells, and largely work on the surface proteome. Which means the measurements related to changes in translation are indirect - quantified based on changes on the surface proteome (seen with labelling), and not the entire proteome.

(2) A major limitation - which will confound any interpretation- is the need to use methionine-free media; this can be a problem beyond protein synthesis/met incorporation since this will almost instantly deplete SAM pools in cells. There is little data provided on the impact of using this approach on SAM pools (time kinetics, how quickly SAM pools are affected, how much of the impact on metabolism comes from purely that, etc).

This is important to establish because (i) of the continuous, very high flux of SAM -> SAH (eg. in the folate pathway, other methylations), and a constant need for SAM synthesis from methionine. This information will set the limits of capabilities of this method, as well as help delineate how much you can interpret results related to metabolic states between compared cells/states, etc. The labelling process is ~2 hours, while effects on SAM can be seen within minutes of methionine starvation in media, in metabolically active cells.

(3) What is the effect on overall adenylate charge/ATP due to shifting to methionine-free media + label addition? How does it vary between the cells tested (suspension vs adherent)? This should be established before data related to Fig. 2. Does it correlate with extent of AHA incorporation?

The primary conclusion that this method suitably reflects overall changes in energetics comes from the titration of 2DG/glycolytic inhibition.

(4) Relatedly, if this label incorporation experiment is carried out (for ~2 hrs), and subsequently there is a washout/replacement with fresh, methionine-supplemented medium, (how quickly) do the cells recover and restore their energetic allocations?

(5) One possible advantage of a system like this can be to address questions in single cells/study cell metabolic heterogeneity. However, these are best done if the attaching moiety has a (selective) fluorescence increase and/or other read-out that can be quantitatively obtained at a single cell level. Largely, using AHA or HPG effectively only leads to bulk estimates (which can be sub-sorted towards single-cell estimates indirectly). This means that this method cannot really be used to study cell-cell metabolic heterogeneity effectively - compared to far simpler approaches, for example using a mitochondrial potentiometric dye with high fluorescence, or reporters for glycolytic activity, etc. This would also be related to Figure 5 - at best, this approach may complement existing approaches towards identifying heterogeneous sub-populations of cells.

However, I do agree that MARBL is flexible, stable, and can be internally normalised and used through flow-based platforms. It can supplement existing approaches to perturb bioenergetics, and also supplement existing approaches to understand metabolic state in live cells, particularly in suspension cells.

Reviewer #2 (Public review):

Summary:

Delacruz et al. describe a new method, called "MARBL" (Methionine Analogues for Ratiometric Bioenergetics in Live cells) to measure metabolic activity in single cells. The concept is similar to the SCENITH (anti-puromycin flow cytometry) assay to measure energy metabolism by measuring protein translation activity, yet offers, in theory, two advantages: 1) it keeps cells alive for downstream biological assays and 2) it is a ratiometric measurement, measuring both baseline translation and translation in the presence of metabolic inhibitors to correct for inherent cell-to-cell translation differences.

Specifically, this method takes advantage of two click-chemistry-active methionine analogs, and then clicks fluorophores onto newly-synthesized surface proteins that have incorporated these analogs to measure translational activity. One methionine analog is given to cells for 2-4 hours to measure baseline translational activity, then metabolism is blocked using 2-deoxyglucose and oligomycin and the second methionine analog given to measure "metabolically-linked" translation activity. The authors establish this technique and show that mouse T cells polarized as pathogenic Th17 cells are more translationally active ("resilient") compared to non-pathogenic Th17 cells, and when sorted, the resilient cells produce more interferon-gamma. This latter finding requires live cells after the metabolic measurement assay, showcasing findings that are inaccessible to the SCENITH assay.

Strengths:

The approach used is conceptually clever. It is appealing to measure metabolic/translational activity and to then be able to carry out further assays on sorted cell populations with different degrees of metabolic activity. This would indeed represent a useful advance.

