Changes in soil bacterial community structure in a short-term trial with different silicate rock powders

Background The use of rock powders in soil has emerged as a nature-based technology to improve soil properties relevant to crop development and for atmospheric carbon dioxide removal (CDR) via enhanced rock weathering (ERW). Although modeling this process is crucial, the soil microbiome has been identified as the main reason why several experimental and field results do not fit the geochemical and kinetic theoretical models. Here, the hypothesis that the bacterial community structure is modulated by the application of different silicate rock powders was tested. One phonolite, three basalt variations and one granite, as well as KCl treatments, were applied to a Ferralsol cultivated with Brachiaria in short-term pedogeochemical experiments and assessed after 1 (1M), 4 (4M) and 8 (8M) months. Results The main changes in soil bacterial structure were observed at 8M and found to be modulated according to rock type, with petrochemistry and mineralogy acting as the main drivers. The content of microbial biomass carbon tended to decrease over time in the Control and KCl treatments, especially at 4M, while the rock treatments showed constant behavior. The sampling time and treatment affected the richness and diversity indices. The Si, Ca and Fe from mafic minerals were the main chemical elements related to the soil bacterial changes at 8M. Conclusions The type (acidity) of silicate rock powder modulated the soil bacterial community (SBC) in a pot experiment with tropical soil. The specificity of the SBC for each rock type increased with time until the end of the experiment at 8 months (8M). The carbon content in the microbial biomass was lower in the rock powder treatments in the first month (1 M) than in the control and KCl treatments and was equal to or higher than that in the 8 M treatment. This result illustrates the challenge of modeling rock powder dissolution in soil since the soil medium is not inert but changes concurrently with the dissolution of the rock.


