Microbial communities in paddy soils: differences in abundance and functionality between rhizosphere and pore water, the influence of different soil organic carbon, sulfate fertilization and cultivation time, and contribution to arsenic mobility and speciation

Abstract Abiotic factors and rhizosphere microbial populations influence arsenic accumulation in rice grains. Although mineral and organic surfaces are keystones in element cycling, localization of specific microbial reactions in the root/soil/pore water system is still unclear. Here, we tested if original unplanted soil, rhizosphere soil and pore water represented distinct ecological microniches for arsenic-, sulfur- and iron-cycling microorganisms and compared the influence of relevant factors such as soil type, sulfate fertilization and cultivation time. In rice open-air-mesocosms with two paddy soils (2.0% and 4.7% organic carbon), Illumina 16S rRNA gene sequencing demonstrated minor effects of cultivation time and sulfate fertilization that decreased Archaea-driven microbial networks and incremented sulfate-reducing and sulfur-oxidizing bacteria. Different compartments, characterized by different bacterial and archaeal compositions, had the strongest effect, with higher microbial abundances, bacterial biodiversity and interconnections in the rhizosphere vs pore water. Within each compartment, a significant soil type effect was observed. Higher percentage contributions of rhizosphere dissimilatory arsenate- and iron-reducing, arsenite-oxidizing, and, surprisingly, dissimilatory sulfate-reducing bacteria, as well as pore water iron-oxidizing bacteria in the lower organic carbon soil, supported previous chemistry-based interpretations of a more active S-cycling, a higher percentage of thioarsenates and lower arsenic mobility by sorption to mixed Fe(II)Fe(III)-minerals in this soil.


Introduction
Arsenic accumulates more in rice than in other crops, posing health concerns at a global le v el (Mehar g et al. 2009 ).In aer obic soil en vironments , most of the metalloid is immobilized as arsenate in Fe (o xyhydr)o xides, while in flooded rice soils it is released mainly as arsenite by reductive dissolution.
Micr oor ganisms methylate inorganic arsenic species to the less toxic mono-(MMA) and dimethylated (DMA) oxyarsenates, which are also taken up by the plants from pad d y soil porewater (Meharg and Zhao 2012 ).Recently, thiolated arsenic forms have also been detected, both in pad d y soil pore water (Wang et al. 2020b ) and in rice grains (Colina Blanco et al. 2021 ).The reactions controlling the extent of arsenic dissolution and conversion into different chemical species depend on soil geochemical and physical factors and are often microbially mediated.Particularly, the plant rhizosphere in the pad d y fields is characterized by steep gradients of redox conditions and physicochemical characteristics (pH, organic matter content and redo x-sensiti ve elements, such as arsenic, sulfur and iron) that shape microbial community, even at microscale le v el.
Water management of the rice pad d y w as sho wn to str ongl y affect arsenic biogeochemistry b y fav oring specific microbial populations, which can actively convert the different metalloid oxidation states.In rice field soil, continuous flooding promotes the presence of arsenic-solubilizing ferric iron-and arsenatereducing bacteria (Zecchin et al. 2017a(Zecchin et al. ,b , 2019 ) ), while in aerobic rice field soil the predominance of ferrous iron-and arseniteoxidizing bacteria leads to arsenic immobilization on the solid phase, lo w ering its concentrations in the pore water and in rice grains (Xu et al. 2008, Arao et al. 2009, Zecchin et al. 2017b, Li et al. 2019 ).
Besides water management, sulfate fertilization is a promising tool to decrease arsenic contamination in rice grain, acting both at the plant (i.e.synthesis of phytochelatins) and at the soil level (Dahlawi et al. 2018, Zou et al. 2018, Chen et al. 2019, Hu et al. 2007, 2021, Fang et al. 2023 ).The decr eased concentr ation of arsenic in the pore water of sulfate-amended rice pad d y soil was positiv el y r elated to the presence of rhizospheric dissimilatory sulfate-r educing micr oor ganisms (DSRM) (Jia et al. 2015 ) that, by producing sulfide in anoxic conditions at circumneutral pH, contribute to the r emov al of arsenic by secondary iron sulfides (Hu et al. 2007, Burton et al. 2014, Xu et al. 2019 ).Part of sulfide is used by sulfur-oxidizing bacteria (SOB), which contribute to the production of elemental sulfur (S 0 ) in rice paddies (Stubner et al. 1998, Zhou et al. 2002, Friedrich et al. 2005, Hamilton et al. 2014 ).Moreov er, sulfide and S 0 ar e hypothesized to r eact abioticall y with either arsenite or methylated arsenates, to yield different inorganic and methylated thioarsenates (Planer-Friedrich et al. 2015, Fan et al. 2018, Wang et al. 2020b ).Wang et al. ( 2020b ) suggested that soil organic carbon (C) content plays an important role in the biogeochemistry of arsenic by fueling microbial activity.In their study, the authors observed that sulfate addition caused a stronger decrease of dissolved arsenic coupled to higher percentage of methylation and thiolation in a low C soil compared with a high C soil.The hypothesis was that in the high C soil, reducing conditions lead to FeS mineral formation, a r elativ el y lar ge r emov al of r educed sulfur fr om the por e water and less active sulfur-cycling.By contrast, the lower C content caused less pr onounced r educing conditions [with higher Eh and less Fe(II) in the por e water], with consequentl y a higher conversion of sulfide to S 0 and finally sulfate and increased adsorption of arsenic to mixed v alence ir on miner als .T he oxidized sulfur would then be av ailable a gain for ne w or ganic C driv en r eduction, promoting an active sulfur cycling.
The localization of arsenic, sulfur and ir on biogeoc hemical r eaction sites in the soil/r oot/por e water rice pad d y system is an important but still overlooked aspect.In fact, it is not clear whether arsenic, sulfur and iron transformations occur in solution or in the solid phases and which types of microbial populations are crucial in regulating these reactions .Moreo ver, while the composition of the microbial communities inhabiting different soil/plant compartments (i.e.bulk soil, rhizosphere soil, rhizoplane, endosphere) have been revealed by several authors (Somenahally et al. 2011, Zecchin et al. 2017a,b , Jia et al. 2014, Das et al. 2016 ), to date, the micr obial comm unities living in rice pad d y por e water hav e ne v er been c har acterized, and their composition and r ole in element cycling is still unknown.A pr e vious study (Tian et al. 2021 ) suggested that in wetlands the water table le v el is positiv el y r elated to micr obial species ric hness and div ersity in the por e water.In light of recent issues with water scarcity, which are driving the consider ation of nov el water-saving a gr onomic r eg imes, the ecolog ical equilibrium of k e ystone arsenic-, sulfur-and ir on-cycling micr obial species in the pore water can be altered.In order to clarify if compartmentalization is a major driver of microbial communities involved in arsenic, sulfur and iron biogeochemistry in rice paddies, in the present study we c har acterized bacterial and archaeal populations inhabiting the original unplanted soil, the rhizosphere soil and the porewater of two rice pad d y soils with differ ent or ganic C content, non-fertilized and fertilized with sulfate, and tested whether the compartment effect leads to stronger differences in comparison with other factors such as sulfate fertilization, soil type (with low and high organic C) and cultivation time.Vice versa, the possible influence of the different microbial communities on the geochemical parameters was statistically evaluated to determine the role of specific microbial populations in arsenic, sulfur and iron cycling, focusing on total arsenic mobility and speciation, specifically thiolation and methylation.