Weaknesses:

In principle, one key benefit of this technique is that cells can be used for biological assays after the metabolic measurement. Indeed, this would represent a valuable tool in the field.

However, in this technique, cells are subjected to methionine deprivation, addition of non-natural methionine analogs, click chemistry, and high doses of toxic metabolic inhibitors 2-deoxyglucose and oligomycin. Indeed, the authors show in Figure S5F that 1/3 more of the post-MARBL cells die relative to cells not subject to this technique (60% viability in unclicked control, 40% in MARBL-measured cells). This data suggests that cells after this technique may be stressed and not reflective of the biological function of unmanipulated cells. More controls on viability and cell function (e.g. cytokine production) at more time points after the MARBL assay would have been valuable to address this issue.

Another weakness of the paper is limited benchmarking against established metabolic assays in the field. The main assays used currently in the field are SCENITH and Seahorse. The authors do not compare their findings to SCENITH. They do compare their results to Seahorse, but the data shown don't address the key question: how does energy production measured by Seahorse, say in unmanipulated vs 2dg+oligomycin-treated cells, compare to the MARBL measurement? (Instead, they show a calculated "glucose dependence" metric in cells subjected to low vs high inhibitor dose, not showing the underlying data or cells that didn't receive an inhibitor).

Author response:

We are grateful to the reviewers and eLife editors for providing thoughtful and constructive commentary on our manuscript. The major concerns raised during review center on (A) the broader impact of methionine depletion on cellular metabolism beyond protein synthesis, including potential effects on cellular bioenergetics, and B) the extent to which assay conditions may induce cellular stress that influences downstream functional measurements. We will address both points during the revision period through experiments we have outlined below, most of which are already underway. 

Related to (A), we will define the metabolic consequences of methionine deprivation on cellular metabolism with greater precision and are currently optimizing a third non-canonical alkyne amino acid, β-ethynylserine (β-ES), for labeling surface-exposed proteins in living cells. β-ES is a clickable threonine analogue that is efficiently incorporated into the proteome in the presence of physiological threonine and therefore does not require metabolic deprivation (1). We are in the process of optimizing β-ES for the MARBL workflow as a substitute for homopropargylglycine (HPG), which reduces the duration of methionine deprivation from six hours to two hours. Thus, β-ES eliminates the SAM-depletion concern for the baseline-translation half of the MARBL ratiometric measurement and constitutes a genuine methodological advance beyond the original submission. 

Related to (B), we plan to generate additional data benchmarking MARBL against Seahorse and other established assays for measuring cellular bioenergetics, while also performing phenotypic and functional cellular characterization at key stages of the workflow. Experiments that are planned during the revision period are described in our point-by-point response below.

Public Reviews:

Reviewer #1 (Public review):

The idea behind this paper is to have an alternate, reliable and quantitative approach to assess cell-cell metabolic heterogeneity. This study tries to achieve that using translationally-coupled energetic responses to metabolic stress. This is interesting because, in general, most quantitative measurements of metabolic outputs are 'bulk' and average for many cells. To overcome this, many recent studies use some read-outs of translation (presuming that translation is the single major energy sink in cells - however, this is objectively correct only in rapidly proliferating cells). That said, the authors take an interesting approach - to use two clickable methionine analogs, and assess baseline vs metabolically coupled translation within the same cell.

The highlight is the methodology development where two distinct, clickable CMAs are used (to replace methionine in proteins). The labelling and approach are clever, and can be useful if carefully used. But there is going to be a challenge in using this, since the depletion of methionine itself (required for labelling), and a bias towards incorporation in some proteins (because these reagents are not highly permeable) will make it challenging to obtain precise metabolic state information - which otherwise can be obtained directly and far more precisely using a combination of other methods (ATP/flux measurement, respiratory capacity, translation rates etc). I therefore only make broader comments in my review below - to help structure this study better, clearly identify key limitations (and there are several that are clearly seen), and better clarify what MARBL may be useful for.

