Introduction

Coastal ecosystems are of considerable ecological, economic and social significance as they house 41% of the world’s global population (as of 2003), while contributing 77% of global ecosystem services value1. Of the globe’s diverse coastal ecosystems, blue carbon (BC) habitats—including mangroves, salt marshes, and seagrasses—have garnered considerable interest due to their high carbon (C) burial rates, wealth of ecosystem services, and vulnerability to sea level rise, land use change, and peat collapse2,3,4,5. BC contributes approximately 5.3 Pg of C to total global C stocks6 and has varying chemistry due to differences in sources (allochthonous, autochthonous), turnover/reworking, burial rates, and in situ conditions7.

Globally, 0.16–0.39% of BC habitats are lost each year8,9,10, and their degradation may release 0.15–1.02 Pg of CO2 to the atmosphere4. In addition to gaseous C losses, they are important exporters of dissolved organic matter (DOM)11,12,13,14,15 and particulate organic matter (POM)16,17,18 to coastal waters. In post-industrial times, eutrophication has increased algal biomass in coastal waters19, likely increasing algal derived DOM as well. This confluence of BC-derived OM with fresh algal exudates may influence the turnover of previously sequestered BC through the priming effect (PE)20,21.

The PE, or interactive effects on organic matter (OM) degradation in the presence of mixed substrates, has been well described in soils22,23,24,25,26. While there exists convincing evidence of the PE in aquatic systems in both lab20,27,28,29,30 and field studies21,23,31,32,33, there is no consensus as to its occurrence, prevalence, and underlying mechanisms34,35,36,37,38. Positive (enhanced OM degradation), neutral, or negative (suppressed OM degradation) PEs have been reported, and the direction and magnitude of PEs can vary in aquatic systems39. While many substrates may be involved with aquatic PEs, here we focus on peat, which is typically believed to be a refractory substrate, and algal biomass as the more reactive substrate. Using these two substrates as an example, the mechanisms underlying PEs may include either (1) co-metabolism of peat and algal-derived DOM by a mixed microbial community; (2) nutrient mining of the peat DOM to subsidize microbial activity, whereby limiting nutrients are mined from the substrate to support growth; and/or (3) simultaneous consumption of peat and algal DOM by one microbial group20,22. It has been suggested that the term priming is not a single mechanism, but a concept that describes these different mechanisms, and investigations into the mechanisms of priming will further assist in elucidating the variability of priming studies in aquatic ecosystems26,38,39.

Eutrophication and net primary production have increased on inner continental shelves and estuaries in recent decades40,41, and the co-occurrence of BC-derived OM with more reactive algal biomass is likely conducive to the PE, which may ultimately result in a loss of BC to CO2. While shelves tend to be net CO2 sinks19, estuaries are generally net heterotrophic (i.e., net ecosystem metabolism results in CO2 efflux), although there are notable exceptions, particularly fjords42,43,44. However, the PE may be one process contributing to BC turnover and CO2 efflux in estuarine systems20, leading to our overarching research question: when BC is exported to coastal waters, is it re-mineralized to CO2, subsequently contributing to a positive climate feedback; and if so, what microbial processes and molecular changes are involved in this response? Recent meta-analyses have highlighted the need for molecular level approaches to address questions about OM reactivity across ecosystem scales38,39,45, and have highlighted that the variability in reports of aquatic priming may be driven by the metric used to measure priming, be it CO2 released, bacterial respiration, or mass differences of DOC or POC38. Here, we apply a suite of tools to evaluate changes in molecular composition and microbial pathways associated with the priming effect. However, it should be noted that measurements such as metabolomics and metatranscriptomics are important tools21,39, but do not provide information about rates of CO2 released or DOM loss, per se35,46.

Here, we address our overarching research question using a laboratory experiment with different substrates, to characterize rapid (i.e., 28-h) molecular and microbial changes that occur under differing substrates. Four treatments were established: a seawater control (control), a BC leachate prepared from mangrove sediments sampled at a depth of 1.35 m (peat), a 13C-labeled algal leachate (algae), and a combined peat and algal leachate (primed). Each treatment received an estuarine inoculant collected from a site adjacent to where the mangrove sediment core was collected. One challenge in priming studies is that many do not use labeled substrates or include controls with only stable OM and no control with labile OM38, which we sought to address with our experimental design. Our specific hypothesis is: if the PE affects the turnover of BC in coastal waters, then we would expect a greater loss in dissolved organic carbon (DOC), and concomitant CO2 and dissolved inorganic carbon (DIC) production in the primed treatment. To address this hypothesis, pre and post incubation samples were analyzed for (1) bulk measures of DOM, POM, and inorganic C; (2) low molecular weight organic compound (<1000 Da) composition via liquid chromatography mass spectrometry (LC-MS) and Fourier-transform ion cyclotron resonance mass spectrometry (FT-ICR MS); and (3) the differential expression of microbial transcripts to evaluate changes in metabolic pathways.