Experimental set-up
The rice growing experiment was carried out at the Rice Research Center (Ente Nazionale Risi, ENR) in Castello d'Agogna (Pavia, Ital y).The mesocosms wer e set up in the open air in 0.83 m 2 plastic tanks filled with 30 cm of soil from two distinct pad d y fields located in Cascina Fornazzo and Cascina Veronica (Pavia, her eafter r eferr ed to as "Fornazzo" and "Ver onica" soils, r espectiv el y).Fornazzo and Veronica soils were taken as re presentati ves of high and low C soils, being c har acterized by 47 and 20 g kg −1 of organic C, respectively (Wang et al. 2020a ).Arsenic concentrations were similar between the two soils with 5.6 and 5.8 mg kg −1 , r espectiv el y, whic h is below the Italian national limit for public use soil (20 mg kg −1 , D. Lgs . 152/ 2006 ).Furthermore , Fornazzo soil had slightly higher total S and Fe(II) contents in comparison with Veronica soil (see Supplementary Table 1 for the complete c har acterization of the two soils).Absolute concentrations of dissolved total S and Fe(II) were lower in Veronica than in Fornazzo pore water ( Supplementary Table 2 ), which reflected on the one hand the differences in total S and Fe contents in the two soils ( Supplementary Table 1 ).Ho w e v er, the pr oportion of Fe mobilized from soil to pore water was similar for both soils, while the proportion of S mobilized fr om Ver onica soil w as lo w er than from Fornazzo soil ( Supplementary Table 3 ), reflecting a higher ov er all redox potential in Veronica soil, as described before (Wang et al. 2020b ).
Rice plants ( Oryza sativa var.Selenio) w ere w ater-seeded and culti vated under contin uous flooding for the whole life cycle, using non-sterile tap water provided with a garden hose.Before seeding, mesocosms were fertilized with 100 kg ha −1 of either ammonium sulfate [(NH 4 ) 2 SO 4 ] or urea (CH 4 N 2 O) as control nitrogen fertilizer.Further fertilization was applied at the tillering stage with 30 kg ha −1 and at the booting stage with 50 kg ha −1 of either urea or ammonium sulfate, according to the usual agronomic pr actices.For eac h type of fertilization (i.e.ammonium sulfate vs ur ea/contr ol) and for each soil (i.e.Fornazzo vs Veronica), three r eplicates wer e set up.The physicoc hemical anal yses wer e performed in the pore water over time by Wang et al. ( 2020a ) and the r esults ar e summarized in Supplementary Tables 2 and 4 .

Rhizosphere soil separ a tion and DNA isolation
To analyze the microbial communities inhabiting the rice rhizospheric compartment, rhizosphere soil (i.e.soil strictly attached to the roots) and pore water were collected during stem elongation, flo w ering and the dough stage (corresponding to approximately 60, 80 and 100 days after seeding, r espectiv el y).These three rice life stages are considered crucial for both the development of rhizospheric micr obial comm unities on expanding roots (Edw ar ds et al. 2018 ) and for arsenic uptake, which is highest during flo w ering (Zheng et al. 2011 ).The original unplanted soil was sampled for the c har acterization of the starting microbial comm unity.For eac h experimental r eplicate, thr ee plants wer e collected and pooled in one composite sample.Roots were shaken in tetrasodium pyrophosphate and the rhizosphere soil was separ ated fr om r oots according to Zecc hin et al. ( 2017b ).Por e water was sampled with 15 μm-pore size Rhizon samplers (Rhizon SMS 5 cm, Rhizosphere, Wageningen, The Netherlands) and planktonic cells were collected on cellulose acetate filters (0.2 μm pores) with a vacuum pump.DN A w as isolated from all samples using DNeasy Po w erSoil kit (QIAGEN, Hilden, Germany), accor ding to the manufacturer's instructions .T he quality of the isolated DN A w as c hec ked under UV light by a gar ose gel electr ophor esis on a 1% Tris-acetate-EDT A (T AE) a gar ose gel stained with GelRed (Biotium, CA, USA).

Illumina 16S rRNA genes libraries
F rom DN A isolated from the original soils, rhizosphere soil and pore water samples, bacterial and archaeal 16S rRNA genes were sequenced with primers 341F/806R (5 -CCTA CGGGA GGCA GCA G-3 /5 -GGA CTA CHV GGGTWTCTAAT-3 ) and 344F/806R (5 -CCCT A YGGGGYGC ASC AG-3 ), r espectiv el y (Rago et al. 2017 ).Sequencing was performed on 1 and 0.1 μg of DNA for rhizosphere soil and pore water, respectively, at the DN A Services (DN AS) facility, Resear ch Resour ces Center (RRC), University of Illinois at Chicago (UIC, USA).Raw reads wer e pr ocessed and anal yzed with QIIME2 ( https://qiime2.org/, Bolyen et al. 2019 ).The D AD A2 workflow (Callahan et al. 2016 ) was used to r emov e barcodes and sequence adapters, filter high quality non-c himeric r eads, cluster the r eads in single amplicon sequence variants (ASVs) and pick one re presentati ve sequence for each ASV.Alpha diversity was estimated upon rarefaction of the datasets.Microbial species richness was determined by calculating the number of observ ed micr obial species and using the Chao1 richness estimator (Chao, 1984 ), while microbial species e v enness w as estimated accor ding to Pielou's algorithm (Pielou 1966 ).The taxonomy of r epr esentativ e sequences was assigned using the SILVA SSU r efer ence dataset version 138 ( https:// www.arb-silva.de/ ).The taxonomic classification was performed using a naïve Bayes classifier optimized for the primers used in the sequencing pr ocess (Pedr egosa et al. 2011, Bokulich et al. 2018 ).ASV tables were obtained to determine the r elativ e abundance of each taxon in the samples.Re presentati ve sequences were aligned with mafft (Katoh and Standley 2013 ) and phylogenetic analysis of the re presentati ve sequences was performed with FastTree (Price et al. 2010 ).

