The research, published in the journal Biogeosciences by Youssef Saadaoui of the Helmholtz-Zentrum Hereon and colleagues at the University of Hamburg and partner institutions, set out to quantify what mathematicians call aggregation bias in marsh carbon flux estimates. The team’s central finding is striking in its simplicity. When models collapse the wildly variable oxygen conditions of the marsh rhizosphere into a single average value, they overestimate microbial carbon dioxide production by roughly 12 percent on average. Because the amount of carbon a marsh stores is essentially the difference between what plants put into the soil and what microbes breathe out, a systematic overestimate of respiration translates directly into an underestimate of the marsh carbon sink.

To understand why this happens, it helps to picture the underground world of a salt marsh. Oxygen in these soils is anything but uniform. Marsh grasses such as Spartina pipe air down through internal channels in their roots, releasing it into the surrounding sediment through a process known as radial oxygen loss. This creates tiny oxic halos around root surfaces. Just millimetres away, however, oxygen plunges to nearly zero, consumed by microbes and by the chemical re-oxidation of reduced compounds such as dissolved sulfide, iron sulfides, and ammonium. Tides, drainage, sediment texture, and plant physiology all churn this mosaic, so that within a few centimetres a microbe might inhabit anything from a well-aerated hotspot to an effectively airless pocket.

Most ecosystem and land-surface models, however, do not see this patchwork. They represent belowground redox conditions with a single bulk value, an average oxygen level or a uniform moisture scalar applied across an entire grid cell or plant type. The trouble is that microbial respiration does not respond linearly to oxygen. Instead, it follows a saturating, Michaelis-Menten-type curve: when oxygen is scarce, a little more of it produces a large boost in decomposition, but once oxygen is plentiful, additional amounts barely matter. When a curved response like this is evaluated at an average input rather than averaged across the full range of inputs, the result diverges from reality, a classic consequence of Jensen’s inequality, a mathematical principle familiar to ecologists since the early 1990s.

Saadaoui and colleagues turned this abstract principle into a concrete number for marsh soils. They calibrated a widely used enzyme-kinetic respiration model, the Dual Arrhenius and Michaelis-Menten framework, against long-term laboratory incubations of marsh soils collected from the Elbe estuary in Germany. In those incubations, soils of varying organic carbon content were held at controlled temperature and moisture for up to 465 days while carbon dioxide production was tracked by gas chromatography. Crucially, the incubations served only to pin down the carbon-side kinetics of the model; oxygen effects were then manipulated separately in the aggregation experiments, with temperature and substrate availability held fixed so that any difference between model configurations could be attributed purely to how oxygen was treated.

The team then ran the same calibrated model in two configurations. In the uniform configuration, respiration was evaluated at the arithmetic mean of a prescribed rhizosphere oxygen distribution, generated from realistic variation in water content and air-filled porosity driven by tidal flooding and drainage. In the heterogeneous configuration, respiration was averaged across the full oxygen distribution, capturing the fact that some microbes live in oxygen-rich zones while others barely receive any. For the gas-phase oxygen proxy, the uniform approach produced fluxes consistently higher than the heterogeneous average, with a mean bias of about minus 12 percent, meaning the single-average treatment overestimated aerobic carbon dioxide release by roughly a tenth. The overestimate ranged between about 6 and 13 percent depending on the microbial oxygen affinity assumed.

The study also revealed when the bias matters and when it does not. The key quantity is the ratio of mean oxygen concentration to the half-saturation constant, the oxygen level at which microbes achieve half their maximum respiration rate. When mean oxygen sits far below this threshold, the respiration response is nearly linear and averaging errors vanish; the dissolved-phase oxygen proxy, whose concentrations are hundreds of times smaller than the gas-phase equivalent, produced a negligible bias of about 0.05 percent. When oxygen is effectively saturating, the response flattens and the bias again shrinks. The largest errors arise in the intermediate regime, where respiration is strongly limited by oxygen but not starved of it, and where spatial variability is substantial. Narrowing the oxygen distribution, as in a benchmark representing nearly saturated soils, reduced the bias to half a percent.

A diagnostic analysis confirmed the mechanism. By holding the mean oxygen constant and systematically increasing its variance with symmetric two-point mixtures, the researchers showed that the bias grows nearly linearly with variance, exactly as a second-order Taylor approximation based on the curvature of the oxygen-limitation term predicts. The bias remained negative across every distribution shape tested, including lognormal fields, indicating that the effect is a structural property of concave oxygen limitation rather than an artefact of any particular assumption. In practical terms, this means modellers can anticipate the error from just three pieces of information: the mean oxygen concentration, its variance, and a calibrated half-saturation constant.

The implications ripple outward to global carbon accounting. Published estimates place global salt-marsh carbon burial at roughly 10 to 21 teragrams of carbon per year, though the full range of published accumulation-rate estimates spans from about 0.9 to more than 31 teragrams. If a bias of around 12 percent were broadly representative of tidal marshes, correcting for it would raise inferred global sequestration by on the order of 1 to 2.5 teragrams of carbon per year, equivalent to roughly 0.1 to 0.25 gigatonnes over a century. The authors are careful to frame these as order-of-magnitude extrapolations rather than forecasts, and the corrected flux remains small beside fossil fuel emissions. Still, for blue carbon inventories and coastal greenhouse gas budgets, where marshes are counted among nature’s most efficient carbon sinks, a systematic error of this size is far from trivial.

