Investigation of iron-reducing and iron-oxidizing bacterial communities in the rice rhizosphere of iron-toxic paddy field: a case study in Burkina Faso, West Africa

ABSTRACT Iron (Fe) toxicity in rice is one of the serious problems in some paddy fields in West African areas. Microbial community structures involved in the redox cycle of Fe have not been revealed in the Fe-toxic paddy fields. The present study investigated the bacterial community structure and the abundance of Gallionellaceae, Geobacteraceae, and Anaeromyxobacteraceae, as the representative indicator bacteria of Fe oxidizers and reducers, in the bulk and rhizosphere soils and rice roots of a Fe-toxic paddy field in Burkina Faso (BF)in 2017–2019. Thosein a paddy field in Anjo, Japan (AN) were also analyzed for comparison. The amplicon sequencing analysis revealed that the BF rhizosphere soil was characterized by typical anaerobic bacterial groups like Firmicutes and Deltaproteobacteria, including several potential Fe reducers. The relative abundance of Gallionellaceae, lithotrophic Fe oxidizers, in the BF rice roots was significantly lower than that in the AN rice roots. Quantitative PCR analysis showed that the ratios of Gallionellaceae to Geobacteraceae and to Anaeromyxobacteraceae were higher in the rice roots than in the soils irrespective of the fields. However, the ratios of Gallionellaceae to Geobacteraceae were lower in the BF soils and roots than in the AN soils and roots. The ratios of Gallionellaceae to Anaeromyxobacteraceae in the BF soils were also lower than those in the AN soils. These findings indicated the relative predominance of Geobacter- and Anaeromyxobacter-related Fe reducers over Gallionellaceae-related Fe oxidizers in the rice rhizosphere of the BF field, corresponding well to the circumstances of Fe-toxic soil: higher Fe(II) amounts in the soil. Since Fe(II)-oxidizing activity at rice roots is an important factor as a primary defense system against Fe(II) in the soil solution, the ratios of Gallionellaceae to Geobacteraceae and to Anaeromyxobacteraceae may serve as an indicator of potential Fe(II)-oxidizing activity of rice rhizosphere. Further studies focusing on the activity of Fe oxidizers and Fe reducers at rice roots under effective cultivation practices and in various types of Fe-toxic paddy fields will help to promote a better understanding of the Fe-toxic soil circumstances and to establish sustainable rice cultivation in the Fe-toxic soils.