Background
Remineralization is an agricultural practice based on applying milled/crushed rock particles to the soil surface or at small depths to increase soil fertility and plant development [1].The mechanisms by which plant growth improves when silicate rocks are applied to soil have not yet been completely elucidated.However, substantial data on the benefits of rock powder application in tropical agriculture have been obtained, especially for carbonate (limestone) and sulfate rocks (gypsum) [2,3].Despite the complexity of soil chemical dynamics in weathered soils under tropical or subtropical conditions, the agricultural use of silicate rock powders as soil amendments is strongly recommended.However, prior to its use, it is important to consider that the results are dependent upon several factors, such as the rock characteristics (e.g., chemical and mineralogical composition, particle size distribution), soil type, soil organisms, and plant species [4], because silicate anions tend to form less soluble solid phases than carbonate and sulfate anions.
Thus, it is reasonable that remineralization will take direct or indirect advantage of microbial activity, such as siderophore production, biofilm formation, enzyme activity, and various types of redox reactions [5][6][7].Indeed, considering soil formation, biological activities (i.e., living organisms) accelerate the transfer of chemical elements from the lithosphere to the biosphere [8] and are one of the five factors involved in soil formation.
Rock powders are coproducts of several mining activities, such as aggregate production for construction and dimensional rocks.The powder generated by processing these rocks accumulates in the mining backyard in most cases, potentially causing problems in the management of the mining workflow and even environmental concerns, such as the source of air suspended and breathable particles [9].
The Brazilian territory is vast and is occupied by 63 M ha of cropland and 172 M ha of pastureland [10]; therefore, even the partial use of remineralizers as a soil amendment could have a substantial impact on saving natural resources and atmospheric carbon dioxide removal (CDR) [11].Furthermore, the adsorption of contaminant effluent, a reduction in nitrogen emissions, and an increase in plant resistance to abiotic stress [4,12] are concurrent benefits of rock powder use.However, it is important to think about the life cycle assessment of the use of rock powders beyond the agronomical benefits, considering the distance between the quarry and the place of application since these mining coproducts should not be transported too far [13].
The use of registered rock powder (remineralizer) has been regulated by Federal Laws in Brazil since 2013 [3].Therefore, the correct use of rock powders as a soil amendment for agriculture has tremendous potential to boost the circular economy on a national scale.A major obstacle in the use of remineralizers is the lack of causal models to determine the outcomes of their use in agricultural and ecological scenarios [14,15].To help to fill this gap, it is paramount to identify how their use affects soil biological functioning, highlighting their impact on soil microbial community structure and soil microbiological attributes.To date, little is known about how bacterial community structures respond when different silicate rock powders are applied to soil, especially their response over time in short-term trials.
Therefore, in this investigation, the hypothesis that the soil bacterial community (SBC) structure responds differently to the type of silicate rocks applied to the soil in short-term experiments based on the mineralogical and chemical composition of these rocks was tested.
The soil and treatments were thoroughly mixed inside plastic bags before being dispensed into 3 L plastic pots.Five igneous rocks were chosen according to their Si content (acidic, intermediate and basic), mineralogy and petrochemistry and due to their K-bearing mineral content.The rocks were comminuted into a jaw crusher and then into a motorized mortar and pestle mill (Marconi, model MA890).Rock powders were prepared to have similar particle size distributions by combining 50% (m/m) very fine sand (VFS, from 0.10 mm to 0.05 mm), 25% medium sand (MS, from 0.5 mm to 0.25 mm) and 25% very coarse sand (VCS, from 2 to 1 mm).These fractions were chosen to maximize the ratio between the mass of the rock particles and the mass of the soil particles.Seven treatments were used: Control (C), KCl, phonolite (Ph), basalt 1 (B1), basalt 2 (B2), basalt 3 (B3), and granite (Gr).All basalts were collected from sites on the Serra Geral Formation in São Paulo State, where they are usually tholeiitic and contain small amounts of 2:1 phyllosilicate [20].Phonolite and granite were collected from Minas Gerais State.
All treatments, except C and KCl, were amended with a single dose of 21 Mg ha −1 rock powder (considering the 0-20 cm layer and a soil density of 1.3 Mg m −3 ) at the beginning of the experiment.KCl (0.3 Mg ha −1 ; reagent grade, Merck) was applied as recommended by Novais et al. [21] for Brazilian soils as a function of the exchangeable K levels determined by soil fertility analysis.Treatment C followed the same procedure but without the addition of any amendments.
The experiment followed a randomized block design with 4 replicates, resulting in 28 experimental units (EUs).Because soil sampling was destructive, it was necessary to make three copies of the 28 EUs totaling 84 EUs so that one set of 28 EUs could be disassembled at each of the three time points reported here.We named each of these copies a replica, and they were spatially randomized among the other replicas.Therefore, one replica was disassembled after 1 month (1M replica), the second replica was disassembled after 4 months (4M), and the last replica was disassembled after 8 months (8M).After disassembling each replicate, aliquots of the soils were immediately stored at − 80 °C for further molecular analysis or at 4 °C for microbial biomass carbon and enzyme activity analyses.
The pots were sown with approximately 15 seeds of the tropical grass Urochloa brizantha (Hochst.Ex A. Rich.)STAPF.(syn.Brachiaria brizantha cv.Marandu), and after the first month, the most vigorous plant was selected, and the others were removed.The pots were maintained with 20% moisture on a dry mass basis [22].Once a month, an extra 1 L of water was added, creating a leaching solution that was collected at the bottom of the pots.This procedure allowed us to evaluate the potential leaching of cations from the pots.When the plants reached approximately 40 cm in height, they were cut 15 cm above the soil surface to permit regrowth [23].There was a total of 2 cuts during the 8 months of the experiment.

Chemistry and mineralogy of the soil and rocks
Soil chemical analyses were performed according to the standard methods for Sao Paulo State [24].Briefly, soilexchangeable bases (Ca, Mg, K) and P were extracted using the resin method.A 1 mol L −1 KCl solution was used to extract Al, while Ca phosphate and hot water were used to extract S and B, respectively.Cu, Fe, Mn and Zn were extracted using diethylenetriaminepentaacetic acid (DTPA).Na and Si were extracted using Mehlich 1 and 0.01 mol L −1 CaCl 2 (Additional file 1: Table S1).
Mineralogical analyses of the soil and the rocks were carried out in a Rigaku Miniflex II XRD benchtop system using CuKα radiation at 30 kV and 15 mA coupled to a graphite monochromator and a spinning sample holder.The samples were irradiated from 3 to 60° 2θ at 0.01° 2θ s −1 .X-ray diffraction (XRD) data were processed with "Match!" software (Cristal Impact, [25]), and minerals were identified using the COD reference database (Crystallographic Open Database) [26] and the mineralogical tables in Chen [27] and Brindley and Brown [28].
For total chemical analysis, soil and rock samples were air dried, homogenized and 10 g milled to a particle size of < 200 μm.Subsamples (0.25 g for rock, 0.1 g for soil) were subjected to two different analyses.One aliquot was fused with lithium metaborate (LiBO 2 ) in a muffle furnace and dissolved in an acid solution (C 4 H 6 O 6 and HNO 3 ).The elemental contents were determined by optical emission spectrometry (ICP-OS).Another aliquot was analyzed after multi-acid digestion (HCl, HNO 3 , HF and HClO 4 ), and the elemental content was determined by inductively coupled plasma-mass spectrometry (ICP-MS).This analysis was performed to quantify the total elements in each rock powder, linking them to the minerals identified in the mineralogical analysis (XRD) and the dissolved elements in the pot leachate.
Petrographic slides were made by selecting a particular area on each rock sample.Each sample was impregnated with industrial Araldite, and the resulting block was sawed and glued to a glass slide and then brought to a thickness of 0.03 mm (30 µm).The slides were observed under a Nikon Eclipse 50iPol petrographic microscope with transmitted light and plane-polarized with image capture to be described.This analysis was performed to characterize the rocks in terms of the types of minerals present (to support the XRD data) and to observe the shape and size of the rock minerals, which also differed among the three basalts tested.