Quantification of microorganisms involved in arsenic, sulfur and iron tr ansforma tions b y RT-qPCR
To further analyze microorganisms putatively involved in arsenic cycling in rice rhizosphere and in pore water, the 16S rRNA genes of total bacteria and archaea and genes encoding the A subunit of arsenite oxidase ( aioA ), arsenate reductase ( arsC ), the A subunit of dissimilatory arsenate reductase ( arrA ), arsenite methyltr ansfer ase ( arsM ) and the A subunit of dissimilatory bisulfite reductase ( dsrA ) were amplified and quantified by real-time qPCR (RT-qPCR).Furthermore, 16S rRNA genes of the microorganisms belonging to ir on-r educing Geobacteriaceae and Shewanellaceae and to iron-oxidizing Gallionellaceae were quantified.Details of primer pairs and protocols used in this study can be found in Supplementary Table 5 .For eac h r eaction, 10 ng of template DN A w as mixed with primers and Titan HotTaq EvaGeen ® qPCR Mix (Bioatlas, Estonia), in a total volume of 20 μL.The thermal pr otocols wer e carried out on a QuantStudio TM 3 System (Thermofisher, Waltham, MA, USA).The correct size of qPCR amplicons was c hec ked by a gar ose gel electr ophor esis.Standard curv es wer e created by the amplification of the selected target from plasmid DNA ( Supplementary Table 5 ).The abundance of the quantified functional genes was expressed as relative abundance by normalization to total bacterial and ar chaeal 16S rRN A genes, while the 16S rRNA genes of iron-cycling bacteria were normalized only to total bacterial 16S rRNA genes.

Sta tistical anal ysis
The statistical analyses of Illumina 16S rRNA gene library data were performed using QIIME2 and the R program, v. 3.6.0(R Core Team 2015 ), pac ka ge v egan v ersion 2.5-5 (Oksanen et al. 2020 ).
With the R base pr ogr am, one-way anal ysis of v ariance (ANOVA), Tuk e y's b, Duncan and t-test at P ≤ 0.05 were used for comparisons in the analysis of the alpha diversity, of the abundance of micr oor ganisms r elated to arsenic, sulfur and ir on cycles and of the qPCR amplifications .T he alpha diversity was analyzed by gathering the samples in different groups to evaluate the "compartment effect" (i.e.original unplanted soil vs rhizosphere soil vs pore water), the "soil type effect" (i.e.Fornazzo vs Veronica), the "sulfate amendment effect" (i.e.control vs sulfate) and the "time effect" (i.e.stem elongation vs flo w ering vs dough).
To compare bacterial and archaeal diversity among the samples, w eighted UniF rac distances were calculated from rarefied ASV tables and principal coordinates analysis (PCoA) was performed (Lozupone et al. 2005, Hamady and Knight 2009, Halko et al. 2010 ).Significantl y differ ent gr oups of samples defined by the "compartment effect", the "soil type effect", the "sulfate amendment effect" and the "time effect" wer e identified, a ppl ying the perm utational anal ysis of v ariance (PERMANOVA, perm utations = 999), using the QIIME2 pipeline (Anderson 2001 ).
Significant differences in the abundance (i.e.differential abundance) of bacterial and archaeal families and genera retrieved with 16S rRNA genes Illumina sequencing due to the different soils and to sulfate application were tested using the quasilikelihood F-test implemented in the R pac ka ge EdgeR v ersion 3.11 (Robinson et al. 2010, R Core Team 2015 ).
To highlight statistically significant positive and negative interactions among bacterial and archaeal genera, co-occurrence network analysis was performed by testing the probabilistic cooccurrence model on presence-absence genus tables using the R pac ka ge cooccur v ersion 1.3 (Veec h 2013 , Griffith et al. 2016 ).Positive and negative correlations were tested by grouping original unplanted soil, rhizosphere soil and pore water samples collected according to the compartment (i.e.original unplanted soil vs rhizosphere soil vs pore water), soil type (i.e.Fornazzo vs Veronica), sulfate amendment (i.e.control vs sulfate) and timing (i.e.stem elongation vs flo w ering vs dough).Co-occurrence analysis is based on the presence/absence of each genus in the samples .T he genera that were not present in all the replicates of at least one sample with at least 20 reads were removed from the analysis.To estimate the number of possible keystone gener a, eac h netw ork w as recalculated by removing one genus and calculating the percentage of lost connections without that genus .T his process was repeated for all genera.
To investigate links between chemistry and microbial populations involved in arsenic, sulfur and iron c ycles, linear P earson corr elations wer e calculated between the r elativ e abundance of the differ ent micr obial populations in the rhizospher e soil and in the pore water and pore water physicochemical parameters [i.e. total arsenic, ferrous iron, total sulfur, methylated arsenic, methylated oxyarsenates , total thioarsenates , inorganic thioarsenates , methylated thioarsenates, total organic carbon (TOC), total inorganic carbon (TIC), pH and Eh] at each time point.
Possible statistically significant correlations between the bacterial and archaeal community compositions, the functional predictions, the physicoc hemical par ameters measur ed in the pore water (i.e. total arsenic, ferrous iron, total sulfur, methylated arsenic, methylated oxyarsenates, total thioarsenates, inorganic thioarsenates , methylated thioarsenates , T OC , TIC , pH and Eh) and qPCR data were evaluated by applying redundancy analysis (RDA; Legendre and Legendre 2012 ) and the Mantel test (permutations = 999), both implemented in the vegan package (Legendre and Legendre 2012 ).Bacterial and archaeal genera abundance data, the r elativ e abundance of micr oor ganisms involv ed in arsenic, sulfur and iron cycles (as indicated in Supplementary Dataset 1 ) and the r elativ e abundance of enzymes involved in arsenic, sulfur and iron cycles (as indicated in Supplementary Datasets 2 and 3 ) were Hellinger-transformed to calculate Bray-Curtis dissimilarities, while the physicochemical and qPCR data were logtransformed to calculate Euclidean dissimilarities (Legendre and Gallager 2001 ).

