Characterization of the sediment and soil samplesFor our experiments with natural samples, we chose lake sediment and forest soil samples with similar endogenous Fe content (~50 mg Fe kg−1 dry sample), excavated from Missisquoi Bay in the U.S. state of Vermont and an Ultisol from the Calhoun Experimental Forest in the U.S. state of South Carolina, respectively (Fig. 1a). Missisquoi Bay is a eutrophic section of Lake Champlain that undergoes diel and seasonal oxic-anoxic cycles39 (Fig. 1a, b). These redox cycles promote dynamic formation of Fe oxides, which are implicated in the mobility and fate of the bioavailable sediment P, albeit the mechanism is not well understood. In this lake environment, P transformation is attributed to enzymatic and biological activity, but little is known about the potential conversion of Porg to Pi by Fe oxides (Fig. 1b). In our complimentary forest setting, the soils at the Calhoun Experimental Forest contain redoximorphic features and undergo redox fluctuations in response to rainfall and organic carbon pulses, leading to periodic dissolution and reformation of Fe minerals40 (Fig. 1b).Fig. 1: Characterization of the sediment and soil samples.a (left) Location and (right) topography of (top) Missisquoi Bay (N44°59’33” W73°8’20”) and (bottom) Calhoun Critical Zone Observatory (N34°36’33.012” W81°43’40.62”) for lake sediment and forest soil sampling sites, respectively. Scale bar in (a) represents 500 km in geographical location and 10 km in topography visualization; arrow and orange vertical line specify sampling location. Maps were created using QGIS and topography data from the U.S. Geological Survey. b Illustration of the current view for the abiotic fate of Porg in (top) lake and (bottom) soil; legend for (b): organic phosphorus (Porg; white circle), inorganic phosphorus (Pi; light red circle), dissolved iron (Fe) (Fe2+; black text), Fe oxide mineral (light brown hexagon). c Mineral content in the (top) sediment and (bottom) soil samples: ferrihydrite (brown), hematite (red), goethite (peach), biotite (yellow), glauconite (gold), albite (gray), quartz (white), and illite or kaolinite (blueish gray). In (c) mineral composition was determined by combining X-ray fluorescence, X-ray diffraction, and Fe X-ray absorption near-edge structure spectroscopy data (SI, Fig. S1-S5). Source data are provided as a Source Data file.We combined multiple characterization techniques to determine the specific minerals in the sediment and soil samples that would be responsible for adsorption and catalytic reactivities towards Porg compounds: X-ray fluorescence (XRF) to determine the major elements present [Supplemental Information (SI), Fig. S1]; X-ray diffraction (XRD) to identify the major crystalline mineral phases (quartz, micas, feldspars, clays, or Fe oxides) in accordance with the elemental composition from the XRF analysis (SI, Fig. S2); and, Fe K-edge XANES spectroscopy to resolve different forms of Fe phases such as Fe oxides of low crystallinity (i.e., ferrihydrite) and high crystallinity (i.e., hematite, goethite), an Fe phosphate mineral (i.e., vivianite), and an Fe-rich mica (i.e., biotite) (SI, Fig. S3-Fig. S5, and Table S1). While our Fe XANES data accounted for low-crystallinity Fe-mineral phases, our XRD data do not account for the possible presence of low-crystallinity silicate or aluminosilicate phases in the natural samples. With respect to the possible presence of low-crystallinity aluminum (Al)-containing minerals, we obtained 27Al nuclear magnetic resonance (NMR) spectra of the sediment and soil samples to probe for the expected five-coordinated Al in the reactive sites of these minerals41,42,43. Our NMR data only detected tetrahedral and octahedral Al as would be found in clays and micas, which can be identified by XRD; no five-coordinated Al was detected (SI, Fig. S6). Furthermore, as will be discussed in the next section, there was minimal to no catalytic reactivity of silicate or aluminosilicates towards the different Porg compounds (Fig. 2). No Mn-bearing minerals were included in the mineral