Microbial biomass carbon (MBC) and enzyme activities
MBC was extracted from the soil using the fumigationextraction method and the content was determined by titration according to Vance et al. [29].Briefly, 10 g of soil was adjusted to have 40% of its maximum water holding capacity and then placed in an evacuated desiccator.For each soil sample, both ethanol-free chloroform-fumigated and nonfumigated samples were left in the dark for 24 h.Then, the soils were extracted with 0.5M K 2 SO 4 solution (4:1, v/w) to determine the carbon content by titration with 0.033M (NH 4 ) 2 Fe(SO 4 ) 2 (H 2 O) 6 (Mohr's salt).
Acid phosphatase (EC 3.1.3.2),β-glucosidase (EC 3.2.1.21),and arylsulfatase (EC 3.1.6.1)activities were evaluated due to their direct roles in phosphorus, carbon, and sulfur cycling in the soil, respectively, and were determined following the methodology proposed by Tabatabai [30].Briefly, for acid phosphatase, 1 g of soil was incubated with modified universal buffer (MUB) at pH 6.5 along with the p-nitrophenyl-phosphate (PNF) substrate at 37 °C for 1 h, and extraction solution composed of 0.5M CaCl 2 and 0.5M NaOH was added.β-Glucosidase assays were performed with MUB at pH 6 with p-nitrophenyl-β-d-glucoside (PNG) as the substrate, and arylsulfatase assays were performed with 0.5M sodium acetate buffer at pH 5.8 with p-nitrophenyl sulfate (PNS) as the substrate.In each case, the samples were incubated, the extraction solutions were added, the samples were filtered through a No. 2 Whatman filter, and the color intensities were measured at 410 nm using a spectrophotometer (EZ Read 400, Biochrom).Enzyme activities were determined based on a standard curve developed with a ρ-nitrophenol solution.

DNA extraction and SBC structure assessment
Total DNA was extracted from 0.4 g of each of the 84 soil samples using a commercial kit (PowerSoil DNA Isolation, MoBio, Carlsbad, CA, USA) following the manufacturer's instructions.The integrity of the DNA was verified using gel electrophoresis (1% agarose) prior to performing PCR assays, and the quantity of DNA was measured using a Qubit fluorimeter (Invitrogen, Carlsbad, CA).
Aliquots of PCR amplicons (approximately 200 ng) were digested with the restriction enzyme Hhal (10 U/µl) (Thermo Scientific) at 37 °C for 3 h.The digested material was then precipitated with 2 µl of 3M Na-acetate, 2 µl of 125 mM EDTA and 50 µl of absolute ethanol and centrifuged at 4000 rpm for 30 min.The precipitated DNA was washed with 70% ethanol and dried by centrifugation.The DNA was suspended in formamide Hi-DiTM (Applied Biosciences, Foster City, CA) and analyzed on an ABI Prism 3500 automatic sequencer (Applied Biosystems, Life Technologies).A threshold of 50 units of fluorescence was adopted to remove the background of the samples, and the results were transformed into a relative abundance matrix of the peak areas.
The T-RFLP technique provided us with a matrix based on the number and peak heights or areas of the terminal restriction fragments (TRFs), which were separated with an automated DNA analyzer [34].As with all techniques in molecular biology, when using the fingerprinting technique, there are several bottlenecks regarding full-scale application for some ecological approaches, such as the evaluation of functions and species composition [35].However, the use of T-RFLP is not obsolete and is still highly valuable for studying microbial community dynamics [36][37][38][39].Over the past decade, highthroughput sequencing has transformed soil ecology studies by enhancing our ability to discern taxonomic identities, particularly when designated sequence data are accessible.However, as demonstrated by Lindström et al. [40], both T-RFLP and high-throughput sequencing methodologies effectively capture similar spatiotemporal variations, providing comparable insights into the functional and taxonomical details of the microbial communities in the soil.