The study also carries a broader lesson for Earth system science. As mechanistic, oxygen-explicit respiration schemes migrate into regional and global models, the way oxygen is aggregated within each coarse grid cell becomes a first-order control on predicted fluxes, not a secondary detail. The researchers offer modellers a lightweight remedy: a curvature-based correction factor derived from the local oxygen variance, which their tests show can remove most of the bias when variability is modest. Where variability is large or poorly constrained, propagating a plausible range of oxygen variance into flux uncertainty is the more honest option. They further caution that defining the uniform baseline in terms of mean water content rather than mean oxygen can, in extreme cases, even flip the apparent sign of the effect. Realising the full promise of these diagnostics, the authors note, will require more high-resolution microelectrode and planar optode measurements of rhizosphere oxygen across tidal elevations, plant species, and seasons, along with better constraints on microbial oxygen kinetics in marsh soils. For now, the message is clear: beneath the quiet surface of a marsh, the invisible architecture of oxygen is quietly reshaping how much carbon we think these ecosystems can hold.

Subject of Research: Aggregation bias in modelled marsh soil carbon dioxide fluxes caused by millimetre-scale rhizosphere oxygen heterogeneity

Article Title: Quantifying aggregation bias in marsh carbon flux estimates caused by rhizosphere oxygen heterogeneity

Article References: Saadaoui, Y., Beer, C., Mueller, P., Neiske, F., Becker, J. N., Eschenbach, A., & Porada, P. (2026). Quantifying aggregation bias in marsh carbon flux estimates caused by rhizosphere oxygen heterogeneity. Biogeosciences, 23(19), 6803-6816. https://doi.org/10.5194/bg-23-6803-2026

Image Credits: AI Generated

DOI: 10.5194/bg-23-6803-2026

Keywords: tidal marshes, blue carbon, rhizosphere oxygen, heterotrophic respiration, aggregation bias, Jensen's inequality, Michaelis-Menten kinetics, DAMM model, carbon flux, soil incubation, Elbe estuary, biogeochemical modelling

Aldolase catalyzes the fourth step of glycolysis, cleaving fructose-1,6-bisphosphate into two three-carbon sugars, glyceraldehyde-3-phosphate and dihydroxyacetone phosphate. Because these triose phosphates feed both the lower half of glycolysis and the glycerol backbone used to build lipids, the enzyme sits at a critical junction of central carbon metabolism. Most of what is known about aldolase comes from medical research, particularly cancer biology, where studies have shown that blocking aldolase can push glucose-6-phosphate into the pentose phosphate pathway, increasing the production of NADPH, the reducing power that cells use for biosynthesis. The Xiamen University team, led by Yiting Zhang and Xueping Ling, asked whether the same logic could be applied to a microbe whose entire commercial value depends on how much reducing power and carbon skeleton it can pour into fatty acid synthesis.

To find out, the researchers constructed a knockout strain by replacing the aldolase gene, designated ALDOA, with a bleomycin resistance cassette through homologous recombination. They amplified roughly 420-base-pair sequences flanking the gene, ligated them into the pBlue-zeo plasmid around the resistance marker, and introduced the construct into wild-type cells by electroporation. Transformants were selected on plates containing 50 milligrams per liter of bleomycin, and PCR sequencing of the homologous arms confirmed that the native gene had been successfully replaced. The resulting strain, called ΔALDOA, was then compared with the wild type over a 168-hour shake-flask fermentation using glucose-based medium.

The immediate consequences of the deletion were negative for bulk productivity. Final biomass in the knockout strain was 9.6 percent lower than the wild-type maximum, and total lipid production fell by 23.2 percent. During active growth the deficits were even larger: at 96 hours the mutant’s cell dry weight was up to 40.61 percent below the wild type, and at 72 hours its lipid yield lagged by as much as 49.34 percent. Glucose consumption told a similar story. The wild-type strain essentially exhausted the sugar supply by 120 hours, whereas the knockout had used only about half of it at that point and needed until 168 hours to finish. Intriguingly, the mutant actually consumed glucose faster than the wild type in the earliest phase of cultivation, apparently ramping up overall metabolism to compensate for the metabolic bottleneck, before its impaired glycolysis caught up with it in the middle and later stages.

The fatty acid composition, however, moved in a very different direction. At 120 hours the proportion of saturated fatty acids in the knockout strain was about 17.4 percent lower than in the wild type, driven mainly by reductions in the C16:0 and C14:0 species. Polyunsaturated fatty acids, including DHA, docosapentaenoic acid, and eicosapentaenoic acid, rose by 28.0 percent overall, with DHA itself up 22.9 percent and EPA up a remarkable 75.0 percent. When the team broke the lipid pool down into triglycerides and phospholipids, they found that polyunsaturated fatty acids are mostly stored in triglycerides, at 35 to 48 percent of that fraction, while phospholipids carry 10 to 15 percent and contain no detectable EPA at all. The knockout boosted the polyunsaturated content of triglycerides specifically, with DHA rising 27.4 percent at 72 hours and 20.8 percent at 120 hours, while the main saturated species, C16:0, dropped by 46.5 percent and 36.0 percent at the same time points.