Results and discussion

Stable isotope values indicate rapid allocation of C and N from DOM to POM

The three potential OM sources in this study (peat, algae, and inoculant) had distinct POM stable isotope values. The POM of the estuarine inoculant had a δ13C value of −13.94‰ and a δ15N value of 6.77‰; the peat source had a δ13C value of −22.61‰; and the 13C-labeled algal substrate were highly enriched for both δ13C (1,613,152‰) and δ15N (1975‰). Interestingly, at the initial timepoint, POM δ13C and δ15N values were also highly enriched in the primed (δ13C 34812‰ ± 2026; δ15N 452‰ ± 23.04) and algal treatments (δ13C 35562‰ ± 2257; δ15N 415‰ ± 26.50), likely indicating that there was a rapid allocation of the enriched algal DOM into the particulate fraction. While there is not sufficient data to determine the mechanism of the rapid transfer of enriched C and N into the particulate fraction, it may have been driven by rapid sorption of amino acids and other N-rich compounds onto POM47,48 during the establishment of the incubation, or rapid microbial utilization of the 13C labeled algal DOM. This was corroborated by enriched 13C-DIC and 13C-CO2 values seen in these treatments at the initial timepoint (Fig. 1 and Table 1) and similar rapid microbial utilization of dual-labeled amino acid probes has been previously reported49.

Fig. 1: Hypothesized processes occurring with different substrates.
figure 1

Conceptual diagram of hypothesized processes occurring during the experiment. Trends in specific parameters are indicated with arrows to indicate an increase (up) or decrease (down) in each value, and processes are shown with curved arrows. Stable isotope values indicated that rapid allocation of DOM to POM occurred in treatments with algal DOM. The algal treatment had greater CO2 released, but treatments with peat had greater DOC lost, greater RNA concentrations, similar CO2 released, and decreased abundance of pathways for the biosynthesis of N-rich hydrophobic compounds. These findings indicate that while certain metrics, such as DOM composition and PC:PN ratios converged at the end of the experiment, there were nuanced changes in DOM and POM dynamics, and that the addition of a peat substrate may promote microbial biomass production relative to CO2 production.

Full size image
Table 1 Initial and final measurements of particulate δ15N and δ13C values, particulate nitrogen (PN), particulate carbon (PC), particulate organic C (POC), particulate carbon: particulate nitrogen (PC:PN), pH, dissolved inorganic C (DIC), δ13C values of DIC (δ13C-DIC), dissolved organic C (DOC), and total and δ13C values of CO2.
Full size table

Between initial and final time points, all treatments experienced a decrease in pH, an increase in total CO2 and 13CO2, a loss of DOC, and an increase in DIC (Table 1), indicating active respiration over the course of the experiment. As expected, treatments with the 13C-labeled algal substrate had significantly greater increases in 13CO2 (ANOVA, F(3, 11) = 48.28, p < 0.001; 95% CI [0.83, 1.00]; Tukey HSD p < 0.001, 95% CI [−192.83, −103.42] (control - algae), and 95% CI [−165.10, −82.32] (peat - algae); p < 0.001, 95% CI [70.09, 159.50] (control - primed), and 95% CI [48.99, 131.77] (peat - primed)) and δ13CO2 (KW, Chi square = 12.22, p = 0.01, df = 3; Dunn’s post hoc p = 0.0029 (algae-control); p = 0.0044 (algae-peat)) and δ13C-DIC (KW, Chi square = 12.39, p = 0.01, df = 3; Dunn’s post hoc p = 0.0038 (algae-control and algae-peat)) relative to unlabeled treatments. The algal treatment had significantly greater increases in total CO2 relative to the control (KW, Chi square 9.03, df = 3, p = 0.03; Dunn’s post-hoc p = 0.0029), but the increase in total CO2 from the primed treatment was not significantly different from the peat treatment (KW, Chi square 9.03, df = 3, p = 0.03; Dunn’s post-hoc p = 0.6926 (primed-peat)) (Table 2 and Fig. 2). The similar release of CO2 from the peat substrate is interesting, as peat is typically believed to be a refractory substrate. Particulate N (PN), particulate organic carbon (POC), and particulate carbon (PC, defined as POC and particulate inorganic C), increased in all treatments, but no significant difference in the change of PN, POC, and PC was observed between treatments (PN: KW Chi square = 5.03, p = 0.17, df = 3; POC: Chi square = 5.28, p = 0.15, df = 3; PC: KW Chi square = 5.77, p = 0.12, df = 3). Correspondingly, all treatments also exhibited a decrease in DOC, and while there was no significant difference in DOC loss between treatments (one-way ANOVA, F(3, 12) = 1.13, p = 0.377), the greatest decrease was seen in the primed and peat treatments (61.1% and 54.2% loss, respectively). Contrary to our original hypothesis, the decrease in DOC in the primed treatment did not result in a corresponding significant increase in CO2 or DIC relative to the algal treatment (total CO2: KW, Chi square 9.03, df = 3, p = 0.03; Dunn’s post-hoc p = 0.2059, DIC: one-way ANOVA F(3, 12) = 0.80, p = 0.519; 95% CI [0.00, 1.00]) suggesting that under priming conditions, microbial communities may favor biomass production relative to CO2 production when respiring organic substrates50. While mineralization of OM is necessary for biomass production, bacterial growth efficiency can vary substantially in aquatic systems based on OM inputs; for example, bacterial growth efficiency can vary from as low as 4% to as high as 55% when terrestrial and wetland OM inputs are highest51. Nevertheless, other mechanisms, such as the flocculation of peat DOM into POM52,53, may also be an important mechanism to explain these findings. However, the concentration of RNA was significantly greater in the primed treatment (KW, Chi squared = 10.963, p = 0.011, df = 3; Dunn’s post-hoc p = 0.0087 (algae – primed), p = 0.0012 (control-primed)), suggesting that microbial biomass was greater, and supporting the concept that microbial communities are favoring biomass production relative to CO2 production under priming conditions (Table 1).