Introduction
Rice is a staple food and feeds half of the world's population.In general, paddy fields are kept under flooded conditions during most of the rice cultivation period with irrigation water or rainwater and drained before rice harvesting.As a result, the circumstances in the soil change between anoxic, reduced conditions and oxic, oxidized conditions (Kyuma 2004;Ponnamperuma 1972;Takai and Kamura 1966).
Iron (Fe) toxicity in rice is one of the serious problems in some areas of South and South-East Asian countries and sub-Saharan African countries (Becker and Asch 2005;Kirk et al. 2022;Sahrawat 2004).The disorder is caused by the excess uptake of ferrous Fe (Fe(II)) ion by rice, which is generated in the soil under anoxic conditions, frequently in association with a poor nutrient status of the soil (Kirk et al. 2022;Sahrawat 2004).Reactive oxygen species are overproduced in rice plants with the excess amounts of Fe(II) ion by the Fenton reaction, which brings about oxidative damages to the cell components, e.g., membranes, DNA, and protein, and physiological disorders, such as the decrease in the photosynthetic rate.Finally, the 'bronzing' of rice leaves emerges as a result of cell death (Aung and Masuda 2020;Onaga, Dramé, and Ismail 2016).
The damages of rice by Fe toxicity show large variations depending on the region, soil type, cropping season, and severity and duration of Fe toxicity occurrence.Becker and Asch (2005) classified the site circumstances of Fe toxicityemerging paddy field soil (Fe-toxic paddy field soil) into three clusters: young acid-sulfate soil in the coastal region (cluster I), marshes and highland swamps with clayey Ulti-and Histosols (cluster II), and inland valley swamps with sandy soils (cluster III).The typical concentration ranges of Fe(II) in the soil solution (Fe(II) aq ) that cause the emergence of Fe toxicity are largely different among the clusters (500-2500, 300-900, and 20-600 mg Fe(II) l −1 for the clusters I, II, and III, respectively).The cluster III-type paddy fields, as well as the cluster I, are major Fe toxic paddy fields that cover ca. 5 Mha globally, mostly the inland valley swamps in the humid zone of West Africa and the highlands in Madagascar (Supplementary Fig. S1; Becker and Asch 2005).Although the concentration range of Fe(II) aq is not always high and the disorder of Fe toxicity emerges locally and patchy, the disorder frequently occurs even in the early growth stage of rice since Fe(II) and Fe(III) are inflowed by the interflow and surface flow from surrounding area depending on the precipitations.As a result, considerable losses of rice yield (30-70%) have been reported, and the complete failure of rice cultivation is possible when severe circumstances coincide in the early growth stage of rice (Becker and Asch 2005).
To overcome this serious problem, previous studies have attempted to reveal the emerging and tolerance mechanisms of Fe toxicity in rice and develop adaptation strategies (Becker and Asch 2005;Kirk et al. 2022;Rasheed et al. 2020;Sahrawat 2004;Zahra et al. 2021).Several Fe-tolerant rice cultivars have been screened so far (Konaté et al. 2022;Mahender et al. 2019;Melandri et al. 2021;Sikirou et al. 2015Sikirou et al. , 2018)).However, those countermeasures have not been always effective.Since the emerging mechanisms of Fe toxicity are complicated and further studies are still needed to elucidate the mechanisms, especially under field conditions.
Meanwhile, Fe(II) can be oxidized in the rice rhizosphere, which conduces to the formation of Fe plaque on rice roots (Khan et al. 2016).Oxidation of Fe(II) is a primary factor to prevent the excess uptake of Fe(II) by rice roots (Becker and Asch 2005).Fe(II) oxidation is mediated by not only abiotic reactions but also microbial reactions, and Fe(II)-oxidizing bacteria (FeOB) are classified into four physiological groups; acidophilic FeOB, microaerophilic FeOB, anaerobic, nitratereducing FeOB, and anaerobic, phototrophic FeOB (Kappler et al. 2021).Recently, suggestive evidence for the involvement of microaerophilic, lithotrophic Fe(II)-oxidizers, belonging to the family Gallionellaceae, in the oxidation process of Fe(II) of paddy soil has been reported (Khalifa et al. 2018;Nakagawa et al. 2020;Naruse et al. 2019;Watanabe et al. 2013Watanabe et al. , 2021)).The rhizosphere of wetland plants including rice is one of the suitable habitats for microaerophilic Fe(II)-oxidizers (Emerson, Weiss, and Megonigal 1999;Maisch et al. 2019;Neubauer, Emerson, and Megonigal 2002;Weiss et al. 2003).Therefore, Gallionella-related Fe(II)-oxidizing bacteria (Gallionellarelated FeOB) potentially play an important role in the alleviation of Fe toxicity by oxidizing Fe(II) in the rice rhizosphere and can serve as an indicator of the Fe-oxidizing capacity.However, the ecology of Gallionella-related FeOB in Fe-toxic paddy field soils has not been revealed yet.
Elucidation of the features of the bacterial community involved in the redox cycle of Fe in Fe-toxic paddy fields gives us further insights into the circumstances of Fe dynamics in the fields and will be helpful to alleviate Fe toxicity of rice in the fields.However, no study investigated the interrelations between FeRB and FeOB in the paddy fields in the past.To unravel microbiological features in Fe-toxic paddy fields as an initiative study, we investigated the bacterial community structure of rice rhizosphere, especially focusing on Geobacter-and Anaeromyxobacter-related FeRB and Gallionella-related FeOB communities as representative indicator bacteria of the Fe-toxic circumstances, in a paddy field located in an inland valley bottom area in Burkina Faso, West Africa, as a case study.The site is categorized as the cluster III type and severe Fe toxicity has frequently emerged (Konaté et al. 2022).

Investigated field
Farmer's paddy fields (plots A and B) in Burkina Faso, West Africa, were investigated in 2017-2019.The fields were located in the Kou valley, ca.30 km far from Bobo Dioulasso, a western city of Burkina Faso (BF field), where irrigated rice cultivation has been continued since the 1960s.Fe toxicity of rice frequently emerged in this area (Konaté et al. 2022;Sikirou et al. 2015).Rice cultivation has been generally carried out twice (April or May to July and July or August to October) during the wet season (April to October) depending on the precipitations.The soil was classified as Dystric Gleysol (Narteh and Sahrawat 1999) with a Clay loam texture (Konaté et al. 2022;Otoidobiga et al. 2016).
As a contrast to the bacterial community structure in the BF field, the bacterial community in experimental paddy fields (plot E1 and H7), where no Fe toxicity of rice has been observed, in the Aichi Agricultural Research Center (Anjo, AN field) in Japan were also investigated.A paddy-upland rotation system (year by year) has been introduced in the investigated plots, and rice has been cultivated in the summer season (April or May to August or September).The soil was classified as the Stagnic Cambisol with a Light Clay texture.
The field practices, fertilizations, sampling dates, and general physicochemical properties of the soils in the BF and AN fields are summarized in Table 1 and supplementary Supplementary Tables S1 and S2.Fe(II) aq concentrations in the irrigation water, open ditch water around the plot, surface water, and soil solution, and the contents of acetate-extractable Fe(II) at the sampling occasions, as an indicator of Fe toxicity risk of rice, are shown in Supplementary Table S3.

