Spatial Partitioning of Glucose Metabolism Supports Tissue-Specific Metabolic Programs In Vivo

  1. Ian J. Gonzalez
  2. Aaron D. Wolfe
  3. Benjamin Clark
  4. Michael Hanna
  5. Gabrielle E. Geise
  6. Quishi Sun
  7. Snusha Ravikumar
  8. Anastasia Tsives
  9. Sarah E. Emerson
  10. Stephan Siebel
  11. Richard Kibbey
  12. L. Safak Yilmaz
  13. Albertha J.M. Walhout
  14. Daniel Colón-Ramos

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

Tissues exhibit metabolic heterogeneity that tailors conserved pathways to distinct physiological demands, yet how this heterogeneity is achieved in vivo remains poorly understood. Here, we use Caenorhabditis elegans to investigate tissue-specific requirements for glucose-6-phosphate isomerase (GPI-1), a conserved reversible enzyme that links glycolysis and the pentose phosphate pathway (PPP). Tissue-specific metabolic-network modeling predicted differential glycolytic and PPP flux potential across adult tissues and identified tissue-specific biases in GPI-1 reaction directionality. Genetic disruption of gpi-1 produced germline defects consistent with impaired PPP-associated anabolic metabolism and somatic defects consistent with impaired glycolysis, indicating that GPI-1 supports distinct metabolic functions across tissues. We further discovered that two GPI-1 isoforms are differentially expressed and localized: GPI-1A is broadly expressed and cytosolic, whereas GPI-1B is enriched in the germline and localizes to endoplasmic reticulum–associated compartments. Isoform-specific perturbations revealed distinct requirements for GPI-1A and GPI-1B in somatic glycolysis and reproductive physiology. These findings implicate isoform-specific subcellular localization as a possible contributor to the partition of the functions of a conserved reversible enzyme, enabling tissue-specific anabolic and catabolic metabolism in vivo .

Article activity feed

  1. Review Commons

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

    Learn more at Review Commons

    Reply to the reviewers

    __Reviewer 1 Minor Points : __

    “The authors generate and analyze pfk-1.1 and gpd-1 mutants, which help support the roles of GPI-1 in PPP and glycolysis and the resulting movement and germline phenotypes. However, these data are largely absent from the discussion, and it would help to incorporate them there.” __Response and action: __We will expand upon the role that these enzymes play in delineating tissue-specific utilization of different metabolic pathways and how they contextualize GPI-1 function in the discussion section. “The authors note that the GPI-1B endomembrane clusters are observed in some embryos. From presented images it also appears that oocytes also have heterogeneous presence of clusters. If this is correct, it would be useful for the authors to comment on potential consequences of this or is the proposed function earlier in the germline where these structures also observed in the rachis. “ __Response and action: __We thank the reviewer for this comment. We indeed observe structures that originate in the rachis of the animal beginning at the L4 stage persist through some embryos. We believe that this persistent localization may be associated with the prolonged role and need for GPI-1 throughout development. We will add to the discussion section paragraph that details these structures, expanding on the possible role of these structures across different gamete developmental stages.

    “In the Microscopy methods section a note to add a references is included "(cite HYlight paper)" __Response and action: __This typo has been corrected in line 773. “Brood size analysis, sentence fragment "We note that when these mutants are grown in conditions that vary" Response and action: This sentence fragment has been removed from line 852.

    Reviewer 2 Major Points

    “The current experiments do not separate the effect of subcellular localization from other isoform-specific properties. A central conclusion is that the different subcellular localization of GPI-1A and GPI-1B contributes causally to their functional specialization. The GPI-1B swap strain is an informative reagent, but the experiment does not isolate localization as the relevant variable. GPI-1A and GPI-1B differ in amino-acid sequence, and the swap could alter protein abundance, translation, stability, catalytic activity, dimerization, interactions with other proteins, or access to substrates. Therefore, the observation that the swap supports swimming while incompletely restoring neuronal FBP and lactate responses is consistent with an effect of localization, but is not sufficient to demonstrate it. The authors should quantify GPI-1 abundance in the same somatic cells or tissues used for the functional measurements. Calibrated fluorescence from endogenously tagged proteins would be more informative than a whole-animal Western blot, which could be dominated by expression in unrelated tissues. GPI activity should also be compared across the relevant isoform and swap strains. The localization-defective GPI-1B CCM allele could provide a valuable separation-of-function test, but the introduced substitutions could themselves affect folding or catalytic function. The authors should therefore establish that the CCM protein is expressed at a comparable level and retains GPI activity before interpreting a physiological effect as resulting from altered localization. If this allele is functional, testing at least one central metabolic or physiological phenotype, such as the neuronal HYlight or FiLa response, swimming under mitochondrial inhibition, or reproductive output, would directly address the proposed mechanism. These experiments are essential only if the authors wish to conclude that localization controls GPI-1 function. If they are not technically feasible, the title, abstract, Results, and Discussion should consistently state that isoform-specific localization is associated with, and may contribute to, the observed functional differences.” __Response and action: __The reviewer correctly points out that, although we identify clear differences in the subcellular localization of the two isoforms in vivo, the sequence differences between GPI-1A and GPI-1B could also alter their intrinsic biochemical properties, independently of their localization, and thereby contribute to their distinct functional effects. To address this possibility, we collaborated with Dora Tang (Saarland University) and Anthony Hyman (EMBL) to express and purify GPI-1A and GPI-1B for direct biochemical comparison in vitro. GPI-1A could be purified in a soluble form, whereas GPI-1B exhibited pronounced aggregation under the conditions tested, preventing a reliable comparison of enzymatic activity between the two isoforms. Thus, these experiments did not allow us to determine whether the isoforms differ intrinsically in catalytic activity. We therefore agree that our current data do not conclusively establish subcellular localization as the sole causal determinant of the functional differences between GPI-1A and GPI-1B. This limitation does not alter the in vivo findings that the isoforms differ in expression, localization, and functional capacity, but it does affect how strongly we can connect these observations mechanistically. We will therefore revise the manuscript throughout to distinguish the demonstrated association between isoform-specific localization and function from the proposed causal model, and will explicitly discuss the possibility of additional isoform-specific biochemical properties in the Limitations section of the Discussion.