composition analysis because neither environmental sample exhibited a quantifiable amount of Mn (SI, Fig. S1).Fig. 2: Catalytic and adsorption reactivities of minerals towards different organic phosphorus (Porg) structures.a, b Transformation of adenosine triphosphate (ATP)-P (150 µM or 4.6 mg L−1) after 7 day reaction with mica, illite, quartz, kaolinite, ferrihydrite, hematite, or goethite at either 1 g L−1 (e.g., Quartz 1) or 4 g L−1 (e.g., Quartz 4). Transformation of (c) adenosine monophosphate (AMP)-P (50 µM or 1.5 mg L−1), (d) phytate-P (300 µM or 9.0 mg L−1), and (e) glucose-6-phosphate (G6P)-P (50 µM or 1.5 mg L−1) after 7 day reaction with mica, quartz, illite, kaolinite, ferrihydrite, hematite, or goethite (1 g L−1). Data are represented as mean values +/- SD where n = 3 independent samples except for n = 9 for ATP blank and n = 6 for AMP blank. Aqueous inorganic P (Pi) and Porg species were determined by ultraviolet-visible absorption spectroscopy and high-resolution liquid chromatography mass spectrometry analysis, respectively; particulate Pi and Porg species were determined by P K-edge X-ray absorption near-edge structure spectroscopy. Color codes: dissolved Porg reactant (white), dissolved Porg product (yellow), particulate Porg (brown), dissolved Pi (red), particulate Pi (red stripes). Source data are provided as a Source Data file.Taken collectively, our analysis revealed that the sediment sample contained primarily glauconite (a mica, 26%), followed by quartz (21%), illite (a 2:1 clay, 16%), albite (a feldspar, 16%), and biotite (a mica, 8%); about 12% of the sediment contained Fe oxides identified as ferrihydrite (~10%) and goethite (2%), with a minor fraction (<1%) present as vivianite (Fig. 1c). The soil sample contained primarily kaolinite (a 1:1 clay, 69%) and quartz (12%), with the remaining fraction constituted of Fe oxides as ferrihydrite (10%), hematite (4%), and goethite (4%); vivianite was also present as a minor fraction (<1%) (Fig. 1c). Our sediment characterization was in agreement with a previously reported mineral composition for the sediment sample44, albeit the Fe content was greater than a previously reported value39. For the soil, the ratios of Fe oxide minerals in the soil mineral composition were in close agreement with the reported values from extracted Fe content, but the absolute values were not directly comparable to previous 57Fe Mössbauer spectroscopy data45. These discrepancies with previous data were likely due to differences in sample preparation or characterization techniques, such as the different scales of ordered minerals probed by Fe XANES versus Mössbauer spectroscopies and characterization of separate 5 µm soil sections versus bulk sample. In sum, based on our analysis, we found that about 80% (or more) on a per-mass basis of both sediment and soil samples was comprised of silicate minerals of different types (quartz, micas, feldspars, clays), and <20% constituted the Fe oxide fraction (Fig. 1c). Next, we explored the reactivities of these different mineral types in the environmental samples towards Porg compounds.Exceptional adsorption and catalysis by Fe oxidesWe performed reactions of various Porg compounds with the different mineral types identified in the natural samples (at 1 g L−1 or 4 g L−1): biotite (a mica), quartz, kaolinite (a 1:1 clay), illite (a 2:1 clay), ferrihydrite (a low-crystallinity Fe oxide), hematite (a crystalline Fe oxide), and goethite (a crystalline Fe oxide) (Fig. 2; SI, Fig. S6–Fig. S10). In our first set of experiments, we chose ATP (ATP-P, 150 µM or 4.6 mg L−1) as our representative Porg that contains both phosphoanhydride (P-O-P) and phosphoester (C-O-P) bonds that are ubiquitous in the reservoir of Porg compounds derived from metabolism and biomass of plants and microbes. We used high-resolution LC-MS to quantify particulate and aqueous species of Pi and Porg for the ribonucleotide reactants and products in solution (i.e., ATP, ADP, AMP, adenosine), a visible light absorption spectroscopy method for solution Pi, and synchrotron-based P K-edge XANES