Abstract

   Cadmium (Cd), a heavy metal, is negatively associated with plant
   growth. AMT (ammonium transporter) genes can confer Cd resistance and
   enhance nitrogen (N) uptake in soybeans. The potential of AMT genes to
   alleviate Cd toxicity by modulating rhizosphere microbiota remains
   unkonwn. Here, the rhizosphere microbial taxonomic and metabolic
   differences in three genotypes, i.e., double knockout and
   overexpression lines and wild type, were identified. The results showed
   that GmAMT2.1/2.2 genes could induce soybean to recruit beneficial
   microorganisms, such as Tumebacillus, Alicyclobacillus, and
   Penicillium, by altering metabolites. The bacterial, fungal, and
   cross-kingdom synthetic microbial communities (SynComs) formed by these
   microorganisms can help soybean resist Cd toxicity. The mechanisms by
   which SynComs help soybeans resist Cd stress include reducing Cd
   content, increasing ammonium (NH[4]^+-N) uptake and regulating specific
   functional genes in soybeans. Overall, this study provides valuable
   insights for the developing microbial formulations that enhance Cd
   resistance in sustainable agriculture.

   Subject terms: Applied microbiology, Microbial communities

Introduction

   Cadmium (Cd) is a ubiquitous heavy metal with a widely distributed and
   known toxicity^[36]1, and contamination of agricultural land by Cd is
   primarily caused by human activities, such as industrial wastewater
   discharge, disposal of large quantities of metal wastes and sewage
   sludge, and pesticide misuse^[37]2. Cd is more mobile than other heavy
   metals in plants, leading to its increased accumulation and subsequent
   induction of severe toxicity^[38]3,[39]4. This is caused by cation
   deficiency, inhibition of the biosynthesis of chlorophyll or other
   essential substances, and increased oxidative damage, such as
   structural and functional cellular degeneration and destruction of
   biomolecules, which are closely linked to plant growth and
   development^[40]5.

   The composition and functional activities of the plant rhizosphere
   microbiome are dynamically regulated under various environmental
   stresses, providing a critical foundation for plant adaptation and
   health^[41]6–[42]9. Interestingly, some microorganisms can reduce the
   toxicity of Cd to plants and themselves by adsorbing, complexing,
   enzymatically converting, and redox-converting Cd ions^[43]10,[44]11.
   For instance, Klebsiella mobilis CIAM 880 has the ability to form
   complexes with unbound Cd ions, which reduces their bioavailability of
   Cd to barley plants and mitigates the toxic effects^[45]12.
   Enterobacter bugandensis can also reduce Cd accumulation in wheat
   grains by bioprecipitation and extracellular adsorption^[46]9,[47]13.

   Plant functional genes, including those involved in biotic and abiotic
   stress response and nutrient uptake and transport, could modulate the
   secretion of root exudates and some ionic signals, which significantly
   influence the rhizosphere microorganisms and are essential for
   enhancing plant stress tolerance and nutrient use
   efficiency^[48]14–[49]17. Conversely, the activity of rhizosphere
   microorganisms can induce systemic tolerance in plants by releasing
   metabolites, which can affect host genes expression, and alter
   phytohormone secretion^[50]17,[51]18. For example, Pseudomonas
   aeruginosa and Burkholderia gladioli are able to reduce Cd uptake and
   expression of metal transporter genes in tomatoes plants, leading to
   improved growth and photosynthetic pigmentation^[52]19. NRT1.1B,
   encoding a nitrate transporter in rice, is involved in the attraction
   of bacteria that are enriched in O. sativa subsp. indica rice. These
   bacteria can improve rice growth under organic nitrogen (N) conditions
   using the synthetic community^[53]20. Our recent work has shown that
   loss of sst gene function in rice can affect rhizosphere microbes by
   altering plant metabolites, such as salicin and arbutin, which in turn
   can help the host to resist salt stress^[54]21.

   Ammonium transporters (AMTs), involved in ammonium (NH[4]^+-N) uptake
   and transport in most plants, promote the uptake of N sources by
   plants^[55]22,[56]23. Previous studies have investigated the role of
   AMTs in promoting NH[4]^+-N transport to alleviate salt stress^[57]24.
   A growing body of evidence suggests that AMTs-targeted NH[4]^+-N is
   more sensitive to Cd concentration than NO[3]^--N. For example,
   NH[4]^+-N can inhibit Cd translocation from roots to shoots, thereby
   protecting Arabidopsis thaliana from Cdtoxicity^[58]25,[59]26. Recent
   research^[60]27 has also found in Solanum nigrum L. that the
   transcription level of Cd transport-related genes is regulated by
   NH[4]^+-N signaling to prevent the Cd accumulation and flux, compared
   to NO[3]^--N. The researchers also suggest that NH[4]^+-N can fix Cd in
   the cell wall component, thereby improving Cd tolerance. However, the
   potential of GmAMT2.1/2.2 genes to alleviate external Cd toxicity by
   regulating metabolites and microbiota in the soybean rhizosphere
   remains to be determined.