Table 2 Percent change in each measure over the course of the experiment.
Full size table
Fig. 2: Changes in bulk measures over the incubation.
figure 2

a Plot of the relative percent change in total CO2, b dissolved organic carbon (DOC), c the δ13C value of dissolved inorganic carbon (DIC) and d the carbon to nitrogen ratio (PC:PN) over the course of the incubation period. Smaller, lighter points are values from each replicate, and larger points are mean values with standard deviation. Significant differences from a Kruskal–Wallis test, with Dunn’s post-hoc test are shown with letters. A positive change indicates an increase over the course of the incubation; a negative change indicates a decrease over the course of the incubation. No change (zero) is indicated with the horizontal red line.

Full size image

At the end of the experiment, the particulate material of the algae and primed treatments had highly enriched δ13C (12102 ± 2316‰, and 9832 ± 2448‰, respectively) and δ15N values (476 ± 120‰, and 390 ± 70‰, respectively). The highly enriched label was used to capture changes in δ13C values in the headspace and DIC, and was based on experiments from Ward et al.21. Although both treatments were highly enriched, the direction of their change differed for δ15N values, as the primed treatment exhibited a ~14% decrease in δ15N, while the δ15N value of the algal treatment exhibited a ~ 15% increase over 28 h (Fig. 2 and Table 1). This may indicate different mechanisms of N sorption and/or cycling between these two treatments, which are summarized in Fig. 1. The decrease in δ15N in the primed treatment is likely from microbial assimilation of dissolved N compounds (fractionation factor ~ −27 to 0‰)54, preferential incorporation of 14N into microbial biomass55, and incorporation of peat-derived N into POM (peat ~3‰56 versus algae 1975‰). This is supported by the decrease in PC:PN in the primed treatment, indicating greater incorporation of N into the POM of this treatment. Whereas the increase in δ15N values in the algal treatment may indicate further sorption of the highly enriched (1975‰) algal-DOM onto POM.

The microbial community and DOM composition detected via ESI(−) mode converged over the course of the incubation

The most abundant class of bacteria was the Gammaproteobacteria for all treatments. Members of the Gammaproteobacteria are important drivers of coastal C cycling21,57,58, some of whom, such as Acinetobacter, have been used as model organisms in previous priming experiments59. By the end of the experiment, certain taxa, such as members of the Acidobacteria, Aquificae, and Chloroflexi, were lost, indicating a rapid (28 h) shift in the composition of the active microbial community. At the end of the experiment, the microbial community of each treatment converged to a highly similar community composition dominated by Gammaproteobacteria, despite differences in substrates. When the functional profile of the microbial community was assessed by evaluating differences in clusters of orthologous groups (COGs) for each treatment, we found that most COGs were associated with Proteobacteria60, further indicating that Proteobacteria were not only numerically abundant, but also a functionally important phylum for the turnover of DOM in the coastal zone.

In addition to convergence in microbial community composition, PC:PN values and DOM data acquired in FT-ICR MS and LC-MS indicated that there was an overall convergence in stoichiometry and in DOM composition by the end of the experiment. Although there were differences between initial PC:PN ratios of the POM (3.40–5.96), by the final timepoint PC:PN values converged to ~5 (4.49–4.93; Table 1). To achieve this convergence in PC:PN ratios, the peat treatment experienced a significant increase in PC:PN (43.2% increase), relative to the algae and primed treatments, which experienced significant decreases in PC:PN (12.9 and 20.1%, respectively; KW Chi square = 7.3485, p = 0.06; df = 3; Dunn’s post hoc p = 0.0489 (peat-algae), p = 0.0097 (peat-primed); Table 2, Fig. 2). Based on the feature diversity of DOM detected via ESI(−) FT-ICR MS, samples at the initial timepoint clustered separately primarily by treatment, while samples at the final timepoint clustered closer together, indicating a convergence of the overall composition of the DOM pool by the end of the experiment (Fig. 3a). Similar trends were found with LC-MS analyses in both ionization modes, except for the control at the final timepoint which showed a different response in ESI(+) (Fig. 3b, c). ESI(−) has been used in numerous studies for DOM characterization21,61,62, because oxygen containing compounds typically found in DOM readily ionize in this mode. However, the finding in ESI(+) suggests that compounds ionized in different electrospray ionization mode (ESI(−) or ESI(+)) may exhibit different biogeochemical dynamics. Biomolecules and compounds such as amino acids and proteins are more likely to be ionized in positive mode. Instead, acidic groups, such as carboxylic acids, which may be directly synthesized by organisms or formed via DOM oxidation are more likely to ionize in negative mode. These findings highlight the caution needed when comparing data obtained at different ionization modes as distinct compounds ionize more efficiently in one of the modes63,64. Overall, we found that PC:PN values and the composition of microbial communities and DOM differed in the initial timepoint, but then converged to similar PC:PN values, and a similar microbial community and DOM composition by the final timepoint. This suggests that, when presented with differing substrate types, microbial succession and DOM reworking resulted in a more homogeneous microbial community composition and DOM chemical composition, which may be associated with the communities reaching stoichiometric homeostasis65,66,67. Although stoichiometric homeostasis is not consistently demonstrated by autotrophs, heterotrophs, including bacterial heterotrophs, are generally more homeostatic65. While mixed microbial assemblages can have varying degrees of homeostasis depending on factors such as phosphorus availability68, axenic cultures exhibit greater homeostasis than mixed communities, and members of the Gammaproteobacteria have been shown to be homeostatic68,69. As our samples were dominated by this class, it is likely that the community convergence was also coupled with stoichiometric homeostasis70, however, we lack sufficient evidence to unequivocally determine if this is the case.