spectroscopy for the relative fractions of particulate Pi and particulate Porg associated with Fe-bearing minerals; particulate Pi on non-Fe minerals was determined via mass balance based on the concentrations of organic and inorganic products in solution after reactions with a Porg reactant. We note that the XANES analysis can determine the relative organic fraction versus inorganic fraction of the total particulate P, but does not discriminate between possible different Porg compounds in the particulate Porg fraction14.Following the reaction with ATP, quartz and the aluminosilicates (mica, kaolinite, illite) exhibited minimal to no catalytic reactivity whereby aqueous Pi or particulate Pi accounted for <4% of the reacted Porg; in terms of adsorption reactivity, less than 20% of the reacted Porg was found as particulate Porg with these minerals (Fig. 2a). In contrast, the reactions with the different Fe oxides (ferrihydrite, hematite, or goethite) yielded substantial dephosphorylation or adsorption of the Porg reactant, characterized by up to 31% as aqueous Pi, 12–36% as particulate Pi, and up to 46% as particulate Porg (Fig. 2a). In regards to the fate of the reacted P (i.e., sum of transformed and adsorbed Porg-derived P) across the Fe oxide minerals, ferrihydrite exhibited the highest adsorption with greater than 75% of reacted P as particulate Porg, while goethite was the most catalytically active, with 53–79% of the reacted P as collectively adsorbed Pi and aqueous Pi (Fig. 2a). Despite a greater than 10-fold difference in the specific surface area of ferrihydrite (230 m2 g−1) compared to goethite (16 m2 g−1), there was a near 3-fold higher fraction of particulate P species on goethite than on ferrihydrite (p < 0.05) (Fig. 2a), suggesting that mineral surface chemistry rather than specific surface area governs the extent of catalytic reactivity, a worthwhile research avenue to pursue in future investigations.Catalytic reactivity of Fe oxides in mineral mixturesSince Fe oxides are present in heterogeneous mineral mixtures in natural samples, we also investigated how the presence of other minerals (ferrihydrite, quartz, or kaolinite) would influence the high catalytic reactivity of goethite towards ATP (Fig. 2b). We prepared mineral mixtures based on the composition of the natural samples (Fig. 1c). First, with a 4:1 ferrihydrite:goethite mixture, we found that the amount of particulate Pi generated in the mixture was equivalent to that of goethite alone (p = 0.21) while there were an 8-fold increase in particulate Porg (p < 0.001) and no Pi in solution, indicating that the higher adsorption reactivity of ferrihydrite relative to goethite overwhelmed the reactivity in the ferrihydrite:goethite mixture (Fig. 2b). Second, when comparing the goethite-only condition to 4:1 quartz:goethite and kaolinite:goethite mixtures, the same amount of dissolved Pi was produced from the Porg reactant (p ≥ 0.75) but only 20% of the particulate Pi fraction remained (p < 0.01), and this was accompanied by a 6-fold increase in particulate Porg (p < 0.01) (Fig. 2b). Our data thus implied that, in terms of the production of solution Pi or particulate Pi, Fe oxides would likely retain their catalytic reactivity in mixed-mineral matrices in environmental samples.Catalytic reactivity of Fe oxides with different Porg typesAmongst the chemical diversity of Porg types found in biomolecules, phosphomonesters are widely found in soils36,37,38,46,47. To probe Fe oxide reactivity with these other types of Porg, we performed reactions involving each of the three Fe oxides (ferrihydrite, hematite, and goethite) with three phosphomonoesters: AMP (AMP-P, 50 µM P or 1.5 mg P L−1), G6P (G6P-P, 50 µM P or 1.5 mg P L−1); and phytate (phytate-P, 300 µM P or 9.0 mg P L−1) (Fig. 2c–e). To compare the Fe oxide reactivity with the other mineral types in the environmental samples, we also performed experiments of the phosphomonoester compounds reacted with quartz and the aluminosilicates (mica, kaolinite, illite) (Fig. 2c–e).In contrast to the ATP reactions, all the Pi derived