Introduction

Peatlands store ~ 30% of global soil carbon (C)1, equivalent to 70–157% of the C currently in the atmosphere2,3. Predicting peatland C transformation and loss under continued climate warming is thus crucial to producing accurate climate forecast models. Yet, two key gaps persist in our knowledge of C gas emissions from peat bogs dominated by Sphagnum moss, a peatland keystone species. First, the mechanisms controlling the extraordinarily slow organic matter decomposition rates in bogs (which, being slower than photosynthesis, causes net C sequestration4,5) are not fully understood. While the anoxic, cold, acidic, nutrient-depleted conditions typical of bogs contribute to their resistance to decomposition6,7, decades of research suggest that Sphagnum spp. possess unique characteristics that also have significant inhibitory effects8,9,10,11,12 via mechanisms that remain controversial (i.e., release of phenolic compounds and alleged tanning agents8,12,13,14). Second, measured CO2:CH4 production ratios during bog peat decomposition often significantly exceed the 1:1 ratio that is derived from biotic anaerobic redox decomposition models15—which assume that in peatlands (i) cellulose-like organic material with an overall carbon oxidation state of 0 is the dominant substrate for C respiration; (ii) biotic metabolisms are limited primarily to fermentation and methanogenesis; and (iii) the rate at which methanogenesis produces CH4 is comparable to the rate at which fermentation produces CO2, since fermentation is inhibited by its own production of H2 and therefore relies on methanogenesis to serve as a H2 sink in order to proceed. Electron balance and consideration of oxidation state also indicate that CO2 (oxidation state =  + 4) and CH4 (oxidation state = − 4) should be produced in equimolar amounts during the decomposition of cellulosic material (oxidation state = 0). A more detailed discussion of this problem can be found in Conrad15 and Wilson et al.16 (see also the supplement: ‘Explanation of the Predicted 1:1 CO2:CH4 Production Ratio). Addressing this discrepancy in CO2:CH4 emissions is crucial to our understanding of peatland-climate interactions, since CH4 has a much higher warming potential than CO217.

We hypothesized that both the slow decomposition rates and high CO2:CH4 production ratios observed in bogs could be attributed, in part, to non-microbially mediated Maillard reactions—a type of non-enzymatic browning (NEB) involving the interaction between amine and carbonyl containing compounds18. Maillard reactions within the peat matrix would effectively slow decomposition rates by transforming primary amines into less bioavailable N-containing aromatics19, which could slow biotic decomposition rates20 by N-limitation, suppressing microbial oxidative enzyme synthesis21,22,23, and/or shifting microbial enzymes towards those that degrade refractory versus labile SOM22. Maillard reactions would also elevate CO2:CH4 ratios via abiotic CO2 expulsion which, in contrast to microbially-mediated fermentation reactions, would not necessarily be inhibited by the accumulation of hydrogen and so would not require the concomitant production of CH4 to maintain favorable hydrogen concentrations to proceed15,24. Maillard reactions have been shown to occur naturally in sediments at low temperatures relevant to peat bogs (10 °C)25. Although typically assumed to be slow in such settings, Maillard reactions are consequential on the ecosystem scale in some systems particularly when catalyzed by Fe and Mn25. While Fe and Mn are characteristically depleted in ombrotrophic bogs26 (< 1 µM Mn; < 100 µM Fe at our bog site; Table S1), the high concentrations of galacturonic acid (GalA) in Sphagnum spp. (14–21% per gram dry weight of biomass27,28,29)—a potent Maillard reagent30—could similarly stimulate such reactions (Fig. 1) despite the dearth of Maillard-catalyzing metals. In addition to the high GalA content of Sphagnum spp., Maillard reactions are plausible in bogs because (i) they contain a diverse suite of amine-bearing compounds (including Sphagnum-derived proteins/amino acids31) that could participate in N-sequestering NEB reactions and (ii) N-sequestration is a parsimonious explanation for the depth-related increase in overall N content, aromatics, and N-alkyls, in bog peats32.