(1) These reagents used for MARBL are not highly cell permeable/transported into cells, and largely work on the surface proteome. Which means the measurements related to changes in translation are indirect - quantified based on changes on the surface proteome (seen with labelling), and not the entire proteome.

We appreciate the reviewer’s careful consideration of the MARBL labeling strategy and the opportunity to clarify an important feature of the method. We interpret this comment to mean that the detection reagents used in MARBL are not cell-permeable, rather than that the methionine analogues AHA and HPG are inefficiently transported into cells. HPG and AHA are transported through the ubiquitously expressed sodium-dependent neutral amino acid transporter SLC1A5 (2), and are incorporated into the proteome by intracellular translational machinery. The bias identified by the reviewer arises from the click detection step, not from metabolic labeling. We agree with the reviewer that MARBL measures the appearance of a subset of newly synthesized proteins where the amino acid analogue is accessible at the cell surface rather than the nascent production of the entire proteome. However, this is an intentional feature of MARBL and represents a key technical advance. It enables MARBL to generate a translation-dependent signal in live cells without requiring destructive interventions like fixation and permeabilization to access the entire proteome, as is required for other approaches such as BONCAT, THRONCAT, CENCAT, and SCENITH (1,3–5). Importantly, we empirically demonstrate that the surfaceaccessible fraction of the nascent proteome provides sufficient signal for quantitative measurements of translation that respond predictably to inhibition of protein synthesis as well as to metabolic perturbations that alter cellular energetics (Main Figure 1D-G, 2C). MARBL is also readily compatible with fixation and permeabilization, allowing the same labeling strategy to quantify analogue incorporation when a measure of total protein synthesis is desired and there is no need for the recovery of live cells. In the revised manuscript, we will include data directly comparing MARBL surface labeling with total nascent protein synthesis measured following fixation and permeabilization to show that these signals track with one another.

(2) A major limitation - which will confound any interpretation- is the need to use methionine-free media; this can be a problem beyond protein synthesis/met incorporation since this will almost instantly deplete SAM pools in cells. There is little data provided on the impact of using this approach on SAM pools (time kinetics, how quickly SAM pools are affected, how much of the impact on metabolism comes from purely that, etc).

This is important to establish because (i) of the continuous, very high flux of SAM -> SAH (eg. in the folate pathway, other methylations), and a constant need for SAM synthesis from methionine. This information will set the limits of capabilities of this method, as well as help delineate how much you can interpret results related to metabolic states between compared cells/states, etc. The labelling process is ~2 hours, while effects on SAM can be seen within minutes of methionine starvation in media, in metabolically active cells.

The reviewer is correct in pointing out that cellular SAM pools are dynamic and adapt rapidly to methionine availability within hours of its deprivation or add-back (6,7). In the current version of the manuscript, we observe an ~100-fold decrease in intracellular SAM with 2 hours of methionine-free incubation in Jurkat cells, corresponding to the same time window over which we label with AHA in MARBL (currently presented as a min-max scaled heat map in Supplementary Figure 1A-B). We agree this needs to be presented more clearly as a potential limitation and quantified more explicitly. To address this, we plan to (i) reformat the existing SAM/ SAH/ methionine kinetics as absolute peak areas (Author response image 1A-C), (ii) add a novel dual-isotope simultaneous measurement of ATP turnover and SAM turnover across a panel of human cell lines to define the extent to which changes in SAM metabolism during the MARBL labeling workflow are likely to influence cellular energy demand (using ¹³C₅-methionine and H₂¹⁸O), and (iii) add β-ES as a threonine-based alternative to HPG for measuring baseline translation that substantially reduces the duration of methionine deprivation required for MARBL. β-ES is a clickable, alkyne-modified threonine analogue that is incorporated into newly synthesized proteins by mammalian cells in complete medium (1,4). Compared to methionine, which is a major part of the methionine cycle and 1-carbon metabolism that are required for cell proliferation and survival, threonine plays a less extensive role in mammalian intermediary metabolism. We plan to optimize and validate β-ESàAHA as well as AHAàβ-ES dual labeling and determine whether this modified MARBL workflow preserves the dynamic range and metabolic responsiveness of the HPGàAHA approach as in the original submitted manuscript. These experiments are currently underway, and we have already found that β-ES produces robust surface signal above background within 2-4 hours of labeling using our established MARBL protocol in the presence of normal threonine levels (Author response image 2A-D). We are currently optimizing the surface labeling protocol to further reduce the duration of baseline β-ES incubation in the revised manuscript. We are unable to eliminate methionine depletion entirely from the MARBL workflow. This is because endogenous methionyl-tRNA synthetases possess drastically higher affinities for canonical methionine (8–10), which prevents the incorporation of methionine analogues into proteins. However, these planned experiments will better define the impact of methionine limitation while also providing an alternative MARBL implementation that restricts methionine withdrawal to the shorter AHA-labeling window.