from the reacted AMP with goethite and ferrihydrite was retained as particulate Pi while aqueous Pi was absent, a significant finding that was made possible here due to the application of the XANES technique (Fig. 2c). On the one hand, the catalytic reactivity of ferrihydrite was higher for AMP than for ATP, as reflected by the 20-fold increase in the particulate Pi fraction (p < 0.001) (Fig. 2a, c). On the other hand, the catalytic reactivity of goethite was less for AMP than for ATP as characterized by a 3-fold decrease in particulate Pi fraction (p < 0.01) accompanied by no change in the particulate Porg fraction (p = 0.29) and no measured solution Pi (Fig. 2a, c). Hematite did not display any adsorption or catalytic reactivity towards AMP (Fig. 2a and c). These findings here are consistent with a previous report15 of 2-fold to 6-fold greater reactivity of goethite than hematite for p-NPP dephosphorylation, albeit direct comparison between p-NPP (a synthetic Porg) and AMP (a natural biomolecule) is not appropriate due to the smaller molecular weight of p-NPP versus AMP (263.1 g mol−1 versus 324.23 g mol−1) and the difference in their chemical structures such as the presence of a sugar base and two heterocyclic nitrogenous rings in AMP whereas p-NPP has one benzyl ring. Interestingly, after all the ATP reacted with goethite was transformed or adsorbed in our aforementioned experiments with ATP and goethite, 20–24% of the initial ATP-P remained as AMP-P in solution and no ADP was detected (SI, Fig. S11). The results with the AMP-goethite experiment implied that this accumulation of AMP in the ATP-goethite experiment was due to the lower reactivity of goethite for AMP relative to ATP (Fig. 2c; SI, Fig. S11). Notably, the silicate and aluminosilicate minerals did not exhibit any adsorption or catalytic reactivity towards AMP (Fig. 2c).With respect to phytate, the silicate and aluminosilicate minerals were all found to adsorb phytate (from 5% to 50% of the total reacted phytate-P) and, rather than an Fe oxide, the clay illite adsorbed the most phytate (Fig. 2d). However, similar to the results with AMP and ATP, catalytic reactivity towards phytate was only obtained with Fe oxides (Fig. 2d). Relative to controls, phytate-derived particulate Pi was higher by 8–14% (p < 0.01) with hematite and by 24–32% with goethite (p < 0.001); the particulate Pi with ferrihydrite, however, corresponded to the adsorption of solution Pi from the control experiment (Fig. 2d). As with AMP, the most significant adsorption for G6P was with ferrihydrite, with 47–78% of the reacted G6P found only as particulate Porg (Fig. 2e). Some G6P adsorption (4–12%) was observed with two of the aluminosilicates (mica and kaolinite) and one of the other Fe oxides (goethite) (Fig. 2e). None of the investigated minerals catalyzed G6P dephosphorylation (Fig. 2e). Only data with ferrihydrite and goethite were consistent with reports by Wan et al.16 of higher aqueous Pi production from AMP than from G6P and phytate. Even accounting for particulate Pi, we obtained lower reactivity of AMP and G6P with hematite than reported by Wan et al.16. Compared to our experiments, these previous experiments16 were performed with 60% less hematite (on a per-mass basis), 20-fold higher concentration for G6P and AMP, and 3-fold higher concentration for phytate (Fig. 2c–e). Therefore, the discrepancy could be due to possible lower reactive sites in hematite compared to ferrihydrite and goethite, thus necessitating higher Porg concentration to observe reactivity.In sum, our data revealed that the silicates and aluminosilicates either had minimal to no catalytic reactivity or exhibited variable extent of adsorption reactivity towards the different compounds. Only the Fe oxides were found to catalyze appreciable to significant dephosphorylation of Porg compounds containing phosphoester and phosphoanhydride bonds. Importantly, the Fe oxide-catalyzed reactions seemed to be dependent on both the mineral surface chemistry and the type of Porg species. As proposed previously14,28, we posit that the differences