Fig. 1
Fig. 1
Full size image

Three postulated pathways for galacturonic acid-based (“GalA”) CO2 emission and/or N sequestration, invoked by Bornik et al.30 and Wegener et al.19. Products/reactants are designated with blue letters and reaction pathways are indicated with red numbers. Pathway 1 involves isomerization of GalA (a) to D-tagaturonic acid (b) and subsequent decarboxylation and dehydration that leads to the production of norfuraneol (c). Pathway 2 produces 4,5-unsaturated 4-deoxy-L-arabinose (d) via decarboxylation and dehydration. Pathway 3 requires the addition of an amine functional group (e), which facilitates imine formation (f). Subsequent dehydration results in compound (g), which rearranges to form an Amadori compound (h). This undergoes keto-enol-tautomerization to form (i). Subsequent decarboxylation results in (j), which is followed by dehydration, leading to the formation of pyrolic (k) or pyrimidine (l) rings. The cumulative impact of pathway 3 (Maillard reaction) is immobilization of bio-available N (which results in enzymatic suppression) as well as abiotic CO2 expulsion. Checkmarks (green) illustrate reaction steps consistent with the FTICR-MS analysis in incubated bog peat in this study (Fig. 4c).

We hypothesized that additional NEB pathways involving GalA alone could also contribute to the high CO2:CH4 production ratios observed in bogs, as two such pathways which are relevant to the pH range studied herein have been identified that also result in CO2 expulsion (Fig. 1; pathways 1–2). Previous high-temperature (≥ 100 °C) studies have demonstrated that GalA-amine interactions are more efficient than GalA-only interactions19, leading us to hypothesize that Maillard-type GalA-amine reactions (Pathway 3, Fig. 1) produce more CO2 than GalA-only reactions (Pathways 1–2, Fig. 1) in the bog peat.

We employed a series of incubation experiments to test these hypotheses (Table S2). We first investigated whether abiotic reaction of GalA (with and without amine-bearing compounds present) results in CO2 production at low temperatures relevant to bogs. We then examined the extent to which NEB contributes to CO2 production potential and whether it is associated with N-sequestering GalA-amine interactions in bog peat. Finally, we tested potential confounding factors—including unintended biotic involvement, extracellular enzyme activity (EEA), and sterilizing treatment effects—via additional treatments and sterility assessments.

Investigating the potential for GalA to produce CO2 at low temperatures

We incubated synthetic NEB substrates for 58 days at 23 °C, the approximate mean high temperature during the growing season at the well-characterized Stordalen Mire peat bog31 (where we also collected bog peat for incubations). NEB substrates included (i) a solution of 1 M GalA—relevant to pathways 1–2 (Fig. 1)—and (ii) a solution of 1 M GalA + supersaturated amino acids (sodium glutamate, or glu, and aspartic acid, or asp; hereafter, “Amines”)—relevant to pathway 3 (Fig. 1; Maillard pathway). Amino acids (glu + asp) were selected for this study because, while any compound with a primary or secondary amine is thought to react via this mechanism, the readily obtainable amines associated with amino acids were considered ideal for observing amine interactions over short timescales. All solutions were filter-sterilized at 0.2-µm33 which we will later demonstrate was sufficient to remove unintended biotic involvement. CO2 was produced within 1 day in all solutions containing NEB reactants (time series data included in Figs. S1 and S2, as well as in Auxiliary file S2), confirming that these substrates do react to produce CO2 at peatland-relevant temperatures. This finding significantly broadens the list of potential environments that could undergo NEB-based CO2 production18,19,25,30.

CO2 production from synthetic solutions of GalA + Amines was significantly higher than in GalA-only solutions (~ fourfold, p < 0.05; 4.1 ± 0.2 × 10–2 µmoles CO2 mL−1 d−1; Fig. 2; timeseries data included in auxiliary files S2–S3 and Figs. S1–S2), consistent with previous findings that N-sequestering GalA-Amine interactions (Maillard reactions; Pathway 3, Fig. 1) are kinetically faster than NEB reactions of GalA alone19 (Pathways 1–3, Fig. 1). To confirm that the higher CO2 production in the GalA + Amines solution was due to their reaction and not to the addition of some unknown amine-based CO2 production pathway, we incubated Amine-only solutions (alongside deionized water negative controls). CO2 production from Amine-only solutions was negligible (> 300-fold lower than GalA + Amine solutions; p < 0.05; Fig. 2) and was below detection for the negative controls (6.0 ± 0.2 × 10–6 µmoles mL−1 d−1; Fig. 2). To prevent confounding chemical reactions associated with pH alteration, we did not alter the pH of the GalA + Amine solutions (pH = 3.6) which was lower than the pH of the natural bog porewater of 4.2 ± 0.231. However, these reactions are thought to be positively correlated with pH19, thus if they are shown to occur at the lower pH, then this should be a conservative estimate of the occurrence of this reaction. All other synthetic solutions discussed herein (GalA-only, Amines-only, and DI-control) were pH-corrected by the addition of NaOH or HCl to match the GalA + Amine solution (pH = 3.6). The effects of additional uncorrected pH sub-treatments are discussed in the supplement (Fig. S1).

