Introduction

Peatlands are critical ecosystems holding at least double the carbon (C) content of Earth’s standing forests1. The catotelm, a peatland layer characterized by low temperatures, anaerobic conditions, and recalcitrant soil organic matter (SOM), maintains slow decomposition rates2,3. Yet, as temperatures rise, peatlands face an alarming threat: acceleration of organic matter decomposition relative to carbon fixation and exacerbating climate change. This shift could unleash greenhouse gases (GHGs) through rapid degradation of subsoil C. Conversely, elevated CO2 concentrations may alter plant communities and potentially increase the input of dissolved substrate belowground by stimulating photosynthetic production, thereby increasing C inputs and primary production4. However, the potential climate effects of new C inputs from plants are unclear. A recent study by Wilson et al.5 revealed that warming increased microbial activity and GHG production, primarily driven by greater labile C inputs from plants of peat bogs, especially in the surface layers. However, the complex relationship between these environmental shifts, the influx of carbon, microbial behavior, and GHG emissions in peatlands remains a complex and undetermined topic. Further research is therefore essential to assess the peatland’s carbon balance in relation to climate feedback.

The decomposition of soil organic matter (SOM) is influenced by substrate chemistry, microbial dynamics, and environmental factors6,7. Deep soils contain old C, which is often considered to be inert and resistant to microbial degradation3. However, changes like increased rooting depths or oxygen exposure can revive microbial activity4. Previous research from our group has demonstrated that microbial communities in deep peat bogs, like the S1 Bog at the Marcell Experimental Forest (MEF), rely heavily on recent photosynthate (labile carbon) transported from the surface via vertical advection, suggesting that carbon cycling at depth at the MEF is not isolated from surface processes3. Therefore, predicting how deep soil carbon will respond to climate change requires a clear understanding of how changing environmental conditions, organic matter inputs, microbial communities, and their metabolic activity all interact. While much is known about microbial control of carbon cycling in surface soils, the dynamics of deep soil ecosystems under a changing surface environment remain largely unexplored8. In this regard, soil priming, the process of fresh plant-derived labile organic matter facilitating the breakdown of native organic carbon in deep soil layers, plays a crucial role. Soil priming was first documented nearly a 100 years ago9, and since has been a prime research interest due to its importance in mitigating the effects of climate change10. However, despite the wealth of research on the priming, especially with respect to the influences of soil characteristics11,12,13,14,15, substrate inputs8,16, microbiome properties17,18,19, even the long-term climate factors such as precipitation20 and warming21, the role of specific microbial taxa on these priming regulatory processes is still poorly understood21,22,23.

Microbial ecology aims to understand the role of microorganisms mediating ecosystem function, necessitating the direct measurement of microbial metabolic processes24. Stable isotope probing (SIP) has emerged as a powerful tool for identifying actively metabolizing populations25,26,27. However, investigations integrating DNA-SIP with metabolomics (in particular, Stable isotope assisted metabolomics - SIAM) to simultaneously characterize microbial community structure and function in complex ecosystems are scarce, if even existing. Unlike traditional metabolomics, which can be limited by difficulties in interpreting changes in metabolite pools due to the combined effects of production and consumption28,29, SIAM metabolomics offers a more direct approach. By feeding isotopically-labeled substrates, such as 13C-glucose, to the system and tracking the incorporation of these labels into newly formed metabolites, we can directly trace metabolic activity at the cellular and organismal levels30,31. Further, while past isotope tracing studies often focused on a limited number of metabolites especially in clinical studies, advancements in technology and data processing have enabled a broader metabolite coverage, facilitating a more holistic understanding of system-wide metabolic profiles of active microbial populations and their contributions to biogeochemical cycling31,32.

Here, we employed a novel DNA SIP-SIAM approach to assess the impact of labile carbon addition on soil-microbe dynamics and carbon flow in Northern peatlands (See Methods: synopsis). This approach involved tracing the incorporation of 13C-glucose into different microbial communities and their microbial metabolites. Peat soils from the surface (acrotelm < 0.30 m) and deep (catotelm > 1 m) layers of the S1 bog in the Marcell Experimental Forest were incubated across three conditions: (a) 13C-glucose amendment, (b) 12C-glucose amendment, and (c) no amendment (Supplementary Fig. 1). The unlabeled 12C-glucose treatment was used for the majority of analyses to track general metabolite and microbial community responses, while the 13C-labeled glucose condition enabled targeted tracking of metabolites and organisms actively incorporating the isotopic label. Further, the two depths were chosen to investigate microbial responses in contrasting environments: an oscillating redox zone with readily degradable sugars such as glucose and a deeper zone with limited oxygen and more recalcitrant organic matter3.

We hypothesized that the addition of labile carbon substrates, such as glucose, would reveal the vulnerability of deep C stores to changing root inputs and rhizodeposition composition. Specifically, we anticipated that (i) glucose supplementation would activate distinct metabolic pathways in deep peat microbial communities compared to surface peat, driving shifts in microbial composition and diversity over time; (ii) microbes inefficient at metabolizing glucose would decline, while taxa capable of direct glucose uptake or benefiting from metabolic byproducts would increase in abundance; and (iii) energy gained through microbial degradation of glucose would prime decomposition of complex organic matter through co-metabolism, increasing microbial attack and the breakdown of hard to decompose organic matter (e.g.; lignin, cellulose, aromatics, and long-chain hydrocarbons). Ultimately, these dynamics are expected to accelerate organic matter decomposition, resulting in higher GHG production rates in deep peat layers upon glucose amendment, relative to untreated controls (Fig. 1). By integrating DNA-SIP with SIAM, we provide a powerful molecular toolset for mechanistically linking the active microbial populations to their metabolic functions during priming of older, recalcitrant soil carbon pools upon labile substrate addition. This multi-omics approach enables direct connections between observed microbial/metabolic biomarkers and greenhouse gas emission potentials. Such integrated insights into microbial priming processes are critically needed to improve parameterization of microbial response mechanisms and better constrain carbon cycle feedbacks within predictive climate models as peatland ecosystems face warming and other global change pressures.

Fig. 1: Microbial breakdown of recalcitrant compounds in deep peat upon labile carbon additions.
figure 1

Climate change can lead to increased plant productivity, thus vascular plants covering peatlands could add more labile root exudates into the subsoil. These substrates could be transported into thermodynamically constrained deep peat layers, which contain massive amounts of the world’s C. These labile carbon addition revives the microbial activity, changes microbial dynamics, and increases the breakdown of recalcitrant Soil Organic Matter (SOM), leading to release of Greenhouse Gases (GHGs) into atmosphere.