in catalytic reactivity may stem from the binding conformations of different Porg compounds on the surface of the Fe oxide minerals.Evolution of Particulate Pi from reacted Porg in natural samplesOur findings with the pure minerals stressed the exceptional adsorption and catalytic reactivities of the Fe oxides relative to the other minerals identified in the environmental samples. Notably, we found that particulate Pi, which was largely ignored in previous investigations, was an important product of the reaction of Fe oxides with Porg species bearing a phosphomonoester bond or a phosphoanhydride bond. Furthermore, our data with mineral mixtures implied that the Fe oxide-mediated catalysis would likely remain prominent even in the heterogeneous mineral matrix of environmental samples. To evaluate this, we performed Porg reactions with natural sediment and soil samples.First, we determined the starting P content in the natural samples: the sediment sample contained, on average, 1334 mg P per kg, with 49% as particulate Porg and 51% as particulate Pi; the soil sample only had 0.43 mg P per kg, with 56% as particulate Porg and 44% as particulate Pi (Fig. 3a). Relative to the starting P content in the natural samples, reacting ATP-P (300 µM or 9.3 mg L−1) with the sediment and soil samples (at 1 g L−1) generated excess particulate P of, on average, 2.7 mg P per g of the sediment sample and 4.1 mg P per g of the soil sample (Fig. 3a, b). Importantly, of this ribonucleotide-sourced particulate P, 20 – 50% was particulate Pi and 52 – 79% was particulate Porg (Fig. 3a, b). Therefore, following reactions with ATP, the sediment sample had a near 3-fold increase in the amount of both Porg and Pi in the particulate fraction and the soil sample had 3 orders of magnitude higher of particulate Porg and Pi (Fig. 3a). Interestingly, of the total Pi evolved from the Porg reactant, only about one-third was solution Pi while nearly two-thirds remained as particulate Pi fraction, thus highlighting a significant underestimation of the produced Pi would result in the absence of particulate P speciation (SI, Fig. S12). Biologically mediated Porg dephosphorylation was not expected to be significant in our natural samples due to long-term storage (~4 years) of both samples and low carbon loading (<0.2% g C g−1 soil) particularly for the soil sample. Nevertheless, we tested the possibility of residual microbial or enzymatic reactions in the natural samples by performing experiments with an antimicrobial agent or an enzyme denaturing agent, respectively (SI, Fig. S13). We determined that these biotic reactions accounted for only 0–5% and 23 – 26% of the total reactivity in the sediment and soil samples, respectively (Fig. 3a, b). Taken collectively, our findings bring attention to the occurrence of a pool of abiotically generated particulate Pi from mineral-mediated Porg transformation that has been hitherto unaccounted for in environmental matrices. Next, we probed which of the minerals within the heterogeneous matrix of the natural samples may be responsible for the Porg reactivity.Fig. 3: Particulate inorganic phosphorus (Pi) generated from ribonucleotides reacted with natural samples.For (a) sediment and (b) soil samples: (top) Contribution of abiotic reactivity (gray) and (bottom) distribution of particulate organic phosphorus (Porg; brown) and particulate Pi (pink) before (reference, ref) and after 7 d reaction with Porg as adenosine triphosphate (ATP) (ATP-P, 300 µM or 9.3 mg L−1) with 1 g L−1 of the dry sediment or soil sample. Error bars represent (top) standard deviation of 3 independent replicates for abiotic contribution or (bottom) errors in X-ray absorption near-edge structure spectroscopy model fitting for particulate P speciation. c, d μ-X-ray fluorescence mapping of count intensity for iron (Fe) (blue, max intensity = 500 for c or 900 for d) and P (red, max intensity = 25,000) in (c) sediment or (d) soil before (left) and after (right) 7 d reaction with ATP. In (a) and (b) background particulate P species in the reference sediment and soil samples