Fig. 2
Fig. 2
Full size image

CO2 production rates (in µmoles CO2 mL−1 d−1) for four 0.2-µm filtered deionized (DI) water solutions containing various combinations of non-enzymatic browning reactants (n = 3 for each). The four treatments were: DI Water (negative control), Amines (positive control; super-saturated solutions of L-aspartic acid and sodium glutamate), 1 M GalA (1 M galacturonic acid), and 1 M GalA + Amines (1 M galacturonic acid + super-saturated L-aspartic acid and sodium glutamate). Variants on these treatments (not pictured) are indicated in Table S3.

Testing the occurrence of NEB in natural bog peats

To test whether NEB reactions can occur from the natural constituents of the organic matter available in bog peat, we collected peat from a bog site within Stordalen Mire (Sweden) and incubated it at 23 °C as slurries both live and following sterilization (via gamma irradiation). CO2 production rates from gamma-irradiated peat slurries were 13 ± 5% of rates in untreated bioactive slurries (2.2 ± 0.5 × 10–2 µmoles mL−1 d−1; Fig. 3; Fig. S3; Auxiliary file S2). These results were consistent with the hypothesis that abiotic CO2 production can occur in bogs. To further determine whether this bog-derived abiotic CO2 production could be due to the Maillard pathway, we examined changes in the dissolved organic matter (DOM) pool of gamma-irradiated and live peat slurries using FTICR-MS. Over the course of the incubation, the control peat slurries underwent significant increases in nitrogenated aromatic compounds as indicated by their lowered H/C ratios (Fig. 4A). Further, the fraction of N compounds in aromatic, condensed formula also increased during the incubation period for the control incubations (Fig. 4B). (Notably, the total number of N-containing compounds did not change significantly over time for either the control—mean = 451; p = 0.3—or the gamma-irradiated—mean = 57 ± 12; p = 0.3—treatments). Both results are consistent with GalA-based Maillard reactions.

Fig. 3
Fig. 3
Full size image

CO2 production rates in incubated porewater (PW) and peat by treatment (n = 3). CO2 production rates are in µmoles CO2 mL−1 d−1, where “mL” references the volume of porewater in each incubation. Bioactive porewater was filtered to 0.7-µm in the field. “0.2-µm” indicates 0.2-µm filtration in the lab prior to the incubation set up. “10-kDa” signifies porewater filtered to 10 kDa to remove hydrolytic enzymes. “DI” is deionized water; “GIR” is gamma-irradiated.

Fig. 4
Fig. 4
Full size image

Shifts in dissolved organic matter (DOM) over time during the gamma-irradiated incubations. Panel (a) compares the H/C ratio versus O/C ratio of compounds over time in the control incubations. The size of the symbols indicates relative intensity of each formula normalized to the total signal intensity of the sample. The reduced H/C ratio from day 0 to day 50 indicates increasing aromaticity. Panel (b) shows a significant increase (p < 0.05) in the proportion of condensed N compounds relative to the total number of N-containing compounds, expressed as a percentage, in the control (live) incubations over time. Notably, the total number of N-containing compounds did not change significantly over time in either the control or gamma-irradiated incubations. Panel (c) compares the proportion of transforms consistent with the galacturonic acid (“GalA”) NEB reaction relative to the total number of observed transforms, expressed as a percentage, in the gamma-irradiated samples and the control (live) incubations. The total number of GalA-involved transforms (the numerator in the percentage calculation) is displayed in white text at the bottom of the bars in panel (c). Transforms consistent with four steps in reaction 3 (Fig. 1) are shown. These include: GalA; GalA-H2O (3-I through 3-III); GalA-H2O-CO2 (3-I through 3-VI); and GalA-H2O-CO2-3H2O (3-I through 3-VII). For all but the final (righthand) panel, the percent transforms are significantly (p < 0.05) higher among gamma-irradiated versus control samples (p values included above the bars in black text).