Full size image

In summary, we used metabolomics and genomics, leveraging 13C-labeling and three complementary metabolomics techniques, to obtain insights into how labile carbon inputs like glucose can prime the breakdown of vast, previously stable deep peat carbon reserves through distinct metabolic and microbial community shifts. The energy supplemented by glucose-stimulated taxa capable of direct glucose utilization or benefiting from byproducts, enhanced their activity, and primed the breakdown of complex aromatics and older organic matter, accelerating carbon cycling over time and CO2 release. As a powerful multi-omics approach combining DNA-SIP and SIAM, we could definitively distinguish active glucose metabolizers from microbes driving priming of native carbon pools based on isotopic labeling patterns - a capability unattainable through either method alone. This enabled pinpointing the specific microorganisms and metabolic pathways involved in glucose labile substrate transformations versus recalcitrant carbon decomposition. Tracking isotopically labeled metabolites illuminated how supplemental glucose flowed through microbial metabolic networks to stimulate older carbon mineralization, offering molecular details about these priming mechanisms in peatlands. While previous work showed labile carbon can induce priming across soil systems assessed through different approaches including GHG production33,34, Mass spectroscopy35, NMR36, isotope tracing37,38, and microbiome characterization39, our integrated DNA-SIP/SIAM approach is the first to probe these dynamics in peatlands at the coupled microbial population and metabolite level, especially leveraging on the metabolite sizes, across two different biologically distinct soil depths. The microbial/metabolic biomarkers identified provide novel indicators for monitoring and predicting greenhouse gas emissions as climate change destabilizes deep peat carbon stores through altered root inputs that relieve thermodynamic constraints. While our study focused on anaerobic conditions that dominate our study site, we acknowledge that water-table fluctuations in peatlands can create periods of aerobic conditions, particularly in surface layers. Future studies examining priming effects under aerobic conditions would complement our findings and provide a more complete understanding of peatland carbon dynamics across varying hydrological conditions.

Overall, this multi-omics investigation elucidated how changing resources restructure the peatland microbiome to accelerate deep carbon processing, offering critical process-level understanding to improve model parameterization of microbial priming responses and carbon cycle feedback as these ecosystems undergo global change pressure. Such advancements are vital for better constraining ecosystem carbon cycle feedback as peatlands face warming, elevated CO2, and other global change drivers destabilizing their vast recalcitrant carbon reservoirs.

Results and discussion

Glucose addition induces distinct metabolic shifts that are stratified by peat depth

Comprehensively capturing the diverse metabolite pool in complex soil environments poses challenges that cannot be adequately addressed by any single analytical platform. Therefore, we employed a novel integrated metabolomics approach utilizing three complementary analytical methods 1H NMR, GC MS, and LC MS/MS. This multi-platform strategy emphasizes the inherent biases of each technique in preferentially detecting different metabolites based on their mass (Fig. 2A), thereby providing an unprecedented level of comprehensive metabolite characterization crucial for evaluating microbial processes in complex soil systems like peatlands. Notably, this approach identified a total of 658 unique metabolites, with 34 observed using 1H NMR, 110 identified by GC MS, and 514 detected through LC MS/MS. The detailed breakdown of these metabolites and their downstream processing can be found in the methods section.

Fig. 2: Overview of the metabolites detected across the three analytical platforms employed and their respective ordinations.
figure 2

Panel (A) displays a merged histogram of the exact masses (amu) of metabolites detected across Nuclear Magnetic Resonance (1H NMR), Gas Chromatography-Mass Spectrometry (GC MS), and Liquid Chromatography-Tandem Mass Spectrometry (LC MS/MS) analytical platforms. The total number of metabolites from each method used in subsequent analyses is indicated in parentheses in the legend. Panel (B) depicts the Principal Coordinate Analysis (PCoA) ordinations generated using Bray-Curtis distances for each analytical method. The variation explained by the first two coordinate axes is mentioned under each plot. Samples are colored based on sampling depth and shaped according to timepoints, as shown in the legend below. The statistical significance of sample separation based on soil depth (surface vs. deep), treatment effect (no glucose vs. glucose), and time (days 7–70) was assessed using permutational multivariate analysis of variance (PERMANOVA) through the adonis2 function in R112. Only the significant components and interactions are presented next to each ordination plot, as per the legend. Individual PCoA ordinations for each depth are provided in Supplementary Fig. 2.

Full size image

Subsequently, we assessed the differences in overall metabolic profiles across depths (surface vs. deep), treatments (unlabeled glucose vs. no addition), and timepoints (initial vs. final) (Fig. 2B). As detailed below, depth was the primary factor driving differences in metabolite composition between samples across all three analytical methods (Fig. 2 Ba, Bb, Bc, p.value < 0.001 and R2 = 0.38, 0.49,0.50, respectively). This vertical stratification of metabolic profiles aligns with previous observations of depth-dependent variations in physical, chemical, and microbial properties in peat soils3, likely reflecting the contrasting geochemical and biological regimes shaping carbon transformation in these discrete zones. While the surface acrotelm experiences fluctuating redox conditions and inputs of relatively labile plant compounds, the deeper catotelm exists as a permanently anoxic, nutrient-limited environment dominated by complex, recalcitrant organic matter accumulation over centuries to millennia. As such, surface and deep peat harbor fundamentally distinct metabolic landscapes finely attuned to their divergent biogeochemical drivers.

Key discriminatory metabolites highlight depth-specific responses to glucose

We conducted a depth-specific assessment to uncover the effect of unlabeled glucose addition on soil processes (Supplementary Fig. 2). For surface samples, time was the major factor in governing ordination across all three analytical platforms. This is likely due to the inherent sugar-rich nature of surface peat, which reduces its responsiveness to glucose addition3. This observation is consistent with Basiliko et al.40, who also reported only a minor priming effect of synthetic root exudate additions, on surface peat40. Additionally, our analysis of changes in bulk organic matter composition properties revealed organic matter with higher Gibbs Free Energy (GFE) values in control deep samples compared to their surface counterparts (Supplementary Fig. 3). Organic matter with high GFE represents less bioavailable compounds, as these are the materials that require higher energy input for microbial decomposition. This is especially true in anoxic systems, where the lack of oxygen as a terminal electron acceptor forces microbes to rely on less efficient terminal electron acceptors. As a result, microbes in these deep, anoxic peat layers will preferentially degrade organic compounds that yield higher amounts of energy or require less energy investment for catabolism2. Furthermore, these thermodynamic constraints could severely inhibit the activity of phenol hydrolases41, leading to slowed degradation of deep C42.

As hypothesized, glucose addition had a greater influence on deeper layers due to their distinct organic matter composition and thermodynamic constraints, with the latter referring to the lack of favorable free energy change required to drive microbial reactions. These observations are in line with previous work where the addition of fresh carbon inputs into subsoil has resulted in the mineralization of thousands years old SOM43, attesting to the pivotal role of fresh carbon inputs in destabilizing deep soil carbon44. However, for the deeper layers, we observed the effect of glucose addition (treatment effect) to vary across the different metabolite size ranges (masses) (Supplementary Fig. 2). For instance, 1H NMR ordinations showed strong separation (R2 = 0.38 and p < 0.001) based on the treatment effect, suggesting a significant impact of glucose on low molecular weight (MW) metabolites (<150 Da). GC MS metabolites (~100–450 Da) exhibited a borderline significant separation (R2 = 0.16 and p < 0.1), while LC MS/MS data (~300–1000 Da) showed no statistical grouping of samples based on the glucose treatment.

