The study area is located north of Puducherry city (formerly Pondicherry), around Auroville, on the Coromandel Coast in southeastern India (Fig. 1A). The climate of this region is classified as Tropical Savannah with a dry winter (Aw in the Köppen-Geiger classification), with an average temperature of ≈ 28.6 °C, ranging between 24.8–30.8 °C (January and May, respectively). The region receives an average rainfall of ≈ 1225 mm yr⁻¹, with the majority occurring during the northeast monsoon between October and December, and a smaller contribution for the southwest monsoon between July and August (Duraisamy Rajasekaran et al., 2024).

The characterisation of CaOx crystals was conducted through microscopy observations, while the quantification was performed using enzymatic oxalate kit analyses. Biominerals were imaged and characterised using a scanning electron microscope coupled with an energy-dispersive X-ray spectrometer (SEM-EDX; Tescan Mira LMU and Penta-FET 3x detector). Tweezer-selected plant material was loaded onto alloy stubs using adhesive stickers and AuPd sputter coated (Cressington 108 auto) for 1 min. Images were captured at 5 or 10 kV acceleration voltage with an average working distance of 12 mm. Quantitative EDX analyses were performed at 20 kV, and peaks were analysed using the Oxford Instruments Aztec software.

The CaOx content of soil or biomass were quantified colourimetrically using oxalate enzymatic assay kits (LIBIOS). CaOx was extracted according to Certini et al. (2000). Briefly, 0.1 g of ground plant material or 2 g of ground soil was combined with 5 mL 1 M HCl and shaken on a rotary table for 16 h at 150 rpm. The extractions were then centrifuged (1560 g for 5 min), separating the supernatant, before diluting (1:5), and neutralising the extracts (pH = 5–7). After this, the procedure followed the kits instructions, and the CaOx content was measured on a PerkinElmer UV/VIS Lambda 365 spectrophotometer at 590 nm. Measurements were made relative to anhydrous oxalic acid standards, which were measured at the beginning and end of each run. In addition, analytical replicates and blanks were measured to check for analytical drift, reproducibility, and background contamination, respectively. A molar correction factor was applied based on the molecular weight of CaOx monohydrate, which was the most observed crystal habit in the microscopy (142.11 g mol⁻¹), to convert extracted values to CaOx % oven dry weight (D.W.) of plant material.

Soil analyses results were corrected according to their hygroscopic moisture content. Concentrations or contents are thus reported based on oven dried soil weight (105 °C; van Reeuwijk, 2002). Blind replicates were measured in all sample analyses (> 20 %) to ensure the reliability of the results. See Sect. S3.2 for the reasoning regarding each specific method.

Soil pH_KCl was measured potentiometrically with a glass-body combination electrode (Thermo Scientific^™ Orion Star AIII) at a 1:2.5 ratio in 1 M KCl (Pansu and Gautheyrou, 2006). The extraction of exchangeable cations (Ca²⁺, Mg²⁺, Na⁺, K⁺, and Al³⁺) was conducted using 0.0166 M cobalt hexamine extraction (Aran et al., 2008). This method was selected because it displaces exchangeable cations while minimising dissolution of soluble phases such as CaCO₃ and CaOx. A 2 g portion of the sample was mixed with 40 mL of 0.0166 M cobalt hexamine solution and shaken on a rotary table at 120 rpm for 1 h. The mixture was then filtered (Whatman No. 42), diluted (1:25) in 2 % HNO₃, and measured on a Perkin Elmer Optima 8300 inductively coupled plasma-optical emission spectrometer with an internal Sc standard to correct data for analytical drift. The total element composition of ground and furnaced (Solo 111-13/10/30) samples, fused into lithium tetraborate pellets (PANalytical Perl X3 Fuser), was quantified using X-ray fluorescence (PANalytical Axiosm spectrometer). The spectrometer was equipped with a 4 kW Rh X-ray tube and calibrated with 21 international silicate rock reference materials. The results were corrected for loss of ignition. The bulk mineral composition of soil samples was analysed using X-ray diffraction (X-TRA Thermo-ARL Diffractometer) with the methods described in Adatte et al. (1996). The powder holders were prepared with pressed ground soil, and diffractograms were transformed using the Thermo Fischer WinXRD 2.0-6 program. The mineral peaks were then converted into percentages according to Adatte et al. (1996).