Di v ersity of bacterial and archaeal communities
Illumina sequencing of 16S rRNA genes produced in total 330 206 and 551 058 high quality bacterial and arc haeal r eads, r espectiv el y ( Supplementary Table 6 ).On av er a ge, the rhizospher e soil sho w ed a higher number of ASVs than the pore water.This difference was mor e pr onounced in the bacterial vs arc haeal libr ary and mor e pronounced in the higher organic C soil Fornazzo vs the lo w er organic C soil Veronica ( Supplementary Table 6 ).Accordingly, bacterial and ar chaeal 16S rRN A genes biomarkers were higher in the higher organic C soil in both rhizosphere soil and pore water (data not shown).
Both bacterial species richness (Chao1 index) and e v enness wer e significantl y lo w er in the pore w ater with respect to the original unplanted soil, while an opposite trend was observed for Arc haea, whic h wer e significantl y ric her and mor e uniform in the pore water samples compared with the original unplanted soil and with the rhizosphere soil (Fig. 1 A, P ≤ 0.05).A soil type effect was observed for both Bacteria and Archaea in all compartments, eac h following differ ent patterns ( Supplementary Figur e 1A , P ≤ 0.05).Archaeal Chao1 index negativ el y r esponded to sulfate amendment, being lo w er in all rhizosphere soil and pore water samples where sulfate was supplied, compared with the contr ols ( Supplementary Figur e 1B , P ≤ 0.001).In the rhizospher e soil, both bacterial and archaeal Chao1 index significantly decreased, while in the pore water the trend was more variable ( Supplementary Figure 1C , P ≤ 0.05).
PCoA analysis based on weighted UniFrac revealed a significant "compartment effect" in both bacterial and archaeal communities ( P ≤ 0.01, Fig. 1 B).When analyzing the beta di versity di viding soil (i.e.original unplanted soil and rhizosphere soil) and pore water samples, a significant "soil type effect" was observed in both bacterial and archaeal communities in all compartments, while sulfate amendment and time effects were significant only in soil samples ( P ≤ 0.05, Supplementary Figure 2A and B ).

Composition of rice rhizosphere bacterial and arc haeal comm unities
Soil and pore water samples showed highl y differ ent composition in both bacterial and archaeal communities, evidencing a strong compartment effect (Fig. 2 ).In soil samples, the predominant bacterial phyla were Proteobacteria , Actinobacteriota (formerly Actinobacteria ) , Firmicutes , Acidobacteriota (formerly Acidobacteria ) and other unc har acterized Bacteria (r elativ e abundance 20-30%; Fig. 2 A).In the pore water, uncharacterized Bacteria (relative abundance 40-60%), Proteobacteria and Patescibacteria were the most abundant, and sulfate amendment increased the r elativ e abundance of Epsylonproteobacteraeota (former class Epsilonproteobacteria ) with the concomitant decrease of Patescibacteria .The compartment effect was evident also within Proteobacteria , being more abundant in soil samples, with the exception of Gammaproteobacteria that had their highest abundance in the pore water of Veronica soil ( Supplementary Figure 3 ).
Differ ential abundance anal ysis was performed at the genus le v el in all compartments at the flo w ering stage to evaluate the "soil type effect" and the "sulfate amendment effect".In soil and por e water samples, differ ent gener a wer e significantl y affected by the soil type and sulfate fertilization (Fig. 3 ).Most of the genera [i.e.35 genera belonging to the Acidobacteriota , Actinobacteriota , Bacteroidota (formerly Bacteroidetes ), Cyanobacteria , Firmicutes , Nitrospirota (formerly Nitrospirae ), Alphaproteobacteria , Gammaproteobacteria , Methanomicrobia , Thermoplasmata and Nitrososphaeria ] wer e significantl y driv en by soil type r ather than by sulfate fertilization (Fig. 3 , P ≤ 0.05).Sulfate fertilization significantly decreased the abundance of unc har acterized Elsterales in Veronica and of Methanoregula in Fornazzo in the rhizosphere soil, while in the pore water uncharacterized Campylobacterales and Burkholderiaceae , Ferritrophicum , Methylomonas , Methanobacterium and Candidatus Nitr osotalea significantl y incr eased in sulfate-amended samples ( P ≤ 0.05).
Co-occurr ence network anal ysis (Fig. 4 ) r e v ealed that the compartment, the soil type, sulfate amendment and timing significantly affected specific correlations between microbial genera.In fact, (1) the number of connections was higher in unplanted original soil and in rhizosphere soil than in pore water (i.e."compartment effect"), (2) a higher number of nodes and connections wer e pr esent in lo w er C Veronica soil than in higher C Fornazzo soil (i.e."soil type effect"), (3) less connections wer e observ ed in sulfate-amended samples (i.e."sulfate amendment effect") and (4) the number of connections increased over time (i.e."time effect") (Fig. 4 A, Supplementary Dataset 4 ).
To e v aluate the pr esence of possible k e ystone genera, data concerning the number of connections that were lost when each genus was r emov ed (r esults shown in Supplementary Dataset 5 ) wer e compar ed with the number of genera that show a high number of connections ( Supplementary Figure 4 ), with the  proportion of genera responsible for the loss of at least one connection and the percentage of the maximum number of lost connections ( Supplementary Figure 5 ).The networks based on the rhizosphere soil, the pore water and the stem elongation sho w ed the lo w est proportion of gener a r esponsible for the loss of at least one connection ( Supplementary Figure 5 ), suggesting the presence of a lo w er number of potential keystone genera in these samples if compared with the others.Ho w ever, in the netw orks with the highest number of nodes and connections (i.e.Fornazzo, Veronica, control, sulfate, flo w ering and dough), the maximum number of lost connections from genera removal is lo w er with respect to the other networks .T his might indicate that the trophic networks and the ecological niches in the samples c har acterized by the same soil type and by the same fertilization type and the ones established at the flo w ering and dough stages are likely more stable, pr obabl y because the same functions can be completed by differ ent micr obial gener a. Ther efor e, the loss of one genus does not compromise the presence of a specific microbial function in the ecosystem, due to functional redundancy.
The highest number of positive and negative connections was observed between Alphaproteobacteria , Patescibacteria and Actinobacteriota , follo w ed b y Thaumarchaeota , Acidobacteriota , Bacteroidota and Chloroflexi (Fig. 4 B).The genera with the highest number of positive connections were mostly affiliated to archaeal phyla, such as Thaumar chaeota , Euryar chaeota and Crenar chaeota , together with unc har acterized members of the Acidobacteriota family Blastocatellaceae ( Supplementary Dataset 6 ).Nitrospirota sho w ed the highest number of connections in proportion to the number of genera present in the phylum ( Supplementary Figure 6 ), explained by the presence of only one uncharacterized genus within the order Thermodesulfovibrionia , significantl y positiv el y r elated to Proteobacteria genera (i.e.Rhizobiales genera, Myxococcales genera, Sphingomonas , Comamonas , Desulfobacterium and Acinetobacter ).
In the sulfate-amended samples, a lo w er number of connections was mostly ascribable to a lo w er number of connections related to all archaeal phyla, concomitant to a higher number of connections of the genera Desulfobacterium , Comamonas and Pseudomonas when sulfate was applied ( Supplementary Dataset 6 ).