Sampling of soil and rice roots
Bulk soil, rhizosphere soil, and rice root samples were taken from the investigated fields.
In the BF field, rice plants were classified into two categories depending on the severity of Fe toxicity symptoms (moderate to severe symptom [MS] vs. no or slight symptom [NS]) by checking the severity of leaf bronzing (Supplementary Table S1).In 2017 and 2018, MS and NS samples were collected from two plots (plots A and B) with each 2 or 4 replications.Plot A is ca. 10 m distant from plot B, but the Fe toxicity symptom in plot A was more severe than that in plot B (Supplementary Fig. S1).In 2019, we selected MS and NS rice plant samples from plot B with 4 replications each since the growth stage of rice was different between plots A and B in 2019.All the samples were taken in the second rice cultivation period of the year.
In the AN field plots, soil and rice root samples were taken twice a year: the tillering (TL) and flowering (FL) stages with each one or two replications (Supplementary Table S2).
The sampling was carried out in the same manner on all the sampling occasions.A soil block (area ca.20 × 20 cm, depth 15 cm) with a rice hill was dug out from the plot.Then, some portions (ca 0.5 kg in total) of the four corners of the block were gathered in a plastic bag to make a composite soil sample as bulk soil, except for the bulk soil sample in 2017, in which a composite soil sample was made from the two replicates of the block.Subsequently, adjacent soils were removed by hand as much as possible, and the soil in the vicinity of rice roots, rhizosphere soil, was collected from a rice hill in a plastic bag.Roots were carefully washed with water and cut into 5-mm lengths.The samples were mixed well, and a portion of the fresh soils and roots was used for the most-probable number (MPN) enumeration of FeRB population.Another portion was frozen at − 20°C until DNA extraction.

Quantification of Fe-reducing bacteria by most-probable number method
The viable number of cultivable FeRB in the soil and root samples was estimated by the MPN method.Approximately 5 g of fresh soil and root samples were dispersed with 45 mL of 0.5% MgCl 2 (primary solution) by vigorously shaking for 15 min, followed by serial dilutions of the primary solution in test tubes.The diluted soil solutions (10 2 -10 8 dilutions) were inoculated in a medium with three replications prepared in the test tubes or five replications in the wells of 96-well deep plate.One litter of the medium consisted of 20 g glucose, 5.0 g sodium acetate, 50 mg MgSO 4 ⋅7 H 2 O, 250 mg K 2 HPO 4 , 250 mg KH 2 PO 4 , 50 mg NaCl, 5.0 mg Na 2 MoO 4 ⋅2 H 2 O, 5.0 g CaCO 3 , 0.01 mg biotin, 0.01 thiamine hydrochloride, 0.01 mg p-aminobenzoic acid (Hammann and Ottow 1974).As a source of Fe, 1.0 g of ferrihydrite, which was prepared as described by Straub, Kappler, and Schink (2005), was added to the medium.9 mL or 1.5 mL of the medium were dispensed in the test tubes or the wells of plates.The test tubes were closed with butyl rubber and screw cap before incubation, but the gas phase was not replaced with N 2 gas.We preliminarily confirmed that the number of cultivable FeRB under the examined conditions was not different from the number when the gas phase was replaced with N 2 gas (data not shown).The 96-well deep plates were sealed with a gaspermeable sheet and incubated in a plastic container with AnaeroPack Kenki (Mitsubishi Gas Chemical, Tokyo, Japan), which contains an oxygen scavenger, to make anaerobic conditions.After 5 days of incubation at 28-30°C, the proliferation of FeRB was judged by confirming the formation of Fe(II) in the culture solution with the colorimetric method by 1,10-phenanthroline.

DNA extraction
DNA was extracted from the soil (0.5 g) and root (0.2 g) samples with FastDNA SPIN Kit for Soil (MP Biomedicals, Santa Ana, U.S. A.) to analyze the microbial communities.ZR Bashing Beads Lysis Tube (0.1-and 0.5-mm Beads, Zymo Research, Irvine, U.S. A.) was used instead of Lysing Matrix E tube in the kit.A beads tube containing a sample and lysis buffer of the kit was horizontally shaken for 1 min with a portable shaker, TerraLyzer (Zymo Research).The obtained crude DNA solution was purified following the kit's instructions.DNA was finally dissolved in 100 μL of TE buffer.Then 25 μL of DNA stable plus (Biomatrica, San Diego, U.S.A.) was added to the DNA samples to avoid the degradation of DNA during the transportation.DNA samples were further purified with OneStep PCR Inhibitor Removal Kit (Zymo Research) before PCR amplification.