Leachate and plant tissue analyses
Leaching solutions were collected, and the volume, pH, electrical conductivity (EC) and temperature were immediately measured.Then, the pH of each leachate was adjusted to 3 with 3% nitric acid (added dropwise) to prevent precipitation and microbial growth, and the leachate was stored in 15 mL plastic tubes at 5 °C until ICP-OES analysis [41].The aboveground plant tissue was collected and dried at 60 °C, the biomass weight was measured on a dry basis, and aliquots (0.25 g) were put into Teflon vessels and subjected to microwave acid digestion in a Mars Xpress (CEM) with 2 mL of hydrogen peroxide and 4 mL of nitric acid.The heating gradient was as follows: to 80 °C over 3 min, 150 °C over 10 min, 180 °C over 10 min and finally 180 °C over 5 min.The extract volume was adjusted to 35 mL with ultrapure water before analysis by ICP-OES [42] to quantify the absorbed elements (this paper presents the K content only).

Data analyses
One-way ANOVA was used to evaluate the K, Ca and Na concentrations in the leachate and plant tissue to verify their significance after the assumptions for ANOVA (homogeneity and normality) were met.Wald tests were used for the soil microbiological attributes (microbial biomass carbon and enzyme activities) in R software (version 3.6.3)(R Core Team 2016).
Changes in the SBC structure were evaluated based on the Bray-Curtis distances of the relative abundance of the peak area matrix obtained from the T-RFLP.The Shannon-Weaver and richness indices of the TRFs were calculated according to Zhang et al. [43].
Although ecological diversity indices are frequently applied to fingerprinting methods, multiple taxa can generate the same TRF, and rare TRFs can be excluded by a relative abundance threshold [44].To determine the differences in the SBC structures between different treatments amended with silicate rocks, we performed nonmetric multidimensional scaling (NMDS) coupled with PERMANOVA (ANOSIM function in R, permutations = 999) considering the relative abundance of the peak area matrix transformed into log (x + 1) [45].We carried out global redundancy analysis (RDA) coupled with a forward selection function to verify the most representative mineralogical properties (see Additional file 2: Table S2).Diversity, richness, and PERMANOVA were performed using the vegan package in R software [46].
A structural equation model (SEM) was adopted to explore how the most representative mineralogical rock attributes obtained from the NMDS (r 2 > 0.75; Table 2D) influenced the SBC structure at the most impacted sampling time (8M).For this purpose, the C and KCl treatments were not considered because they were not amended with silicate rocks.First, we used NMDS to determine the first axis scores, which were used as indicators of bacterial community structure (composite variable) [47][48][49].We used a minimum set of parameters to assess the model fit, including the comparative fit index (CFI), root mean square error of approximation (RMSEA), and Tucker-Lewis index (TLI), using the benchmark values according to Fan et al. ( 2016).The modeling process was performed using the lavaan [50] and semPlot [51] packages.In our SEM, we plotted the std.all, which determines the standardized coefficient for the path.As a result, coefficients above 1 can be observed.On the other hand, when employing the std.lv, which standardizes to the latent factors, coefficients typically range from -1 to 1.The original data and scripts used are provided in the Additional file.