Phospholipidomics provided a mechanistic explanation rooted in how DHA travels through the cell. Phosphatidylcholine, the most abundant phospholipid in Schizochytrium at roughly half of the total, is known to be the first docking site for newly synthesized DHA before the fatty acid is transferred to triglycerides for storage. In the knockout strain, the proportion of phosphatidylcholine was 11.5 percent higher than the wild type at 72 hours and 7.5 percent higher at 120 hours, and lysophosphatidylcholine, its precursor in the glycerophosphorylcholine acylation pathway, reached 4.8 times the control level at 72 hours. Meanwhile, phosphatidylglycerol, a phospholipid tied to cell growth, collapsed by 92.2 percent at 72 hours, and phosphatidylinositol, phosphatidylserine, and phosphatidic acid all declined as well. Because these species are built through the Kennedy pathway from dihydroxyacetone phosphate-derived precursors, their depletion directly reflects the loss of aldolase activity upstream.

Gene expression analysis reinforced the picture. Transcripts of the fatty acid synthase gene, which produces saturated fatty acids, fell in the mutant, while the chain length factor gene of the anaerobic polyketide synthase cluster, responsible for elongating polyunsaturated chains, rose sharply in the middle and later stages of fermentation. The gene for glucose-6-phosphate dehydrogenase, the gateway enzyme of the pentose phosphate pathway, followed a similar pattern, indicating enhanced flux through that route and a greater supply of NADPH for fatty acid desaturation and elongation. Transcripts of phosphatidic acid phosphatase and diacylglycerol acyltransferase, the enzymes that convert phosphatidic acid into triglycerides via the Kennedy pathway, decreased, whereas the genes for phospholipid diacylglycerol choline phosphate transferase, or PDAT, and lysophosphatidylcholine acyltransferase, or LPCAT, were strongly upregulated. Together these shifts suggest the mutant builds its triglycerides increasingly through the phosphatidylcholine route, shuttling DHA from phospholipid storage into neutral lipid reserves.

Metabolomics by gas chromatography-mass spectrometry identified 35 differential metabolites that mapped the rewiring across glycolysis, the tricarboxylic acid cycle, the pentose phosphate pathway, and amino acid and sterol synthesis. Sugar alcohols associated with the oxidative pentose phosphate route surged, with arabitol up 4.06-fold and ribitol up 7.84-fold at 120 hours. The knockout strain also accumulated protective metabolites: inositol, a signaling molecule against oxidative stress, rose threefold, and mannitol, a hydroxyl radical scavenger, increased 1.23-fold, while glycine rose to supply precursors for glutathione synthesis. Flow cytometry using a fluorescent probe confirmed that reactive oxygen species were lower in the mutant at both measured time points. Since polyunsaturated fatty acids are vulnerable to free-radical oxidation, the reduced oxidative environment plausibly helps preserve the DHA the cells do make, complementing the increased reducing power available for synthesizing it in the first place.

The authors propose a coherent metabolic model: deleting ALDOA weakens glycolysis and the TCA cycle, slowing growth, and starves the Kennedy pathway of dihydroxyacetone phosphate, cutting total lipid and phospholipid output. In compensation, carbon is pushed into the pentose phosphate pathway, raising NADPH for polyunsaturated fatty acid synthesis, while the cell reinforces the glycerophosphorylcholine acylation route and PDAT-mediated transfer so that more DHA is bound to phosphatidylcholine and then moved into triglycerides. The trade-off is clear, with less biomass but a richer omega-3 profile per cell. The team notes that future work using CRISPR or RNA interference to tune, rather than abolish, aldolase expression could improve DHA accumulation without the growth penalty, offering a new metabolic engineering strategy for one of the world’s most important microbial oil producers.

Subject of Research: Aldolase gene disruption and carbon flux redistribution for DHA synthesis in Schizochytrium

Article Title: Functions of aldolase in lipid synthesis of Schizochytrium sp. by gene disruption to switch carbon metabolism

Article References: Zhang, Y., Wu, X., Guo, X., Li, K., Lu, Y., Lin, X., & Ling, X. (2024). Functions of aldolase in lipid synthesis of Schizochytrium sp. by gene disruption to switch carbon metabolism. Blue Biotechnology, 1(1), Article 17. https://doi.org/10.1186/s44315-024-00014-6

Image Credits: AI Generated

DOI: 10.1186/s44315-024-00014-6

Keywords: Schizochytrium, aldolase, DHA, polyunsaturated fatty acids, glycolysis, pentose phosphate pathway, carbon flux, metabolic engineering, phospholipids, triglycerides, NADPH, marine biotechnology