Fig. 3: High-resolution mass spectrometry data.
figure 3

a A non-metric multidimensional scaling (NMDS) plot for FT-ICR MS data (stress = 0.02) analyzed with electrospray negative ionization mode (ESI(−)); b principal coordinates analysis (PCoA) plot for LC-MS data in ESI(−); and c PCoA plot for LC-MS data in positive ionization mode (ESI (+)). Faded symbols represent individual samples. Solid symbols are the average value for each sample group. Error bars are based on one standard deviation. NMDS of ESI(−) FT-ICR MS data was constructed based on feature presence/absence and Jaccard similarity distances. LC-MS PCoA was based on feature abundance. Despite the different analytical methods, FT-ICR MS and LC-MS data both showed a composition shift in DOM from initial (TP0) to final (TP1) conditions. The initial timepoint (TP0) was primarily separated based on treatments, while the treatments clustered closer to each other at the final timepoint (TP1), except for the control at the final timepoint (Control TP1) in ESI(+), suggesting an overall convergence in DOM composition.

Full size image

Despite the similarities between treatments at the final timepoint, treatment differences in community composition were seen within the Gammaproteobacteria. Specifically, when compared to the control and peat treatment, the algal and primed treatments had a greater abundance of Alteromonadales60, which was also a dominant order in a prior study investigating the mobilization of marsh-derived DOM21, suggesting that this order is important for mobilizing algal- and peat-derived DOM in the coastal zone.

Although microbial communities and DOM composition converged, functional pathways differed

Although the microbial community structure converged at the end of the experiment, functional differences were still observed between the treatments. Multi-dimensional scaling (MDS) plots based on KEGG orthologs (KOs) indicated that the samples clustered by timepoint and treatment (Fig. S1). At the end of the experiment, the seawater control and peat treatments had more similar functional profiles based on KOs, while the algae and primed treatments were more similar to each other, which is in agreement with the patterns in microbial community composition (i.e., greater Alteromonadales in the algae and primed treatments). Based on heatmap clustering and z-scores, the initial timepoint clustered separately from the final timepoint, indicating that there was a change in the functional profile over the course of the 28-hour incubation for all treatments (Fig. S2). Furthermore, at the final timepoint, the control and peat treatments clustered separately from the primed and algal treatments (Fig. S2).

Microbial functional profiles, based on COGs with >700 mean estimated gene copies, indicated that dominant COG pathways differed by treatment (Fig. 4). The primed treatment had fewer dominant COG pathways (6 dominant pathways), while the algal treatment had the greatest number of dominant COG pathways (14 pathways), indicating reduced functional diversity within the primed treatment. Treatments that did not have the peat substrate (i.e., algae and control treatments) had greater relative abundance of COG categories related to the biosynthesis of branched-chain amino acids (isoleucine, leucine, and valine), fatty acids and purines. This indicates that pathways for the biosynthesis of N-rich, hydrophobic compounds were more abundant in treatments that did not have peat-derived DOM. Sorption of hydrophobic amino acids has been shown to be an important mechanism influencing the allocation of amino acids in other systems48. When considering the particulate δ15N data described earlier, it is plausible that N was rapidly allocated into POM48 in treatments with the algal-derived DOM. Interestingly, there were other similarities seen between treatments with peat (i.e., primed and peat treatments). Specifically, these treatments had the greatest relative loss in DOC, similar amounts of total CO2 released, increases in DOM S/C ratios (Fig. 5), and significantly higher RNA concentrations. These results indicate that there may be some characteristic of the peat-derived DOM that influences the accessibility or turnover of algal DOM when both substrates are present. Taken with our other results, this suggests that while total CO2 did not increase under priming conditions, it’s likely that the addition of a peat substrate may promote microbial biomass production relative to CO2 production (Fig. 1).

Fig. 4: Relative abundance of COG pathways.
figure 4

Clusters of orthologous groups (COG) pathways that were present with 700 gene copies or greater for all treatments and timepoints. Treatments that did not have peat (i.e., control and algal treatments) had greater relative abundances for pathways associated with the biosynthesis of nitrogen rich hydrophobic compounds, such as amino acid biosynthesis (isoleucine, leucine, and valine), fatty acid biosynthesis and purine biosynthesis.