Inferred microbial functionalities and biomarkers related to arsenic, sulfur and iron biogeochemical cycles
Micr obial functionalities r elated to arsenic, sulfur and iron cycles in the analyzed samples were inferred on the basis of the gener a r etrie v ed with Illumina sequencing of 16S rRNA genes ( Supplementary Dataset 1 ).
All the r etrie v ed gener a involv ed in arsenic, sulfur and ir on c ycles w er e in gener al mor e abundant in the rhizospher e soil than in the pore water ( Supplementary Figure 7A ), indicating that microbial populations of this compartment contribute mostly to those elemental cycling.These outcomes suggest that the compartment was the str ongest driv er if compared with soil type, sulfate amendment and timing.A significant effect was exerted by the soil type on DAsRB/DFeRB and on AsOB in the rhizosphere soil, and on pore water SOB and F eOB (Fig. 5 ).T he two versatile genera Bacillus and Geothermobacter able to perform dissimilatory r espir a-tion of both arsenate and ferric ir on wer e the only contributors to the group DAsRB/DFeRB ( Supplementary Dataset 1 ).
Some of the bacterial and archaeal genera that were involved in arsenic, sulfur and iron and methane cycles (i.e.included in Supplementary Dataset 1 ) were also involved in positive and/or negative correlations according to the co-occurrence analysis .T hese genera are highlighted in Supplementary Dataset 6 .Specifically, MA belonging to the genera Methanomassiliicoccus and Methanoculleus sho w ed more than 100 positive connections with other genera ( Supplementary Dataset 6 ).A number of dir ectl y and indir ectl y arsenic-cycling bacterial genera showed a high number of positive connections, with Clostridium , Mesorhizobium , Bradyrhizobium , unc har acterized Thermodesulfovibrionia , Desulfobacterium and Comamonas among the most connected ones ( Supplementary Dataset 6 ).
To implement the information on microbial functions inferred by the presence of specific microbial genera in the 16S rRNA gene libr ary, pr edicted enzymes were investigated by Tax4Fun2 and specific gene biomarkers were quantified by RT-qPCR at the flowering stage.
The ubiquitous arsenate detoxification system ARS (i.e.arsenate reductase ArsC, arsenite efflux pump ArsB) was detected in the pore water and in rhizosphere soil, where arsC was of the order of 10 4 and 10 8 copies per g of mL/dry soil, r espectiv el y, r eflecting the ability of the Arsenate reductase to use soluble arsenic (Fig. 6 , Supplementary Figure 7B ).Arsenite oxidase coded by aioA gene was present only in rhizosphere soil at 10 8 copies per g of dry soil, together with arsenite methyltr ansfer ase arsM , whic h was also retrie v ed in the pore water of higher organic C soil Fornazzo (Fig. 6 , Supplementary Figur e 7B ).Her e , the higher C content might ha ve favored the presence of methylated groups.Genes encoding the F igure 4. Co-occurrence netw ork anal ysis of original unplanted soil, rhizospher e soil and por e water samples.Significant positiv e and negativ e correlations ( P ≤ 0.05) were investigated by grouping the samples according to the compartment (original unplanted soil vs rhizosphere soil vs pore water), soil type (Fornazzo vs Veronica), sulfate amendment (control vs sulfate) and timing (stem elongation vs flo w ering vs dough), resulting in six different networks (A).The number of positive connections between classes/phyla were investigated to highlight possible syntrophic or antagonistic interactions, as well as the establishment of ecological niches (B).

Figure 5.
Relative abundance of genera putatively involved in arsenic, sulfur and iron transformation, and in methanogenesis in rhizosphere soil and pore water samples collected at the flowering time point.The genera included in this analysis were selected according to the literature, as well as to the presence of functional marker genes in their genome deposited in NCBI, as cited in Supplementary Dataset 1 .

Figure 6.
Relative abundance quantified by RT-qPCR of genes involved in arsenic and sulfur transformation (A) and of bacterial families involved in the iron cycle (B), in the rhizosphere soil and pore water of Fornazzo and Veronica soils, with and without sulfate amendment, at the flo w ering stage.Data of arsenic-and sulfur-transforming genes were normalized with respect to the total number of 16S rRNA gene copies of Bacteria and Archaea, while data of iron-cycling bacteria were normalized to total Bacteria only.r espir atory arsenate reductase (i.e.ArrA) and the anaerobic arsenite oxidase (i.e.ArxA) were not retrieved by Tax4Fun2, nor by R T-qPCR, accor ding to pr e vious e vidence in rice pad d y soil from the same area (Zecchin et al. 2017a ).
Sulfur cycle-related enzymes showed a compartmentdependent pattern, with sulfur oxidase (SoxAB) and thiosulfate r eductase mor e abundant in rhizospher e soil samples, and enzymes involved in dissimilatory sulfate r espir ation (DsrAB) and r espectiv e biomarker dsr being absent in the pore water (Fig. 6 , Supplementary Figure 7B ).Sulfate amendment increased the abundance of enzymes involved in dissimilatory sulfate r espir ation (DsrAB) in the rhizosphere sample of the lo w er carbon soil Veronica, and SoxAB in the pore water (Fig. 6 , Supplementary Figure 7B ).
A "compartment effect" was observed for the r elativ e abundance of most targets, with aioA and dsr only detected in rhizosphere soil samples, and arsC and arsM more abundant in the rhizosphere soil compared with the pore water (Fig. 6 ) Regarding the iron cycle , DF eRB of the families Geobacteraceae and Shew anellaceae wer e mor e abundant in the por e water compared with the rhizosphere soil (Fig. 6 ), with Shewanellaceae significantl y mor e abundant in Ver onica samples ( P ≤ 0.05).Sulfate amendment significantl y incr eased the abundance of Gallionellaceae in Ver onica rhizospher e soil, and of Geobacteraceae in both Fornazzo and Ver onica, compar ed with the controls ( P ≤ 0.05).
In gener al, micr obial functions wer e mor e r epr esented in the rhizosphere compartment with respect to the soil pore water.This habit reflects either the higher bacterial biodiversity that characterizes the soil compartment, and the fact that a large part of the pore water microbial community comprised uncharacterized ASVs.For such ASVs the prediction of functions might fail to give a correct picture.