Amplicon sequencing analysis of bacterial 16S rRNA genes
The composition and diversity of the bacterial community in the rhizosphere soil and root samples were analyzed with the Illumina MiSeq (Illumina, San Diego, U.  (ver. 2021(ver. .11, Bolyen et al. 2019) and the EzBioCloud 16S rRNA database (Yoon et al. 2017).In brief, the primer region of each read was removed with the cutadapt (Martin 2011), followed by filtering, denoising, merging of forward and reverse reads, and removing of chimera sequences with DADA2 (Callahan et al. 2016).The truncated position for the DADA2 analysis was 270 and 170 bp of forward and reverse reads, respectively, at which the first quartile and median values of quality score were 31.25 and 37 for forward reads and 34 and 38 for reverse reads, respectively.The obtained representative amplicon sequence variants (ASVs) were classified with a trained classifier based on the EzBioCloud 16S rRNA database, in which mitochondrial and chloroplast sequences retrieved from Silva 138 SSURef NR99 full-length sequences (Quast et al. 2013) were populated to the database (Robeson et al. 2021).Then, the ASVs assigned to mitochondria, chloroplast, and taxa other than Bacteria were filtered out, at which the frequency of ASVs in each sample was 8,073-22,365 with a median of 15,750.Since rarefaction analysis showed that the sequencing depth sufficiently covered the richness of each sample, the ASV numbers of each sample were normalized to the minimum number of ASVs (i.e., 8,073), followed by the calculation of the relative abundance of each taxon and α-and β-diversity indices.
The raw sequencing data of Miseq were deposited in the SRA database under the BioProject accession number PRJDB13364 with DRA013791.

Quantification of 16S rRNA genes by real-time PCR
The copy numbers of 16S rRNA genes of Gallionellaceae, Geobacteraceae, and Anaeromyxobacteraceae, Fe-oxidizers and Fe-reducers, in the bulk soil, rhizosphere soil, and rice root samples were quantified with Thermal Cycler Dice Real Time System II (TaKaRa, Kusatsu, Japan) and TB Green Premix Ex Taq (TaKaRa).The copy number of bacterial 16S rRNA genes was also quantified in the bulk and rhizosphere soil samples.The standard reference of Gallionellaceae and total bacteria was prepared from 16S rRNA gene amplicons of Ferrigenium kumadai An22 (Khalifa et al. 2018) and Escherichia coli K12, as described by Naruse et al. (2019) and Nakagawa et al. (2020).A 16S rRNA gene fragment as the standard reference of Geobacteraceae and Anaeromyxobacteraceae was obtained from a clone library which was derived from a Japanese paddy field soil (Matsuba Y. pers.comm.).The details of the reaction mixture and amplification programs for the quantitative PCR were described in Supplementary Table S4.

Statistical analysis
The richness (the number of observed ASVs), Chao1, Shannon, and evenness of bacterial community were estimated as α-diversity indices.Principal coordinate analysis (PCoA) was carried out, based on the unweighted and weighted UniFrac distances (Lozupone and Knight 2005).Linear discriminant analysis effect size (LEfSe) analysis (Segata et al. 2011) was carried out with the matrix data of relative abundance at the family level to identify distinct bacterial taxa in samples.Significant differences in the αdiversity indices, the logarithmic ratios of 16S rRNA gene copy numbers between Gallionellaceae, Geobacteraceae, and Anaeromyxobacteraceae, and the logarithmic MPN of FeRB among the samples were examined by the Mann-Whitney U test or the Kruskal-Wallis test with the Bonferroni correction.