Concurrently, we tracked changes in CO2 and dissolved organic carbon (DOC) concentrations over time, following the addition of labeled and unlabeled glucose. Glucose addition stimulated SOM decomposition, leading to increased CO2 release into the gas phase (p < 0.05, Supplementary Fig. 4A), carbon release into solution (p < 0.05, Supplementary Fig. 4B), and enhanced CO2 production rates in both the surface and deep peat samples (p < 0.05, Supplementary Fig. 4C). This observation suggests that deep microbiota is likely to overcome thermodynamic limitations due to the presence of glucose, which was especially evident in low MW metabolites, which include end products of glucose fermentation, as explained by the microbial activation model of priming45.

However, an overwhelming majority of the studies on soil organic matter priming is restricted to a particular layer of soil, which inadvertently hinders the differential impact of these processes at different soil layers46 along their organic matter3,47 and microbial composition48,49 gradients as discussed above. Hence, to identify distinctive metabolic trends between surface and deep soil layers of control (no glucose) and unlabeled glucose-treated samples, we applied Orthogonal Projections to latent Structures Discriminatory Analyses (OPLS-DA)50. Accordingly, a comparison of surface and deep control samples, resulted in 91 uniquely discriminatory features, accounting for 33.2% of total discriminatory metabolites (n = 274) (GitHub workbook 1). On the other hand, for glucose-treated samples, we identified 152 unique features that differed between surface and deep layers, accounting for 45.4% of total metabolites (n = 335). A total of 183 metabolites were discriminatory for both comparisons, highlighting the dominant influence of the soil depth (Fig. 3A), where 178 metabolites showed similar abundances across soil layers. Among these, simple sugars and their derivatives such as 1,4-xylobiose, D-ribose, and D-gluconic acid were more abundant in surface layers for both treatments, attesting to the typical carbohydrate rich nature of the surface peat soil as observed before3. In addition, root exudates such as phenolic compounds51, putrescine52, and glycerol53 were also significantly more abundant in surface samples for both treatments while myo-inositol54 was significantly abundant only for the control treatment in surface samples. These observations collectively underscore the sugar-rich nature of surface peat samples, likely attributed to surface vegetation.

Fig. 3: Significant metabolic changes and associated pathways upon glucose treatment in deep samples.
figure 3

Panel (A) summarizes the results of Orthogonal Projections to Latent Structures—Discriminatory Analysis (OPLS-DA), performed independently for control and glucose treatments discriminating sampling depth across each metabolite datasets. Results are represented using a DiVENN diagram101 to highlight unique and shared discriminatory metabolites identified for control and treatment groups. Each dot represents a metabolite which is colored based on the abundance; surface abundant (surface > deep: green) and deep abundant (surface <deep: brown). Shared discriminatory metabolites with differences in their abundances across two analyses (surface > deep for control but surface <deep for glucose, vice versa), were colored yellow. The abundance of unique discriminatory metabolites (for the control and glucose-amended group) detected with NMR and GC MS methods are shown in the heatmaps. Each metabolite in the heatmap was scaled using a z-score to enhance the visualization of abundance distribution. Using 1H NMR, the consumption of 13C labeled glucose was tracked separately, and the 13C enriched metabolites identified were cross referenced to the above depth-discriminatory metabolites, irrespective of the treatment. These metabolites are denoted with a star. Panel (B) shows the differential pathway shifts for a selected set of pathways based on metabolites that showed an absolute log2 fold change in abundance greater than 1 between treatments, for aggregated timepoints. Bars attest to the number of matched metabolites to the respective KEGG reference pathway63, colored based on their comparative abundance.

Full size image

On the other hand, lipids such as lignoceric acid, octadecanenitrile, arachidic acid, behenic acid, and serratene were notably abundant in the deep layers for both treatments. This observation is consistent with the recalcitrant nature of these compounds, making them less susceptible to microbial degradation3. Interestingly, byproducts of glucose fermentation, such as short-chain fatty acids (lactate, butyrate, and acetate), were abundant only in the glucose-amended deep peat, suggesting enhanced fermentation. In addition, the increased abundance of benzoate in the deep layers exclusively for the glucose treatment is consistent with the degradation of benzene-containing compounds in the deep layers upon glucose addition. Such benzoate degradation pathways produce CO255, which is also consistent with our CO2 measurements and investigations into microbial carbon preferences (Supplementary Fig. 4). Furthermore, these observations are in line with deep soil priming effects observed before upon supplementation with labile C inputs56,57,58, which is concurrently supported by our CO2 accumulation and porewater DOC measurements, both of which increased upon glucose supplementation (Supplementary Fig. 4A, B).

Furthermore, the surface samples initially showed a higher contribution of priming to total CO2 production (27%) on day 7, which declined to 18% at longer incubation times, suggesting a transition to alternative carbon sources as the initial sources were depleted, which is in line with the observations made by Liao et al.58. Deep samples declined from 35% to 6% priming, emphasizing evolving microbial dynamics (Supplementary Fig. 4E). The differing priming trends over time, between surface and deep samples provide insight into how microbial communities at different depths use the added glucose. In line with our hypothesis, deep samples showed higher initial priming, indicating a greater impact of glucose in these carbon-limited layers. In surface soils, microbes likely have more regular access to fresh carbon inputs from plant litter, as discussed earlier. Thus, they may preferentially use the easily accessible glucose initially to degrade complex, but relatively fresh substrates, contributing to the higher 27% priming effect early on. As this labile glucose is consumed, surface microbes transition to their typical mix of carbon sources, reflected in the priming effect declining to 18%. In contrast, microbes in deeper samples are carbon limited as deep peat is enriched with energetically rich, but difficult to degrade compounds3,59. This is further evident in the DOC values, where deep peat harbors lower DOC concentrations than surface samples (Supplementary Fig. 4B). The addition of glucose allows microbes to reinvest the energy created from glucose fermentation into degrading these energy-rich but recalcitrant metabolites overcoming their thermodynamic limitations as indicated by the strong initial priming (35%). As the energy yield from the glucose fermentation pathways become depleted, microbes have less energy to invest in recalcitrant compound degradation and priming rapidly declines to just 6% in deep samples. To balance their redox status and continue to produce energy, microbes likely resort to producing butyrate, which sinks electrons60, and acetate, which generates ATP through substrate-level phosphorylation61, indicated by the elevated butyrate and acetate levels in the deep peat (Fig. 3A). The persistent extensive priming in surface samples (18% vs. 6% in deep) by day 70 suggests microbes there retain a preference for complex but fresh substrates from surface inputs whereas deep soil microbes take advantage of transient priming to catalyze recalcitrant deep carbon.