    • “The evidence does not currently establish germline-specific PPP metabolism or reverse GPI-1 flux. The modeling predicts reverse GPI-1 flux in the germline, and the reproductive phenotypes and metabolomics are consistent with a role for GPI-1 in anabolic metabolism. However, the metabolomics was performed using whole animals. It therefore cannot determine whether the PPP-associated metabolite changes arise in the germline, the soma, or through systemic responses to developmental delay, infertility, embryonic lethality, redox stress, or altered communication between tissues. In this regard, the authors should distinguish clearly among three different levels of evidence:
    • i) a predicted tissue-specific flux state from the network model;* ii) changes in whole-animal metabolite abundance; and iii) direct measurements of pathway flux. At present, the study provides the first two, but not the third. If the authors wish to retain a strong claim that the metabolomic phenotype originates in the gonad, tissue-resolved targeted metabolomics would be informative. One possible approach would be rapid mechanical extrusion and isolation of adult gonads, followed by targeted analysis of a prespecified set of PPP, glycolytic, nucleotide, and redox metabolites. Because dissection itself may rapidly alter metabolism, this experiment would require appropriate controls: isolated gonads, the corresponding somatic carcasses, a processed recombination control in which separated gonads and carcasses are mixed back together, and intact animals processed in parallel. The dissection medium should either be retained or tested for released metabolites, and the interval between incision and metabolic quenching should be minimized, standardized, and reported. At least three independent biological replicates would be needed, with the number of animals determined empirically from the sensitivity and reproducibility of the assay rather than fixed in advance. This experiment would localize changes in metabolite pools but would still not directly measure PPP flux. If tissue-resolved metabolomics is not feasible, the germline-specific PPP conclusion should be presented as an inference supported by modeling, reproductive phenotypes, and whole-animal metabolite measurements.” __Response and action: __We agree with the reviewer that direct tissue-resolved measurements of PPP metabolites or flux in the germline would be highly informative. However, to our knowledge, quantitative metabolomic analysis of central-carbon metabolites from isolated C. elegans tissues has not yet been established, and there is currently no genetically encoded cell-biological sensor that directly reports PPP flux in vivo. Recent studies have achieved mass-spectrometry-based proteomics from manually dissected gonads and intestines, and spatial approaches have begun to resolve broader metabolic signatures within the animal, but tissue-resolved measurements of labile glycolytic and PPP intermediates, and particularly metabolic flux, remain an important technical frontier in C. elegans. This technical limitation is one reason that our study integrates complementary approaches, including tissue-resolved metabolic-network modeling, genetics, live metabolic sensors, whole-animal metabolomics, and quantitative physiology, to address tissue-specific metabolism in vivo. We agree, however, that these approaches provide different levels of evidence and do not constitute a direct measurement of germline PPP flux. We will therefore revise the manuscript to distinguish clearly between tissue-specific flux predicted by the model, changes in whole-animal metabolite abundance, and direct measurements of metabolic flux. We will also emphasize in our edits that identifying this limitation is itself important in the context of the conceptual problem uncovered by our study. Our findings show that metabolic pathway usage and enzyme localization can differ substantially across tissues in vivo, underscoring why tools capable of measuring pathway activity with tissue and subcellular resolution are needed. By revealing these differences, the study not only advances a model for tissue-specific metabolic specialization but also defines an important methodological gap for the field. “The metabolic-network predictions require uncertainty analysis and complete computational disclosure. The metabolic modeling provides the foundation for the tissue-specific interpretation of the study. The manuscript describes the use of MERGE, CatEx, iMAT++, eFPA, relaxed biomass requirements, sink reactions, exchange constraints, a penalty for selected boundary reactions, and FVA-derived directional constraints. However, the reader cannot determine from the current description how robust the GPI-1 directionality prediction is to these choices. The authors should deposit the complete computational workflow, including: i) the single-cell or pseudo-bulk input matrices; ii) cell-type assignments and aggregation procedures; iii) the exact version of iCEL1314; iv) all modified reactions, sink reactions, exchange bounds, and hard constraints; v) CatEx thresholds and activity assignments; vi) the complete objective functions and solver settings; vii) the exchange penalty and the rationale for its value; viii) all scripts used for iMAT++, eFPA, FVA, and figure generation; ix) and the complete reaction-level input and output tables. The FVA range for GPI-1 should be reported for each tissue. It would be important to know whether reverse flux is required by the feasible solution space, strongly favored across alternative solutions, or simply permitted in one optimized solution. Sensitivity analyses should determine whether the germline prediction persists after reasonable changes to expression thresholds, exchange penalties, biomass relaxation, and other major assumptions. Throughout the manuscript, modeled flux directionality should be described as a prediction rather than evidence that tissue-specific carbon flux was measured experimentally. The biological context mismatch should also be acknowledged: the modeling is based on Day 1 adult expression profiles, whereas some of the supporting phenotypes occur during larval development or embryogenesis.” __Response and action: __We will deposit all metrics, quantitations and workflows related to these findings. We will also make sure the language in the text is clear that these are predictive models and not direct measurements of flux. “Replication, statistical analysis, and interpretation of the biosensor experiments require revision** **
    • i) The manuscript does not consistently provide the number of animals, cells, independent experiments, or biological replicates for the swimming and biosensor measurements. These values should be reported in every figure legend, together with the definition of the experimental unit. The authors should state whether animals were obtained from independent cultures, how many experimental days were included, whether acquisition and analysis were blinded, and what exclusion criteria were applied.*** **
    • ii) The HYlight and FiLa experiments contain paired and longitudinal measurements from the same animals before and after hypoxia. A one-way ANOVA does not account for within-animal pairing or genotype-by-time interactions. The data should be analyzed using an appropriate paired endpoint analysis, a repeated-measures or mixed-effects model, or a prespecified summary measurement such as change from baseline or area under the curve. Individual paired values or trajectories should be shown where practical. Multiple-comparison procedures, effect sizes, exact p values, and the meaning of the shaded regions in time-course plots should also be reported.*** **
    • iii) HYlight and FiLa are valuable reporters, but they measure FBP- and lactate-sensitive fluorescence ratios rather than flux directly. Dynamic changes are consistent with altered glycolytic activity, but the manuscript should avoid describing either reporter as a direct measurement of glycolytic flux unless this interpretation is independently validated.*** **
    • iv) The C. elegans-adapted FiLa construct also requires clearer validation and description. The authors should explain its expected response direction, expression control, and normalization procedure. Most directly, FiLa should be tested in the ldh-1(ola514) deletion background, which removes the sole C. elegans lactate dehydrogenase ortholog and has been shown to reduce the signal of an independent neuronal lactate reporter. Alternatively, the authors could use their existing pfk-1.1(ola458) knockout, which is expected to reduce glycolytic lactate production. In either case, basal and hypoxia-induced FiLa responses should be compared with wild type using paired individual-animal measurements.*” __Response and action: __We will edit the manuscript to make sure both Methods and Figure legends include all relevant data about experimental and methodological parameters as described by the reviewer. We will also make sure it is clear from our narrative that biosensor data is not a direct readout of flux. We will also include a negative control for the lactate sensor in the form ldh-1 or pfk-1.1 mutants to calibrate the sensor as suggested. *“The metabolomics analysis requires additional statistical rigor and access to the complete data The metabolomics section does not clearly state the number of biological replicates, the number or identity of excluded samples, or whether the investigators were blinded during data processing. These details are essential. The authors report normalization to the sum of targeted metabolites, selection of the detection mode with the "best intensity" when a metabolite was detected more than once, and pathway analyses using nominal p values. These procedures require additional justification. Sum normalization can generate compositional effects when mutations produce widespread metabolic changes. The criterion used to choose among detection modes should be prespecified and reported for every affected metabolite. Given the large number of metabolites tested, adjusted p values or an appropriate false-discovery-rate procedure should be provided.