Higher Fe(II) concentrations with poor soil fertility in BF paddy field
The soils of BF field contained more than 17 and 10 g kg −1 dry soil of Fe d (dithionite-etractable Fe) and Fe o (oxalate-extractable Fe), respectively, and the values of CEC were no greater than 11 cmol c kg −1 dry soil (Table 1).The percentages of cation saturation (<18%) and the amounts of available phosphate (<18 mg-P kg −1 ) were quite low, compared with those in the AN soils.Fe(II) aq concentrations in the soil solutions in the BF soils ranged from 36 to 132 mg l −1 with the averaged values of 73.2-107 mg l −1 , while those in the AN soils were very low (Supplementary Table S3).The Fe(II) aq concentrations in the BF soils were within the range of Fe(II) aq concentrations in the cluster III (20-600 mg l −1 , Becker and Asch 2005), although no significant difference was observed irrespective of the severity of Fe toxicity.The pH values of the air-dried soil were 4.8-5.7,and those in the fresh soils were 6.2-6.3 (Table 1).The pH values of the irrigation and surface water in the BF field were 6.0-6.4 and 5.7-6.4,and Fe(II) aq concentrations in the irrigation water were low (<2 mg l −1 ).That is, most Fe(II) was likely generated in the soil.These circumstances of the BF field were a typical case of cluster III and indicate a high risk of Fe toxicity of rice with a poor nutrient status of the soil (Becker and Asch 2005).The color of the BF rice root samples was dark brown, while that of the AN root samples was relatively light brown (Supplementary Fig. S2)

Trends of higher abundance of cultivable FeRB in the BF paddy field
The MPN enumeration of cultivable FeRB showed that approximately 10 3 -10 5 cells g −1 dry soil and 10 2 −10 6 cells g −1 dry roots were detected in the BF rhizosphere soil and BF rice roots, respectively.The abundance in the AN field was 10 3 -10 4 cells g −1 dry soil and 10 2 −10 4 cells g −1 dry roots with an exception for rhizosphere soil.Although no significant difference was observed between the BF and AN fields, trends of higher cell numbers were observed in the BF field irrespective of the bulk and rhizosphere soils and roots (Figure 1).The phylogenetic analysis of bacteria grown in the medium showed that Clostridium, Azotobacter, and Aeromonas predominated in the medium from the BF rhizosphere soil, while Clostridium and Paenibacillus were the predominant groups in the medium from the AN rhizosphere soil (data not shown).

Distinct bacterial community developed in the BF paddy field
Bacterial communities in the rhizosphere soil and rice roots were analyzed with the Illumina MiSeq.Rhizosphere soil harbored diverse bacterial communities as the α-diversity indices were higher in the rhizosphere soil than in the rice roots fields (Figure 2).In both the BF and AN fields, and both the rhizosphere soil and roots, Proteobacteria was the most predominant  phylum, which occupied more than 30% and 50% of the community in the rhizosphere soil and rice roots, respectively (Supplementary Fig. S3a).Especially in the BF roots, Betaproteobacteria was predominated, in which Comamonadaceae was the most abundant group.In the rhizosphere soil, Chloroflexi, Acidobacteria, and Actinobacteria were also major taxa.PCoA showed that the community structure of the bacterial communities was distinct between the rhizosphere soil and roots and between the BF and AN fields (Supplementary Figs.S3b, S3c).Although the rice variety and growth stage in the BF field were not fixed in the 3 years' investigation, the influence of these factors on the bacterial community structures was limited, compared with the site differences between the BF and AN fields.Noticeable differences were not observed in the community structure between the MS and NS samples in both the rhizosphere soil and roots of the BF field.
LEfSe analysis identified several bacterial taxa as a distinct taxon between the BF and AN rhizosphere soils and between the BF and AN rice roots (Supplementary Tables S5, S6), some of which include potential Fe reducers, e.g., Pseudomonadaceae, Burkholderiaceae, and Comamonadaceae.Among them, the members in Deltaproteobacteria and Firmicutes, which are the representative bacterial taxa in flooded paddy fields and contain wellknown Fe-and sulfate-reducers and fermenters, and Gallionellaceae in Betaproteobacteria, are of note to describe the features of the bacterial community in the Fe-toxic paddy field soil (Figure 3).
The relative abundance of several potential Fe-reducing bacterial groups including Anaeromyxobacteraceae, Geobacterceae, Clostridiaceae, Bacillaceae, and sulfate-reducing bacterial group was higher in the BF rhizosphere soil than those in the AN rhizosphere soil (Figure 3a).On the other hand, the relative abundance of Gallionellaceae was higher in the AN roots than in the BF roots (Figure 3b).These results showed that the FeRB members prevailed in the BF rhizosphere soil, while the Gallionella-related FeOB community did not predominate in the BF rice roots.However, notable features in the bacterial communities were not observed between the MS and NS samples in both the soil and roots (Supplementary Tables S7 and S8).