The total CaCO₃ equivalent content was determined using the acetic acid dissolution technique (Loeppert et al., 1984) and potentiometric analysis (Thermo Scientific^™ Orion Star AIII). This involved measuring the change in pH of the acetic acid extractants, relative to CaCO₃ standard extracts, which had been mixed with analytical-grade quartz and vortexed for 10 s (Rowley et al., 2020). A 2 g portion of ground soil was combined with 25 mL 0.4 M acetic acid and shaken at 250 rpm for 16 h on a rotary shaker. Soil organic C and total N contents were quantified using an elemental analyser (ThermoScientific FLASH 2000). Inorganic C was removed from the samples through fumigation techniques (Harris et al., 2001). An inhouse reference soil was measured throughout the runs to check for reproducibility (< 5 % relative standard deviation).

Analysis and visualisation of the data were conducted in R (RCoreTeam, 2024). Sequencing data were analysed and plotted for visualisation with the vegan package (Oksanen et al., 2024). To enhance clarity in the heatmap visualisation, rare taxa were filtered out by removing genera for which maximum relative abundance was < 1 %. Following this, hierarchical clustering of samples was performed using complete linkage (full dataset) or average linkage (filtered/abundant genera) and the Bray-Curtis distance, which generated a column dendrogram for genus organisation within the heatmap. Additionally, alpha and beta diversity analyses were conducted and are covered in the Methods S3.6 section. Non-metric Multidimensional Scaling (NMDS) analysis was then performed on taxon composition with k= 2 dimensions. To statistically test for significant differences in community composition between sample types, a permutational multivariate analysis of variance (PERMANOVA) was computed using the adonis2() function in the vegan package with 1000 permutations.

Soil and vegetation plots were generated using the ggplot2 package (Wickham, 2016). Alpha diversity metrics, plant, and soil variables were analysed for normality using the Shapiro-Wilk test, which indicated that the data did not follow a normal distribution. Consequently, non-parametric tests were applied throughout the study. Kruskal-Wallis tests were used to assess significant differences in median values among independent groups (soil sample type, organic matter type, species, and tree size categorized by DBH). For significant results, the Wilcoxon rank-sum tests (Mann–Whitney U tests) were used for pairwise comparisons to identify specific group differences. To account for multiple comparisons, p-values were adjusted using the Benjamini-Hochberg method, to control the false discovery rate. All statistical tests were performed with a significance threshold of α= 0.05, and mean values for statistically relevant groupings are presented with ±1 standard deviation.

Field observations of the studied individuals are summarised in Table S3. Biomass values calculated using the published allometric equations were skewed by variations in DBH. Consequently, conversion of contents to absolute quantities based upon these biomass values are only reported in the supplementary information. The highest allometrically-calculated biomass was for a L. tetraphylla individual (L4 = 3688 kg), while one S. emarginatus individual had the lowest biomass (Sa2 = 90 kg). The biomass of mature tree species differed in the following order: L. tetraphylla > S. emarginatus > A. heterophyllus > D. ebenum. All selected species exhibited effervescence in response to 10 % HCl, indicating the presence of CaCO₃ (Fig. S2). Particularly, strong reactions were observed on the trunks of S. emarginatus (Sa3) and L. tetraphylla (L4), but also in the rhizosphere of L4.

Calcium oxalate crystals of different forms were observed and identified in all organic matter types. Light microscopy performed directly at the field site led to the identification of CaOx in 21 out of the 25 tested TDEF species (Table S2), thus highlighting its widespread presence in the local TDEF ecosystem. CaOx monohydrate crystals (whewellite; CaC₂O₄ ⋅ H₂O) were observed in all samples of the four selected species as simple or twinned druses (Fig. 2A and B) and prismatic crystals (Fig. 2C to E), which were the most observed crystal habits. They also were identified in all the tissue types of the different species. Druse CaOx monohydrate crystals were predominantly identified in leaves and litter (Fig. 2A and B) but not in the bark of the tree species. CaOx dihydrate (weddellite; CaC₂O₄ ⋅ 2H₂O) was rarely detected in L. tetraphylla branches and A. heterophyllus bark samples. Opal and echinate sphere phytoliths were also occasionally observed in A. heterophyllus leaf samples. In accordance with field observations, micritic CaCO₃ was readily identified in S. emarginatus trunk samples (Fig. 2E) and L. tetraphylla trunk and root samples (Fig. 2F). Less common forms of CaCO₃, including needle-fibre and spheric CaCO₃, were observed in S. emarginatus trunk. Overall, biominerals, such as CaOx and CaCO₃, were identified in abundance throughout the different sampled organic matter types of the observed species (Figs. 2 and Fig. S3A and B).