    The authors should deposit: i) the raw mass-spectrometry files; ii) processed peak tables before and after normalization; iii) metabolite identities and confidence levels; iv) retention times and reference standards; v) missing-value handling; vi) QC measurements and exclusion criteria; vii) the complete differential-abundance results, including metabolites not highlighted in the figures; viii) effect sizes, nominal p values, and adjusted p values; and ix) the complete pathway-enrichment outputs. The presentation of approximately 60 highly changing and interconnected metabolites could be biasing the results towards those that form a coherent pathway narrative. The authors should explain how these metabolites were selected and make the full, unselected results available, so readers can assess whether the highlighted PPP and glycolysis patterns are representative. Finally, altered steady-state metabolite abundance should not be described as direct evidence of altered flux.”* __Response and action: __We will deposit all available raw data files and analysis and also include additional methodological and statistical details in the methods section. “The special culture conditions used for pfk-1.1 may affect the central pathway comparison**

    • pfk-1.1 knockout animals sometimes display substantial embryonic lethality and this phenotype can be suppressed by growth on newly prepared NGM plates with thicker OP50 lawns. This observation is biologically interesting, but it also creates an important potential confounding variable because the manuscript uses pfk-1.1 as the glycolysis-specific benchmark against which gpi-1 and gspd-1 phenotypes are interpreted.*** The authors should state clearly:**
    • i) which culture conditions were used for each experiment;***
    • ii) whether all genotypes were grown and assayed simultaneously under identical conditions;***
    • and iii) how frequently pfk-1.1 embryonic lethality was observed under standard versus optimized conditions.*** The complete results obtained under the different dietary or plate conditions should be provided. Ideally, the reproductive comparison should be repeated under at least two clearly defined conditions for all genotypes to establish which conclusions are diet-dependent. If diet-dependent, the text should acknowledge that the apparent distinction between pfk-1.1 and gpi-1 reproductive phenotypes is conditional rather than absolute.** The same food and plate conditions should be used to document the swimming, biosensor, developmental, and metabolomics experiments.” __Response and action: __We will edit the Methods section to clearly outline which experiments utilized which pfk-1.1 culture conditions. In general, save for the culturing conditions used to rear animals for metabolomics analysis (due to requirements of those analyses, which we will now explain in the Methods), the same conditions were used for all animals, for all genotypes and for all experiments. “The interpretation of the single-isoform reproductive phenotypes should be corrected. The interpretation of Figure 8G-H does not appear to follow from the genotypes. GPI-1A-only animals lack GPI-1B. Their delayed onset of egg production supports a contribution of GPI-1B to normal reproductive timing. GPI-1B-only animals lack GPI-1A. These animals display the larger reduction in hatched progeny, which indicates an important requirement for GPI-1A, rather than GPI-1B, for normal brood size under the tested conditions. The current conclusion that GPI-1B is specifically required for both normal brood size and reproductive timing should therefore be revised. The data instead support partially redundant and nonidentical contributions of the two isoforms, with GPI-1B contributing to reproductive timing and GPI-1A making a substantial contribution to total reproductive output. Because the isoform-specific alleles alter alternative starts and exon organization, the authors should also confirm their effects on transcript structure and protein production. Isoform-resolved RT-PCR or RNA sequencing, together with quantitative protein analysis, would help exclude unintended effects on splicing or translation. The possibility that wild-type animals contain GPI-1A/GPI-1B heterodimers, whereas the single-isoform strains cannot, should be discussed. At least one central phenotype of the complete gpi-1 knockout should also be confirmed with an independent allele or a near-endogenous genomic rescue. This is particularly important for the developmental, reproductive, and metabolomic conclusions, which rely heavily on one whole-gene deletion background.”

    __Response and action: __We will revise the Results and Discussion to clarify the roles of the individual GPI-1 isoforms in the reproductive phenotypes. Our data support partially redundant but nonidentical functions for GPI-1A and GPI-1B, underscoring the importance of both isoforms. Critically, however, the requirement for each isoform, and the phenotypes that depend on it, varies across tissues, which is a central message of the study. Briefly, in somatic tissues, GPI-1A is sufficient to support wild-type glycolytic function, whereas GPI-1B is not. In the germline, the relationship is more complex: neither single isoform fully recapitulates wild-type reproductive function. Loss of GPI-1B delays the onset of egg production, supporting a role for GPI-1B in reproductive timing, whereas loss of GPI-1A produces a stronger reduction in total reproductive output. Thus, both isoforms contribute to normal reproductive fitness, but with distinguishable effects on different aspects of germline function. This is consistent with our broader model that the isoforms have distinct but partially overlapping functions. Viewed in the context of the distinct metabolic demands of these tissues, these findings support a model in which two closely related GPI isoforms, with different patterns of expression and subcellular localization, contribute differently to glucose utilization across tissues and thereby help support distinct metabolic programs in vivo. We will edit the results to make this point clearer. Additionally, we will include data not currently in the manuscript where we test an additional null allele of *gpi-1 *and demonstrate failure to upregulate neuronal glycolysis via the HYlight sensor. __Reviewer 2 Minor Comments: __