Full size image
Fig. 5: Fourier-transform ion cyclotron resonance mass spectrometry data.
figure 5

Differences in saturation (hydrogen to C; H/C), oxidation (oxygen to C; O/C), nitrogen (N/C), sulfur (S/C), and phosphorus (P/C) over the course of the experiment. SW Control, SWA Algae, SWP Peat, SWPA Primed, TP0 initial, TP1 final.

Full size image

While most COGs were associated with Proteobacteria, there were interesting differences after COGs were filtered to determine which were unique to each treatment. There were no unique COGs associated with the control treatment, indicating that the control was representative of baseline conditions, but unique COGs were seen for all other treatments. The primed treatment had 14 unique COGs and the algal treatment had 12 unique COGs, including some associated with Bacteroidetes. The peat treatment had 17 unique COGs, the majority of which were associated with Planctomycetes, and COG pathways for the biosynthesis of amino acids (leucine, isoleucine, valine), arginine and nitrogenous bases/nucleobases (purine and pyrimidine). While these were not dominant COG pathways in this treatment (Fig. 4), Planctomycetes may be important taxa in treatments with peat DOM, although they may not incorporate amino acids as readily as Roseobacter (Proteobacteria)49.

Microbial pathways highlight the need for future research into the role of heteroatoms and priming

In addition to C and N, it is likely that differences in substrates influenced the availability of other elements, such as phosphorus (P) and sulfur (S), as heteroatoms have been found to be important components of the DOM pool71. Here, we reflect on patterns in pathways related to heteroatom turnover, but a detailed investigation of these elements was beyond the scope of this study. Phosphate transporters were under-expressed in the algae treatment, suggesting there was reduced P demand in this treatment. However, this was not observed in the primed treatment possibly because microbial communities may have been more P-limited, as previous priming studies have found that the heterotrophic metabolism of recalcitrant OM is in part limited by P72. Correspondingly, based on the FT-ICR MS data, there was very little change in the DOM P/C ratio in the algal treatment, but the primed treatment had a decrease in P/C, possibly due to greater P demand (Fig. 5).

Additionally, Enzyme commission numbers (ECs) associated with assimilatory sulfate reduction, as well as two ECs associated with dissimilatory sulfate reduction and oxidation, were upregulated in the primed treatment relative to the algae treatment. Interestingly, there was also an increase in the S/C ratio in the DOM, particularly in the peat and primed treatments, suggesting turnover, and potential release of peat-derived S compounds. A past study found that there was a rapid turnover of organic sulfur compounds in seawater46, and this further suggests that sulfur metabolism is important to consider when evaluating the turnover of BC derived DOM in the coastal zone. While recent studies in the soil priming literature have begun to find evidence for the importance of S in priming73, the evaluation of heteroatoms such as P and S in aquatic priming has been under-studied.

Insights and future directions for studying priming mechanisms

Recent meta analyses have highlighted that the term priming is a concept that encompasses several mechanisms, and that targeted labeling experiments with high resolution techniques are needed to elucidate the mechanisms driving priming38,39,43. Additionally, it has been recently emphasized that the study of OM degradation needs to span across ecosystems to form unifying theories of OM decomposition45. Here, we applied concepts that originated in soil sciences to draw linkages with freshwater and marine studies and performed a laboratory study of the priming effect in coastal waters and its associated molecular and microbial changes. This work, which in large part derives from a series of lab and field priming experiments conducted over the past eight years20,21,28,30,32,34,35,59 reveals that molecular techniques such as those used here can provide valuable information on DOM composition and microbial functions associated with DOM turnover and beyond. Stable isotopes indicated that there was likely rapid allocation and uptake of algal-derived DOM into POM. We found that while the microbial community, stoichiometry, and overall DOM compound diversity as determined by high resolution mass spectrometry converged by the end of the experiment, microbial pathways and heteroatom ratios determined by high resolution mass spectrometry indicated that there were nuanced changes associated with the utilization of different substrates, which are summarized in Fig. 1. Contrary to our hypothesis, the primed treatment did not have the greatest increase in CO2 or DIC relative to the algal control. However, the primed and peat treatments exhibited a greater loss of DOC, and greater RNA concentrations, which is possibly due to microbial communities shifting from CO2 production to biomass generation in these treatments. Additionally, treatments with peat (i.e., peat and primed) had less abundant pathways associated with the biosynthesis of nitrogen (N)-rich compounds. These findings indicate that assessing the mechanisms of priming can be difficult to resolve even with the combination of controlled laboratory experiments and high-resolution techniques. However, such combined tools continue to prove useful to the new challenges in 21st century biogeochemistry, such as increasing eutrophication and climate change74, and serve as hypothesis generating tools that can help inform the development of unifying theories of OM turnover across ecosystems45.