are noted as “Sediment ref” and “Soil ref”, respectively. In (c) and (d) the scale bars (shown in white) represent 300 µm. Source data are provided as a Source Data file.Association of evolved Pi with Fe oxides in natural samplesThrough µ-XRF mapping, we found that Fe was distributed throughout the sediment and soil samples and that P appeared to be co-localized with Fe in distinct regions (Fig. 3c and d); we confirmed the lack of strong correlation of P with calcium (Ca) (SI, Fig. S14). Subsequently, we obtained extensive bulk P K-edge XANES data to determine the mineral types associated with the particulate P generated from the reacted ribonucleotide by conducting linear combination fitting (LCF) using XANES spectra obtained from separate experiments of Porg or Pi with the different mineral components identified in the natural samples (Fig. 4; SI, Table S6 and Table S7). It was not possible to use the XANES spectra to distinguish the specific P species associated with the different aluminosilicates nor the specific Fe oxide minerals associated with either particulate Pi or particulate Porg species (SI, Fig. S15). Specifically, we were able to employ the LCFs of the XANES spectra to distinguish Pi associated with Ca using apatite as a reference, P (without discriminating between Pi or Porg) associated with silicates and aluminosilicates, Pi associated with Fe in Fe oxides, Porg species associated with Fe in Fe oxides, and Pi in Fe-Pi clusters using vivianite as reference (Fig. 4a–e). We determined the fate of newly generated Pi and Porg following the ATP reaction with the natural samples by accounting for the total particulate P in excess of the background amount in the reference samples and characterizing the differences in the fractions of Pi and Porg associated with specific minerals in natural samples before and after reactions with ATP (Fig. 4f, g). The following were characteristic features in the XANES spectra that enabled the LCFs to capture distinct P speciation in the natural samples reacted with Porg: differences in the rising edge of the white line for Porg versus Pi (Fig. 4c, d), the shift in the peak of the white line with different fractions of Pi versus Porg bound to Fe oxides (Fig. 4d), and the pre-edge region of silicate-bound P (4e).Fig. 4: Specific association of organic phosphorus (Porg)-derived inorganic phosphorus (Pi) with iron (Fe) and Fe oxides in sediment and soil samples.a, b Bulk P K-edge X-ray absorption near-edge structure spectroscopy (XANES) data (black line) of the (a) sediment and (b) soil samples after 7 d reaction with the Porg reactant (adenosine triphosphate; ATP); the model fits (red line) are from linear combination fitting (LCF) of reference spectra (gray lines) of Pi or Porg reacted with the different mineral types identified in the samples (see Fig. 1c). c, d, e Close-up of spectral regions of the XANES data obtained with: (c) the reference Porg (ATP, prepared as ATP disodium salt hydrate) powder, the reference Pi (prepared as disodium phosphate), and the soil reacted with Porg reactant; (d) a representative Fe oxide (FeOx) mineral (goethite) reacted with ATP (Porg) for 75 min or 7 days, the same mineral reacted with Pi for 7 days, and the soil reacted with the Porg reactant; (e) a representative aluminosilicate (Al-silicate) mineral (kaolinite) reacted with ATP (Porg) for 75 min or Pi for 7 days, and the soil reacted with the Porg reactant. Color key: soil reacted with ATP (black line); Porg reference or Porg bound to a mineral (yellow line); Pi reference or Pi bound to a mineral (red line); and Porg bound to an Fe oxide (light blue dashed line). f, g Speciation of the excess particulate P content (103 mg P kg−1 dry sample) generated following 7 d ATP reaction with (f) the lake sediment sample or (g) the forest soil sample. Data in (a), (b), (f) and (g) are detailed in SI Table S2, Table S3, Table S4, Table S5, Table S6, and Table S7. In (f) and (g) error bars represent model fitting error range from the LCF of XANES spectra from a single bulk P K-edge XANES measurement. Source data are