    “The evidence most directly supports localization of GPI-1B to ER-associated endomembrane structures. Statements claiming definitive ER membrane localization should be qualified unless endogenous germline colocalization or quantitative immuno-EM analysis is provided. Colocalization with ER markers should be quantified across independent cells and experiments. The analysis should include the number of animals or cells examined, the method used to calculate colocalization, and appropriate controls for overexpression.” __Response and action: __We support ER-associated localization of GPI-1B using three independent lines of evidence: (1) immuno-electron microscopy, (2) colocalization with ER markers in vivo across tissues, and (3) colocalization with ER markers in cultured cells. We will strengthen this section by adding quantitative colocalization analyses, reporting the number of cells and animals examined, and providing the relevant methodological details and controls. We will also revise the text to make explicit how these complementary lines of evidence support our conclusion and, where appropriate, qualify the localization as ER-associated rather than confined to the ER membrane. “GPI-1B localization to lipid droplets is described as occurring in "some cells." The frequency of this observation should be reported and quantified. If it is rare or highly variable, it should not be presented as a general property of the protein.” __Response and action: __We will clarify the incidence of GPI-1B colocalization with lipid droplet markers. “The Methods contain the placeholder "cite HYlight paper." The correct reference and a complete description of the FiLa analysis should be inserted.” __Response and action: __This typo has been corrected in line 773. “A dedicated data and code availability section should be added. In addition to the modeling and metabolomics files discussed above, the authors should deposit source data for every graph, uncropped image data, representative raw microscopy stacks, biosensor time series, swimming videos or movement tracks, and the code or macros used for image and behavioral analysis.” __Response and action: __We will deposit all available data files for each experiment into a publicly available data repository. “Figure legends should state the exact n, whether error bars represent standard deviation, standard error, or confidence intervals, and how many independent experiments were performed. Exact p values should be provided whenever practical.” __Response and action: __We will update all figure legends to include exact n, error bar meaning and number of independent experiments. We will also include a table of exact p values. “The manuscript should indicate whether representative images from different genotypes were displayed using identical acquisition settings, ratio bounds, background subtraction, and lookup-table limits. Any genotype-specific rescaling should be stated explicitly.” __Response and action: __We will include a statement in the methods section about how data acquisition and post-acquision analysis was performed for each image. “The manual mending of WormLab tracks should be described in greater detail. The authors should report whether the analyst was blinded to genotype, how frequently tracks were edited, and the criteria used to decide when editing was appropriate.” __Response and action: __We will expand the description of the manual WormLab track mending in the methods section. “Terms such as "necessary and sufficient," "demonstrate," "drives," "controls carbon allocation," and "partitions flux" should be used only when the corresponding causal relationship is experimetally tested. In several places, "supports," "is associated with," or "is consistent with" would represent the data more accurately.” __Response and action: __We will review all implemented uses of the flagged language from this reviewer comment and when appropriate, replace with the suggested phrases above. “Please verify the figure references throughout the Results. For example, the discussion of the complete gpi-1 knockout HYlight phenotype in the single-isoform section appears to refer to Figure 2A, although the relevant metabolic sensor data are presented elsewhere.” __Response and action: __We will carefully review the figure correspondence throughout the results for any typos. “The possible effects of developmental stage should be discussed. The mutants differ in developmental rate and reproductive state, and comparisons described as Day 1 adults may not represent physiologically equivalent animals if staging was based only on chronological time.” __Response and action: __We will clarify our methods section to state that experiments with animals of Day 1 adults were not only based on chronological time, but also made use of developmental landmarks. “The manuscript should clarify whether hypoxia and sodium azide are being treated as equivalent forms of mitochondrial inhibition. They impose related but not identical physiological stresses, and conclusions derived from one should not automatically be extended to the other.” __Response and action: __We will edit the results and methods to be explicit about the difference between the mitochondrial inhibition methods. “The authors should distinguish GPI-1's reversible biochemical reaction from the committed oxidative PPP step. Referring to GPI-1 as a "PPP enzyme" may be useful in the context of reverse cyclic PPP flux, but this usage should be clearly defined to avoid implying that GPI-1 is part of the canonical oxidative PPP.” __Response and action: __We agree that GPI-1 is not part of the canonical oxidative PPP and does not catalyze a committed PPP reaction. We will revise the text to avoid language that could imply that GPI-1 is a canonical oxidative PPP enzyme. “The strain table and allele schematics are helpful, but the final verified DNA sequence for every engineered allele should be provided in a supplementary file or public repository.” __Response and action: __We will provide the DNA sequences as a supplementary file and deposit all relevant strain genotypes/sequences in the Caenorhabditis Genetics Center, a publicly accessible repository. “The manuscript would benefit from a final model figure that separates what is directly demonstrated from what remains inferred. For example, isoform expression and localization could be shown as established findings, whereas tissue-specific PPP flux and localization-dependent carbon partitioning should be labeled as proposed mechanisms.” __Response and action: __We will add a final model figure that highlights key findings from the manuscript. __Reviewer 3 Major points: __

    “In Figure 1, the authors use snRNA-seq data for modeling metabolic pathways and although they find a candidate that they were able to validate, it is important to point out for the readers that transcript levels do not always correlate with protein levels of the enzymes. Therefore, this type of predictive modeling should consider it as a limitation.” __Response and action: __We thank the reviewer for this comment and agree with this important consideration. Though our modeling data is validated with *gpi-1 *mutants, our reported flux metrics are computational predictions based on transcriptomic data and not direct flux measurements with accompanied protein concentration measurements. We will qualify that these modeling studies do not infer and are not based on protein level of the modeled metabolic enzymes and discuss this limitation in more detail with an added section in our discussion section. “While most of the paper contrasts somatic cells and the germ line, the logic behind performing many of the experiments (such as Hylight) in neurons alone is unclear. Indeed, it is not clear why muscle is not considered given that the swimming assay that is used a readout of glycolysis is presumably largely driven by muscle metabolism.” __Response and action: __Our goal of including both readouts in neurons and muscles for somatic cell function was to demonstrate across multiple tissues the importance of gpi-1-regulated glycolytic function. While our expertise is in examining sensors at the neuronal level, expanding these studies to other tissues (such as muscles and the germline) is an important future direction and discussion point to elaborate on. We will include additional rationale for specifically examining neurons with sensor-based approaches and elaborate further on the importance of expanding these approaches to other tissues within both the results and discussion section. “Many of the phenotypes in the paper are not quantified, such as Fig. subcellular localizations of the isoforms (eg. Fig. 6 and 7). I believe that determining what fraction of isoform B localizes to the cytosol is important because of the observed functional redundancy with isoform A.” __Response and action: __We will quantify the relative level of each isoform in the cytosolic vs subcellular fraction, and additionally provide quantification for colocalization studies.