Tracing 13C-enrichment unveils glucose utilization pathways

This represents the first study to comprehensively track the fate of isotopically-labeled glucose metabolites in a complex peatland soil system by integrating three complementary analytical platforms 1H NMR, GC MS, and LC MS/MS. This multi-pronged metabolomics approach enabled clear delineation of which metabolites originated directly from the 13C-glucose amendment versus those generated from priming of older carbon pools. We traced the origin of differentially abundant metabolites by analyzing their 13C enrichment, linking them to 13C-glucose addition. For this, separate samples amended with 13C-glucose under identical conditions were examined. 1H NMR identified 13C-enriched low MW metabolites, while LC MS/MS tracked 13C incorporation into high MW metabolites, tracing the fate of added carbon. 1H NMR revealed that 10 low MW metabolites were 13C-enriched after glucose addition (Supplementary Table 1). Notably, acetate and butyrate, which were previously identified as significantly more abundant in deep control samples, were found to be labeled. This 13C-enrichment of acetate and butyrate further confirms that deep microbial communities are carbon substrate-limited, and that glucose addition activates fermentation pathways in the deep peat layers, as discussed above. Additionally, among the other discriminatory metabolites identified earlier as differentially abundant, we also observed 13C-enrichment in methanol, ethanol, and valerate, indicating their potential origin from the 13C-glucose.

In contrast, LC MS/MS analysis did not detect any 13C-enrichment in larger MW compounds. This novel integrated metabolomics strategy provided unprecedented insights into glucose metabolic routing. The 13C-enrichment of low molecular weight fermentation products like acetate and butyrate confirms deep microbial communities are carbon-limited and glucose activated fermentation pathways. The predominance of fermentation is corroborated by our previous research which showed that due to the paucity of terminal electron acceptors in peat soils, CO2 is produced primarily by fermentation and terminal decomposition pathways are tuned to the supply of fermentation intermediates62. Thus, the energy from glucose utilization likely primed the microbial degradation of larger, older carbon compounds in the deep peat layers45. Contrastingly, no 13C-labeling in higher mass compounds suggests the added glucose fueled catabolic respiration/fermentation rather than anabolism, priming degradation of larger recalcitrant carbon pools as corroborated by previous evidence of peat decomposition being tuned to fermentation intermediates.

Mapping metabolites to pathways elucidates microbial functional shifts

While most priming studies have examined pathways through the lens of greenhouse gas emissions or broad microbial community changes, our integrated multi-omics approach provides unprecedented molecular insights into the specific metabolic pathways and mechanisms underlying how microbes biochemically respond to and process labile carbon inputs like glucose. By combining DNA-SIP to identify active microbial populations with stable isotope metabolomics to map labeled and unlabeled metabolic transformations, we could delineate key pathways activated upon glucose amendment across the peatland depth profile. Accordingly we mapped metabolites that exhibited large changes (Log2FC > 1), to Kyoto Encyclopedia of Genes and Genomes (KEGG) mapping63 to compare between control and glucose samples in each layer across aggregated time points. This analysis identified 15 key pathways that showed notable changes in the number of mapped metabolites across treatments and time (Fig. 3B). While surface samples had comparatively fewer pathway matches across treatments and time, in the deep samples, initial time points had more matches originating from control-abundant metabolites, with later time points overwhelmingly shifted to matches originating from glucose-abundant metabolites (notably for the aromatic compound degradation pathway). This observation further supports the persistence of glucose-induced priming effects in the deep samples, as observed before34,64,65,66. Similarly, Zhang et al.36 also reported that the added glucose induces notable changes over time in both labile and recalcitrant organic compounds in native SOM in deep soil incubations in addition to priming the decomposition of aromatic C–O36. Furthermore, the investigation of the mapped metabolites in aromatic degradation pathway revealed that the major differential contributors between control initial and glucose final pathway hits were acetate and propionate, which may partially contribute to the accumulation of short-chain fatty acids (SCFAs) observed earlier for the glucose samples (GitHub workbook 2). Similar trends were observed for other KEGG reference pathways including carbon metabolism66, amino acid metabolism pathways including phenylalanine, tyrosine67, and methane metabolism68, agreeing with prior observations made across different soil systems. Our metabolomic analysis revealed glucose-derived methanol and ethanol, both known substrates for methanogenesis69. These compounds, together with the observed acetate enrichment in glucose-amended deep samples, suggest that labile carbon additions may stimulate methanogenic activity in deep recalcitrant layers. The accumulation of these methanogenic precursors, particularly in deep peat where conditions favor methanogenesis, highlights the potential for labile carbon inputs to influence methane production in these systems. Further investigations are needed to quantify the extent of this effect on methane emissions from deep peat layers.

These observations signal a significant activation of the system, likely tied to shifts in microbial community function and composition due to the carbon-limited nature of the deep microbiome under normal (control) environmental conditions.

Microbial utilization of 13C-labeled glucose explored through DNA-SIP in deep soil

Parallel to metabolomic observations, we also observed microbial compositional, and diversity shifts upon glucose addition, especially for deep microbial communities (Supplementary note 1). Furthermore, we observed a prominent shift in microbial-metabolome congruency across soil depths, suggesting drastic changes in microbiome-metabolome profiles in deep peat, upon receiving glucose (Supplementary note 1, and Supplementary Fig. 5). Thus, we employed DNA-SIP to identify specific microbes directly utilizing the added 13C-glucose in deep soil layers.

Using this approach, we identified five and twelve different taxa at early and late time points, respectively, that were significantly enriched in the heavy (13C-containing) fraction (GitHub workbook 3; sheet 1). These taxa are considered as “direct 13C assimilators”, due to their direct assimilation of 13C-glucose or its byproducts. Interestingly, members of the Koribacteraceae family were prominent across both time points, suggesting their early and persistent utilization of 13C. These bacteria are known for degrading xylan, a common component of plant cell wall70. Additionally, we identified Burkholderia71 and Corynebacterium72, both well-established aromatic compound degraders in soil. Among the late 13C assimilators, we observed Polynucleobacter, a bacterium previously linked to the release of stored organic carbon in peatlands73 and an aromatic compound degrading bacterium74, attesting to its pivotal role in C release in these environments75. Furthermore, Syntrophorhabdaceae members, known for degrading aromatic compounds such as phenol76, and Holophagaceae members with similar functions in peatlands77 were also identified. Thus, these taxa are likely candidates for the direct assimilators of the added glucose, contributing to the observed priming effect in deep soil layers. Furthermore, Telmatospirillum, identified as a 13C assimilator, has been isolated from a boreal fen which has been observed to grow chemoorganotrophically on several organic acids and glucose under anoxic conditions, such as those found in the deep peat78. Therefore, in addition to direct consumption of glucose39, this organism is capable of utilizing various fermentation byproducts like acetate, propionate, and ethanol78, which we also observed in our 13C enriched NMR metabolites (Supplementary Table 1 Information). Collectively, these findings highlight a diverse range of microbes directly and indirectly utilizing the added 13C-glucose in the deep peat layers. However, further studies are warranted to elucidate their specific pathways in breaking down recalcitrant substrates, upon labile C input additions in these environments.