Predominance of the members of Geobacteraceae and Anaeromyxobacteraceae in the BF paddy field
The abundance of Gallionellaceae, Geobacteraceae, and Anaeromyxobacteraceae, Fe-oxidizers and Fe-reducers, and total bacteria were estimated by qPCR as representative indicator bacteria of the communities of Fe oxidizers and reducers (Supplementary Fig. S4).To compare the predominance of FeRB or FeOB among the samples by excluding the influence of DNA extraction efficiency as much as possible, the ratios of Gallionellaceae to Geobacteraceae (Figure 4a) and   Gallionellaceae to Anaeromyxobacteraceae (Figure 4b) were compared among the samples as an indicator of the redox status of the fields.
In both the BF and AN fields, all the ratios were significantly higher in the roots than in the rhizosphere soil, indicating that rice roots are preferable habitats for Gallionellarelated FeOB community.However, not only in the bulk and rhizosphere soils but also roots, the ratio of Gallionellaceae to Geobacteraceae in the BF field was significantly lower than those in the AN field (Figure 4a).The ratio of Gallionellaceae to Anaeromyxobacteraceae was also significantly lower in the bulk and rhizosphere soils of the BF field than those in the AN field (Figure 4b).In the BF field, significant differences were not observed between the MS and NS in all the cases.Significant positive correlations were observed between the copy numbers of 16S rRNA genes of Gallionellaceae, Geobacteraceae, and Anaeromyxobacteraceae in the BF soils, while no significant correlation was observed between those in the AN soils (Figure 5a).In the rice roots, a positive correlation was observed between the 16S rRNA gene copy numbers of Gallionellaceae and Geobacteraceae both in the BF and AN roots, while a significant positive correlation between Gallionellaceae and Anaeromyxobacteraceae was observed only in the AN roots (Figure 5b).

Discussion
Fe toxicity in rice is a serious problem in several countries (Becker and Asch 2005;Kirk et al. 2022;Sahrawat 2004), but the features of bacterial community in Fe-toxic paddy field has not yet been revealed.The present study investigated the bacterial community in a Fe-toxic paddy field in the Kou valley, Burkina Faso in 2017-2019, as an initiative case study, especially focusing on Gallionellaceae, Geobacteraceae, and Anaeromyxobacteraceae as the representative indicator bacteria of Fe oxidizers and reducers.The bacterial community in a paddy field in Japan (AN field) was investigated as an outgroup to characterize the community in the BF field.Although the field conditions, e.g., soil properties, rice varieties, and rice growth stages, which may affect the redox status in the soil and the range of radial oxygen loss (ROL) from rice roots, were not fixed in the BF field in the three years' investigations, the influences of these factors on the bacterial community were limited, compared with the bacterial community in the AN fields, which also included the samples at the tillering and flowering stages.In the BF field, Geobacter-and Anaeromyxobacter-related FeRB were prevailing over Gallionella-related FeOB.Other potential Fe reducers such as Clostridiaceae, Bacillaceae, and sulfate-reducing bacterial group also characterized the bacterial communities both in the BF soil and roots and tendency of higher viable number of FeRB in the BF field than in the AN field was observed.These findings suggested that distinct bacterial communities have been developed in the BF field, which are well corresponding to the circumstances of Fe-toxic soil: higher Fe(II) amounts in the soil.Masuda et al. (2017) quantified the copy number of 16S rRNA genes of Geobacteraceae, Anaeromyxobacteraceae, and total bacteria in eight paddy field soils in various locations in Japan.The ratios of Geobacteraceae and Anaeromyxobacteraceae to total bacteria in the eight paddy field soils (0.68-6.1 and 0.58-2.2%,respectively [n = 8]) were significantly lower than those in the BF bulk and rhizosphere soils obtained in the present study (average 1.8-41 and 5.6-62%, respectively [n = 12]; Mann-Whitney U test, p < 0.01).Their study also supported our observations regarding the relative predominance of Geobacteraceae and Anaeromyxobacteraceae in the investigated BF fields.
The present study is the first report that showed the predominance of various Fe reducers in the cluster III-type Fe-toxic paddy field.