    Reviewer 3 Minor Comments

    __“__The volcano plots of pathway analysis in Fig 3 are difficult to follow. Other forms of representations might be suitable. I also found Fig. S2C and D with their direct comparisons in individual genotypes to be more informative than the panels in Fig. 3.” __Response and action: __We will place the volcano plots and pathway analysis currently featured in Fig S2 A-D in the positions of Figure 3 A and B. This will show a direct comparison between each individual genotype (gpi-1 KO animals/pfk-1.1 KO animals) relative to wild type animals and will make the conclusions more clear. We will also place the comparison of the mutants to each other in the supplemental figure to demonstrate that these two mutants ultimately differ in accumulated and depleted metabolites. “Fig. 3E and F lack statistical analysis.” __Response and action: __We will include statistical comparisons of the starting baseline and ending point in hypoxia for the panneuronal HYlight worms across genotypes. “EM images in Fig. 6 are not intuitively clear to an unexperienced reader, better examples or more annotations is needed” __Response and action: __We will include annotation to the existing figure to make it clear where the immunogold beads are and what structures we are identifying as ER membranes. “Quantification of colocalization with lipid droplets in Fig. S4 would be informative” __Response and action: __We will quantify co-localization for the lipid droplet figure in S4 in addition to Figures 6 and 7.

  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

    In this study, the authors investigated tissue-specific metabolism of glucose in the worm (C. elegans). Through a combination of modeling and the available transcriptome data, they found that the pentose phosphate pathway (PPP) and specifically the enzyme GPI which sits at the junction of glucose and PPP is differentially active in the germline versus somatic cells. The authors then used an array of elegant genetic experiments along with optical and behavioral readouts to show that the worm has two isoforms of the GPI enzyme that are differentially expresses in these tissues, have disticnt subcellular localization, and favor PPP or glycolysis. Through replacement and rescue experiments,they show that while the isoforms are partially functionally redundant, A is more important for somatic function such as swimming, while B is critical for reproductive fitness. Overall, this is a well-done and rigorous study with exciting findings that demonstrates the functional diversity of metabolic pathways in different tissues. I do have a few comments to strengthen their conclusions:

    1. In Figure 1, the authors use snRNA-seq data for modeling metabolic pathways and although they find a candidate that they were able to validate, it is important to point out for the readers that transcript levels do not always correlate with protein levels of the enzymes. Therefore, this type pf predictive modeling should consider it as a limitation.
    2. While most of the paper contrasts somatic cells and the germ line, the logic behind performing many of the experiments (such as Hylight) in neurons alone is unclear. Indeed, it is not clear why muscle is not considered given that the swimming assay that is used a readout of glycolysis is presumably largely driven by muscle metabolism.
    3. Many of the phenotypes in the paper are not quantified, such as Fig. subcellular localizations of the isoforms (eg. Fig. 6 and 7). I believe that determining what fraction of isoform B localizes to the cytosol is important because of the observed functional redundancy with isoform A.

    Minor comment:

    1. the volcano plots of pathway analysis in Fig 3 are difficult to follow. Other forms of representations might be suitable. I also found Fig. S2C and D with their direct comparisons in invidual genotypes to be more informative than the panels in Fig. 3.
    2. Fig. 3E and F lack statistical analysis.
    3. EM images in Fig. 6 are not intuitively clear to an unexperienced reader, better examples or more annotations is needed
    4. Quantification of colocalization with lipid droplets in Fig. S4 would be informative

    Significance

    The study provides a rigorous analysis of tissue-specific expression of a glycolytic enzyme GPI, and provides a mechanism for catabolic vs. anabolic fate of glucose in gonads as compared to somatic cells. In other words, the study provides a significant conceptual advance in revealing a new mechanisms for how metabolic pathways are differentially regulated/utilized in different tissues in the body based on their unique needs. The study is particularly strong in its application of state-of-the-art genetic tools and optical sensors for metabolites. It also provides a testable model of how GPI might be regulated in mammalian systems. But it does leave open the question of how distinct subcellular localization of GPI isoforms may regulate its enzymatic function. I do however consider this beyond the scope of the current study. Overall, this study would be of broadly interesting to the metabolism field.

    My expertise is in neurometabolism, synaptic function, and neuronal cell biology.

  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

    Summary

    In this manuscript, Gonzalez et al. investigate how glucose-6-phosphate isomerase, a conserved reversible enzyme positioned between glycolysis and the pentose phosphate pathway, supports different metabolic requirements across tissues in vivo. Using Caenorhabditis elegans, the authors combine tissue-resolved metabolic-network modeling based on adult single-cell RNA-seq data with CRISPR-generated mutants, whole-animal metabolomics, behavioral and reproductive assays, genetically encoded FBP and lactate sensors, and imaging of endogenously tagged GPI-1 isoforms.

    The modeling predicts differences in glycolytic and PPP capacity among tissues and suggests preferential reverse GPI-1 flux in reproductive tissues. Whole-gene disruption of gpi-1 produces reproductive and developmental phenotypes that resemble some effects of gspd-1 loss, together with somatic glycolytic defects resembling those of pfk-1.1 loss under mitochondrial stress. The authors further identify two GPI-1 isoforms with distinct expression and localization patterns. GPI-1A is broadly expressed and predominantly cytosolic, whereas GPI-1B contains a short N-terminal domain that promotes localization to ER-associated endomembrane structures and is enriched in the germline. Single-isoform strains and an isoform-swap strain reveal differences in the capacity of GPI-1A and GPI-1B to support reproduction, swimming, and neuronal FBP and lactate responses to hypoxia.

    This is a creative, technically broad, and potentially important study. The combination of whole-animal genetics, single-cell metabolic imaging, network modeling, and isoform-level cell biology is a major strength. The data convincingly establish that the two GPI-1 isoforms differ in expression, localization, and their ability to support particular physiological and metabolic readouts. At the same time, the current experiments do not yet establish the main mechanistic conclusion that subcellular localization itself partitions GPI-1 function between germline PPP metabolism and somatic glycolysis. Several central conclusions also rely on predicted rather than measured flux or on whole-animal, instead of tissue-specfic, metabolite pools; and these are not properly acknowledged in every case. Some statistical or analytical procedures require additional clarification. Most of these concerns could be resolved through a combination of focused experimental controls, reanalysis, complete data disclosure, and qualification of the claims.

    1. The current experiments do not separate the effect of subcellular localization from other isoform-specific properties

    A central conclusion is that the different subcellular localization of GPI-1A and GPI-1B contributes causally to their functional specialization. The GPI-1B swap strain is an informative reagent, but the experiment does not isolate localization as the relevant variable. GPI-1A and GPI-1B differ in amino-acid sequence, and the swap could alter protein abundance, translation, stability, catalytic activity, dimerization, interactions with other proteins, or access to substrates. Therefore, the observation that the swap supports swimming while incompletely restoring neuronal FBP and lactate responses is consistent with an effect of localization, but is not sufficient to demonstrate it.