Furthermore, we investigated the microbial communities using OPLS-DA to compare the abundances of significantly different ASVs between the unlabeled glucose and 13C-glucose treatments for each soil depth. This analysis considered the documented phenomenon of isotopic fractionation, where microbes may exhibit preferences for utilizing labeled 13C or unlabeled glucose79. Additionally, it accounted for potential changes in cross-feeding interactions within the microbial community. Following OPLS-DA, we compared these discriminatory taxa with the previously identified direct 13C assimilators. Interestingly, seven members showed significant changes in abundance between fractionated glucose and 13C treatments in the deep samples only. This suggests that these 13C assimilators experienced substantial population shifts in response to the 13C-glucose addition (see GitHub workbook 3; sheet 2). The overlap between these taxa and those identified through isotopic fractionation analysis suggests that isotopic effects, alongside the direct glucose effect, both likely contributed to the observed microbial community differences in the deep soil layers.

Thus, to resolve these microbes, we cross-compared all observed direct 13C assimilators across all deep samples to check their occurrence in control and glucose samples. The absence of these members in both glucose and control deep samples could reflect these members being absent in the respective treatments, which might be due to the natural variation of the sample used. Thus, we theorized that the 13C assimilators that are present in all three sample types provide strong evidence of their natural occurrence in deep microbiota and in addition to being the prime candidates for the observed changes. These included soil bacteria such as Streptococcus80, members of Pedosphaerales81, Burkholderia82, Polynucleobacter74, OD1 (Newly assigned Parcubacteria)83. In addition, ASVs that occur in both glucose and 13C-glucose samples without being present in the control samples would likely attest to their lower abundance in natural deep microbiota, which might be influenced collectively by the presence of glucose. These included soil bacteria such as Prevotella84, Veillonella85, Corynebacterium72, Syntrophorhabdaceae76, and Holophagaceae77. Thus, these observations collectively present an extra layer of reasoning for these direct assimilators to be considered vital members in glucose-induced priming of the deep peat compared to those absent in unlabeled glucose samples. Individual plots highlighting each of the direct assimilators across all samples at different timepoints are provided elsewhere (GitHub workbook 3; sheet 3). Collectively, these members emphasize the distinction between surface and deep microorganisms in glucose metabolism, which we were able to directly validate using DNA-SIP. The nuanced microbial responses to distinct depth conditions are evident in the limited number of genera directly engaging in glucose metabolism within deeper samples. Furthermore, using multi-faceted insights we could identify the potential key members responsible for deep microbiota 13C assimilation and subsequent priming. This underscores how the utilization of glucose by different organisms may be influenced by their genetic potential and adaptability to survive within the oxygen-deprived deep peat layers.

Methods

Synopsis

We conducted three distinct experimental incubations: control (no glucose), unlabeled glucose (12C glucose), and 13C labeled glucose (13C glucose) for 70 days across two different peat layers; surface (acrotelm) and deep (catotelm), for 70 days (Supplementary Fig. 1). Metabolomic and CO2 measurements were taken throughout, and the microbiome sampling was carried out at the beginning and the end of the incubation experiment. Metabolomic and microbial analyses were carried out using control and 12C glucose-amended samples. 13C labeled glucose incubations were used for DNA-SIP and SIAM metabolomics to infer specific microbes and metabolites associated with glucose amended peat samples. We assumed that that both labeled and unlabeled glucose incubations would yield similar responses, given that microbial reactions would primarily be influenced by time and depth rather than isotopic fractionations. Nevertheless, as elaborated in the main text, we observed some discernible effects of labeling on microbial communities. It’s noteworthy to underscore that these findings do not undermine the main results of the study, as glucose addition indeed triggered decomposition of old carbon. The potential impact of isotopic incorporation may primarily affect certain taxa enriched only with 13C glucose addition, highlighting the nuances in microbial responses to labeled substrates (GitHub workbook 3).

Study site

Soil samples were collected from the S1 bog (N 47°30.476′; W 93°27.162′) in the Marcell Experimental Forest (MEF) in northern Minnesota in July 2017. This site is characterized by a high coverage of vascular plants along with Sphagnum. Peat cores were sampled below the water table (anoxic conditions). Core sections from 2 depths (0–20 cm and 90–110 cm) were used for incubation. These depths were chosen because our previous work has shown vertical stratification with depth3,47 along the peat column in (a) microbial community composition, (b) metabolic potential, and (c) SOM physical and chemical properties. Surface samples (0–20 cm) are dominated by inputs from vegetation. The deep zone (>90 cm) is characterized by an accumulation of recalcitrant compounds and static decomposition. Directly after collection samples were shipped on dry ice to the Pacific Northwest National Laboratory. Samples were stored frozen until analysis.

Incubation setup

Approximately 20 g of defrosted peat from selected depths (0–20 cm and 90–110 cm), collected from the S1 bog, were added to pre-weighed 125 mL vials and 20 mL of anoxic Milli-Q water was then added to create peat slurries. Each vial was sealed, flushed with N2 gas for 30 mins, and shaken. We conducted the experiment with six replicates for each treatment across the two depths: control (no addition), unlabeled glucose addition, and labeled glucose addition. Initially, we started with a total of 42 vials (7 vials multiplied by 2 depths and 3 treatments). These vials underwent a pre-incubation period of two weeks to ensure the depletion of any remaining oxygen and removal of terminal electron acceptors, confirming an anaerobic state. Following the pre-incubation, at time 0 the vials were flushed with N2 gas62,86. Subsequently, 13C6-labeled (Sigma-Aldrich: 110187-42-3) and 12C6-unlabeled glucose (Sigma-Aldrich: 50-99-7), were added to two sets of 7 vials each. These vials were then incubated in dark at 22 °C for 70 days. The concentration of added glucose was determined based on the percentage of carbon (C) in the peat soil, ensuring an equivalent of 1% of the total carbon concentration. A glucose solution was prepared by dissolving the powder in milliQ water and added to match 1% of the soil carbon content, estimated around 45% for surface samples and approximately 50% for deeper samples, derived from prior research.