Kendrick mass analysis was used to identify the occurrence of individual steps within the GalA + amine pathway in the control incubations to find evidence consistent with the occurrence of pathway 3 (Fig. 1). Kendrick mass analysis involves mathematically transforming the FTICR-MS m/z data to alternate mass scales by defining the desired chemical moiety as having an integer atomic mass value and then using that conversion factor to scale the remaining peaks into the new mass scale (sensu34). The Kendrick mass (KM) of a peak is then defined as KM = observed mass * Fnominal mass/Fexact mass, where F is the transform corresponding to the addition/removal of the functional group of interest. The difference between the resulting exact and nominal masses is the Kendrick mass defect (KMD). In the newly calculated masses, peaks with the same KMD indicate molecular formulae that differ by integer multiples of the chemical moiety (transform) investigated. From this we infer that the two compounds are linked by the addition or removal of the functional group (explained in detail in16 [references therein]). For example, if CO2 is the chemical moiety under consideration, after transforming the data, if there are two peaks with identical KMD it is considered that the higher mass peak was decarboxylated to the lower mass peak. We applied this data transform to the samples based on a list of 180 chemical moieties assembled from the literature previously identified as occurring in peat samples that represented potential microbial decomposition strategies (and abiotic reactions of interest, see16). When two peaks within the spectrum under consideration had KMD values within 0.000001 mass units and an identical z* score, those peaks were counted as a transform pair.

For each of the 180 transforms that we investigated, we counted the total number of such transform pairs in each sample. Using this approach, we were able to identify compounds involved in KMD pairs that are consistent with various steps of the GalA reaction pathway 3 (Fig. 1). These included addition of GalA (exact mass = 194.042655; Fig. 1, pathway 3-I), addition of GalA followed by loss of 1 water (exact mass = 176.03209; Fig. 1, pathways 3-I through 3-III), addition of GalA followed by loss of 1 water and 1 CO2 (exact mass = 132.0422589; Fig. 1, pathways 3-I through 3-VI); and addition of GalA followed by loss of 1 CO2 and 4 waters (exact mass = 78.010565; Fig. 1, pathways 3-I through 3-VIII). Finding two peaks separated by these exact masses (within 0.000001 mass units) was taken as an indication that the two compounds were involved in the identified step of the GalA pathway.

The total number of pairs in each of these steps was presented as the proportion of the total number of transform pairs identified in each sample. That is, we took the number of pairs involved in each GalA pathway step and divided it by the total number of transform pairs involved in any of the 180 transforms from our list (and multiplied by 100 to achieve a percentage). This data was presented as a percentage to control for differences in the total number of transforms identified between the live samples and killed controls. Because the killed controls presumably no longer include transforms from the biotic pathways of decomposition, but the gamma-irradiation might stimulate more abiotic reactions to occur, it is unclear whether the killed controls or the live experiments would have more total transforms in them. It is important to note that a single peak can be involved in multiple transforms, from which we infer that there are different pathways by which it can be degraded. Thus, the total transform counts need not sum to unity.

We found evidence for four steps in the Maillard GalA-amine reaction (Fig. 1, Pathway 3, reactions 3-I, 3-III, 3-VI, and 3-VII/3-VIII), which are associated with both CO2 production and bioavailable N sequestration (Fig. 4C). Each of the observed Pathway 3 transforms except for 3-VII were significantly higher in the gamma-irradiated versus bioactive peat incubations (p ≤ 0.05; Fig. 4C). This finding is consistent with the occurrence of Maillard reactions in both sterilized and live peat incubations. Furthermore, the observation that Maillard reactions generally contributed more to C transformations in sterilized peat than in bioactive peat confirms that Maillard reactions are an active abiotic process in field-derived peats at environmentally relevant temperatures.