The enzymatic oxalate kits revealed that, across all plant samples, the average CaOx content of the different organic matter types was 4.4 ± 3.2 % D.W. (Table S5). Although CaOx was measured in soils from the hollowed-out trunk of Sa3 (336 mg kg⁻¹) and D3 (40 mg kg⁻¹), CaOx was not detected in other soil samples. Among plant tissues, bark samples exhibited higher CaOx contents (7.83 ± 4.6 % D.W.), with contents decreasing sequentially as follows: bark > leaves ≥ branches ≥ litter ≥ roots (Fig. 3A). Significantly higher CaOx contents (3.42 ± 1.14 % D.W.; Wilcoxon test, p-value = 0.01) were observed in smaller trees (< 30 cm DBH) compared to larger individuals (> 30 cm DBH; 2.69 ± 1.06 % D.W.). At the species level, the bark of D. ebenum displayed a significantly higher CaOx content (13.4 ± 3.29 % D.W.) than other organic matter types from the same species. In A. heterophyllus, branches (5.82 ± 2.37 % D.W.) and bark (4.91 ± 1.92 % D.W.) also had higher CaOx content than other sample types from the same species (Fig. 3B, Table S5).

The absolute quantity of CaOx in each species, based on the allometrically-calculated biomass and the proportion of the organic matter types, followed the same order as the sampled trees' biomass: L. tetraphylla > A. heterophyllus > S. emarginatus > D. ebenum (Fig. S4 and Table S5).

From the initial 31 320 ASV counts obtained from LotuS2, 30 345 were assigned to the Bacteria kingdom, meaning that only 3 % were not assigned to a group. This supports the fact that the region used is prevalently found in this kingdom. In total, 86 % of these bacterial ASVs were assigned to Actinomycetota and 13.8 % to Pseudomonadota (Figs. S5 and S6). The most prevalent taxon in the dataset, irrespective of sample type, was an agglomerate of Streptosporangiaceae (uncultured), belonging to the Actinomycetota phylum, with approximately 50 % relative abundance (Fig. S7). Furthermore, the most frequent observed genera in the whole dataset were Saccharopolyspora, Nonomuraea, Planotetraspora, Planobispora, and Kutzneria (Table 1, Fig. S7).

Root and adjacent soil samples exhibited the lowest median alpha diversity values (Shannon and Simpson) and pairwise Wilcoxon rank-sum tests revealed that there was a significant difference only between root and litter samples (p-value = 0.02; Fig. S8A). Control soils displayed the highest median value in the Shannon diversity index, with a significant difference observed solely between root and control soil samples (p-value = 0.03; Fig. S8B). The Simpson diversity index showed relatively stable median values across sample types, and again, only root and control soil samples presented a significant difference (p-value = 0.03; Fig. S8C). The NMDS analysis of taxon lists revealed clear clustering of samples by tree species and sample types (Fig. S9) and the PERMANOVA analysis indicated that all pairwise comparisons of samples by types or tree species were significantly different (Table S6).

The frc gene sequence analysis is summarised in a heatmap (Fig. 4), which highlights the distribution of taxa as a function of sample type. In this analysis, an unclassified genus of the family “Streptosporangiaceae uncul.” was removed to enhance figure clarity as it consistently represented ≈ 50 % of the ASVs across different sample types (Table 1), thereby dominating the heatmap colour-scale. The analysis showed that litter samples clustered together, with different representative taxa for L. tetraphylla as compared to D. ebenum and S. emarginatus, supporting the NMDS analysis (Fig. S9). Litter samples from L. tetraphylla were associated to Micromonospora, uncultured Burkholderiales, Methylobrevis, and Tardiphaga. On the other hand, D. ebenum and S. emarginatus litter samples were associated to Xanthobacter, Hartmannibacter, Hansschlegelia, Methylorubrum, and Methylobacterium. Then, soil samples showed characteristic taxa, with Saccharopolyspora present at higher relative abundances compared to other sample types. Interestingly, control soil samples (except for one sample) were found to be uniquely characterized by a specific assemblage of putative oxalotrophs, the genera Afipia and Nitrospira. Finally, Pandoraea, Rhabdonatronobacter, and an uncultured Rhodocyclales were specifically associated with roots samples, while Methylobrevis was largely associated with roots, along with the litter of L. tetraphylla (Fig. 4).