Correlation among microbial di v ersity, functionality and environmental parameters
The correlation among the microbial community composition and functionality in the samples and the main physicochemical par ameters measur ed in the por e water (i.e .total arsenic , total sulfur, ferr ous ir on, total thioarsenates, inor ganic thioarsenates, meth ylated thioarsenates, meth ylated oxyarsenates , T OC , TIC , pH and Eh), as well as the qPCR quantifications , were in vestigated by RDA analysis (Fig. 7 ) and the Mantel test ( Supplementary Tables 7 ,  8 and 9 ).
In both the rhizosphere soil and the por e water, the beta diversity of bacterial and archaeal communities was significantly driv en by TIC, ferr ous ir on and total sulfur concentrations (all higher in Fornazzo samples; P ≤ 0.05, Fig. 7 A, Supplementary Table 7 ).TOC significantl y sha ped the bacterial and archaeal communities in the rhizosphere soil ( P ≤ 0.05, Fig. 7 A, Supplementary Table 7 ).pH was significantly correlated with the bacterial communities in both the rhizosphere soil and the pore water ( P ≤ 0.05, Fig. 7 A, Supplementary Table 7 ).In the rhizosphere soil, the bacterial community was significantly shaped by total arsenic concentration ( P ≤ 0.05, Fig. 7 A, Supplementary Table 7 ).On the other hand, the arc haeal comm unity living in the pore water was significantly related to the concentration of methylated thioarsenates ( P ≤ 0.05, Fig. 7 A, Supplementary Table 7 ).Shewanellaceae 16S rRNA and arsM genes quantified by qPCR were significantl y r elated to the beta diversity of both bacterial and archaeal communities in both the rhizosphere soil and pore water samples, being more abundant in Veronica samples ( P ≤ 0.05, Fig. 7 A, Supplementary Table 7 ).In the pore water, both bacterial and arc haeal comm unities wer e significantl y r elated to Gallionellaceae 16S gene copy number that was higher in Veronica samples ( P ≤ 0.05, Fig. 7 A, Supplementary Table 7 ).
The RDA between inferred functionalities and physicoc hemical par ameters sho w ed a similar pattern in the rhizosphere soil, were the abundance of microbial species (i.e .DAsRB/DF eRB, AsRB , AsOB , AsMB , FeOB and SOB) and enzymes (ArsC, ArsB, ArsM, ArsH, DsrAB and Sox) involved in arsenic, sulfur and iron cycles were shaped by TIC , TOC , total sulfur and ferrous iron and wer e r elated to arsM gene copy number ( P ≤ 0.05, Fig. 7 B and C, Supplementary Tables 8 and 9 ).On the other hand, the distribution of the specific genera in the pore water samples was shaped by total sulfur and significantl y r elated to arsC and arsM gene copies ( P ≤ 0.05, Fig. 7 B, Supplementary Table 8 ), while Tax4Fun2-inferr ed enzymes wer e significantl y r elated to total sulfur and ferr ous ir on, as well as to the concentration of methylated arsenic ( P ≤ 0.05, Fig. 7 C, Supplementary Table 9 ).
To further investigate whether the dynamics in the microbial populations observed in this study support previously reported differ ences in c hemistry betw een the tw o soils, P earson linear corr elation tests wer e performed between differ ent por e water c hemistry parameters (i.e. total arsenic, total sulfur, total thioarsenates , inorganic thioarsenates , methylated thioarsenates , methylated oxyarsenates , T OC , TIC , pH and Eh) and rhizospheric and pore water arsenic-, sulfur-and iron-cycling microbial populations.
While arsenic did not show any correlation with microbial populations involved in arsenic, sulfur and iron cycles, Fe(II), total S, TIC and TOC wer e significantl y corr elated with differ ent microbial populations, showing different trends ( Supplementary Table 10 ).Specifically, Fe(II) was negatively correlated with rhizospheric AsOB and DSRB, while total S, TIC and TOC were negativ el y corr elated with rhizospheric DAsRB/DFeRB, AsOB and DSRB but positiv el y corr elated with rhizospheric FeOB.Inter estingl y, rhizospheric SOB were positively correlated with different thiolated and methylated As species (i.e. total thioarsenates, inorganic thioarsenates , methylated thioarsenates).P ore water F eOB were negativ el y corr elated with Fe(II) and total S, but positiv el y corr elated with methylated arsenic.Rhizospheric DAsRB/DFeRB, AsOB and DSRB, and pore water FeOB, are significantly negatively driven by Fe(II) and total S, which are lo w er in Veronica soil.This outcome is in accordance with pr e viousl y shown data (in Fig. 5 ), where these populations wer e mor e abundant in Ver onica soil.Mor eov er, these data suggest a link between SOB and arsenic thiolation and between FeOB and methylated arsenic.