provided as a Source Data file.In accordance with the observed higher reactivity of the Fe oxides relative to the other minerals in the reactions with pure minerals (Fig. 2a, b), we found that both particulate Porg and particulate Pi generated after reactions of the sediment and soil samples with the ribonucleotide reactant were associated primarily with Fe and Fe-oxide fractions in the sample matrix (Fig. 4f, g). On a dry sediment basis, about 72% of the ribonucleotide-derived particulate P was found in clusters of Fe-complexed Pi (1066 ± 118 µg g−1) and particulate Porg associated with the Fe oxide fraction (912 ± 502 µg g−1); the remaining was P associated with aluminosilicate minerals (651 ± 97 µg g−1), and Pi in Ca-phosphate mineral (110 ± 22 µg g−1) (Fig. 4f). Based on our experiments with pure minerals (Fig. 2a), the accumulation of Fe-Pi clusters in the sediment sample matrix was attributed to strong Fe binding of Pi catalytically generated from the high abundance of particulate Porg associated with Fe oxides (Fig. 4f). In contrast to the lake sediment, the particulate P generated from the reacted ribonucleotide with the soil sample was nearly completely associated with the Fe oxide fraction, whereby ~28% was Fe oxide-bound Pi (1124 ± 279 µg g−1) and the remaining 72% was Fe oxide-bound Porg (2758 ± 103 µg g−1); there was a relatively minor amount of P associated with silicates (181 ± 127 µg g−1) (Fig. 4g). Based on the different mineral compositions of sediment versus soil samples and informed by our experiments with pure minerals (Fig. 2a), we posit that the observed differences between the natural samples in the mineral associations of the ribonucleotide-sourced P was due to the different forms of Fe oxides (hematite-free versus hematite-rich Fe-oxide fractions), silicate types (quartz-enriched versus clay-enriched), and available Ca (nearly 12-fold higher Ca content in the sediment sample) (Table S8).Rates of Fe oxide-mediated catalysis versus soil enzymesOur data thus far point to the production of both aqueous Pi and particulate Pi during Fe oxide-mediated Porg dephosphorylation, the environmental relevance of which needs to be considered in relation to reported enzymatic rates (Fig. 5; SI, S11 and S12). To this end, with the three different Fe oxides (goethite, hematite, ferrihydrite) identified in the natural samples, we performed kinetic experiments to obtain the production rates of aqueous Pi and particulate Pi species as a function of reacted ATP concentrations (25–400 µM), and subsequently determined two apparent maximal dephosphorylation rates (Vmax): one Vmax for the production rate of dissolved Pi and one Vmax for the production rate of particulate Pi (Fig. 5a, b; SI, Fig. S16). While the goethite-catalyzed reaction generated both dissolved and particulate Pi, only particulate Pi was found in appreciable quantities with both hematite and ferrihydrite (Fig. 5b). The Vmax for particulate Pi with ferrihydrite (2.59–4.24 µmol Pi h−1 gmineral−1) was up to 4-fold higher than with goethite (0.455–1.70 µmol Pi h−1 gmineral−1, p < 0.05) and up to 20-fold higher than with hematite (0.164–0.576 µmol Pi h−1 gmineral−1, p < 0.05) (Fig. 5b). This marked difference in the production rate of particulate Pi is consistent with the well-known higher adsorption reactivity of low-crystallinity Fe oxides relative to crystalline Fe oxides48,49,50. With goethite, the Vmax for Pi in solution (2.70 – 4.65 µmol Pi h−1 gmineral−1) was nearly 5-fold greater than the corresponding Vmax for the particulate Pi measured on the goethite surface (p < 0.05), indicating that goethite exhibited more of an enzyme-like behavior compared to the other Fe oxides (Fig. 5b).Fig. 5: Dephosphorylation kinetics of iron (Fe) oxides and environmental relevance.a Overview of the different phosphorus (P) species monitored during dephosphorylation reaction of an organic phosphorus (Porg) reactant with the different Fe oxides; using adenosine triphosphate (ATP) as the Porg reactant, we monitored ATP in solution (Porg, reactant, aqueous), particulate Porg species (Porg, particulate), Porg products in solution (adenosine diphosphate, ADP, and adenosine monophosphate, AMP; collectively Porg products, aqueous), generated inorganic phosphorus (Pi) bound to the mineral (Pi, particulate), and generated Pi in solution (Pi, aqueous). b Apparent maximum rate (Vmax, µmol Pi h−1 gmineral−1) of Pi, aqueous (red) or Pi, particulate (light orange) generated during ATP dephosphorylation by goethite, hematite, and ferrihydrite; No Pi,aqueous was generated with ferrihydrite (N/A = Not Available). Box plots represent the lower and upper 95% confidence intervals and center value of the Vmax obtained from the model fit to the experimental kinetics data. c Apparent total turnover number (Total kcat, h−1) of goethite, hematite, and ferrihydrite; each kcat value was calculated by normalizing the sum of the Vmax values by the density of Pi binding sites (µmol Pi gmineral−1) for each mineral (gray). Box plots represent the lower and upper 95% confidence intervals and center value. d Rate of Fe oxide-contributed dephosphorylation (µmol Pi h−1 gsoil−1, in log10 scale) as a function of increasing Fe oxide fraction in soil, estimated from the combined apparent Vmax values shown in (b). In (a) aqueous Pi was determined by UV-vis absorption spectroscopy; particulate Pi was determined by P K-edge X-ray absorption near-edge structure spectroscopy. In (d) the dark gray line indicates the average for reported enzymatic dephosphorylation rates, 11.6 ± 0.8 µmol Pi h−1 gsoil−1; shown with the gray box is the full range of reported enzymatic rates in soils around the globe reported by Margalef et al., 201749. Source data are provided as a Source Data file.As a way of normalizing the Vmax value for each mineral, we performed Pi adsorption experiments to determine the site density for Pi binding on each mineral surface (SI, Fig. S17). While goethite exhibited nearly 3-fold higher capacity for Pi binding than hematite (26.0 ± 1.2 µmol Pi g−1 versus 9.1 ± 0.9 µmol Pi g−1, p < 0.001), ferrihydrite had the highest capacity for Pi binding (184.1 ± 6.0 µmol Pi g−1) at 7 and 20 times higher than for goethite (p < 0.001) and hematite (p < 0.001), respectively (SI, Fig. S17). By using the mineral-dependent Pi binding site density to normalize the combined Vmax for particulate Pi and dissolved Pi, we determined the total turnover number for Porg dephosphorylation (kcat) by each mineral (Fig. 5c). Further highlighting the higher catalytic efficiency of goethite compared to the other Fe oxides, we found that the total kcat for goethite was 2-fold to 4-fold higher than the kcat for hematite (p < 0.01) and up to 9-fold greater than the kcat for ferrihydrite (p < 0.001) (Fig. 5c). As we have already pointed out, mineral-dependent reactivity did not appear to be due to differences in surface area. Future research on the surface mechanisms underlying this abiotic catalysis will need to address the dependence of the observed difference in catalytic turnover on both the surface chemistry and mineral structure of each mineral.Here we evaluated the environmental relevance of our findings by comparing reported global values for phosphatase enzyme activity in soils51 to our Fe oxide-mediated dephosphorylation activity as a function of soil Fe oxide content estimated from our total Vmax values (Fig. 5d). Even at 2% fraction of Fe oxide content in soil, we found that the estimated rate of Fe oxide-catalyzed dephosphorylation (0.04–0.06 µmol Pi h−1 gsoil−1) was above the minimum rate reported for enzymes in soils (0.01 µmol Pi h−1 gsoil−1)51 (Fig. 5d). We further estimated that a soil Fe oxide content of ~20% or greater to be of significance for obtaining abiotic rates that would be approximately within one order of magnitude or less of the averaged value of soil phosphatase rates (11.6 ± 0.8 µmol Pi h−1 gsoil−1)51 reported in soils globally (Fig. 5d). In accordance with this estimation, we did observe the association of ribonucleotide-sourced Pi primarily with Fe oxides in our heterogeneous soil sample, which had adequate Fe oxide content (~18%) (Figs. 1d and 4g).