Circumstances of the investigated field and emergence of Fe toxicity
The investigated BF field is located in an inland valley swamp area.Most of Fe(II) in Fe toxic soil is probably generated via Fe(III) reduction in the soil, but interflow of Fe(II) aq from groundwater via water catchment area, and inflow of Fe(III) oxides via surface water likely bring about the enrichment of Fe (Keïta 2015;WARDA 2002).Since our investigation was carried out in the wet season, we speculate that the soil transport from the surrounding area into paddy fields due to heavy rains, which contains a high amount of Fe(III) oxides and some carbon source, followed by Fe(III) reduction in the surface soil (plowed layer), is a probable factor to bring about further accumulation of Fe(II) in the soil.Since the CEC of the BF soil is much lower than that of the AN soil, the generated Fe(II) is kept as Fe(II) aq in the soil solution.These circumstances may bring about the severe Fe toxicity in rice in the BF field.
On the other hand, the concentrations of Fe(II) aq in the soil solution observed in the present study (the averaged values of 73.2-107 mg l −1 in the soil solution; Supplementary Table S3) were within the range of Fe toxicity emergence for the cluster III paddy field soils (20-600 mg Fe(II) l −1 (Becker and Asch 2005),; but not very high, compared with the critical concentrations for 'true Fe toxicity' (>300 mg l −1 ; Sahrawat 2004).Sahrawat (2004) mentioned that 'pseudo-Fe toxicity' (<300 mg l −1 ) is induced under the poor nutrient status of the soil.The pH of the fresh soil was not acidic, but the amounts of exchangeable cations and available phosphate were quite low in the BF field (Table 1).Otoidobiga et al. (2016) showed that fertilization increased the yield of both tolerant and susceptible rice cultivars in a pot experiment using paddy soil from the Kou valley.Therefore, not only Fe(II) aq concentrations but also the poor nutrient status of the soil are the main factor of the emergence of Fe toxicity in the BF field.
The degree of Fe toxicity symptoms (MS and NS) was depending on the plot and year; a severer disorder of Fe toxicity was observed in 2017 even at the tillering stage (Supplementary Fig. S1).Besides, the different degrees of toxicity symptoms were observed even under the same cultivation condition (same rice variety and same plot in 2019).These variations are not atypical cases in the cluster III paddy fields (Becker and Asch 2005) likely because of the local hydrological features.However, the bacterial community structure and the abundance of FeOB and FeRB were not different depending on the severity of Fe toxicity (Supplementary Tables S7 and S8).These results suggested that a potential risk of Fe toxicity is always present in the BF field.Not only the Fe(II) aq concentrations, poor nutrient status of soil, and microbial community, but also probably other factors, e.g., differences in the growth both during and after the nursery stage, roots damage by transplanting, and changes in Fe(II) aq concentrations and nutrient status at microsites and growth stage, might have affected the sensitivity of rice to Fe(II), resulting in the patchy emergence of Fe toxicity symptom.Further investigations are needed to reveal the conditions of emerging Fe toxicity.

FeRB community vs. FeOB community in the rice rhizosphere of investigated field
Even if Fe toxicity has emerged under the influences of several factors, the activity of FeRB and FeOB communities in the rice rhizosphere soil and rice roots must be an important factor for the emergence of Fe toxicity.The present study showed that the ratio of Gallionellaceae to Geobacteraceae was higher in the rice roots than those in the rhizosphere soil in both the BF and AN fields.However, the ratio in the BF rice roots was lower than that in the AN rice roots (Figure 4a).Amplicon sequencing analysis also showed that the relative abundance of Gallionellaceae in the BF rice roots was lower than that in the AN rice roots (Figure 3b).These results suggested that rice roots are preferable habitats for Gallionella-related FeOB community, but the redox conditions of the rice roots in the BF field were more reduced than those in the AN field, as the color of the BF rice roots was dark brown (Supplementary Fig. S2), indicating somewhat unfavorable for Gallionella-related FeOB.
Recently, Maisch et al. (2020) analyzed Fe plaque formation and Fe(III) reduction in the rhizosphere of rice.They showed that the ROL from rice roots, especially root tips, increased under light conditions, while the ROL largely decreased under dark conditions, suggesting the photosynthetic activity of rice affects the redox conditions of the rice rhizosphere.They further showed that up to 29% of Fe plaque was easily reduced to Fe(II) aq by a Geobacter sp.strain under anoxic conditions.In the present study, a positive correlation between the copy numbers of 16S rRNA genes of Gallionellaceae and Geobacteraceae was observed not only in the soils but also in rice roots (Figure 5), suggesting a close linkage of Fe reduction and oxidation in the vicinity of rice roots.
These findings enable us to speculate that Fe(II) oxidation always competes with Fe(III) reduction in the vicinity of rice roots, but the Fe(II) oxidization rates in the rhizosphere decrease in the night, at which microbial Fe(III) reduction actively occurs, followed by the remobilization of Fe plaque (i.e., increase in Fe(II) aq ).As a result, rice roots are exposed to a high concentration of Fe(II) aq , especially in the BF field, and rice absorbs an excessive amount of Fe(II) ions.The leaves are damaged, and the photosynthetic activity decreases, which brings about a further decrease of ROL from rice roots (i.e., falls in a vicious cycle).If rice plants receive serious damage at the initial growth stage, the growth of rice will not recover thereafter, resulting in significant losses (up to 100%) of rice yield.
Although a similar daily cycle must occur even in the AN field, the relative predominance of Gallionellaceae was observed in the AN roots.The CEC of the AN soil is higher than that of BF soil.The nutrient status also influences the oxidizing and/or excluding power of Fe(II) by rice roots (Kirk et al. 2022).Under these circumstances of rice plants, Fe(II) aq concentrations in the rhizosphere are kept low and the risk of Fe toxicity may be low.These speculations are just desk theory, but further investigations may reveal the detailed mechanisms of redox reactions of Fe in the rice rhizosphere and their relation to Fe toxicity of rice.