Discussion
To the best of our knowledge, the analyses performed in the present study revealed for the first time that pad d y field pore water harbors specific microbial populations that are distinct from the ones inhabiting the original unplanted soil and the rhizosphere soil, thus confirming that the compartment, likely characterized by different physico-chemical properties, is the major driv er in sha ping these differ ent ecosystems.In fact, pr e vious studies on rice pad d y soil compartments were only focused on unplanted, bulk or rhizosphere soil, and demonstrated the existence of a "rhizosphere effect" due to the presence of a high amount of root exudates coupled to O 2 leaking from root aerenchyma that together fuel microbial organic matter degradation, respiration and fermentation processes (Lynch and Whipps 1990, Revsbech et al. 1999, Liesack et al. 2000, Demoling et al. 2007, Marschner 2011, Wörner et al. 2016, Huaidong et al. 2017, Ding et al. 2019 ).
Unexpectedl y, arc haeal div ersity was found to be higher in the pore water with respect to the original unplanted soil and to the rhizosphere soil.While a sharp separation of the rhizosphere soil samples from the bulk soil was ensured by the protocol follo w ed (in accordance with Zecchin et al. 2017b ), we cannot ensure that the sampled pore water was derived exclusiv el y fr om the rhizospher e ar ea.Hence, in the por e water samples, micr oor ganisms deriving from the anoxic bulk soil area were likely included.One hypothesis could be that a pr oportionall y higher number of anaerobic and oligotr ophic arc haeal species are present in the pore water compartment compared with the bacterial community.This possible explanation might be found in the lifestyle of pore water arc haeal species, whic h is, ho w e v er, still poorl y infer able because most of the r etrie v ed gener a wer e unc har acterized.These F igure 7. Redundanc y analysis (RDA) of bacterial and archaeal communities in rhizosphere soil and pore water from Fornazzo and Veronica soils with and without sulfate amendment (A), of the distribution in the samples of microbial populations involved in arsenic, iron and sulfur c ycles accor ding to dataset 1 (B), and of the distribution of the functionalities involved in arsenic iron and sulfur cycles according to Tax4Fun2 analysis (C).Red arrows indicate physicochemical parameters and genes quantified with RT-qPCR that are significantly correlated with the microbial community composition of the samples, according to the Mantel test.
outcomes suggest the importance of further investigating microbial communities in rice pad d y pore water in order to clarify the role of uncharacterized microbial taxa in element cycling.
While the rhizosphere soil was dominated by Proteobacteria , Acidobacteriota , Actinobacteriota and methanogenic Archaea typically found in the rice rhizosphere (Bao et al. 2016, Ding et al. 2019 ), the pore water microbial communities were more related to aquatic ecosystems, and mostly hosted unc har acterized microbial species.Among these, members of the phylum Patescibacteria, including microbial taxa formerly assigned to the "Candidate Phyla Radiation" (CPR) group (Parks et al. 2018 ), were dominant.These micr oor ganisms ar e widel y distributed in aquatic subsurface en vironments , where they were suggested to have a fermentative lifestyle and to be associated with autotr ophic ir onand sulfur-cycling micr oor ganisms (Herrmann et al. 2019 ).The pore water archaeal community hosted a significantly higher proportion of members of the DPANN arc haeal super phylum (i.e.Diapherotrites , Parv arc haeota , Aenigmarc haeota , Nanoarc haeota and Nanohaloarc haeota ) compar ed with the rhizospher e soil.This superphylum includes a variety of still poorly characterized smallsized micr oor ganisms with div erse metabolic featur es that ar e supposed to be widespread in the environment (Moissl-Eichinger et al. 2018 ).The presence of DPANN in a gricultur al soils, including rice paddies, was r ecentl y r eported, ho w e v er, without further discussion (Wan et al. 2021, Cho et al. 2022 ).Members of the phylum Nanoarchaeota include putative sulfide-oxidizers that live as obligate endosymbionts of Crenarchaeota (St. John et al. 2019 ), and might play a crucial role in sulfur cycling in rice paddies and in plant detoxification from reduced sulfur compounds.
Inter estingl y, both Illumina sequencing and qPCR indicated that the r elativ e abundance of most of the micr oor ganisms involved in arsenic, sulfur and iron c ycles w er e in gener al mor e abundant in the rhizosphere soil compared with the pore water, suggesting that organic matter, the surface of soil particles and miner als hav e a crucial role in mediating microbial reactions in rice field soil, thus influencing the element biogeochemical cycles as reported before (Hoffman et al. 2021, Crundwell 2013 ).Ho w e v er, because se v er al unc har acterized bacterial and arc haeal gener a wer e r etrie v ed in the por e water, the r elativ e importance of por e water microbial communities with respect to rhizosphere soil in element cycling should be confirmed by further in vestigations .
Ov er all, the r esults underline that the "soil type" is defined by a complex of physico-chemical parameters (i.e.mainly organic C content, iron and sulfur) that were the main drivers in the taxonomic and functional shaping of the microbial communities, rather than sulfate addition.In fact, within the strong influence determined by the compartment, the two soils were originally c har acterized by distinct micr obial comm unities, with differ entially abundant bacterial and archaeal genera, and over time the differentiation was maintained both in the composition and in the ecological networks despite a similar a gr onomic mana gement.Mor eov er, the soil type strictly defined specific physicochemical par ameters (or ganic and inor ganic C, ir on, sulfur and pH) that selected specific microbial populations directly or indirectly involved in arsenic cycling.This confirms the r ele v ance of por e water C content and redox potential in shaping the rhizosphere micr obial populations involv ed in arsenic biogeoc hemistry in rice paddies, as hypothesized in pr e vious studies (Somenahally et al. 2011, Zecchin et al. 2017a, Yang et al. 2018, Ma et al. 2014, 2020, Dai et al. 2020, Hossain et al. 2021 ).Microbial communities in the rhizosphere soil and in the pore water wer e differ entl y influenced b y pore w ater C. In fact, while pore w ater TIC and TOC were significantl y r elated to both the phylogenetic and functional diver-sity of the rhizosphere microbial communities, these parameters wer e onl y weakl y or not r elated to the por e water micr obiome .T he presence of several uncharacterized microbial genera in the pore water might have biased these outcomes.In this regard, the soil type had a certain but rather minor role.In fact, the lo w er organic C content in Veronica soil corresponded to a lo w er micr obial pr oliferation in comparison with Fornazzo soil, as demonstrated in the present study by absolute qPCR quantification, but this difference was not reflected by a lower number of microbial species.This might indicate that not only the concentration but also the quality of organic C substrates inherited from the soil is important for microbial community shaping.Some works suggest that the soluble C released from the added rice straw is rapidly utilized, suppl ying easil y degr adable electr on donors that may prime micr obiall y catal yzed r eductiv e dissolution of soil ir on miner als; ho w e v er, it is the pr ogr essiv e r elease of pr e viousl y ir on-stabilized organic C that feeds the microbial communities during the whole growing season (Said-Pullicino et al. 2016 ;Ye and Howrath, 2017 ).