Methods

Experimental design and bulk measurements

Leachates were prepared using 13C labeled Spirulina biomass (Sigma Aldrich; 605972) and peat collected from a mangrove swamp located on the west coast of Florida, from a section that was 136 cm down-core. The depth of this core was selected to evaluate the turnover of buried refractory BC in deep mangrove sediments, and detailed analysis of the bulk geochemistry down-core can be found in ref. 75. Homogenized, freeze-dried peat or algae was mixed with MilliQ water, stirred in the dark at 25 °C for 24 h, using the same approach as ref. 21. After 24 h, leachates were filtered using a 0.22 μm Durapore PVDF filter membrane (SterivexTM; MilliporeSigma®, St. Louis, MO, USA) prior to establishing the incubations. Incubations were established in 5 L bottles with 1 L of headspace, with 4 replicates for each treatment and timepoint. An unfiltered estuarine water sample was collected adjacent to the mangrove swamp on the day the incubations were started and transported to the laboratory on ice (<3 h). This water was used as an inoculant and was added at a 10% volume of the incubation. All incubations received 4 nM ammonium chloride and 0.3 nM sodium phosphate. Each treatment received 143 mL of substrate, following the approach of Ward et al. (2019), however, this approach resulted in differences in the amount of DOC added, and as such, only relative differences between treatments are discussed throughout the text. The control received 143 mL of MilliQ water (no DOC added), the peat treatment received 136.1 mL of the peat leachate and 9.6 mL of MilliQ water (989 μM DOC added), the algae treatment received 6.9 mL of the algal leachate and 136.1 mL MilliQ water (107 μM DOC added), and the primed treatment received 6.9 mL of the algae leachate and 136.1 mL of the peat leachate (1095 μM DOC added). To account for differences in initial DOC concentrations, for each treatment, we report the initial and final values (Table 1), differences between the initial and final values (Table S3), and relative percent change between the initial and final values (Table 2). After substrates were added, bottles were closed using caps with three-way luer valves. After all treatments were established (~ 1 h), headspace gas samples (60 mL) were collected for the initial timepoint analysis, and the initial timepoint samples were destructively sampled for all analytes as described below, within ~4 h.

Incubations were conducted at 25 °C in the dark for 28.5 h to minimize the effects of photochemistry and photoautotrophs on OM turnover. Due to the required volumes (~ 4 L) for the suite of analyses, it was not feasible to include additional time points over the course of the experiment due to limitations in processing time for the number and volume of samples. As such, only the initial and final timepoints were studied. At the initial and final time point, headspace gas samples were collected using a three-way luer-valve and a 60 mL syringe. Incubations were destructively sampled, and pH was immediately measured with a benchtop pH meter prior to sample filtration. After collection at the initial and final time points, water samples were filtered in parallel through 25 mm and 47 mm glass fiber filters (GF/Fs, nominal pore size 0.7 μm). For each sample, a 0.2 μm polycarbonate filter membrane was also collected in parallel (MilliporeSigma®, St. Louis, MO, USA), flash frozen and stored for RNA extraction. Aliquots of filtered water samples were collected for DOC, total dissolved nitrogen (TDN), DIC and the δ13C value of dissolved inorganic carbon (δ13C-DIC). Samples for DOC and TDN were frozen prior to analysis, while vials for total DIC and δ13C-DIC were filled until overflowing, then capped with no headspace, and immediately stored at 4 °C. The remaining filtered water was frozen at -20 °C until solid phase extraction (SPE) was conducted. Here, we define DOM as the semi-polar fraction that is extractable with SPE using PPL cartridges according to ref. 76. SPE eluates were split into two aliquots, which were respectively analyzed for chemodiversity via FT-ICR MS and LC-MS Orbitrap.

Bulk geochemical data

Gas samples were analyzed for total CO2, and 13CO2 within 24 h of collection using a Picarro Cavity Ring-Down Spectrometer G2201-i isotope analyzer (Picarro Inc., Santa Clara, CA, USA). One sample’s data was discarded because the injection was interrupted during analysis. Filtered water samples for total DIC and δ13C-DIC were analyzed at the University of Florida’s stable isotope laboratory within two weeks using an AutoMate-UIC CO2 Coulometer and a Gasbench-DeltaPlusXL. Filtered water samples were analyzed for DOC and TDN using a Shimadzu TOC-V analyzer at the University of Florida’s Organic Geochemistry Laboratory. 25 mm GF/Fs were immediately frozen, freeze dried, and post-dry weights were recorded for measurement of total suspended solids (TSS). Filters were then split and one half was acidified according to Harris et al.77 to be analyzed for particulate organic carbon (POC) and δ13C, while the other fraction was not acidified and analyzed for particulate carbon (PC), particulate nitrogen (PN), and δ15N using a using a Carlo Erba NA1500 CNS elemental analyzer coupled to an isotope ratio mass spectrometer (Carlo Erba/ThermoFisher Scientific™, Waltham, MA, USA). The ratio of PC to PN was calculated from PC and PN values. Due to the highly enriched label, enriched samples were analyzed at the University of Florida Soil, Water and Ecosystem Sciences Stable Isotope Laboratory and unenriched samples were analyzed at the University of Florida’s Stable Isotope Laboratory. Enriched and unenriched samples were analyzed on separate instruments, and the enriched samples had lower precision than the unenriched samples. Enriched samples had a precision of 0.4‰ (n = 4) and 0.68‰ (n = 4), and the unenriched samples had a precision of 0.26‰ (n = 3) and 0.05‰ (n = 3) for δ13C and δ15N, respectively.