Scanning electron microscopy enabled the imaging of microorganisms actively interacting with CaOx crystals in the organic matter of the different tree species. Characteristic morphological features, such as cell dimensions, filamentation, and branching patterns, were used to identify the potential groups of organisms that the micrographs illustrated. A clustered form (≈ 20 µm in diameter) of mycelia-forming Actinomycetota (hypha diameter ≈ 1 µm) was identified in branch and bark samples from A. heterophyllus (Fig. 5A). This microorganism was imaged degrading CaOx monohydrate crystals in the A. heterophyllus bark (Fig. 5B), coating themselves with a mixture of primary CaOx and secondary CaCO₃ crystals. Actinomycetota spores were also observed in association with CaOx monohydrate crystals in D. ebenum trunk samples (Fig. 5C). Finally, fungal hyphae (hypha width ≈ 2.5 µm supporting a fungal nature) were also observed in L. tetraphylla root sample (Fig. 5D). The possible presence of an anastomosis clamp indicates that, at least, the fungal hyphae in Fig. 5D likely belong to a Basidiomycota fungus. Overall micrographs supported the frc gene sequencing results, suggesting that Actinomycetota are an important phylum of oxalotrophic organisms in the sampled tree species.

Bulk soil properties (Table S7), total element (Table S8) and mineral contents (Table S9) are reported in the Supplementary Materials. There were small differences in individual soil chemical properties between the sites, but nothing that differentiated the sites based specifically on parent material (Figs. S10 and S11). Soil pH_KCl was slightly higher in the Cuddalore sandstone (5.4 ± 0.7) than in the Alluvium (4.8 ± 0.6), although the difference was not statistically significant (Fig. S12). Soil pH values were higher in the hollowed-out trunks (7.4 ± 0.6) than in the adjacent (5.2 ± 0.6) and the control (5.1 ± 0.7) soils (Fig. 6A), but the adjacent and control soils displayed no statistically significant difference. The effective cation exchange capacity was also higher in the hollowed-out trunks (19.8 ± 4.0 cmol_c kg⁻¹), compared to the adjacent (4.0 ± 3.1 cmol_c kg⁻¹) and the control (4.2 ± 4.9 cmol_c kg⁻¹) soils (Fig. 6B). Na_Exch remained below the detection limit in all samples. This difference in effective cation exchange capacity was mainly driven by a higher Ca_Exch content (Fig. 6D), but also by Mg_Exch (Fig. 6E), and K_Exch (Fig. 6F). All three of the major cations were higher in the trunk (CaExch= 13.5 ± 3.2 cmol_c kg⁻¹; MgExch= 3.9 ± 1.7 cmol_c kg⁻¹; KExch= 2.4 ± 2.1 cmol_c kg⁻¹) than in the adjacent (CaExch= 2.8 ± 2.4 cmol_c kg⁻¹; MgExch= 0.9 ± 0.7 cmol_c kg⁻¹; KExch= 0.2 ± 0.1 cmol_c kg⁻¹) or the control soils (CaExch= 3.1 ± 3.9 cmol_c kg⁻¹; MgExch= 0.9 ± 1.0 cmol_c kg⁻¹; KExch= 0.1 ± 0.1 cmol_c kg⁻¹). While the differences were not significant (p-value = 0.09), Al_Extr showed an opposite trend, decreasing in content moving from the control > adjacent > trunk soils (Fig. 6C). Overall, there were limited differences between the sampled adjacent and control soils, but there were distinct shifts in soil chemical properties of the hollowed-out trunk environment of the TDEF species (D. ebenum, L. tetraphylla, and S. emarginatus).

