Tracing metabolic flux through time and space with isotope labeling experiments

Abstract

Metabolism is dynamic and must function in context-specific ways to adjust to changes in the surrounding cellular and ecological environment. When isotopic tracers are used, metabolite flow (i.e., metabolic flux) can be quantified through biochemical networks to assess metabolic pathway operation. The cellular activities considered across multiple tissues and organs result in the observed phenotype and can be analyzed to discover emergent, whole-system properties of biology and elucidate misconceptions about network operation. However, temporal and spatial challenges remain significant hurdles and require novel approaches and creative solutions. We survey current investigations in higher plant and animal systems focused on dynamic isotope labeling experiments, spatially resolved measurement strategies, and observations from re-analysis of our own studies that suggest prospects for future work. Related discoveries will be necessary to push the frontier of our understanding of metabolism to suggest novel solutions to cure disease and feed a growing future world population.

Graphical Abstract

Introduction

Ensuring global health and nutrition for future generations will require transformative gains in agricultural productivity [3–5] and improved strategies to combat disease [6,7]. Spatial and temporal complexity are at the forefront of challenges that limit success in metabolic engineering and, more generally, restrict our understanding of metabolism. Eukaryotic metabolism is segregated between organs, cells, and subcellular organelles, operating over circadian and developmental time regimes, and must continuously adjust to environmental perturbations that vary in duration and intensity. Unfortunately, most experimental efforts are limited to analysis of isolated components, steady-state conditions, and simplified systems that cannot fully recapitulate the spatial and temporal dynamics that occur in vivo. Isotopic tracers (i.e., stable isotopes such as ¹³C, ¹⁵N, ²H, ¹⁸O, etc. or radioisotopes such as ¹⁴C or ³H) are the primary tools used by researchers to quantitatively track the dynamics of metabolism. Thus, their importance to metabolic studies cannot be overstated, and recent advances in mass spectrometry (MS) technologies (including wide availability of high-resolution and tandem MS instruments) offer a burgeoning opportunity to further extend the scope of isotope tracers to upgrade the information content of metabolomics datasets. Here we examine recent applications of isotopic labeling experiments to assess spatial and temporal aspects of metabolism that would be otherwise inaccessible to measurement.

Assessing temporal dynamics and metabolite turnover

When plants see the light—Temporal dynamics that balance light harvesting and carbon assimilation

One of the most crucial functions for a plant is to maximize photosynthesis for growth and metabolism; however incident light fluctuates with seasonal daylength, cloud cover, sun flecks, and canopy shading. The changes in light intensity are dynamic and unpredictable such that plant response is not optimized for agricultural productivity [13], and daily starch production must account for unforeseen cloud cover. Recent ¹³C and ¹⁴C labeling studies have indicated starch breakdown occurs during the latter part of photoperiods [15], potentially explaining how plants adjust to the unknown cumulative light levels per day. ¹³C tracer studies have described the systemic changes that occur when canopies shade leaves, including induced senescence to maximize resource allocation for growth and survival [16]. At the cellular level, metabolic adaptations to accommodate altered light levels can include seemingly counterproductive paths and futile cycles such as the oxidative pentose phosphate (OPP) pathway. Sharkey and coworkers [10–12] indicate that the oxidative portion of pentose phosphate metabolism (i.e., G6P shunt) could play an important role in stabilizing photosynthesis and would therefore account for a portion of total day-time respiration in leaves, however details about the spatial origin of the G6P supply for the shunt and other aspects of operation remain to be elucidated and might benefit from labeling-based flux analyses.

Identification of futile cycles with isotope tracers

Isotope labeling studies offer tremendous potential for deciphering the operation of metabolic networks [17] and discovering hidden metabolic functions [18]. In particular, the operation of futile cycles can be very difficult to detect without isotope tracers [19]. A futile cycle is a ‘wasteful’ process in which two metabolic pathways run simultaneously in opposite directions but have no overall effect other than to dissipate energy. Prior studies have applied isotope labeling experiments to quantify rates of futile cycling in E. coli [20], B. subtilis [21], and S. cerevisiae [22]. In plants, recent ¹⁴C/¹³C studies suggest that futile cycles help maintain nutrient status under altered environments [23,24], a process that is particularly relevant to future agriculture with finite phosphorus resources and intensive nitrogen fertilizer production. Coordination of glycolytic metabolism with sugar sensing and sucrose cycling is thought to consume large amounts of ATP necessary to balance resource allocation in plants [25], and similar processes to ensure homeostatic regulation of glucose metabolism are also well-studied in mammalian physiology. For example, various futile cycles involving PEP [26], pyruvate [27,28], and glucose [29] have been hypothesized to play important roles in regulating glucose-stimulated insulin secretion by pancreatic islet cells. Direct examination of flux through these cyclic pathways was only possible through the development of novel stable isotope technologies adapted for cultured islets.