Several potential confounding sources of CO2 in the sterilized incubations were assessed including: (i) chemical disruptions imposed by the gamma-irradiation, (ii) extracellular enzyme activity (EEA), and (iii) biotic production due to a lack of sterility. Gamma irradiation is known to be highly oxidative35,36, and FTICR-MS analysis of the DOM revealed a disruption of peat chemistry. Relative to living controls, gamma-irradiated DOM samples had higher O/C ratios, higher nominal oxidation state of carbon (NOSC), lower double bond equivalents, and lower aromaticity indices (Fig. S2). To eliminate these impacts, we repeated sterilized incubations with filter-sterilized (0.2-µm) bog porewater, which does not result in disruptive oxidizing effects to the DOM37. These filter-sterilized incubations also produced appreciable CO2, equivalent to approximately 8% that of untreated bioactive peat (on a per mL porewater basis). The difference relative to that observed in gamma-irradiated peat incubations (8 vs. 13%, p < 0.05) suggests that chemical disruptions from gamma-irradiation may have appreciably contributed to CO2 production in gamma-irradiated incubations. It is also possible that incubating the porewater without peat had lower substrate concentrations; DOM in incubations of peat tends to increase over time probably due to leaching.

Additionally, EEA may have stimulated CO2 production37, including in 0.2-µm filtered incubations38,39,40,41,42,43. To isolate this potential contribution, we incubated bog porewater filtered to 10-kDa (which filters out hydrolytic enzymes, which typically measure ≥ 50-kDa40,41,42,43). There was no significant difference in CO2 production relative to 0.2-µm-filtered porewater (Fig. 3), indicating that EEA did not significantly contribute to the ~ 8%-of-bioactive CO2 production in these incubations. To our knowledge, all prior soil examples of acellular CO2 production via EEA were conducted under aerobic conditions, or in the presence of substantial high-energy-yielding inorganic terminal electron acceptors (i.e., Fe); thus, the lack of apparent EEA-based CO2 production in our anaerobic, nutrient-deplete, 0.2-µm filtered incubations is unsurprising40 [references therein].

To assess for potential biotic CO2 production in ostensibly filter-sterilized bog incubations (10 kDa, 0.2-µm, and 0.7-µm), we (i) performed standard sterility growth assessments (involving additions of tryptic soy broth and fluid thioglycolate medium)44, (ii) attempted to stimulate any biota present in the filtered bog porewater by measuring CO2 production rates after addition of a simplified media of nutrients known to stimulate biotic CO2 production (i.e., glucose + nitrates + phosphates), and (iii) measured isotopic fractionation. With standard growth-based sterility assessments, no visible growth occurred in the 0.2um-filtered samples (Fig. S3). One of fifteen gamma-irradiated peat replicates exhibited slow growth, but its CO2 production did not significantly differ from other replicates, suggesting that contamination occurred either during the destructive sampling after the CO2 production period ended, or during the gas analysis period but without significant impact on CO2 production. With attempted biotic CO2 stimulation (via simplified media addition), no significant additional CO2 production occurred in 10-kDa or 0.2-µm filter-sterilized porewaters (but did in the positive control, of 0.7-µm filtered porewater; Fig. S4), suggesting that sub-micron-scale biotic influences were minimal. Lastly, we measured the apparent fractionation of δ13C during transformation of DOC to CO2 in gamma-irradiated peat incubations. The measured fractionation factor (~ − 7.5‰) is consistent with kinetically driven abiotic reactions. Biotic transformation would have yielded no fractionation if organic matter was oxidized to CO2 without CH4 production; if CH4 was also produced, which it was not, biotic fractionation would have been ~ 30–80‰45. Cumulatively, these results provide strong evidence that the observed CO2 production in the sterilized peat incubations were not due to biotic processes, but the result of abiotic reactions that produced CO2.

NEB pathways critical to peatland-climate interactions. Sphagnum moss is a keystone species, the ecosystem engineer of peat bogs46. Our results suggest that its release of GalA into porewater is likely contributing to abiotic CO2 production, leading to an elevation of the CO2:CH4 production ratio. Additionally, our results suggest that interactions between GalA and amine-bearing compounds (Maillard reactions) lead to bioavailable N sequestration, thereby suppressing enzyme function and slowing overall decomposition rates. Since Sphagnum spp. thrive in already-ombrotrophic environments47, this potential N-sequestering process may also hinder plant competitors. Wetland biogeochemical models typically limit decomposition rate calculations to biotic factors, and often impose a fixed CO2:CH4 production ratio48,49,50; incorporating NEB-derived CO2 and N sequestration would improve representation of wetlands’ contributions to global C balance. Critical to this will be quantifying NEB’s magnitude under different conditions and its temperature response which, given the kinetically controlled nature of NEB, would likely play a strong regulatory role in NEB-climate interactions.