Soil organic C contents were higher in the trunks (42.8 ± 18.6 g kg⁻¹) and adjacent soils (8.0 ± 1.5 g kg⁻¹) than the control soils (4.9 ± 0.5 g kg⁻¹; Fig. 7A), while the total nitrogen contents were higher in the trunk soils (3.3 ± 1.3 g kg⁻¹), but similar between the adjacent (0.6 ± 0.1 g kg⁻¹) and control soils (0.4 ± < 0.0 g kg⁻¹; Fig. 7B). No trend was observed for C:N ratios (p-value = 0.3), moving from the hollowed-out trunk samples (12.3 ± 0.8) to the adjacent (13.7 ± 0.4) and control soil samples (12.8 ± 0.5; Fig. 7C). Meanwhile, the CaCO₃ contents were higher in the hollowed-out trunks (29.1 ± 11.3 g kg⁻¹) than in the adjacent and the control soils, which had a far lower CaCO₃ content that was near the limits of detection with this method and did not show a statistically significant difference (Fig. 7D). Thus, while there was an increase in soil CaCO₃ content within the hollowed-out CaCO₃-coated trunk environment of the trees, there was very little CaCO₃ in the adjacent surficial soil samples, which displayed no significant difference relative to the control soils.

The study identified four active and novel OCP ecosystems associated with three diagnostic TDEF species and one local agroforestry species, all of which produced significant quantities of CaOx. Although CaOx was not readily identified in the tree-adjacent soils, CaOx crystals were abundant across the sampled species in all analysed organic matter types, including in litter around the trees. No statistically significant differences were found between tree species, but smaller trees exhibited proportionally higher amounts of CaOx than larger individuals, consistent with other non-woody plants (Jones and Ford, 1972; Rahman and Kawamura, 2011), and Brosimum alicastrum Sw. (family Moraceae), another species with an active OCP (Álvarez-Rivera et al., 2021; Rowley et al., 2017). In young plants, CaOx is thought to protect developing cells from high cytoplasmic Ca²⁺ concentrations, which can hinder cell expansion and growth. However, CaOx crystals can also contribute to the defence of plants against herbivory (Franceschi and Nakata, 2005; Hudgins et al., 2003; Ward et al., 1997) and CaOx content was higher in external tissues, such as bark, which are more prone to encounter predation. Raphide (needle-like) CaOx crystals are typically associated with herbivore defence (Lawrie et al., 2023), and were not extensively observed in microscopy data, but prismatic crystals have also been reported to contribute to defensive functions in plants (Korth et al., 2006). As younger trees are more vulnerable to herbivory, we hypothesise that they likely invest more energy in producing CaOx as a defensive strategy (Ward et al., 1997; Wiggins et al., 2016), which would explain the higher CaOx contents observed in the biomass of smaller individuals in the TDEF. The widespread production of CaOx observed in the four studied tree species, as well as in a broader range of local species, highlights its significant, yet overlooked role in the Indian TDEF; however, its absence in the adjacent soils points to a disconnect between its aboveground production and belowground persistence.

The absence of CaOx in the adjacent soils could be explained by the oxalotrophic microbial community, which was characterized with HTS of the frc gene (coding for a Formyl-CoA:oxalate CoA-transferase; Khammar et al. 2009). Prior to discussing the environmental relevance of the results, key assumptions of using the frc gene as a marker of oxalotrophy should be acknowledged (Sonke and Trembath-Reichert, 2023). The first assumption is that the enzyme encoded by the frc gene, Formyl-CoA:oxalate CoA-transferase, establishes a direct metabolic link between formate, oxalate, and oxalotrophy. This enzyme plays a key role in the oxalate degradation pathway of many oxalotrophic bacteria (Anantharam et al., 1989; Ruan et al., 1992; Sahin, 2003), such as the obligate oxalotroph Oxalobacter formigenes (Ruan et al., 1992). Yet, there are currently six biological oxalate degradation pathways documented, and not all of them rely on Formyl-CoA transferase (Sahin, 2003; Schneider et al., 2012). An example of which, Methylobacterium extorquens (Schneider et al., 2012) can convert oxalate to glyoxylate using Oxalyl-CoA reductase. Conversely, not all organisms harbouring the frc gene are actual oxalotrophs as this gene appears to be associated to several other metabolisms (Sonke and Trembath-Reichert, 2023). As a result, an analysis based on the frc gene can only provide a partial image of the oxalotrophic guild.