Mineralogy of the soil and rocks
XRD of the sand fraction of the soil showed only quartz, and no easily bioweatherable primary minerals were identified.Kaolinite and gibbsite were the major minerals in the clay fraction.Therefore, the degree of desilication of the soil was between that of monosialitization and that of allitization (Fig. 1A).
We assign, based on the XRD, total chemical and petrography analyses, that the phonolite had mica (Mca), alkali feldspar (sanidine-Sa), feldspathoid (nepheline-Nph), sodium pyroxene (Aeg-aegirine), eudialyte (Eud) and zeolite (natrolite-Ntr), which are K-and Na-rich minerals (Fig. 1C), which contributed to the high contents of K 2 O and Na 2 O in the chemical composition of this rock (Table 1).Mineralogy of the basalts was similar, with dominant plagioclase (Pl) and pyroxene (Pxaugite) (Fig. 1C).However, B1 has more magnetite (Mag) and pyroxene (Px), while B2 and B3 have more olivine (Ol) and zeolite (Hul-heulandite), respectively.The structures and textures of the basalts are diverse.B1 is microcrystalline, while B2 has a coarser texture, and B3 has cavities and amygdales filled with zeolite (Hul-heulandite) (Fig. 2D).Zeolites are very reactive silicates that have a large surface area and cation exchange capacity (CEC).The variations in chemistry (Table 1), mineralogy (Fig. 1B) and structure (Fig. 2) of these basalts were used to test the sensitivity of the SBC structure to small variations in the characteristics in a single rock group.Gr was the only acid rock used (SiO 2 greater than 52%, i.e., Si greater than 24.3%) [52] (Table 1), which explains the abundant quartz (Qz) peak in the XRD pattern and the abundances of plagioclase (Pl) and mica (Mca) (Fig. 1B).

Soil microbiological attributes
The MBC in treatments C and KCl decreased over the evaluation time (p < 0.05), while for treatments with rock powders, with the exception of Gr, there was a decrease from 1 to 4 M, followed by an increase to 8M, restoring the levels to the same values as those in the first evaluation (1M) (Fig. 3A).At 1M, the MBC in the C and KCl treatments was greater than that in the other treatments, and the MBC in the Ph treatment was intermediate between that in the C and KCl treatments and that in the other silicate rocks.At 4M, the MBC was highest in the KCl treatment group and lower in the other groups.However, at 8M, the opposite trend was observed, and KCl had the lowest MBC while B2 had the highest MBC (Fig. 3A).
The enzyme activities varied according to treatment and evaluation time, but a clear pattern among them was not identified.
The activity of β-glucosidase decreased in treatments C and B1 (p < 0.05) over the evaluated periods.The activity of this enzyme at 1M outperformed that in the other times in treatment C, as opposed to that in the Gr treatment.The activity of β-glucosidase in the Gr treatment group increased at 4M but decreased again at 8M (p < 0.05) (Fig. 3B).There was no significant difference among the treatments at 4M.However, at 8M, the KCl and Gr treatments presented the highest and lowest activity, respectively (p < 0.05) (Fig. 3B).
The arylsulfatase activity at 1M was the greatest for treatments B3 and Gr, followed by B2, B1 and Ph, and significantly lower activity was observed for treatments C and KCl (Fig. 3C).The treatments did not differ at 4M.At 8M, Ph had the highest activity of arylsulfatase, and B3 had the lowest.Over time, the arylsulfatase activity continuously decreased in treatments B2, B3 and Gr until the  end of the experiment at 8M, while that in treatment C increased at 4M.The acid phosphatase activity in the KCl treatment group was the highest among all groups at 1M and the lowest in C and B3 at this time (p < 0.05) (Fig. 3D).At 4 M, there was a general increase in acid phosphatase activity in C, Ph, B1 and B3 (see small capital letters in Fig. 3D) compared to those at 1M.In particular, C had the greatest increase in acid phosphatase activity, while B2 had the lowest when comparing all the treatments at 4M.At 8M, the KCl treatment showed the highest acid phosphatase activity, and B1 had the lowest activity.The activity of this enzyme in the basalts consistently lower during the experiment.Despite its presence in a small amount (maximum of 0.61% P 2 O 5 , Table 1), this seemed to be sufficient to impact the phosphorus metabolism of the bacteria.In addition, we observed The richness index of the bacterial community fluctuated (p < 0.05) over time for the KCl and B1 treatments only (see the small capital letter results in Fig. 3E).At 1M, KCl showed the lowest richness and Shannon indices, which increased at the following evaluation times (4M and 8M) along with the exhaustion of this source.In B1, there was a drastic decrease in these indices at 4M, which recovered at 8M.At each time point, the lowest richness index was observed in KCl, while the highest was observed in Gr at 1M.At 4M, the lowest richness index was observed in B1, whereas at 8M, there were no differences between the treatments and the control (Fig. 3E).
In terms of the Shannon index, at 1M, the only difference was the lowest value in the KCl treatment group (p < 0.05), while at 4M, the value in the KCl group was the highest, and that in the B1 group was the lowest.Finally, at 8M, there were no differences among the treatments (Fig. 3F).The Shannon index was the only microbial index evaluated that did not vary throughout the experiment (lack of small capital letters in Fig. 3F).
NMDS analysis revealed that the bacterial community structure changed over time.The greatest variation was observed with KCl treatment at 1M (Fig. 4A), which lacked an adjustment along with the other treatments at 1M.Thus, the KCl treatment results are shown in Fig. 4B, and there was no significant difference among the other treatments (Table 2).
After four months (4M) (Fig. 4C), the initially strong effect of KCl diminished, but the conditions did not completely return to those of treatment C.However, KCl treatment also did not differ significantly from Ph, B3 or Gr.On the other hand, the effects of some of the rock powders started to appear, since the Ph, B1 and B3 treatments clustered (p < 0.05) (Table 2).
At 8M, three clusters were identified: one formed by C, KCl and Ph in the left quadrant; one with all the basalt Gr treatment (Fig. 4D).ANOSIM confirmed the similarity of the bacterial community structure among basalttreated soils B2 and B3; however, B1 differed (p < 0.05).Moreover, no significant difference was found between Ph and B3 (Table 2).