    The authors should quantify GPI-1 abundance in the same somatic cells or tissues used for the functional measurements. Calibrated fluorescence from endogenously tagged proteins would be more informative than a whole-animal Western blot, which could be dominated by expression in unrelated tissues. GPI activity should also be compared across the relevant isoform and swap strains.

    The localization-defective GPI-1B CCM allele could provide a valuable separation-of-function test, but the introduced substitutions could themselves affect folding or catalytic function. The authors should therefore establish that the CCM protein is expressed at a comparable level and retains GPI activity before interpreting a physiological effect as resulting from altered localization. If this allele is functional, testing at least one central metabolic or physiological phenotype, such as the neuronal HYlight or FiLa response, swimming under mitochondrial inhibition, or reproductive output, would directly address the proposed mechanism.

    These experiments are essential only if the authors wish to conclude that localization controls GPI-1 function. If they are not technically feasible, the title, abstract, Results, and Discussion should consistently state that isoform-specific localization is associated with, and may contribute to, the observed functional differences.

    Estimated time investment: approximately 3-6 weeks if the existing strains can be used; potentially 2-3 months if additional strain construction is necessary.

    1. The evidence does not currently establish germline-specific PPP metabolism or reverse GPI-1 flux

    The modeling predicts reverse GPI-1 flux in the germline, and the reproductive phenotypes and metabolomics are consistent with a role for GPI-1 in anabolic metabolism. However, the metabolomics was performed using whole animals. It therefore cannot determine whether the PPP-associated metabolite changes arise in the germline, the soma, or through systemic responses to developmental delay, infertility, embryonic lethality, redox stress, or altered communication between tissues.

    In this regard, the authors should distinguish clearly among three different levels of evidence:

    i) a predicted tissue-specific flux state from the network model;

    ii) changes in whole-animal metabolite abundance;

    and iii) direct measurements of pathway flux.

    At present, the study provides the first two, but not the third.

    If the authors wish to retain a strong claim that the metabolomic phenotype originates in the gonad, tissue-resolved targeted metabolomics would be informative. One possible approach would be rapid mechanical extrusion and isolation of adult gonads, followed by targeted analysis of a prespecified set of PPP, glycolytic, nucleotide, and redox metabolites. Because dissection itself may rapidly alter metabolism, this experiment would require appropriate controls: isolated gonads, the corresponding somatic carcasses, a processed recombination control in which separated gonads and carcasses are mixed back together, and intact animals processed in parallel. The dissection medium should either be retained or tested for released metabolites, and the interval between incision and metabolic quenching should be minimized, standardized, and reported. At least three independent biological replicates would be needed, with the number of animals determined empirically from the sensitivity and reproducibility of the assay rather than fixed in advance.

    This experiment would localize changes in metabolite pools but would still not directly measure PPP flux. If tissue-resolved metabolomics is not feasible, the germline-specific PPP conclusion should be presented as an inference supported by modeling, reproductive phenotypes, and whole-animal metabolite measurements.

    Estimated time investment: approximately 3-4 months if tissue-resolved metabolomics is undertaken; several days if the issue is addressed by appropriately qualifying the conclusions.

    1. The metabolic-network predictions require uncertainty analysis and complete computational disclosure

    The metabolic modeling provides the foundation for the tissue-specific interpretation of the study. The manuscript describes the use of MERGE, CatEx, iMAT++, eFPA, relaxed biomass requirements, sink reactions, exchange constraints, a penalty for selected boundary reactions, and FVA-derived directional constraints. However, the reader cannot determine from the current description how robust the GPI-1 directionality prediction is to these choices.

    The authors should deposit the complete computational workflow, including:

    i) the single-cell or pseudo-bulk input matrices;

    ii) cell-type assignments and aggregation procedures;

    iii) the exact version of iCEL1314;

    iv) all modified reactions, sink reactions, exchange bounds, and hard constraints;

    v) CatEx thresholds and activity assignments;

    vi) the complete objective functions and solver settings;

    vii) the exchange penalty and the rationale for its value;

    viii) all scripts used for iMAT++, eFPA, FVA, and figure generation;

    ix) and the complete reaction-level input and output tables.

    The FVA range for GPI-1 should be reported for each tissue. It would be important to know whether reverse flux is required by the feasible solution space, strongly favored across alternative solutions, or simply permitted in one optimized solution. Sensitivity analyses should determine whether the germline prediction persists after reasonable changes to expression thresholds, exchange penalties, biomass relaxation, and other major assumptions.

    Throughout the manuscript, modeled flux directionality should be described as a prediction rather than evidence that tissue-specific carbon flux was measured experimentally. The biological context mismatch should also be acknowledged: the modeling is based on Day 1 adult expression profiles, whereas some of the supporting phenotypes occur during larval development or embryogenesis.

    Estimated time investment: approximately 1 month.

    1. Replication, statistical analysis, and interpretation of the biosensor experiments require revision

    i) The manuscript does not consistently provide the number of animals, cells, independent experiments, or biological replicates for the swimming and biosensor measurements. These values should be reported in every figure legend, together with the definition of the experimental unit. The authors should state whether animals were obtained from independent cultures, how many experimental days were included, whether acquisition and analysis were blinded, and what exclusion criteria were applied.

    ii) The HYlight and FiLa experiments contain paired and longitudinal measurements from the same animals before and after hypoxia. A one-way ANOVA does not account for within-animal pairing or genotype-by-time interactions. The data should be analyzed using an appropriate paired endpoint analysis, a repeated-measures or mixed-effects model, or a prespecified summary measurement such as change from baseline or area under the curve. Individual paired values or trajectories should be shown where practical. Multiple-comparison procedures, effect sizes, exact p values, and the meaning of the shaded regions in time-course plots should also be reported.

    iii) HYlight and FiLa are valuable reporters, but they measure FBP- and lactate-sensitive fluorescence ratios rather than flux directly. Dynamic changes are consistent with altered glycolytic activity, but the manuscript should avoid describing either reporter as a direct measurement of glycolytic flux unless this interpretation is independently validated.

    iv) The C. elegans-adapted FiLa construct also requires clearer validation and description. The authors should explain its expected response direction, expression control, and normalization procedure. Most directly, FiLa should be tested in the ldh-1(ola514) deletion background, which removes the sole C. elegans lactate dehydrogenase ortholog and has been shown to reduce the signal of an independent neuronal lactate reporter. Alternatively, the authors could use their existing pfk-1.1(ola458) knockout, which is expected to reduce glycolytic lactate production. In either case, basal and hypoxia-induced FiLa responses should be compared with wild type using paired individual-animal measurements

    Estimated time investment: approximately 2-4 weeks for reanalysis, clarification, and presentation of existing data. Additional validation experiments may require 1-2 months.