The outcomes from the present study sho w ed that sulfate amendment suppr essed se v er al positiv e and negativ e corr elations driven by arc haeal gener a, pr obabl y not only in relation to sulfate, but to a general higher nutrient availability in rice rhizosphere in comparison with the unamended mesocosms .T his aspect also emerged in the study of Liu et al. ( 2021 ), where alternating wet-dry c ycles w ere found to be more efficient than sulfate fertilization in decr easing CH 4 pr oduction in Ver onica str a w-amended soils .T he hypothesis is that in soils with high C content, sulfate fertilization might not be crucial for DSRB to outcompete MA.
In the present study, sulfate amendment increased the relative abundance of DSRB, as previously observed by Wörner et al. ( 2016 ), and SOB.Many genera that positively responded to sulfate amendment were uncharacterized bacteria and archaea.For some of these, the presence of sulfur cycling as a crucial metabolic tr ait was inferr ed by pr e vious in vivo studies, as for the DSRB class Thermodesulfovibronia (i.e.phylum Nitrospirota ; Sekiguchi et al. 2008, Anantharaman et al. 2016, Zecchin et al. 2018, Umezawa et al. 2020, 2021, Arshad et al. 2017, 2018, Kato et al. 2018 ) and for the SOB Campylobacterales (Inagaki et al. 2003, 2004, Kodama and Watanabe 2004, Sie v ert et al. 2007, Tan and Foght 2014, Stolz et al. 2015 ).T hese observations , coupled to a r elativ el y high abundance of SOB, support pr e vious hypotheses that sulfur cycling occurs at high rates in rice rhizosphere, and that it is stimulated by sulfate amendment (Pester et al. 2012, Wörner et al. 2016, Zecchin et al. 2018 ).In rice paddies, DSRB might have a crucial role in arsenic thiolation by the production of sulfide, as previously observed for Desulfovibrio desulfuricans in the human gut (DC.Rubin et al. 2014 ).
In the pore water of lower organic C Veronica soil, the obligate FeOB F erritrophicum , r esponsible for ir on plaque formation in wetland plants (Weiss et al. 2007 ), positiv el y r esponded to sulfate fertilization.The significant increase of these iron-related microorganisms in sulfate-amended rice pad d y mesocosms in the lo w er organic C soil contributes to explain the effect of sulfate amendment in decr easing dissolv ed arsenic.Initiall y, this low or ganic C soil already had a significantly higher abundance of DAsRB/DFeRB and DSRB in the rhizosphere soil and of FeOB in the pore water than in the high organic C soil.So, the more marked effect of sulfate amendment observed in Veronica pore water in decreasing dissolved arsenic with respect to Fornazzo might be explained (1) by a higher immobilization of arsenic with enhanced iron plaque production by FeOB, similar to what was pr e viousl y described by Hu et al. ( 2007 ); and/or (2) by a higher co-precipitation or adsorption of arsenite (produced by DAsRB), with secondary iron sulfide minerals (produced by DFeRB and DSRB), or rather, as suggested before (Wang et al. 2020b ), by mixed F e(II)F e(III) minerals (produced by DFeRB coupled to re-oxidation of reduced sulfur).Mor eov er, the significant positiv e corr elation between SOB and total sulfur might contribute to explain the higher thiolation of arsenic in Veronica pore water with respect to Fornazzo, providing locall y higher concentr ations of sulfide for arsenic thiolation and supporting a more active sulfur cycling.It might be hypothesized that all three proposed microbial processes might have occurred in the mesocosms, resulting in the lo w er total arsenic mobility and higher percentage contribution to total arsenic of thioarsenates in Ver onica compar ed with Fornazzo.In the lo w er carbon soil, the presence of SOB and DSRB supports an active sulfur cycle fueled by available sulfur species.The positiv e corr elation between SOB and thiolated arsenic species might be explained by the availability of S 0 and SO 4 2 − to be used as electron acceptors by DSRB that produce sulfide and increment thiolation.On the other hand, in higher carbon soil, the formation of FeS subtracts reduced sulfur fr om por e water thus establishing a less active microbially mediated sulfur cycle, ther efor e lowering thiolated arsenics.
In view of the upcoming water scarcity due to climate change, these outcomes suggest that a lo w er pore water content will pr obabl y hav e dr amatic effects on element cycling mediated by microbial populations both present in the different soil/rhizosphere/water compartments and affected by the redox potential.It was pr e viousl y shown (Zecchin et al. 2017a ) that the micr obial comm unities de v eloped in the rhizospher e soil of rice cultiv ated under aer obic conditions wer e completel y differ ent if compared with the ones inhabiting the rhizosphere soil of continuously flooded rice plants .Hence , a lo w er w ater content in rice pad d y soils is expected to pr ogr essiv el y decr ease the activity of pore water-specific microbial populations such as FeOB to a le v el that will depend on the extent of water scarcity and/or to the type of water management adopted in the different agronomic sc hemes.Mor eov er, because rhizospheric micr obial comm unities ar e significantl y sha ped by the pore water parameters, a decrease of soil water content might in general slow down microbial arsenic, sulfur and iron cycling in rice paddies .T his eventuality should be car efull y consider ed b y w orking on pr edictiv e models that include the outcomes of the present and all pr e vious available data.
Ov er all, the data obtained in this study r e v ealed that the compartment effect in rice pad d y soil is the major driver of microbial diversity and functionality, ultimately affecting the complex interplay between microorganisms in rice pad d y soil and arsenic, sulfur and iron mobility.
The micr obial comm unities inhabiting the rhizospher e soil and the pore water developed over time were completely different and r esponded differ entl y to sulfate amendment.In each compartment, soil type and C content significantly drove the development of the microbial communities .T he effect of sulfate amendment was also compartment-specific.It was linked to iron and sulfur concentrations and, in the low-C soil, promoted iron-oxidizing bacteria, dissimilatory arsenate-, ir on-and sulfate-r educing bacteria, whic h wer e likel y r esponsible for arsenic sequestration by secondary ir on miner als and/or ir on sulfides and its subsequent decrease in the pore water.The higher proportion of arsenic thiolation measured in sulfate-amended compared with unamended soil was found to be related to dissimilatory arsenate-, sulfatereducing bacteria and sulfur-oxidizing bacteria.
These aspects should be considered carefully in the future, when the need to face a pr ogr essiv e dr ought due to climate change will necessarily lead to the adoption of more water-saving a gr onomic sc hemes in rice cultiv ation.

Figure 1 .
Figure 1.Alpha and beta diversity by means of Chao1 index and Pielou's e v enness (A) and of weighted UniFrac dissimilarities of bacterial and archaeal communities in the original unplanted soil, rhizosphere soil and pore water.Comparisons were statistically tested to evaluate the compartment effect according to ANOVA (A; letters indicate significantly different groups; * P ≤ 0.05) and to the PERMANOVA test F (B; * * P ≤ 0.01).

Figure 2 .
Figure 2. Relative abundance of bacterial (A) and archaeal (B) ASVs at phylum and class levels, respectively, retrieved in original unplanted soil, rhizosphere soil and pore water from Fornazzo and Veronica soils without (C) and with (S) sulfate amendment, at different time points (before seeding, stem elongation, flo w ering and dough).

Figure 3 .
Figure 3. Bacterial and archaeal ASVs classified at genus level that significantly responded to soil type and sulfate amendment in the original unplanted soil, in rhizosphere soil and in pore water samples at the flowering time point ( P ≤ 0.05).