Elements in the leachate and plant tissue
Only KCl significantly differed (p < 0.001) from the other treatments in terms of K leached (mg), and the K content in the plant tissue was similar among all the treatments (Table 3).At 8M, the total amount of K leached from the pots after KCl treatment was approximately 30 times greater than the amount leached from Ph treatment,  which is the rock with the greatest concentration of K (Table 3).In addition, the greater K concentration in the KCl solution induced greater Ca leaching, mainly at 1M.

Discussion
In this paper, we tested the sensitivity of the SBC and soil bioindicators to the presence of igneous rock powders (basic, intermediate, and acidic) during an 8-month experimental period.After one month (1M), KCl was the only treatment that caused a change in the SBC structure, as indicated by NMDS analysis (Fig. 4A).Such a drastic change in SBC structure is known to occur as a result of the use of highly soluble fertilizers [53][54][55][56].This change was also observed in the richness and Shannon indices, which were the lowest at 1M in the KCl group (Fig. 3E and F), revealing its impact on the SBC.A similar reduction in richness was observed by Ben Zineb et al. [57] for a treatment with super triple phosphate as opposed to the increase in richness with phosphate rock treatment according to the plant and dose used.This contrasted with the application of silicate rock powders.
In the experiment presented here, this change was possibly caused by the large amount of K and Cl ions solubilized in the system.The concentration of K from a single leaching event at 1M ranged from 0.0 mg L −1 to 31.79 mg L −1 in all but the KCl treatment, which ranged from 12.34 mg L −1 to 1022.00 mg L −1 (data not shown).Therefore, despite K being a macronutrient [58], the K concentration in KCl reached more than 15 times the optimal concentration for bacterial growth, which is approximately 60 mg L −1 [59].
K and Cl ions have similar hydration enthalpies (K + = − 321 J K −1 mol −1 ; Cl − = − 363 J K −1 mol −1 [60]), which leads to the formation of large hydration shells and a more negative soil water potential, decreasing water flow into the cell [61].Microbial cells alter their metabolism to compensate for fluctuations in the osmotic potential of the surrounding solution [62].Therefore, the great change in the SBC in KCl seemed to be triggered by changes in the nutrient and water status of the surrounding environment of the bacteria.
The extremely high concentration of K + in the soil solution in KCl at 1M also induced an increase in the loss of Ca 2+ via the leachate (Table 3).Variations in the availability of Ca also impact the SBC [63].
The large amounts of Na in the Ph (6.74% Na 2 O and 4.76% Na, Table 1) and Gr (4.68% Na 2 O and 3.4% Na, Table 1) rocks were not released fast enough to increase the Na concentration in solution and significantly change the SBC structure (Table 3).Similar to K ions, Na ions have a large hydration shell and very negative hydration enthalpy (− 405 J K −1 mol −1 ).
The covalent character of the Si-O bond implies that silicate rock powders will have a lower solubility and greater buffering capacity than the K-Cl bond with ionic character, which results in greater solubility [64][65][66].This can be shown by comparing the release of K from KCl and Ph.The chemical composition of KCl was 41.5% K (60% K 2 O, m/m), while that of phonolite was 5.45% K (8.0%K 2 O, m/m).The KCl/Ph K content ratio was 7.6 in the solid phase, while in the leaching solution, this ratio was 41.88 at 1M (Table 3).
The high solubility of KCl led to its quick exhaustion, causing the SBC to cluster to the Ph, B3 and Gr groups at 4M (p < 0.05) (Fig. 4C, Table 2), which is interesting because these rocks have higher K concentrations (Ph and Gr) and higher reactivity (Ph and B3) than the other rocks.As leaching continued from 1 to 8M, KCl tended to cluster with the other treatments (Fig. 4D, Table 2).However, at 4M and 8M, while the SBC in the pots treated with rock powders moved farther from C, KCl reverted.This was also observed by Cuhel et al. [67] in a long-term field trial.At the end of our experiment (8M), the richness and Shannon indices did not differ between treatments (Fig. 3E and F), which is in contrast to the results of Mickan et al. [68], in which the effects of soluble fertilizer and rock powder (75 kg ha −1 mix rocks and minerals with ammonia sulfate and potassium sulfate) reduced the microbial diversity after 10 months.However, we need to consider the differences in the dose and composition of the rock powder between the present work and that of Mickan et al. [68].
Concurrently, the amounts of K + and Ca 2+ in the leachate of KCl also decreased due to exhaustion (Table 3), demonstrating the strong influence of soil solution characteristics (such as ionic activity and osmotic potential) on SBC dynamics.