Author response image 1.

Dynamics of intracellular methionine and its derived metabolites. (A-C) LC-MS/MS raw peak areas in Jurkat cells of methionine (A), SAM (B), and SAH (C) after 2, 4, 6, or 8 hours of incubation in Met-free RPMI media supplemented with Met, AHA, or HPG. Abbreviations: AHA = Azidohomoalanine, HPG = Homopropargylglycine, Met = Methionine, SAM = Sadenosylmethionine, SAH = S-adenosyl-L-homocysteine. Statistics: Graphs display mean ± SD (A-C).

Author response image 2.

Extension of live surface labeling protocol using clickable threonine analogue to monitor surface translation. (A) Chemical structures for threonine and its clickable analogue β-ES. (B-C) Representative flow cytometric histograms (B) and corresponding gMFIs of extracellular Azd-647 signal after 1, 2, or 4 hours of β-ES incorporation. (D) gMFIs of Alk-647 (corresponding to AHA) or Azd-647 (corresponding to HPG or β-ES) after 1, 2, or 4 hours of incorporation. Abbreviations: β-ES = β-Ethynylserine, AHA = Azidohomoalanine, HPG = Homopropargylglycine, Met = Methionine, CHX = Cycloheximide, gMFI = Geometric mean fluorescence intensity. Statistics: Graphs display mean ± SD (C-D).

(3) What is the effect on overall adenylate charge/ATP due to shifting to methionine-free media + label addition? How does it vary between the cells tested (suspension vs adherent)? This should be established before data related to Fig. 2. Does it correlate with extent of AHA incorporation?

The primary conclusion that this method suitably reflects overall changes in energetics comes from the titration of 2DG/glycolytic inhibition.

We thank the reviewer for raising these important points, which are related to point #2 above, as both ask how methionine-free labeling conditions alter cellular metabolism. We agree that a more comprehensive set of experiments exploring methionine deprivation will better delimit interpretations that can be drawn from a MARBL assay related to cellular energetics. As described in our response to point #2, we will measure baseline adenylate energy charge across a panel of adherent and suspension cell lines under methionine-replete, methionine-free, and methionine-free +AHA/HPG labeling conditions and determine its relationship to AHA incorporation. We will also perform analogous experiments during β-ES supplementation to determine how this alternative labeling condition affects cellular energetic state and β-ES incorporation. 

We also wish to clarify that the adenylate energy charge measurements in Main Figure 2D and the measurements of newly synthesized ATP by H₂¹⁸O turnover in Main Figure 2E were performed in methionine-free media supplemented with either AHA or methionine. We apologize that this was not made clear in the figure key and legend. In addition, we inadvertently covered the methionine-replete control in the original figure with the key. This has been corrected in Author response image 3A-B and will be updated in the revised manuscript. In Jurkat cells, both adenylate energy charge and ATP synthesis are similar in the presence and absence of methionine.