All statistical analyses for bulk geochemical parameters were conducted in R version 4.3.078. The difference between final and initial timepoints and the relative change (percent change) in each parameter was calculated to determine the change in each variable over the course of the experiment. Significant differences between treatments were evaluated for each parameter using either a one-way ANOVA or a Kruskal–Wallis test, using percent change as the dependent variable. For each parameter, data were tested for the assumptions of ANOVA by using a qqplot, Bartlett test, and Brown Forsyth test79. If data met the assumptions of normality, then a one-way ANOVA was run. If a significant difference was found then Tukey HSD post hoc tests were conducted with the function HSD.test in agricolae package80. If data did not meet the assumptions of normality for an ANOVA, then a Kruskal–Wallis rank sum test was used instead, and if a significant difference was determined, then Dunn’s post-hoc test was used with the function dunn.test in the dunn.test package, using the option altp = TRUE, and an alpha of 0.0581. The statistical test used for each measure is included in Table 2.

Metatranscriptomes

RNA was extracted from filters using a QIAGEN Power water RNA extraction kit (QIAGEN, Venlo, Netherlands). RNA quality and concentration were analyzed at the University of Florida’s Interdisciplinary Center for Biotechnology Research (ICBR) using a 2100 Bioanalyzer system and a 4200 TapeStation system (Agilent Technologies, Santa Clara, CA, USA). Samples with an RNA integrity number (RIN) greater than 7 were selected for sequencing, and total RNA samples were shipped to the Joint Genome Institute (JGI) on dry ice for further processing and sequencing. RNA was extracted for the estuarine inoculant, the algal leachate, and the peat leachate in addition to experimental samples. RNA extracted from the inoculant and peat leachate were of insufficient concentration and quality for sequencing. Additionally, only three samples from the initial timepoint had sufficient RNA extracted for sequencing, resulting in no sequencing data being collected for the initial timepoint of the primed treatment. For the final timepoint, quadruplicate samples were sequenced for each treatment.

Sequence data was generated at the DOE JGI using Illumina technology, and data was processed according to JGI standard operating procedure Metatranscriptome SOP 1066.1. An Illumina library was constructed and sequenced 2 × 151 using the Illumina NovaSeq S4 platform. BBDuk (version 38.73)82 was used to remove contaminants, trim reads that contained adapter sequence and right quality trim reads where quality drops to 0. BBDuk was used to remove reads that contained 1 or more ‘N’ bases, had an average quality score across the read less than 10 or had a minimum length < = 51 bp or 33% of the full read length. Reads mapped with BBMap82 to masked human, cat, dog and mouse references at 93% identity were separated into a chaff file to remove any contaminating reads, per the JGI’s metatranscriptome SOP. Reads aligned to common microbial contaminants were separated into a chaff file. Ribosomal RNA reads were removed and placed into a separate fastq file. Reads containing known spike-ins were removed and placed into a separate fastq file.

The readset was assembled with MEGAHIT v1.1.283. This was run using the following command line options: megahit --k-list 23,43,63,83,103,123 --continue -o out.megahit --12 out.fastq.gz. The input read set was mapped to the final assembly and coverage information generated with bbmap 38.3482. This was run using the following command line options: bbmap.sh nodisk=true interleaved=true ambiguous=random in=out.fastq.gz ref=assembly.contigs.fasta out=pairedMapped.bam covstats=covstats.txt bamscript=to_bam.sh.

Data are available under GOLD study ID Gs0131207 and GOLD analysis project IDs Ga0392337-Ga0392364. The phylogenetic distribution of each sample was retrieved from IMG/MG using MG-RAST84 with a 90% identity cutoff, and estimated gene copies were exported from the IMG/MER portal for further analysis in R version 4.0.0. To minimize the effects of any inter-replicate variability, taxa were filtered for those that were present (i.e., estimated gene counts > 0) in all replicates of a treatment and time point. Then mean and standard deviation of estimated gene copies were calculated for each phylum and order at each treatment and time point in R v. 4.0.0.

Clustered orthologous groups (COGs) were annotated through JGI’s IMG system (IMGAP v5.0.11). The phylogenetic distribution of genes was evaluated based on the distribution of BLAST hits of protein-coding genes in the metatranscriptome datasets. Successive percent identity was evaluated (90% identity was retrieved) and distributions were evaluated based on estimated gene copies for all COG pathways. The phylogenetic distribution of COG pathways was evaluated at the phylum level.

KEGG Ortholog (KO) estimated gene counts were exported from IMG/MG and imported to R as a DGEList object. Differential expression of genes was evaluated using the package edgeR85 with Bioconductor (v. 3.11) and R version 4.0.0. Low counts were removed, and library sizes were recomputed after filtering. Data was then normalized to library size to minimize bias between libraries. MDS and mean difference plots were used to visualize the difference between libraries. After filtering and normalizing the dataset, differential expression analysis was conducted. Differences in expression between treatments at the final timepoint were conducted by using a glmQLFTest. The QL F-test was used because of its strict error rate control. The glmQLFTest output was then filtered for KOs that had a log-fold change not equal to zero, and a p value less than 0.05. Heatmaps of differentially expressed KOs were produced using the function coolmap with limma86 in R version 4.0.0.