The SBC discriminated the major rock types at 8M: the basalts (B1, B2 and B3) clustered in the right upper quadrant of the NMDS (Fig. 4D) and Ph, C and KCl clustered in the left quadrant, while the Gr was in the lower quadrant.In fact, the SEM confirmed that the Si content in the rocks significantly positively affected the bacterial community (Fig. 5), although Fe and Ca were also positively correlated.
Beyond the Si content, the SBC was also sensitive to small variations in rock characteristics (texture, mineralogy and chemistry-Table 1, Fig. 1C and Fig. 2A-C), as inferred from the dissimilarities of the SBCs among the basalt varieties (significant difference at p < 0.05 among B1 at 8M but no significant difference between B2 and B3, Table 2).
The amounts of Si in Ph and in the basalts (B1, B2 and B3) were similar (Table 1).However, these rocks are classified differently because of the conditions under which they crystallize, their chemical composition, and, therefore, the minerals they form.The nepheline in Ph is a feldspathoid mineral that releases Ca, Na, and K faster than the plagioclase (Ca-Na feldspar) present in the basalts.The SBC seemed to be sensitive to such differences since Ph differed from Gr and all the basalts except B3 (Table 2) at the end of the experiment (8 M).This similarity in the SBC between Ph and B3 (Table 2) may be related to the common presence of a small amount of zeolite (heulandite in B3 and natrolite in phonolite; Fig. 2D and F).Zeolite minerals are rich in Ca and Na and have great surface reactivity and porosity.These relationships highlight the importance of understanding not only the total chemistry but also the mineralogy of the rocks since the presence of highly reactive minerals, even in small amounts, may surpass the effects of the less reactive minerals in the rock, even in greater amounts, at least during the initial stages of dissolution.
The SBC structure in the Gr treatment group was similar to that in the basalt group at the beginning of the experiment (1M), as shown by NMDS analysis, but ended the experimental period (8M) (Fig. 4D) distinctly (p < 0.05) (Table 2).In fact, previous works also revealed changes in the soil microbiome composition caused by the presence of rock powder, but they did not compare different rock powders or durations [57,69].
However, we observed clear responses regarding the structure of the bacterial community, whether due to the rapid effect of KCl after 1M or its slow evolution caused by the rock powders, and the effects of these treatments on soil bioindicators were unclear.Several studies have used soil bioindicators to evaluate soil management [70][71][72][73][74] and textural classes [75] in tropical soils.However, studies evaluating the effects of rock powders are scarce, and even so, comparisons are difficult due to the different sources and doses of rocks, soil, plants, enzymes and durations evaluated.In the present study, an exception was the consistent low activity of acid phosphatase in the basalt treatments (B1, B2, and B3).These rocks had between 2 and approximately sixty times more P 2 O 5 (as apatite) in their composition than the other rocks.This enzyme activity was particularly low at 8M (compared to the other treatments at the same time), which may be related to the greater dissolution of apatite at the end of the experiment due both to its low solubility and to the increase in grain porosity and apatite accessibility as the experiment progressed.The greater availability of P seemed to result in lower acid phosphatase activity, which corroborates the literature [76].
The present results showed that, although crop performance was not affected by rock powders in the short term (see K in plant and biomass production in Table 3), the SBC was.Regarding the rock powder treatments, the SBC started showing significant alterations only at 8M.The rock powders did not cause extreme or rapid changes in the bacterial diversity of the soil (Fig. 4A and B) and possibly caused less disruption to the microbial-plant cooperative processes active in the rhizospheric environment, such as nutrient absorption and pathogen protection [77].In fact, Cui et al. [53] observed that the use of NPK decreased SBC richness and increased the abundance of specific groups, decreasing the redundancy and, consequently, the resilience of the SBC.This is related to the large and rapid changes in the soil chemical environment caused by high doses of KCl, as observed in the present study.
This paper contributes to a better understanding of the interplay among rock powders and their impact on the structure of the SBC and bioindicators in the soil.As stated by Swoboda et al. [4] in their review, the duration of a rock powder experiment is important for capturing soil changes.More research is needed to understand the roles of different types of rock powders on the microbiome and how microorganisms affect the bioweathering of minerals.We hope this study will contribute to that end.