Another key issue with investigating less commonly studied genes, such as the frc gene, in environmentally complex samples is the limited availability of reference databases for taxonomic annotation. For this reason, in this study, a database had to be created to enable accurate identification of the oxalotrophic guild. While phylogenetic placement offers a work around to this constraint (Czech et al., 2022), a significant limitation arises from the inherent dependence of the placement on the breadth of available reference sequences in databases. Taxonomic resolution hinges on the alignment's diversity and annotation quality, although it can be limited by uneven clade representation and the gene's sporadic evolution via horizontal transfer or phylogenetic restriction (Czech et al., 2022). Hence, due to these assumptions and limitations, HTS of the frc gene provides a partial and potentially biased snapshot of the oxalotrophic guild in the studied samples. Nevertheless, HTS of the frc gene still represents the best tool available for investigating oxalotrophic microbial communities in complex environmental samples.

Despite these limitations, the HTS analysis identified a range of well-known oxalotrophs (listed in Table S10), including the order Burkholderiales, which contains Oxalobacter formigenes, as well as non-obligate species belonging to the genera Burkholderia (Kost et al., 2014), Cupriavidus (Palmieri et al., 2022), Ralstonia (Sahin, 2003), and Pandoraea (Sahin et al., 2011). Other recognized oxalotrophic genera identified were Xanthobacter (Sahin et al., 2002) and Afipia (Bravo et al., 2015), the latter identified only in a control soil sample. The identification of oxalotrophs in control samples is unsurprising given that most oxalotrophs are not obligate and use CaOx as an alternative C source (Hervé et al., 2016). Thus, oxalotrophs highlighted in control soils, far from a CaOx source, may not be actively metabolising oxalate. The higher alpha-diversity indices in control soil samples could be linked to the lower selective pressure on bacterial communities located away from oxalogenic trees. This is further supported by beta-diversity analysis (PERMANOVA), which showed that all sample types and trees had significatively different oxalotrophic microbial communities. The detection of known oxalotrophs via HTS of the frc gene supports its utility in partially characterizing oxalotrophic microbial communities and highlights that communities in the TDEF are structured and closely linked to tree-derived oxalate inputs.

In addition to known oxalotrophs, the HTS also identified several taxa that have not previously been described as oxalotrophs. For instance, members of the Streptosporangiaceae family were detected, including the genera Nonomuraea, Planotetraspora, and Planobispora, a grouping of aerobic, Gram-positive, chemoorganotrophic Actinomycetota that form a branched, non-fragmenting mycelium (Otoguro et al., 2014). Likewise, genera within the Pseudonocardiaceae family not previously described as oxalotrophs include Kutzneria or Saccharopolyspora (Wei et al., 2023), the latter containing a species shown to be unable to decarboxylate oxalate (Meklat et al., 2014). Other taxa not previously associated with oxalotrophy included Methylobrevis, a methylotrophic genus (Poroshina et al., 2015), and Hansschlegelia, from the Methylocystaceae family known for methane oxidation via the serine pathway (Webb et al., 2014). Nitrospira, a chemolithoautotrophic nitrite-oxidising genus (Bayer et al., 2021) important for nitrification (Daims and Wagner, 2018) was also identified and has been previously shown to grow on formate as a C source, which may explain its presence in our dataset. Nitrogen-cycling taxa, including Hartmannibacter (Suarez et al., 2014) and uncultured Rhodocyclales (Ding et al., 2022; Liang et al., 2023) were also detected by the HTS analysis, but their oxalotrophic potential remains unclear. Ultimately, for all taxa not previously confirmed as oxalotrophs, their role in oxalate utilisation would require further investigation. Some assignments may represent false positives, i.e. taxa harbouring the frc gene but using it in another metabolic pathway (Sonke and Trembath-Reichert, 2023), while others may represent true oxalotrophs, highlighting that the diversity of oxalotrophic bacteria remains largely unexplored.