CO2 measurements and estimations

CO2 gas concentrations and isotopic ratio measurements were analyzed using a gas chromatography-combustion interfaced-isotope ratio mass spectrometer (Finnegan MAT Delta V GC-IRMS), as described before3, at specific intervals: 1, 3, 7, 14, 28, 42, 56, and 70 days. To ensure accurate measurements and account for any additional CO2 from DIC, 1 mL of 43% H3PO4 was introduced to convert DIC to CO2 at the end of the experiment. The addition of acid did not significantly impact headspace CO2 levels, confirming its concentration throughout the study. Following each gas measurement time point, one vial was sacrificed starting from Day 7, leading to a decrease in the number of vials available for subsequent measurements. Specifically, for Day 7 time point, measurements were conducted based on (n = 7) vials, decreasing to (n = 6) vials at Day 14, (n = 5) vials at Day 28, (n = 4) vials at Day 42, (n = 3) vials at Day 56, and finally (n = 2) vials at Day 70. CO2 concentrations in the headspace, reported in parts per million (ppm), were converted to molar amount of CO2 in the whole vial. CO2 production rates were calculated based on the average of three linear regression slopes. The first slope encompassed data points up to day 42, the second slope included points up to day 56, and the third slope included points up to day 70. This approach allowed us to capture the linear distribution of CO2 concentrations over time, ensuring accurate calculation of CO2 production rates. Each obtained slope was then divided by the mass of wet peat in each vial and then converted to molar amounts of CO2 in each vial.

Soil respiration (RSOM) was calculated using the following equation:

$${{{{\rm{R}}}}}_{{SOM}}={{{{\rm{R}}}}}_{t}\times \,\frac{{\delta }^{13}{C}_{t}\,-\,{\delta }^{13}{C}_{g}}{{\delta }^{13}{C}_{{SOM}}\,-\,{\delta }^{13}{C}_{g}}$$

where Rt represents the total CO2 concentration (ppm), δ13Ct is the delta 13C of the CO2, δ13Cg is the naturally abundant delta 13C of the Glucose standard added (see below), and δ13CSOM is the delta 13C of the SOM (see below). Respiration produced due to glucose consumption (Rg) was determined by subtracting RSOM from the total respiration (Rt). Notably, the δ13Cg value, representing the naturally abundant delta 13C of the Glucose standard added, underwent two measurements on the instrument. However, due to a discrepancy in the obtained values for the unlabeled glucose standard, we resolved this by calculating the average of the two measurements, thereby ensuring accuracy in subsequent RSOM calculations.

To understand priming effects, respiration from priming (Rp) was calculated using the following equation:

$${{{{\rm{R}}}}}_{p}={{{{\rm{R}}}}}_{{SOM}}-{{{{\rm{R}}}}}_{S}\,$$

where RSOM represents the soil respiration of the glucose-treated samples (calculated previously), and RS represents the soil respiration of the untreated samples.

After each gas measurement time point, one vial was sacrificed starting from Day 7. Peat soil extracted from the vial was used to determine %C and delta 13C of soil phase peat and to perform DNA stable isotopic probing microbial community analysis (DNA-SIP). Simultaneously, porewater was filtered using 0.45μm pore size filters. The dissolved organic matter (DOM) obtained was subsequently divided for comprehensive metabolomic analysis, and dissolved organic carbon (DOC) concentration measurements, as detailed below.

DOC measurements and estimations

The DOC concentration of the different DOM samples collected from the incubation control and unlabeled glucose-amended jars from two depths (days 7, 14, 28, 42, 56, and 70), reported as mM, was measured by high-temperature catalytic oxidation using a Shimadzu Total Organic Carbon analyzer equipped with a nondispersive infrared detector. Triplicate measurements were done for each sample, and the coefficient of variance was always <2%. Following each measurement time point, one vial was sacrificed starting from Day 7, leading to a decrease in the number of vials available for subsequent measurements similar to above.

Delta 13C and %C

The bulk (non-spatially resolved) δ13C of the samples was analyzed using a Costech Analytical Elemental Analyzer (EA, ECS 4010 CHNSO Analyzer, Valencia, CA, USA) coupled to a Thermo Scientific (Bremen, Germany) Delta V Plus IRMS (Thermo Scientific, Bremen, Germany), at the Pacific Northwest National Laboratory. Peat soil samples (7, 14, 28, 42, 56, and 70 days) were freeze-dried and homogenized using a mortar and pestle. Samples were weighed into tin capsules in preparation for the introduction into the EA. The EA combustion reactor was loaded with cobaltic oxide and chromium oxide catalyst, was held at 1020 °C and the reduction reactor which contained elemental copper, was maintained at 650 °C. In-house glutamic acid isotope standards were calibrated against USGS 40 and USGS 41 (δ13C of −26.39‰ and + 37.63‰, respectively) and used as a basis for a 2-point data correction87. Weight percent carbon (%C) was calculated based on an acetanilide standard (Cosetech Analytical), using peak areas and the sample weights.

Nuclear magnetic resonance (NMR)

180 µL of each DOM sample (two depths, 3 treatments (control, labeled glucose, unlabeled glucose,) and 6 times points (7, 14, 28, 42, 56, and 70 days); n = 36) were combined with 2,2-dimethyl-2-silapentane- 5-sulfonate-d6 (DSS-d6) in D2O (20 µL, 5 mM) and thoroughly mixed prior to transfer to 3 mm NMR tubes.

NMR spectra were acquired on a Varian 600 MHz VNMRS spectrometer equipped with a 5-mm triple-resonance (HCN) cold probe at a regulated temperature of 298 K. The 90° 1H pulse was calibrated prior to the measurement of each sample. The one-dimensional (1D) 1H spectra were acquired using a nuclear Overhauser effect spectroscopy (NOESY) pulse sequence with a spectral width of 12 ppm and 512 transients. The NOESY mixing time was 100 ms, and the acquisition time was 4 s, followed by a relaxation delay of 1.5 s during which pre-saturation of the water signal was applied. Time-domain free induction decays (57,472 total points) were zero filled to 131,072 total points prior to the Fourier transform. Chemical shifts were referenced to the 1H methyl signal in DSS-d6 at 0 ppm. The 1D 1H spectra were manually processed, assigned metabolite identification, and quantified using Chenomx NMR Suite 8.3. Metabolite identification was based on matching the chemical shift, J-coupling, and the intensity of experimental signals to compound signals in the Chenomx and custom in-house databases. Quantification was based on fitted metabolite signals relative to the internal standard (DSS-d6). Signal-to-noise ratios (S/N) were measured using MestReNova 14 with the limit of quantification equal to an S/N of 10 and the limit of detection equal to an S/N of 3. 13C labeling was assessed by 13C satellite analysis from the 1D spectra described above or from a 1D-(13C-edited) HSQC experiment. In several cases, further corroboration of metabolite identity was made using standard 2-D experiments such as 1H / 13 C - heteronuclear correlation (HSQC) experiments or 2-D 1H/ 1H Total Correlation spectroscopy (TOCSY).