Conclusion
Different silicate rock powders caused specific changes in the SBC.At the end of the 8 months (8M), the SBCs clustered according to the rock SiO 2 content (acid rock: granite; intermediate rock: phonolite; and basic rocks: three types of basalts).In addition, variations in mineralogy, chemistry, texture and structure among the basalts led to smaller changes, causing the SBC of Ph to approach the SBC of B3, since both had zeolites, which are very reactive minerals.The SBC in the Gr treatment significantly differed from that in all the other rock powder treatments.In contrast to the rock-amended soils, the SBC in KCl showed rapid and ephemeral changes, manifested not only in the NMDS but also in the richness and Shannon indices.This was the only treatment that had a significant change after the first month (1M), and after eight months (8M), it approached C (control), while the rockamended treatments could still be differentiated from C. The results indicate that the duration of the experiment is an important factor in rock powder management, and thorough analyses of the rock powder and soil used are essential for understanding the various possible changes in the soil-plant-water system.
Ultimately, our research highlights the importance of multiple long-term analyses in microbial geochemistry studies.This study illustrates the challenge of modeling nutrient release, and by extention the carbon dioxide removal (CDR) by enhanced rock weathering (ERW), because rock weathering and the changes in the soil microbiome need to be considered concurrently.This trial also exemplifies the potential of further advanced multidisciplinary studies focusing on the mineral and biological changes that occur simultaneously when rock powder is applied to soils.In this regard, further studies using next-generation sequencing should be valuable.

Fig. 3
Fig. 3 Response of soil microbiological attributes.A Microbial biomass carbon (MBC), B β-glucosidase activity, C arylsulfatase activity, D acid phosphatase activity, E richness and diversity and F Shannon index under the different treatments (C: control; KCl; Ph: phonolite; B1: basalt 1; B2: basalt 2; B3: basalt 3; and Gr: granite) over time: after 1 month (1M), 4 months (4M) and 8 months (8M).Lowercase letters indicate the same treatment over time.Capital letters indicate that the different treatments were compared at the same time by the Wald test (p < 0.05).The absence of any letters indicates that there was no significant difference

Fig. 5 Table 3
Fig. 5 Structural equation model (SEM) representing the complex interrelationships between the most representative (r 2 > 0.75) mineralogical rock traits according to NMDS analysis and their influence on the soil bacterial community structure at 8M.Values associated with solid or dashed arrows represent significant standardized path coefficients with direct and indirect correlations, respectively.Red arrows indicate negative correlations, while black arrows indicate positive correlations.The R 2 values associated with the bacterial structure indicate the proportion of variation explained by correlations with other variables.ns: not significant, *p < 0.05

Table 1
Total chemical compositions of the initial soil and rock samples determined via ICP-OS *Loss on ignition