We note, however, that cells with intact energy-generating systems maintain adenylate energy charge within a relatively narrow range (11). Given this buffering capacity, we do not expect methionine deprivation to produce a large change in adenylate energy charge in most cell types. 

Author response image 3.

Validation of surface translation as a readout of cellular energetics in methionine-free and AHA-supplemented media. (A) Correlation between changes in adenylate energy charge versus normalized Alk-647 Alkyne flow cytometric signal in methionine-free RPMI supplemented with either AHA or Met (Pearson’s R = 0.9044, Pearson’s R² = 0.8179, p = 0.0052). (B) Correlation between changes in newly synthesized ATP versus normalized Alk-647 Alkyne flow cytometric signal in methionine-free RPMI supplemented with either AHA or Met (Pearson’s R = 0.8854, Pearson’s R² = 0.7840, p = 0.008). Abbreviations: AHA = Azidohomoalanine, HPG = Homopropargylglycine, Met = Methionine, CHX = Cycloheximide, gMFI = geometric mean fluorescence intensity, 2DG = 2-Deoxy-D-Glucose, Omy = Oligomycin A. Statistics: Pearson’s correlation coefficient was used for correlation and significance (A-B). The Met + DMSO condition was excluded from the Pearson correlation analysis. Graphs display mean ± SD (A-B).

(4) Relatedly, if this label incorporation experiment is carried out (for ~2 hrs), and subsequently there is a washout/replacement with fresh, methionine-supplemented medium, (how quickly) do the cells recover and restore their energetic allocations?

If we observe substantial changes in adenylate energy charge associated with methionine restriction, as assessed in the experiments planned in Points #2 and #3 above, we will perform methionine add-back experiments to define the kinetics of recovery. In addition, our plan to provide an alternative workflow that replaces HPG with β-ES will reduce the total duration of methionine depletion and further address this concern.

(5) One possible advantage of a system like this can be to address questions in single cells/study cell metabolic heterogeneity. However, these are best done if the attaching moiety has a (selective) fluorescence increase and/or other read-out that can be quantitatively obtained at a single cell level. Largely, using AHA or HPG effectively only leads to bulk estimates (which can be sub-sorted towards single-cell estimates indirectly). This means that this method cannot really be used to study cell-cell metabolic heterogeneity effectively - compared to far simpler approaches, for example using a mitochondrial potentiometric dye with high fluorescence, or reporters for glycolytic activity, etc. This would also be related to Figure 5 - at best, this approach may complement existing approaches towards identifying heterogeneous sub-populations of cells. 

However, I do agree that MARBL is flexible, stable, and can be internally normalised and used through flowbased platforms. It can supplement existing approaches to perturb bioenergetics, and also supplement existing approaches to understand metabolic state in live cells, particularly in suspension cells.

We thank the reviewer for raising this thoughtful point, which we will address with textual changes in the discussion. We note that MARBL provides a quantitative single-cell readout when flow cytometry is used as the analytical endpoint, because the dual-color labeling scheme provides an internal control for baseline translation that accounts for variation in protein synthesis rates independent from cellular energetics. This permits a single metabolic resilience index to be calculated per cell and interpreted either at single-cell resolution or after grouping cells into sub-populations as a bulk estimate. Both approaches have utility depending on the biological question and intended downstream application. We agree with the reviewer that MARBL can be paired with other reporters for metabolism to obtain a more comprehensive understanding of bioenergetic heterogeneity and appreciate the reviewer’s recognition of its value as a complementary approach for studying metabolic state in live cells.

Reviewer #2 (Public review):

Summary: 

Delacruz et al. describe a new method, called “MARBL” (Methionine Analogues for Ratiometric Bioenergetics in Live cells) to measure metabolic activity in single cells. The concept is similar to the SCENITH (anti-puromycin flow cytometry) assay to measure energy metabolism by measuring protein translation activity, yet offers, in theory, two advantages: 1) it keeps cells alive for downstream biological assays and 2) it is a ratiometric measurement, measuring both baseline translation and translation in the presence of metabolic inhibitors to correct for inherent cell-to-cell translation differences.