Gas chromatography mass spectrometry (GC MS)

10 mL of the DOM samples from 24 vials (two depths, 2 treatments (control, unlabeled glucose,) and 6 times points (7, 14, 28, 42, 56, and 70 days)) were first freeze-dried using a Labconco™ FreeZone Freeze Dryer and then derivatized as described previously88. Briefly, 20 µL of 30 mg/mL methoxyamine hydrochloride in pyridine (Sigma-Aldrich) was added to each freeze-dried sample for the protection of carbonyl groups. Samples were vortexed for 20 s, sonicated for 60 s, and were incubated at 37 °C for 90 min with generous shaking at 1000 rpm. After the incubation, 80 µL of N-methyl-N-(trimethylsilyl)trifluoroacetamide (Sigma-Aldrich) with 1% trimethylchlorosilane (Sigma-Aldrich) was added to each sample. The samples were vortexed for 20 s, sonicated for 60 s, and incubated for 30 min at 37 °C with shaking (1,000 rpm). After derivatization, the samples were allowed to cool to room temperature. Samples were transferred to autosampler vials and analyzed by GC MS in random order. Blanks and fatty acid methyl ester (FAME) samples (C8–28) were also included in the analysis for the background referencing and retention time calibration, respectively. Metabolites were separated using a HP-5MS column (30 m × 0.25 mm × 0.25 µm; Agilent Technologies) and analyzed in an Agilent GC 7890A coupled with a single quadrupole MSD 5975 C (Agilent Technologies). For each analysis, 1 µL of sample was injected in splitless mode. The injection port temperature was kept constant at 250 °C for the duration of the analysis. The GC analysis started with an initial oven temperature fixed at 60 °C for 1 min after injection, and then at a steady rate of 10 °C/min, the temperature was increased to 325 °C with the analyses finishing with a 5-min hold at 325 °C. The GC MS raw data files were processed with MetaboliteDetector as reported previously89. Briefly, retention indices (RIs) of the detected polar metabolites were calibrated based on the mixture of FAME internal standards, followed by deconvolution, and subsequent chromatographic alignment of the detected features across all analyses. Identification of the metabolites was carried out by first matching GC MS spectra and RIs against a Pacific Northwest National Laboratory augmented version (>900 metabolites) of FiehnLib library90 containing validated RIs and spectral information. Subsequently, unidentified metabolites were screened against the NIST14 GC MS library by spectral matching alone. To eliminate false identification and reduce the deconvolution errors created during the automated data-processing, all metabolite identifications were individually validated.

Liquid chromatography with tandem mass spectrometry (LC MS/MS)

10 mL of the DOM samples from 24 vials (two depths, 2 treatments (control, unlabeled glucose,) and 6 times points (7, 14, 28, 42, 56, and 70 days)) were first freeze-dried using a Labconco™ FreeZone Freeze Dryer and then reconstituted in 500ul of L HPLC grade methanol. A 21Tesla Agilent FTICR mass spectrometer (MS) equipped with a Waters ultra-performance liquid chromatography (UP-LC) system, was used to collect tandem mass spectrometry (LC MS/MS) spectra of DOM samples in negative ion mode. DOM samples were separated using a Zorbax C18 column (0.5 mm × 150 mm × 5 µm particle size). Samples were injected in a 10 μL volume on the column and eluted with solvent A (5 mM aqueous ammonium formate) and solvent B (5 mM ammonium formate in mass spectrometry grade methanol) with a 60 min linear gradient from 5% to 95% B, followed by isocratic elution at 95% B for 10 min at a flow rate of 0.2 ml/min. The flow from the LC was coupled to the 21 T FTICR-MS using a heated ESI source set to a capillary voltage of 3500 V; sheath, auxiliary, and sweep gas flow rates were 12, 6, and 2 (arbitrary units); and ion transfer tube and vaporizer temperatures were 300 °C and 75 °C, respectively. Mass spectra were collected with 1–2 s transients. MS/MS fragmentation spectra were collected by collision-induced dissociation (CID) of the major features using a collision energy of 40. LC-MS data was processed using the XCMS package in R91. Briefly, CentWave was used for peak picking with a peakwidth parameter of c(5, 70), ppm of 5, snthresh of 7, noise of 1E3, and mzdiff = -0.05. Features with a retention time of less than 5 min and over 33 min were removed from analysis. Peak correspondence was conducted using peak density with a bandwidth value of 30, binsize of 0.25, and with peaks requiring to be present in 10% of samples in a group. Following peak correspondence, retention time adjustment was conducted using Peak Grouping with peaks required to be present in 10% of all samples. Missing peaks using fillChromPeaks using the median width of all chromatographic peaks. Annotations were limited to compound classes, obtained through Canopus92, set to the default settings.

Fourier-transform ion cyclotron resonance-mass spectroscopy (FTICR MS)

High-resolution mass spectra of DOC porewater were collected using a 12 Tesla Bruker FTICR mass spectrometer (MS) by direct injection. These included DOC samples from the two depths, and the control and unlabeled glucose treatment at two different time points (7 days and 70 days). Approximately 100 μL of DOC porewater was mixed with methanol (1:2 ratio) before injection to enhance ionization. A standard Bruker electrospray ionization (ESI) source generated negatively charged molecular ions. Samples were introduced directly into the ESI source. Instrument settings were optimized by using a Suwannee River fulvic acid standard (International Humic Substances Society). Blanks (HPLC grade methanol) were analyzed at the start and end of each day to monitor potential carryover between samples. The instrument was flushed with water/methanol between samples. Ion accumulation times were varied to account for concentration differences. Spectra were averaged from 144 individual scans for each sample and internally calibrated using an organic matter homologous series separated by 14 Da (CH2 groups). Mass measurement accuracy was <1 ppm across 100–1000 m/z and the resolution was ∼240 K at 341 m/z with a 0.8 s transient. Bruker Daltonik Data Analysis converted raw spectra into a peak list by applying a signal-to-noise ratio (S/N) threshold of 7 and a default intensity threshold of 100. Putative chemical formulas were assigned using Formularity software, as described previously93. Formulas were assigned based on: S/N > 7, <1 ppm error, presence of C, H, O, N, S, P, and excluding other elements. All sample peak lists were aligned to ensure consistent formula assignment. Additional rules were a) lowest error and heteroatom formulas chosen, and b) phosphorus assignment requires ≥4 oxygen atoms. The data produced by Formularity including peak masses, peak intensities, and metabolic molecular formulas, was processed using the Metabodirect (v1.0.3) pipeline94. The chemical characterization of peaks within each sample’s ESI FTICR–MS spectrum was evaluated using van Krevelen diagrams. Biochemical compound classes were assigned based on the relative abundance values derived from the counts of carbon (C), hydrogen (H), and oxygen (O). The classification included lipids (0  <  O:C  ≤  0.3 and 1.5  ≤  H:C  ≤  2.5), unsaturated hydrocarbons (0  ≤  O:C  ≤  0.125 and 0.8  ≤  H:C  <  2.5), proteins (0.3  <  O:C  ≤  0.55 and 1.5  ≤  H:C  ≤  2.3), amino sugars (0.55  <  O:C  ≤  0.7 and 1.5  ≤  H:C  ≤  2.2), lignin (0.125  <  O:C  ≤  0.65 and 0.8  ≤  H:C  <  1.5), tannins (0.65  <  O:C  ≤  1.1 and 0.8  ≤  H:C  <  1.5) and condensed hydrocarbons (aromatics; 0  ≤  200 O:C  ≤  0.95 and 0.2  ≤  H:C  <  0.8)47. Gibbs free energy (GFE) was calculated using the nominal oxidation state of carbon (NOSC), a measure of the energy required to oxidize the carbon in each compound. This was calculated based on the compound’s molecular formula, following the approach by AminiTabrizi et al. in 202095.