Sometimes, new futile cycles can emerge as a by-product of metabolic engineering. Liu et al. [30] applied pulse-labeling experiments with [U-¹³C₆]glucose to discover an energy-dissipating phosphorylation/dephosphorylation cycle that became highly active when B. subtilis was engineered to overproduce N-acetylglucosamine (GlcNAc). Identification and removal of the responsible kinase enzyme led to a doubling of GlcNAc productivity due to restored healthy growth of the overproducing strain. Dynamic ¹³CO₂ labeling experiments have been similarly used to quantify ATP-consuming cycles involving photorespiration [14] and pyruvate cycling [2] in photosynthetic species. While these pathways are not futile cycles per se, they are responsible for losses in photosynthetic efficiencies that can range up to 50% or more [31]. The description of these substrate cycles as ‘futile’ may become outdated, as perceived ‘wasteful’ steps likely evolved to serve a meaningful purpose (e.g., as in the case of photorespiration) and possibly to accommodate environmental fluctuations that are not studied in controlled lab experiments aimed at minimizing variability. With increased emphasis now being placed on reducing metabolic burden in recombinant host organisms by eliminating inefficient pathways, it is likely that systematic identification and evaluation of futile cycles will play an ever more important role in metabolic engineering [32].

Dynamic labeling experiments can be used to cluster metabolites based on their turnover rate

A variety of algorithms have been developed to infer network associations and identify co-regulated metabolic modules based on correlations between measured metabolite abundances [33–35]. The mummichog algorithm [36] provided a significant advance by grouping peaks into modules using statistical techniques that are robust to missing or incorrect peak annotations. While many of these methods have proven useful in extracting biologically meaningful information from high-dimensional metabolomics datasets, few if any have attempted to leverage the additional information that is potentially available from isotope labeling studies to reconstruct networks de novo. To test whether metabolic interactions can be inferred from correlation-based analysis of isotope labeling, we investigated a previous dataset where photosynthetic cyanobacteria were labeled with ¹³CO₂ over 20 minutes [2]. Targeted measurements of ¹³C enrichment were obtained using a combination of GC-MS and LC-MS/MS analysis of metabolite extracts. Hierarchical clustering of the atom percent enrichments (APEs) for each target ion (Fig. 1) showed clear discrimination between metabolites that were rapidly labeled (mainly in the Calvin-Benson cycle) versus those that were slowly labeled (in photorespiration and TCA pathways). A few metabolites that lie on the boundary between these two network regions (e.g., malate and glycerate) exhibited intermediate rates of enrichment. This example illustrates how clustering based on dynamic labeling trajectories can be used to effectively group metabolites into local modules that reflect network proximity and/or pathway activity.

Fig. 1. Hierarchical clustering analysis (HCA) of dynamic ¹³C labeling in cyanobacterial metaboltes.

Target ions were clustered based on APE (left; JD Young unpublished). Pathways and metabolites were color-coded in the network diagram according to the HCA outputs (right). Abbreviations can be found in [2].

Extending temporal descriptions to proteins and lipid

Stable isotope approaches like those used to profile turnover rates of small molecules in central metabolism can also be applied to quantify network dynamics in lipid and protein metabolism. For example, the differences in the unlabeled fraction could be used to group amino acids and distinguish auto-, mixo- and heterotrophic metabolism in duckweed [37]. Stable isotopes are now frequently paired with MS, often high-resolution MS (HRMS), to assess protein turnover more comprehensively across the proteome [38,39]. In plants, inorganic ¹⁵N [40,41] or ¹³C [42] have become preferred tracers for studying protein biosynthesis because of: i) the associated challenges with differential amino acid uptake, ii) interference with feedback regulation, iii) amino acid influences on protein turnover rates, and iv) use of amino acids for non-protein purposes [42]. In animals and humans, administration of deuterated water (²H₂O) has been used for several decades as a non-invasive approach to assess protein and lipid turnover rates in vivo. Shankaran et al. [43] recently applied ²H₂O labeling combined with sensitive tandem MS analysis to perform proteome-wide monitoring of skeletal muscle protein synthesis rates over several weeks from blood samples.