Specifically, this method takes advantage of two click-chemistry-active methionine analogs, and then clicks fluorophores onto newly-synthesized surface proteins that have incorporated these analogs to measure translational activity. One methionine analog is given to cells for 2-4 hours to measure baseline translational activity, then metabolism is blocked using 2-deoxyglucose and oligomycin and the second methionine analog given to measure “metabolically-linked” translation activity. The authors establish this technique and show that mouse T cells polarized as pathogenic Th17 cells are more translationally active (“resilient”) compared to nonpathogenic Th17 cells, and when sorted, the resilient cells produce more interferon-gamma. This latter finding requires live cells after the metabolic measurement assay, showcasing findings that are inaccessible to the SCENITH assay.

Strengths:

The approach used is conceptually clever. It is appealing to measure metabolic/translational activity and to then be able to carry out further assays on sorted cell populations with different degrees of metabolic activity. This would indeed represent a useful advance.

We thank the reviewer for recognizing the conceptual strengths of MARBL and its potential to enable downstream analysis of live cell populations with distinct metabolic states.

Weaknesses:

In principle, one key benefit of this technique is that cells can be used for biological assays after the metabolic measurement. Indeed, this would represent a valuable tool in the field.

However, in this technique, cells are subjected to methionine deprivation, addition of non-natural methionine analogs, click chemistry, and high doses of toxic metabolic inhibitors 2-deoxyglucose and oligomycin. Indeed, the authors show in Figure S5F that 1/3 more of the post-MARBL cells die relative to cells not subject to this technique (60% viability in unclicked control, 40% in MARBL-measured cells). This data suggests that cells after this technique may be stressed and not reflective of the biological function of unmanipulated cells. More controls on viability and cell function (e.g. cytokine production) at more time points after the MARBL assay would have been valuable to address this issue.

The reviewer makes important points regarding the conditions needed for the workflow of a MARBL assay that may impact cellular fitness. Perturbational methods, particularly techniques that interrogate bioenergetics, inherently require media-based or pharmacologically induced stress to evaluate cellular responses. Still, we agree that comprehensively profiling the fitness of cells following a MARBL assay and sorting is important since our technique aims to link cellular bioenergetics to functional outcomes. First, we would like to highlight that the MARBL-processed pathogenic and non-pathogenic TH17 cells depicted in Main Figure 5 were rested overnight after Fluorescence-Activated Cell Sorting (FACS) in complete RPMI medium prior to the restimulation assay. This is in line with standard practice to allow cells to recover from the shear stress of sorting prior to subsequent experiments (12,13). In addition, both pathogenic and non-pathogenic TH17 cells maintained the expression of lineage-defining transcription factors throughout the MARBL workflow, as shown by analyzing rested cells stained with antibodies for T-bet (expressed by pathogenic TH17) as well as RORγt (expressed in both cell types) by flow cytometry (Author response image 4A-C). To address this in the revised manuscript, we plan to repeat our non-pathogenic and pathogenic TH17 dual-MARBL and sorting experiment (Main Figure 5A-B) and rest sorted cells in RPMI with IL-2 for longer periods of time (24 or 48 hours) before restimulation for viability and cytokine analysis. These controls will provide more information on cellular fitness throughout the MARBL workflow, and we appreciate the reviewer’s suggestion.

Author response image 4.