Overall, the HTS of the frc gene highlighted that the most prominent taxonomic phylum was Actinomycetota (86 %). This was congruent with separate SEM EDX observations in the bark of A. heterophyllus, which directly imaged mycelia-forming Actinomycetota associated with CaOx (Fig. 5A and B). Actinomycetota were identified in all samples and contain well-known oxalotrophic taxa (Bravo et al., 2013; Robertson and Meyers, 2022; Sonke and Trembath-Reichert, 2023), which, along with Pseudomonadota, are among the most prominent taxonomic groups involved in oxalate metabolism, both in the soil (Actinomycetota) or on plants (Pseudomonadota; Hervé et al., 2016). To conclude, oxalotrophs, predominated by Actinomycetota, were highlighted using HTS of the frc gene and were also imaged directly interacting with CaOx. This suggests that the absence of CaOx in the adjacent soils, despite its widespread production in the TDEF, is likely due to its rapid metabolism by an active oxalotrophic community associated with the investigated tree species in which Actinomycetota play a key role.

Although only three hollowed-out trunk soils were analysed, each presented clear biogeochemical evidence of an active OCP. Notable observations included a localised (relative to the control) increase in soil pH (+1.3 pH units), elevated Ca_Exch content (+8.2 cmol_c kg⁻¹), and precipitation of CaCO₃, both on the trees, and within the hollowed trunk soil (+23 g kg⁻¹ CaCO₃ equivalent). While we cannot rule out the possibility that termites played a role in the accumulation of CaOx-rich organic matter and precipitation of CaCO₃ in the trunks (Clarke et al., 2023; Nel et al., 2025b, a), tests on local termite nests and the trees associated with them did not identify the presence of termite-associated CaCO₃. Driven by an active OCP, Pons et al. (2018) reported a similar pH_H2O increase of +1.5 pH units at the foot of M. excelsa, compared to control soils located 30 m away. Yet, it has also been demonstrated by Cailleau et al. (2004) that M. excelsa is linked with far greater increases in CaCO₃ (+ 300 g kg⁻¹ CaCO₃ relative to control soils) than reported in this study. The difference in CaCO₃ content between these studies could partially be attributed to the larger biomass of M. excelsa, which is an emergent canopy tree (≈ 40 m height) that produces far greater absolute quantities of CaOx, compared to the smaller TDEF trees. These comparisons underscore how tree functional traits, particularly size, biomass, and CaOx production capacity can modulate the strength of OCP-driven changes in different forest systems.

Beyond classical measures of an active OCP, a decreasing trend in extractable Al³⁺ contents was seen, moving from the control soils to the hollowed-out trunk environments of trees. It is well established in soil science that more-soluble cations (Ca_Exch) and easily-weathered minerals (CaCO₃) typically decrease in content with pedogenesis, time, increased rate of weathering, and a higher or positive local water budget (balance between precipitation and evapotranspiration; Jenny, 1941; Doetterl et al., 2025). Meanwhile, less mobile ions and minerals, such as Al³⁺ and oxides remain because they are relatively more resistant to weathering (Adams et al., 2000; Bloom et al., 1979; Reuss et al., 1990). An example of this on a global scale, Slessarev et al. (2016) demonstrated that soil pH is clustered into two separate groups by water budget, governed by CaCO₃ equilibria in drier environments (pH ≈ 8.2) or gibbsite/Al³⁺ hydrolysis in wetter environments (pH ≈ 5.1). Thus, as pedogenesis progresses, soluble minerals and cations such as CaCO₃ are leached from humid soil environments, shifting the soil from a pH range governed by CaCO₃ equilibria to a system dominated by gibbsite/Al³⁺ hydrolysis, increasing Al³⁺ saturation (Chadwick and Chorover, 2001; Rowley et al., 2020; Slessarev et al., 2016). However, an active OCP can act to biogenically reverse this conventional model of pedogenesis on a local scale as the process drives alkalinisation and the precipitation of relatively soluble cations/minerals (Fig. 8C). The observed decline in extractable Al³⁺ from control soils to tree trunks may reflect this shift, indicating a transition away from Al³⁺-dominated systems towards Ca-rich, biologically influenced CaCO₃ equilibria, consistent with localised OCP activity. Yet, the absence of large differences between the biogeochemical properties of adjacent and control soils still requires explanation. In high-rainfall regions, pH alone may not capture subtle buffering effects in surface soils. Future studies could include measures of reserve acidity, which would provide insight into pH buffering dynamics in adjacent versus control soils.