    1. The metabolomics analysis requires additional statistical rigor and access to the complete data

    The metabolomics section does not clearly state the number of biological replicates, the number or identity of excluded samples, or whether the investigators were blinded during data processing. These details are essential.

    The authors report normalization to the sum of targeted metabolites, selection of the detection mode with the "best intensity" when a metabolite was detected more than once, and pathway analyses using nominal p values. These procedures require additional justification. Sum normalization can generate compositional effects when mutations produce widespread metabolic changes. The criterion used to choose among detection modes should be prespecified and reported for every affected metabolite. Given the large number of metabolites tested, adjusted p values or an appropriate false-discovery-rate procedure should be provided.

    The authors should deposit:

    i) the raw mass-spectrometry files;

    ii) processed peak tables before and after normalization;

    iii) metabolite identities and confidence levels;

    iv) retention times and reference standards;

    v) missing-value handling;

    vi) QC measurements and exclusion criteria;

    vii) the complete differential-abundance results, including metabolites not highlighted in the figures;

    viii) effect sizes, nominal p values, and adjusted p values;

    and ix) the complete pathway-enrichment outputs.

    The presentation of approximately 60 highly changing and interconnected metabolites could be biasing the results towards thosethat form a coherent pathway narrative. The authors should explain how these metabolites were selected and make the full, unselected results available, so readers can assess whether the highlighted PPP and glycolysis patterns are representative.

    Finally, altered steady-state metabolite abundance should not be described as direct evidence of altered flux.

    Estimated time investment: approximately 1-2 months for data deposition, reanalysis, and revision of the presentation.

    1. The special culture conditions used for pfk-1.1 may affect the central pathway comparison

    pfk-1.1 knockout animals sometimes display substantial embryonic lethality and this phenotype can be suppressed by growth on newly prepared NGM plates with thicker OP50 lawns. This observation is biologically interesting, but it also creates an important potential confounding variable because the manuscript uses pfk-1.1 as the glycolysis-specific benchmark against which gpi-1 and gspd-1 phenotypes are interpreted.

    The authors should state clearly:

    i) which culture conditions were used for each experiment;

    ii) whether all genotypes were grown and assayed simultaneously under identical conditions;

    and iii) how frequently pfk-1.1 embryonic lethality was observed under standard versus optimized conditions.

    The complete results obtained under the different dietary or plate conditions should be provided. Ideally, the reproductive comparison should be repeated under at least two clearly defined conditions for all genotypes to establish which conclusions are diet-dependent. If diet-dependent, the text should acknowledge that the apparent distinction between pfk-1.1 and gpi-1 reproductive phenotypes is conditional rather than absolute.

    The same food and plate conditions should be used to document the swimming, biosensor, developmental, and metabolomics experiments.

    Estimated time investment: approximately 2-4 weeks if data are available; 2-4 months if repeat experiments are required.

    1. The interpretation of the single-isoform reproductive phenotypes should be corrected

    The interpretation of Figure 8G-H does not appear to follow from the genotypes.

    GPI-1A-only animals lack GPI-1B. Their delayed onset of egg production supports a contribution of GPI-1B to normal reproductive timing.

    GPI-1B-only animals lack GPI-1A. These animals display the larger reduction in hatched progeny, which indicates an important requirement for GPI-1A, rather than GPI-1B, for normal brood size under the tested conditions.

    The current conclusion that GPI-1B is specifically required for both normal brood size and reproductive timing should therefore be revised. The data instead support partially redundant and nonidentical contributions of the two isoforms, with GPI-1B contributing to reproductive timing and GPI-1A making a substantial contribution to total reproductive output.

    Because the isoform-specific alleles alter alternative starts and exon organization, the authors should also confirm their effects on transcript structure and protein production. Isoform-resolved RT-PCR or RNA sequencing, together with quantitative protein analysis, would help exclude unintended effects on splicing or translation. The possibility that wild-type animals contain GPI-1A/GPI-1B heterodimers, whereas the single-isoform strains cannot, should be discussed.

    At least one central phenotype of the complete gpi-1 knockout should also be confirmed with an independent allele or a near-endogenous genomic rescue. This is particularly important for the developmental, reproductive, and metabolomic conclusions, which rely heavily on one whole-gene deletion background.

    Estimated time investment: several days for correcting the interpretation and discussion; approximately 1-2 months for molecular validation or rescue using an existing strain.

    Minor comments

    1. The evidence most directly supports localization of GPI-1B to ER-associated endomembrane structures. Statements claiming definitive ER membrane localization should be qualified unless endogenous germline colocalization or quantitative immuno-EM analysis is provided. Colocalization with ER markers should be quantified across independent cells and experiments. The analysis should include the number of animals or cells examined, the method used to calculate colocalization, and appropriate controls for overexpression.
    2. GPI-1B localization to lipid droplets is described as occurring in "some cells." The frequency of this observation should be reported and quantified. If it is rare or highly variable, it should not be presented as a general property of the protein.
    3. The Methods contain the placeholder "cite HYlight paper." The correct reference and a complete description of the FiLa analysis should be inserted.
    4. A dedicated data and code availability section should be added. In addition to the modeling and metabolomics files discussed above, the authors should deposit source data for every graph, uncropped image data, representative raw microscopy stacks, biosensor time series, swimming videos or movement tracks, and the code or macros used for image and behavioral analysis.
    5. Figure legends should state the exact n, whether error bars represent standard deviation, standard error, or confidence intervals, and how many independent experiments were performed. Exact p values should be provided whenever practical.
    6. The manuscript should indicate whether representative images from different genotypes were displayed using identical acquisition settings, ratio bounds, background subtraction, and lookup-table limits. Any genotype-specific rescaling should be stated explicitly.
    7. The manual mending of WormLab tracks should be described in greater detail. The authors should report whether the analyst was blinded to genotype, how frequently tracks were edited, and the criteria used to decide when editing was appropriate.
    8. Terms such as "necessary and sufficient," "demonstrate," "drives," "controls carbon allocation," and "partitions flux" should be used only when the corresponding causal relationship is experimetally tested. In several places, "supports," "is associated with," or "is consistent with" would represent the data more accurately.
    9. Please verify the figure references throughout the Results. For example, the discussion of the complete gpi-1 knockout HYlight phenotype in the single-isoform section appears to refer to Figure 2A, although the relevant metabolic sensor data are presented elsewhere.
    10. The possible effects of developmental stage should be discussed. The mutants differ in developmental rate and reproductive state, and comparisons described as Day 1 adults may not represent physiologically equivalent animals if staging was based only on chronological time.
    11. The manuscript should clarify whether hypoxia and sodium azide are being treated as equivalent forms of mitochondrial inhibition. They impose related but not identical physiological stresses, and conclusions derived from one should not automatically be extended to the other.
    12. The authors should distinguish GPI-1's reversible biochemical reaction from the committed oxidative PPP step. Referring to GPI-1 as a "PPP enzyme" may be useful in the context of reverse cyclic PPP flux, but this usage should be clearly defined to avoid implying that GPI-1 is part of the canonical oxidative PPP.
    13. The strain table and allele schematics are helpful, but the final verified DNA sequence for every engineered allele should be provided in a supplementary file or public repository.
    14. The manuscript would benefit from a final model figure that separates what is directly demonstrated from what remains inferred. For example, isoform expression and localization could be shown as established findings, whereas tissue-specific PPP flux and localization-dependent carbon partitioning should be labeled as proposed mechanisms.