Expression of lineage-defining transcription factors is maintained post-MARBL processing and fluorescence-activated cell sorting (FACS). (A) Experimental schematic. Ex vivo differentiated pTH17 and npTH17 cells were stained with CD45.2 antibodies conjugated to different color fluorophores, processed via MARBL, mixed at a 1:1 ratio, sorted, and then re-cultured in IL-2-supplemented RPMI media. After resting overnight, the expression of RORγt and T-bet were evaluated by intracellular staining and flow cytometry, distinguishing pTH17 from npTH17 cells based on prior CD45.2 staining. (B-C) Frequency of RORγt (B) and T-bet (C) positivity in murine Th17 cells post-MARBL processing, sorting, and overnight rest in IL-2-supplemented media. Abbreviations: npTH17 = non-pathogenic TH17, pTH17 = pathogenic TH17, HPG = Homopropargylglycine, AHA = Azidohomoalanine, Met = Methionine, 2DG = 2-Deoxy-D-Glucose, Omy = Oligomycin A. Statistics: Graphs display mean ± SD (B-C).

Another weakness of the paper is limited benchmarking against established metabolic assays in the field. The main assays used currently in the field are SCENITH and Seahorse. The authors do not compare their findings to SCENITH. They do compare their results to Seahorse, but the data shown don't address the key question: how does energy production measured by Seahorse, say in unmanipulated vs 2dg+oligomycin-treated cells, compare to the MARBL measurement? (Instead, they show a calculated "glucose dependence" metric in cells subjected to low vs high inhibitor dose, not showing the underlying data or cells that didn't receive an inhibitor).

We appreciate the opportunity to perform additional benchmarking to define how the MARBL signal compares to existing methods for measuring cellular energetics. For our initial validation, we chose to benchmark the single-color MARBL signal against direct LC-MS/MS quantification of adenylate energy charge as well as newly synthesized ATP (Main Figure 2A-E). Although Seahorse and SCENITH are widely used standards in the field, these assays still provide indirect measures of cellular energetics, whereas LC-MS/MS quantifies the high-energy nucleotide pools that dictate cellular energy status. Both the adenylate energy charge and newly synthesized ATP decreased in response to increasing concentrations of 2-deoxy-D-glucose (2DG) and oligomycin A (Omy) treatments that impair ATP regeneration, and this energetic response displayed a linear relationship with the AHA click signal (Main Figure 2D-E). Nevertheless, we agree that additional benchmarking suggested by the reviewer will strengthen the methodological foundation of MARBL and help users understand how MARBL measurements relate to those obtained using more established metabolic assays. For this purpose, it is important to account for differences in what each assay measures. Seahorse resolves oxidative and glycolytic activity through simultaneous measurements of OCR and ECAR, whereas MARBL (as well as SCENITH and CENCAT) integrate the energetic contributions of these pathways into a single translation-dependent readout. This is the reason why we compared MARBL with Seahorse using the glucose-dependence calculation employed by SCENITH in the current version of the manuscript (Main Figure 2FH) (5). As suggested by the reviewer, we will assess how OCR and ECAR measured by Seahorse vary relative to the MARBL signal across different oligomycin and 2-deoxyglucose treatment conditions, providing a more comprehensive view of the bioenergetic responses captured by MARBL. 

References

(1) Ignacio BJ, Dijkstra J, Mora N, Slot EFJ, van Weijsten MJ, Storkebaum E, et al. THRONCAT: metabolic labeling of newly synthesized proteins using a bioorthogonal threonine analog. Nat Commun. 2023 Jun 8;14(1):3367. doi:10.1038/s41467-023-39063-7 PubMed PMID: 37291115; PubMed Central PMCID: PMC10250548.

(2) Pelgrom LR, Davis GM, O’Shaughnessy S, Wezenberg EJM, Van Kasteren SI, Finlay DK, et al. QUAS-R: An SLC1A5-mediated glutamine uptake assay with single-cell resolution reveals metabolic heterogeneity with immune populations. Cell Reports. 2023 Aug 29;42(8):112828. doi:10.1016/j.celrep.2023.112828

(3) Dieterich DC, Link AJ, Graumann J, Tirrell DA, Schuman EM. Selective identification of newly synthesized proteins in mammalian cells using bioorthogonal noncanonical amino acid tagging (BONCAT). Proceedings of the National Academy of Sciences. 2006 Jun 20;103(25):9482–7. doi:10.1073/pnas.0601637103 PubMed PMID: 16769897.