While CaCO₃ precipitation was abundant both aboveground and within the hollowed-out trunk environments of the TDEF species, there were no statistically significant difference between the adjacent and control soils. This is likely due in part to the fact that only surficial soil layers (0–10 cm) were sampled at random locations, both adjacent to, and at a control distance from the 16 trees. A more extensive sampling campaign was beyond the scope of this initial study, partly because the TDEF forests are considered sacred, despite the absence of formal legal protection. However, incremental sampling at multiple depths, including full soil profiles, focusing on the base of the larger trees, and sampling with distance from the rhizosphere, would likely have yielded starker differences in the adjacent and control soil biogeochemical properties. Indeed, significant CaCO₃ precipitation was found associated with the rhizosphere of L. tetraphylla during sampling, supporting this conclusion. Consequently, a more detailed analysis of the change in CaCO₃ with depth would be required to ascertain the exact C budget of the active TDEF OCP systems and the absence of differences between adjacent and control soils could partially be linked to this study's sampling strategy.

However, these results are consistent with those reported by Hervé et al. (2018), who investigated an active OCP associated with T. bellirica in Madhya Pradesh, Central India. The authors found that while CaCO₃ accumulated up to 82 % D.W. in bark, there was relatively little CaCO₃ in the adjacent soils (+10 g kg⁻¹ CaCO₃ relative to control soils). The authors hypothesised that this discrepancy between above- and belowground CaCO₃ was explained by monsoon precipitation (1100 mm, approximately 75 % of annual precipitation in 4 months), which dissolved and leached OCP-precipitated CaCO₃ from the adjacent soils. The present study lends further support to this idea, as while the area is subject to two separate seasons, with dry (inter-monsoons) and wet (monsoons) periods, conditions were slightly moister in Tamil Nadu than in Madhya Pradesh (Duraisamy Rajasekaran et al., 2024; Hervé et al., 2018; Saikranthi et al., 2024). This hypothesis is further substantiated by the fact that there was undetectable level of Na_Exch in the TDEF soils, a soluble monovalent cation with a large hydration sphere, which was likely leached from the system during the rainy periods. Consequently, the low CaCO₃ content of soils adjacent to the oxalogenic TDEF trees was likely a result of the region's humid monsoon conditions. This further suggests that a critical water budget threshold may determine whether an active OCP leads to sustained CaCO₃ accumulation (Slessarev et al., 2016). However, this threshold is unlikely to depend solely on local water budget but could also be shaped by the intensity and frequency of rainfall events, their interplay with temperature and evapotranspiration, and the resultant impacts on CaCO₃ weathering, dissolution, leaching, and reprecipitation dynamics. In addition, CaCO₃ precipitation, inorganic carbon stocks, and bicarbonate export both above and belowground deserves further investigation in any monitoring, reporting, and verification of OCP-systems.

The extent to which OCP systems in the TDEF sequester C will depend on complex interactions between local and regional geochemical reactions (Fig. 8D and E), including (i) the initial source of Ca utilized to sequester CaCO₃ biogenically (Fig. 8A and D), (ii) whether it is carbonic acid that weathers OCP-precipitated CaCO₃, and (iii) the physicochemical or biogenic processes that determine whether leached HCO3- remains in solution or precipitates as a secondary carbonate phases. While local CaCO₃ stocks may decrease due to monsoon leaching, the broader-scale sequestration potential remains significant if the inorganic C remains in solution and is exported from the system, instead of precipitating in soils as secondary CaCO₃ phases (Eq. S2). Percolating soil solutions enriched with dissolved HCO3-, CO32- and Ca_Exch will eventually flow out of the TDEF system into the Bay of Bengal, and could represent an important inorganic C sink (Beerling et al., 2020; Taylor et al., 2021; Zamanian et al., 2016). Nevertheless, key uncertainties remain, including the mechanisms that govern Ca and C fluxes and their sensitivity to climatic variables. To address these, isotopic labelling experiments could now be used to precisely trace the cycle of C from CaOx through CaCO₃ precipitation, and then its leaching under controlled conditions. This study suggests that the OCP might be globally distributed, potentially existing in more humid environments than previously recognized, and represents an understudied soil inorganic C sequestration process.