DNA extraction and DNA stable isotope probing

Total environmental DNA was extracted from peat samples at the end of the incubation experiment (n = 12, 2 depths, 3 treatments, and 2-time points (t = 7 and 70 days)) using the PowerSoil DNA Isolation kits (Qiagen, CA, USA) according to the manufacturer’s protocol with slight modifications. After extraction, DNA quality and quantity were assessed using a NanoDrop spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA) and Qubit HS assay kit (Invitrogen, Carlsbad, CA, USA), respectively, and confirmed by electrophoresis on an agarose gel. Approximately 10 individual DNA extractions per sample were pooled and concentrated using ethanol precipitation to meet DNA-SIP requirements for DNA quantity. Equal amounts of DNA (1 μg per sample) were mixed with CsCl solution (1.714 g/mL) in a heat-sealable OptiSeal polypropylene tube (Beckman Coulter, CA, USA) and centrifuged at 177,000 g for 40 h at 20 °C in an OptimaTM MAX-TL Ultracentrifuge (Beckman Coulter, USA) equipped with a TLA-110 fixed-angle rotor (Beckman Coulter, USA). The centrifuged DNA gradients were fractionated by needle fractionation into 12 or 13 fractions, and the density of each fraction was determined with a digital refractometer (Reichert AR200). DNA was recovered from fractions by precipitation in a polyethylene glycol (PEG) solution with glycogen. The DNA concentrations of the retrieved fractions were measured using Qubit (Invitrogen, CA, USA)96. The enrichment of heavy isotopes in DNA fractions was validated by PCR of SSU rRNA genes as before86. The SIP procedure resulted in significant DNA loss, with approximately 20% recovery. Some heavy fractions had undetectable DNA concentrations by fluorometry, but visible PCR bands by gel electrophoresis indicated active labeled substrate utilization. Though SIP caused DNA loss, detectable amplification demonstrated that the procedure successfully separated labeled DNA. Five ng of unfractionated and fractionated DNA samples were used to assess the diversity and composition of the prokaryotic community composition. A two-step amplicon-based sequencing protocol of the V4 variable region of the SSU rRNA gene was used with 515F and 806R primers and sequenced on an Illumina platform at the Georgia Institute of Technology as previously described86.

Data processing and statistical analysis

Individual metabolomic datasets generated from GC MS and LC MS/MS (control and unlabeled glucose) were prefiltered using prevalence cutoffs of 25% and 50% within each depths, to minimize zero inflation. Accordingly, a total of 658 metabolites across three analytical platforms were used for downstream analysis (Fig. 2A). NMR, GC MS, and LC MS/MS data were then cube-root transformed and median scaled and these normalized datasets were independently used for all downstream analyses. Principal Coordinate Analysis (PCoA) with Bray-Curtis distance was used to obtain metabolomic ordinations (ape)97, and sample separation based on categorical variables were tested using permutational multivariate analysis of variance (PERMANOVA) through the adonis2 (vegan) in R98. Orthogonal Projections to Latent Structures Discriminant analysis (OPLS-DA) with leave one out cross validation50 in R(ropls)99, was used to obtain discriminatory different features across soil depths for individual datasets, using VIP ≥ 1.0 as the filtering criteria100. All identified discriminatory features are provided in GitHub workbook 1. Independently identified significant features were cross compared using DiVENN2.0 for visual comparison101. Differentially abundant metabolites were identified using absolute log2 fold change in abundance greater than 1 between treatments at given time aggregates (initial: days 7,14,28 and final: 42,52,70), and these metabolites were subsequently mapped to the Kyoto Encyclopedia of Genes and Genomes (KEGG) reference database63 in R (KEGGREST)102, to infer associated pathways.

Microbiome data was preprocessed in QIIME2103 using DADA2 pipeline104. Before the analysis, the amplification primers were removed from the raw data using cutadapt105 and Amplicon sequence variants (ASVs) were inferred from DADA2 quality-filtered data. Resulting ASV abundances for all samples (control, 12C and 13C glucose amended both fractionated and unfractionated) were processed following the protocol published by Callahan et al.106, using phyloseq107 and vegan98 packages in R. Paired reads were merged and reads between 251 and 254 bp in length were retained for downstream analysis. Putative chimeras sequences were removed using the removeBimeraDenovo function from the DADA2 package. Taxonomy was assigned using SILVA ribosomal RNA gene Database 132108. Unannotated ASVs at the phylum level or those identified as chloroplast or mitochondria were removed from the dataset. Finally, the ASVs were filtered to remove taxa below prevalence threshold 2. Unique ASVs were aligned to the SILVA v132 reference alignment in Mothur109, and an approximately maximum likelihood tree was constructed using FastTree110. Obtained taxa were agglomerated into class level to generate relative Class abundances (Supplementary Fig. 5), while all other subsequent analyses were performed at ASV level using Chord transformed data111. Shannon diversity was obtained at ASV level prior to transformation. PCA ordinations of microbiome data were generated using the vegan package in R112. Unlabeled glucose and 13C glucose-treated samples were compared at each depth using OPLS-DA as described above to identify differential ASVs associated with isotopic fractionation (GitHub workbook 3). HTSSIP package113 in R was used to determine the direct 13C assimilators of microbial communities. Briefly, fractionated samples were categorized into heavy and light DNA based on density cutoff of 1.73 g/ml and relative abundances of ASVs were compared using HTSSIP, using Benjamini-Hochberg corrected p.value < 0.1 as a significance cutoff. Identified members were further filtered out to retain the members that have been significantly enriched only in the heavier fraction. Following the approach described by Campana et al.114, these ASVs were considered showing direct evidence of incorporating 13C into the community DNA, thus, ‘direct 13C assimilators’. The congruency of microbiome and metabolome datasets was assessed using Co-Inertia analysis (CIA), as done before115, using the made4 package in R112. PCA ordination of unfractionated microbiota samples (control and 12C glucose treatments) and PCoA of corresponding metabolomic samples were used for CIA, where RV attests to the degree of congruency116.

Statistical analyses of CO2 and Dissolved Organic Carbon (DOC) measurements were performed in R112. Our experimental design necessitated destructive sampling to enable comprehensive molecular analyses across time points, though this constrained the number of replicates available at each sampling time. To evaluate treatment effects despite this limitation, we employed independent t-tests to compare control and glucose-treated samples for both DOC and CO2 concentrations, using p < 0.05 as the significance threshold.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.