(4) Vrieling F, van der Zande HJP, Naus B, Smeehuijzen L, van Heck JIP, Ignacio BJ, et al. CENCAT enables immunometabolic profiling by measuring protein synthesis via bioorthogonal noncanonical amino acid tagging. Cell Rep Methods. 2024 Oct 21;4(10):100883. doi:10.1016/j.crmeth.2024.100883 PubMed PMID: 39437716; PubMed Central PMCID: PMC11573747.

(5) Argüello RJ, Combes AJ, Char R, Gigan JP, Baaziz AI, Bousiquot E, et al. SCENITH: A flow cytometry based method to functionally profile energy metabolism with single cell resolution. Cell Metab. 2020 Dec 1;32(6):1063-1075.e7. doi:10.1016/j.cmet.2020.11.007 PubMed PMID: 33264598; PubMed Central PMCID: PMC8407169.

(6) Mentch SJ, Mehrmohamadi M, Huang L, Liu X, Gupta D, Mattocks D, et al. Histone Methylation Dynamics and Gene Regulation Occur through the Sensing of One-Carbon Metabolism. Cell Metab. 2015 Nov 3;22(5):861–73. doi:10.1016/j.cmet.2015.08.024 PubMed PMID: 26411344; PubMed Central PMCID: PMC4635069.

(7) Chen Z, Chen W, Reheman Z, Jiang H, Wu J, Li X. Genetically encoded RNA-based sensors with Pepper fluorogenic aptamer. Nucleic Acids Res. 2023 Sep 8;51(16):8322–36. doi:10.1093/nar/gkad620 PubMed PMID: 37486780; PubMed Central PMCID: PMC10484673.

(8) Kiick KL, Saxon E, Tirrell DA, Bertozzi CR. Incorporation of azides into recombinant proteins for chemoselective modification by the Staudinger ligation. Proc Natl Acad Sci U S A. 2002 Jan 8;99(1):19–24. doi:10.1073/pnas.012583299 PubMed PMID: 11752401; PubMed Central PMCID: PMC117506.

(9) Beatty KE, Liu JC, Xie F, Dieterich DC, Schuman EM, Wang Q, et al. Fluorescence visualization of newly synthesized proteins in mammalian cells. Angew Chem Int Ed Engl. 2006 Nov 13;45(44):7364–7. doi:10.1002/anie.200602114 PubMed PMID: 17036290.

(10) Kiick KL, Weberskirch R, Tirrell DA. Identification of an expanded set of translationally active methionine analogues in Escherichia coli. FEBS Lett. 2001 Jul 27;502(1–2):25–30. doi:10.1016/s0014-5793(01)02657-6 PubMed PMID: 11478942.

(11) De la Fuente IM, Cortés JM, Valero E, Desroches M, Rodrigues S, Malaina I, et al. On the dynamics of the adenylate energy system: homeorhesis vs homeostasis. PLoS One. 2014;9(10):e108676. doi:10.1371/journal.pone.0108676 PubMed PMID: 25303477; PubMed Central PMCID: PMC4193753.

(12) Pollizzi KN, Patel CH, Sun IH, Oh MH, Waickman AT, Wen J, et al. mTORC1 and mTORC2 selectively regulate CD8⁺ T cell differentiation. J Clin Invest. 2015 May 1;125(5):2090–108. doi:10.1172/JCI77746 PubMed PMID: 0.

(13) Roth TL, Puig-Saus C, Yu R, Shifrut E, Carnevale J, Li PJ, et al. Reprogramming human T cell function and specificity with non-viral genome targeting. Nature. 2018 Jul;559(7714):405–9. doi:10.1038/s41586-018-03265