Conclusions and future perspective
In the present study, we characterized the bacterial community structures, especially focusing on Gallionella-related FeOB and Geobacter-and Anaeromyxobacter-related FeRB, in a paddy field of the Kou valley, Burkina Faso, where severe Fe toxicity of rice has emerged.This is the first study that showed the community structures of FeOB and FeRB and their interrelations in a Fe-toxic paddy field.In comparison with the community structure in a paddy field in Japan, anaerobic bacteria including Geobacteraceae and Anaeromyxobacteraceae were prevailing over Gallionellaceae in the BF field.The ratio of Gallionellaceae to Geobacteraceae at the rice roots of the BF field was higher than that in the bulk and rhizosphere soils, but the ratio was lower than that at the AN rice roots.Therefore, it is speculated that rice roots in the BF field were easily exposed to higher Fe(II) aq concentrations.The higher Fe(II) aq concentrations in the soil solution may have been attributed to not only the reduction of Fe(III) oxides and lower CEC in the soil but also the soil transport from the surrounding area, which contains a high amount of Fe(III) oxides, into the paddy field under heavy rain.To overcome these disadvantaged conditions, the selection of site-adapted rice varieties and improvement of the nutrient status of the soil are essential (Becker and Asch 2005;Kirk et al. 2022;Sahrawat 2004).Since Fe(II)-oxidizing activity at rice roots is an important factor as a primary defense system against Fe(II) in the soil solution, the ratio of Gallionellaceae to Geobacteraceae and Gallionellaceae to Anaeromyxobacteraceae may serve as an indicator of potential Fe(II)-oxidizing activity in the rice rhizosphere.
In the present study, only one site of the Fe-toxic paddy field in the cluster III was investigated.Large variations in the emergence of Fe toxicity are observed both at local and regional scales under complex interactions.Further studies focusing on the activity of FeOB and FeRB at rice roots under effective cultivation practices and in various types of Fe-toxic paddy fields will help to promote a better understanding of the Fetoxic soil circumstances and to establish sustainable rice cultivation in the Fe-toxic soils.

Figure 1 .
Figure 1.The number of cultivable Fe-reducing bacteria estimated by the most-probable number method.The numbers in parentheses indicate the number of samples analyzed.bars in the box plots show the minimum and maximum values except for outliers.The both ends and center line of box show the quartile and median values.

Figure 2 .
Figure 2. α-diversity indices of the bacterial community of the rhizosphere soil and rice roots in Burkina Faso (BF, n = 20) and Anjo (AN, n = 12).The in the box plots show the minimum and maximum values except for outliers.The both ends and center line of box show the quartile and median values.Different letters indicate a significant difference (p < 0.05) by the Kruskal-Wallis test with the Bonferroni correction.

Figure 3 .
Figure 3. Distinct members in Deltaproteobacteria, Firmicutes, Gallionellaceae in Betaproteobacteria between the bacterial communities of the rhizosphere soil and rice roots in the BF and an fields, identified by the linear discriminant analysis effect size (LEfSe) analysis (Kruskal-wallis test, p < 0.05; LDA score > 2.0).Closed and open bars indicate the distinct taxa identified from the BF and an fields, respectively.

Figure 4 .Figure 5 .
Figure 4.The ratio of 16S rRNA gene copies of (a) Gallionellaceae to Geobacteraceae and (b) Gallionellaceae to Anaeromyxobacteraceae in the bulk and rhizosphere soils and rice roots in BF (n = 18, 20 and 20) and an (n = 10, 12 and 12).The bars in the box plots show the minimum and maximum values except for outliers.The both ends and center line of box show the quartile and median values.Different letters indicate a significant difference (p < 0.05) by the Kruskal-Wallis test with the Bonferroni correction.

Table 1 .
Chemical properties of the soil samples obtained from investigated fields.BF and AN indicate the fields in Burkina Faso and Anjo, respectively.‡ The pH of fresh soil was 6.2-6.3 in the BF field.§ Fe d and Fe o indicate dithionite-extractable Fe (free Fe) and oxalate-extractable Fe (amorphous Fe).