The total number of reads were between 39,445,860 to 129,699,095 and the number of total genes assembled ranged from 68,723 to 1,047,686 genes (Table S1). The percentage of genes with hits in the clustered orthologous groups of proteins database (COGs) ranged from 47.92% to 71.14%. There was a lower percentage of hits in the Kyoto Encyclopedia of Genes and Genomes (KEGG) Orthology (KO), which ranged from 37.15% to 60.44%. Here, we report differences in both COGs and KOs.

Liquid chromatography mass spectrometry (LC-MS)

One aliquot of each SPE eluate was shipped to PNNL for analysis via Orbitrap Liquid Chromatography Mass Spectrometry. Liquid chromatography (LC) runs were performed on a Thermo Vanquish (Thermo Fisher Scientific, Waltham, Massachusetts, USA) high-performance liquid chromatographer (HPLC). For each sample, DOM compounds were separated in the LC by injecting 10 µL into a Hypersil gold C18 reversed-phase column (150 × 2.1 mm, 3 µm particle size; Thermo Scientific, Waltham, Massachusetts, USA) which was maintained at a constant temperature of 30 °C. Mobile phases consisted of 0.1% formic acid in water (A) and 0.1% formic acid in acetonitrile/water (90/10) (B). At a constant flow rate of 0.3 mL/min, the chromatography gradient began stable at 90% A (10% B) and maintained for 5 min before it changed steadily to 10% A (90% B) during the next 15 min (min. 20) and then held for 2 additional minutes (min. 22). The initial conditions were thus linearly recovered during the next 2 min (min. 24). Washing and stabilization of the column was performed for 11 additional minutes at the initial chromatographic conditions prior to the next analysis (min. 35). The LC was coupled to a LTQ Orbitrap Velos equipped with heated electrospray ionization (HESI) source (Thermo Fisher Scientific, Waltham, Massachusetts, USA) where the high-resolution mass spectrometry (HRMS) measurements of the compounds were carried out. For each ionization mode (positive and negative), the HRMS operated at full-scan mode and a resolving power of 60,000 full width at half maximum (FWHM) at m/z 400. All samples were run in both positive and negative ionization modes and ions between m/z 70 and 800 were detected. A blank sample consisting in methanol:water (80:20) was run every 15 samples for detecting instrument background, according to ref. 21.

LC-MS RAW files of metabolite data, including blanks, were processed with MZmine version 2.5187 (See Table S2 for Mzmine parameters). Negative and positive RAW files were separately processed. Briefly, for each individual sample, baseline of chromatograms was corrected and metabolite features (unique exact m/z with specific retention times [RTs]) were detected and deconvoluted. The generated ion chromatograms for all samples were thus aligned and gap filled before being matched against an in-house metabolite library based on the monoisotopic exact mass and RT values of standards. Peak areas of all detected features were exported to a CSV file.

CSV datasets from negative and positive ionization modes were separately filtered as follows:

  1. (1)

    Blank threshold. For each individual feature, when the number of experimental blanks containing data (integral peak value > 0) was equal to or lower than 40% of the total (2 out of 5 blank samples), data for all experimental blanks was considered zero.

  2. (2)

    Signal to noise (S/N). Individual peak areas with S/N < 5 were considered as zero. Noise level of each individual feature was represented by the average value of the 6 experimental blanks after passing the blank threshold filter.

  3. (3)

    Minimum data. Metabolite features whose peak areas with values > 0 and present in less than 50% of the replicates (2 out of 4) across all groups of samples (SW-TP0; SWA-TP0; SWP-TP0; …) were removed from the dataset. Therefore, only those variables containing peak areas with values > 0 in at least 50% of the replicates of a group of samples remained in the dataset. This filter was used to remove any single signals non-representative of a group of samples (i.e., present in a single sample).

All LC-MS metabolite statistical analyses were performed in R version 4.1.178. One-way ANOVAs were calculated with the aov function, from STATS package78, for each individual detected features to contrast different treatments. Tukey HSD post hoc tests were conducted with HSD.test in agricolae package80. Principal coordinate analyses (PCoAs) were performed on the Bray Curtis distance matrices calculated from the metabolomic fingerprints with the function bcdist from ecodist package88. PCoAs were performed with the cmdscale fundtion of vegan package89.

Direct infusion fourier transform ion cyclotron resonance mass spectrometry (DI-FT-ICR MS)

The second aliquot of each SPE eluate was analyzed via a 21 Tesla FT-ICR MS in the Environmental Molecular Science Laboratory at the Pacific Northwest National Lab. The FT-ICR MS was outfitted with the Thermo Velos Pro as the front end. Samples were automatically infused into a HESI source in negative ionization mode at a flow rate of 4 μL min−1 using an in-house direct infusion cart with the backing solvent at 70% Methanol. Two offline blanks were run after every sample to ensure no carry-over. The data was collected over 22 min. The data was collected by coadding a total of 700 μS from 220 to 1000 m/z. Elemental formulae were assigned using the Formularity software following parameters described in refs. 21,90. Non-metric multidimensional scaling (NMDS) analysis was used to visualize the complex FT-ICR MS data (Fig. 3a). The NMDS was constructed based on feature presence/absence and Jaccard similarity distance in MATLAB.

Reporting summary

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