The technical considerations are slightly different for lipid molecular species that number in the hundreds of thousands and comprise many near-overlapping masses between 500–1000 amu, which challenge instrument resolution. Isotope labeling investigations of lipids [44,45] have mostly involved ¹⁴C [46–50]; however, soft electrospray ionization MS methods maintain intact lipid structures [51] and offer new opportunities for stable isotopes [52]. One of the challenges specifically to quantification of stable isotope labeling is that the m/z ranges of different lipid species can overlap due to varying degrees of unsaturation. A current frontier [53] is to use HRMS to resolve these species based on their mass defects (Fig. 2) and eventually build quantitative flux maps for lipid metabolism [54]. In addition, multi-tracer studies (Fig. 3) can be designed to simultaneously probe multiple intersecting metabolic pathways with a single isotope labeling experiment [55]. The capacity to elucidate incomplete metabolic networks and to quantify macromolecule synthesis and turnover (not just accumulation) that contributes to signaling, regulation and metabolic response remains an opportunity for the use of isotope labeling strategies.

Fig. 2. Nominal & HRMS of lipids.

A) Example of phosphatidylcholine (PC) structure. B) Isotopologue distribution of fatty acids at nominal mass resolution. C) HRMS-derived isotopologue distribution.

Fig. 3. Identification of isotopologue peaks.

Change from a molecular mass [M]⁺ to a mass of [M+2]⁺ can be the result of substitution of various isotopes (i.e., ¹⁵N for ¹⁴N, ¹³C for ¹²C, ¹⁸O for ¹⁶O, ¹⁷O for ¹⁶O, or ²H for ¹H) or can result from the addition of hydrogens to the molecule as in the case of changing the degree of unsaturation in a lipid. Here the differences in mass from substitution of the indicated isotope are shown. The change in mass from addition of two hydrogens (+H₂) is also presented.

Profiling spatial segregation of metabolic functions, fluxes and pool sizes

Subcellular compartmentation – when location matters

Metabolism in eukaryotes is organized at the cellular and subcellular levels and can include significant metabolite channeling described elsewhere [56,57]. Subcellular compartmentalization presents a unique challenge for isotope labeling studies due to the existence of the same metabolic pathways in different organelles, creating compartment-specific ¹³C labeling patterns that are difficult to resolve [58–60]. In such systems, knowledge of major metabolic pathways within subcellular compartments and inter-compartmental transporters is essential [61]. Methods to address organelle compartmentalization involve measuring organelle-specific metabolites as readouts of the metabolism in each compartment [62–67], subcellular fractionation of organelles [68,69], or technologies to image metabolites at the cellular level [70,71]. In some cases, careful selection of target analytes from different cell/tissue types can be used to resolve compartment-specific fluxes [72,73] that reflect cellular or subcellular metabolic heterogeneity. For example, metabolic flux ratio (METAFoR) analysis was used to indirectly calculate compartment-specific labeling patterns of pyruvate and oxaloacetate from measurements of PEP (assumed cytosolic) and oxoglutarate (assumed mitochondrial) [74]. This model subsequently enabled the application of ¹³C flux analysis to several species of yeast including Saccharomyces cerevisiae [75,76]. Further generalizations on this approach (i.e., SUMOFLUX) are based on incorporation of machine learning [77], and through the years a number of compartmentalized metabolic flux maps have been generated as reviewed elsewhere [59,60].

Shlomi et al. [78] recently described a spatial-fluxomics approach for quantifying metabolic fluxes specifically in mitochondria and cytosol of mammalian cell cultures, performing isotope tracing in intact cells followed by rapid subcellular fractionation and quenching of metabolism within 25 seconds, followed by LC/MS-based metabolomics analysis. The method was applied to uncover the surprising result that reductive glutamine metabolism was, in fact, the major producer of cytosolic citrate (rather than glucose oxidation) to support fatty acid biosynthesis under standard normoxic conditions in HeLa cells, and that succinate dehydrogenase-deficient tumor cells reverse the citrate synthase flux to produce oxaloacetate in mitochondria [78]. Reversibility of isocitrate dehydrogenase has been previously described in other lipid-producing systems including brown adipocytes [79], melanomas [80] and developing canola embryos [81] and may reflect conserved principles for carbon use efficiency. Other insights into mitochondrial metabolism have been the result of selective permeabilization of the cytosolic membrane with digitonin, as developed by Nonnenmacher et al. [82], or regression-based methods to indirectly infer differences between cytosolic and mitochondrial metabolism from isotope labeling data. As an example of the latter, Christen and coworkers [83] used a regression approach to estimate the fractional distribution of pyruvate between compartments by assuming that lactate is produced from cytosolic pyruvate and alanine is produced from mitochondrial pyruvate.