    Significance

    General assessment

    This is a creative, technically broad, and potentially important study addressing a fundamental question in metabolic biology: how a conserved reversible enzyme can support distinct physiological needs in different tissues of an intact animal. The strongest aspects are the integration of tissue-resolved modeling with whole-animal genetics and physiology, the endogenous identification of two GPI-1 isoforms, the discovery of an N-terminal localization element in GPI-1B, and the use of single-cell metabolic reporters to examine neuronal responses in vivo.

    The main limitation is that the central mechanistic model currently goes beyond what the experiments directly establish. The data strongly support differences in isoform expression, localization, and functional capacity. They do not yet establish that localization itself causes the differential metabolic effects, nor do they directly measure germline PPP flux or tissue-specific carbon allocation. Other limitations include reliance on whole-animal metabolomics, incomplete reporting of biological replication, insufficient statistical treatment of paired time-course data, uncertainty in the model-derived flux predictions, and culture conditions that may selectively suppress an important pfk-1.1 phenotype.

    These concerns do not diminish the interest of the biological observations, but they should be addressed so that the conclusions accurately reflect the evidence.

    Advance

    The study extends the growing view that conserved metabolic enzymes are not necessarily uniform housekeeping components. Mammalian hexokinase, phosphofructokinase, and pyruvate kinase isoforms have established tissue-specific and compartment-specific properties, but GPI is generally treated as a freely soluble enzyme operating near equilibrium. The identification of a GPI isoform with a short targeting sequence and a distinct in vivo distribution is therefore an interesting conceptual and cell-biological advance.

    The work also provides useful genetic tools for studying GPI function in an intact metazoan and illustrates how modeling, live metabolic sensors, and organismal phenotypes can be combined to generate tissue-specific metabolic hypotheses.

    At present, the advance is strongest as an isoform-localization and functional-association study. A direct separation-of-function experiment showing that relocalization changes metabolism while preserving protein abundance and activity would elevate the work to a more definitive mechanistic demonstration of spatial control over a reversible metabolic node.

    Audience

    The study should be of interest to a broad basic-research audience, including investigators in cellular metabolism, systems biology, cell biology, developmental and reproductive biology, neuroscience, quantitative imaging, and C. elegans physiology. The genetic reagents and live-sensor approaches may be particularly useful to researchers studying tissue interactions and metabolic plasticity in vivo.

    The broader principles may also be relevant to investigators studying metabolic reorganization in cancer, neurological disease, and inherited metabolic disorders, although direct disease relevance remains to be established.

    Reviewer expertise

    My expertise includes C. elegans genetics and physiology, cellular and organismal metabolism, metabolic stress responses, quantitative fluorescence imaging, metabolomics, and the biological interpretation of omics and metabolic-network analyses. I am not a developer of constraint-based metabolic-modeling algorithms and therefore cannot independently audit the mathematical implementation or solver performance, but I can assess the biological assumptions, presentation, and interpretation of the modeling results.

  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, Gonzalez et al. investigate tissue-specific regulation of metabolism. While recent studies suggest that metabolism can vary dramatically across cell types, quantitative measurements in living animals remain challenging, and the mechanisms driving these differences are not fully understood. Using C. elegans as a model, the authors show that differential isoform expression of GPI-1, a key metabolic enzyme sitting at the branch point between glycolysis and the pentose phosphate pathway (PPP), correlates with pathway utilization in a tissue-specific manner in the intact animal. Using newly developed tools for quantitative metabolic imaging together with metabolomics, transcriptomics, electron microscopy, and genetics, the authors make several striking findings. First, they show that the germline preferentially relies on the PPP relative to neurons or muscle. They then find that GPI-1 is expressed as two isoforms, GPI-1A and GPI-1B, in soma and germline. These isoforms differ in subcellular localization and have distinct effects on metabolism, as well as on neuronal and germline phenotypes consistent with disruption of glycolysis or PPP respectively. Further, they identify a specific N-terminal helical region of GPI-1B required for subcellular localization to ER-related membranes in C. elegans in in human RPE-1 cells. The data, findings, figures, and methods are all clearly presented and interpreted.

    Minor comments:

    1. The authors also generate and analyze pfk-1.1 and gpd-1 mutants, which help support the roles of GPI-1 in PPP and glycolysis and the resulting movement and germline phenotypes. However, these data are largely absent from the discussion, and it would help to incorporate them there.

    2. The authors note that the GPI-1B endomembrane clusters are observed in some embryos. From presented images it also appears that oocytes also have heterogeneous presence of clusters. If this is correct, it would be useful for the authors to comment on potential consequences of this or is the proposed function earlier in the germline where these structures also observed in the rachis.
    3. In the Microscopy methods section a note to add a references is included "(cite HYlight paper)"
    4. Brood size analysis, sentence fragment "We note that when these mutants are grown in conditions that vary"

    Significance

    General assessment: Strengths of this study include the use of quantitative phenotypes and multiple complementary methods to interrogate metabolism in an intact animal, including genetics, metabolomics, quantitative imaging reporters, and electron microscopy. The data, findings, figures, and methods are all clearly presented and interpreted. One limitation of the current study is that the reporters used in neurons for glycolysis or reporters for PPP are not yet available for the germline.

    Advance: The ability to explore metabolism in an intact animal reveals striking tissue-specific differences and their consequences for distinct biological pathways. Further, the authors identify a specific isoform and alpha-helical region required to drive these tissue-specific differences.

    Audience: These findings will be of broad interest to those studying metabolism, as well as neuronal and germline biology, and open new avenues for quantitative, tissue-, cell-, and subcellar analysis of metabolism.

  5. Version published to 10.1101/2025.11.09.687253 on bioRxiv