Multi-cellular interactions between different tissues and the involvement of metabolically inactive pools

The distribution of pyruvate across multiple locations is an indication of the central role of this metabolite. In plants it is involved in cycles including C4 photosynthetic metabolism which relies on fine spatial architecture to implement a carbon concentrating mechanism that enhances carbon assimilation. Labeling studies in C4 plants have described: i) the dynamic and spatial complexity of photosynthetic metabolism that takes place across bundle sheath and mesophyll cells [84,85], ii) gradients of intermediates consistent with diffusion driven movement between cells with the possible exception of pyruvate [84], iii) the role of alanine and pyruvate in shuttling nitrogen between cell types [84,85] and v) the capacity for aspartate to be converted to malate for decarboxylation in the bundle sheath [85]. The rapid labeling in the C4 transfer metabolite malate approaches a stable level quickly that is less than 20% of total ¹³CO₂ enrichment because malate (and to a lesser extent other metabolites) exist as pools in multiple locations, some of which are not involved in the C4 mechanism. Unlabeled pools are termed ‘inactive’ and are also observed in C3 metabolism [14,86], indicating that the spatial separation of active metabolism could reflect subcellular features in addition to differing cell types. However interestingly, sedoheptulose 7-phosphate (S7P), that is integral to Calvin-Benson cycle metabolism shows a complete depletion of the unlabeled (M0) fraction within 15 min of ¹³CO₂ pulse labeling [14] and other Calvin-Benson intermediates become fully labeled within 60 min [84]. The decrease in M0 to near zero levels in S7P and other metabolites suggests that these pools are either not present in spatially segregated locations or that there is rapid movement allowing for near equilibrium labeling descriptions between compartments. The latter, which would include other metabolic pathways such as the reversible steps of the oxidative pentose phosphate pathway (OPPP) thought to jointly occur in the cytosol in some tissues [87], is less likely in leaves because precursors such as glucose 6-phosphate are not labeled quickly and would result in slower overall labeling in S7P. The depletion of M0 indicates that the S7P pool is completely turned over and that none of it is inactive within the time frame of the labeling experiment. Further, in silico modeling (Fig. 4) indicates that some carbon positions take many iterations of the Calvin-Benson cycle to label completely because of fluxes through asymmetric bond breaking and forming reactions with multiple precursor pools. Indeed, the property of asymmetry in labeling provides the mechanistic underpinning to distinguish metabolic steps and to quantify pathway fluxes including the Calvin-Benson cycle [8,9].

Fig. 4. Carbon isotopomers for Calvin-Benson Cycle intermediates.

Isotopic labeling for ribulose 1,5-bisphosphate (RUBP) and sedoheptulose 7-phosphate (S7P) were derived through in silico modeling (DK Allen unpublished) using the INCA framework [1] and assuming no photorespiration. Positional carbon enrichments are indicated following the color scheme of the circles that represent individual carbon atoms. S7P carbon atoms 2 and 7 (orange and dark blue) are identical, as are atoms 4 and 5 (yellow and light blue). Qualitative enrichment patterns are consistent with well-known pathway bond rearrangements [8,9] but can vary with flux descriptions. Dotted gray lines and question marks within the metabolic network describe one of several hypothetical versions of the G6P shunt [10–12] which is not included in the metabolite simulations. Abbreviations are based on [14] and canonical textbook descriptions.

Conclusions/Perspectives

As flux analyses continue to progress and take advantage of new instrumentation or labeling experiment design, there remains a need to increase throughput and examine new organs, tissues, and multi-species consortia. In plants, attempts to close this gap emphasize labeling tissue disks [88], cell suspensions [89,90], hairy roots [91], whole plants [92], petioles or hypocotyls [93], stems [94], or shoot tips [95]. In microbial systems, techniques are progressing from studies of axenic cell cultures to co-culture systems [96] and microbiomes [97] studied under conditions that mimic natural environments. Finally, studies of mammalian metabolism are advancing from cell lines to primary cells [29] to in vivo labeling experiments in whole animals [98,99] or human subjects [100,101]. However, there remains a large disconnect between genetic and metabolic investigations where the former emphasizes phenotyping through easily obtained measurements to select desirable qualities that can then be studied in depth through the latter. There also are few studies that link metabolism across developmental time or the spatial attributes of the entire living plant, animal, or microbial system. Thus, significant hurdles that impede our understanding of metabolism in living systems relate directly to spatial and temporal aspects that would benefit from expanded use of isotope tracers. Development of new methods and enhanced instrumentation will enable more intricate studies to reveal novel aspects of plants, animals, and microbial consortia under continually changing biological conditions, which are most relevant to understanding the natural complexity of metabolism.

Highlights.

• Metabolism is segregated at the subcellular and cellular levels and responds in a context-specific manner to environmental and developmental changes that occur over time

• The dynamics of metabolism can be assessed and quantified as metabolic fluxes using stable and radio-isotopes

• Advances in instrumentation and novel experimental designs with isotopes are leading to new insights that will be important to solve global challenges in food security and disease