   Synthetic communities (SynComs) have been increasingly used to study
   the interactions between microbes and their hosts in recent
   years^[61]28,[62]29. The development of high-throughput microbial
   isolation techniques has enabled researchers to efficiently isolate
   more microbial strains^[63]30. The isolated strains were then compared
   with the sequencing results to select suitable candidates for SynComs
   construction. Despite the fact that research on SynComs has often
   focused on bacteria, fungi also play a very important role in helping
   plants resist heavy metal stress^[64]29. In addition, fungi can
   establish symbiotic networks with bacteria^[65]31, which may be more
   effective in helping plants resist stresses. Therefore, it is necessary
   to evaluate the role of different synthetic communities in helping
   soybeans resist Cd toxicity.

   In this study, we hypothesized that the GmAMT2.1/2.2 genes enhance
   NH[4]^+-N uptake in soybeans, and help them to resist Cd toxicity by
   influencing metabolites to recruit beneficial rhizosphere
   microorganisms. To test these hypotheses, high-throughput sequencing,
   and liquid chromatography-mass spectrometry (LC-MS) assays were used to
   identify taxonomic and metabolic differences and their correlations in
   the rhizosphere. In addition, we isolated and identified bacteria and
   fungi in the rhizosphere of different genotypes and constructed
   bacterial, fungal, and bacterial-fungal cross-kingdom SynComs. Finally,
   we verified that these SynComs influence the expression of heavy metal
   tolerance-related genes in soybean roots. The objectives of this study
   were to: (1) investigate how the GmAMT2.1/2.2 genes influence
   rhizosphere microbes by regulating metabolites and N patterns, which in
   turn help soybean to resist Cd toxicity; and (2) determine which SynCom
   consisting of recruited microorganisms is most effective in alleviating
   Cd toxicity in soybean, and to elucidate the corresponding
   physiological and molecular mechanisms.

Result

GmAMT2.1/2.2 are responsible for the Cd toxicity alleviation through
affecting N patterns in soybean

   To identify potential candidate AMTs in response to the Cd toxicity
   alleviation, we examined the transcriptome data of AMT homologous genes
   in the CK and Cd treatment. Among these, two homologous genes
   GmAMT2.1/2.2, which are predominantly expressed in roots (Supplementary
   Fig. [66]1), are significantly up regulated by Cd treatment (Fig.
   [67]1a). To further explore the transcriptional response of
   GmAMT2.1/2.2 to Cd treatment, we analyzed the expression levels of
   GmAMT2.1/2.2 using the soybean plants treated with Cd at different
   concentrations. The expression levels of GmAMT2.1/2.2 exhibited an
   upward trend with increasing Cd concentration and time (except for a
   decrease in GmAMT2.1 expression after 6 h of treatment and a decrease
   in GmAMT2.2 expression after 12 h of treatment) (Fig. [68]1b–e). In
   addition, we generated four lines to confirm the identification and
   function of GmAMT2.1/2.2, including two stable double knockout lines
   (GmAMT2.1/2.2) and two overexpression lines (OXAMT2.2). Mu1 deleted
   5 bp (AGCAT) in sgRNA1 of GmAMT2.1, and deleted 7 bp (CAATGGG) in
   sgRNA2 of GmAMT2.2. Mu2 inserted 1 bp (T) in sgRNA1 of both GmAMT2.1
   and GmAMT2.2, which ultimately lead to Mu1 and Mu2 are loss-of-function
   mutants. In the two overexpressed lines (OX1 and OX2), the expression
   of GmAMT2.2 was up-regulated 1.87- and 2.07-fold, respectively. (Fig.
   [69]1f; Supplementary Figs. [70]1, [71]2).

Fig. 1. GmAMT2.1 and GmAMT2.1 are responsible for alleviating Cd toxicity in
soybean by modulating N patterns.

   [72]Fig. 1
   [73]Open in a new tab

   a Heatmap of the expression of the AMT gene family in soybean under Cd
   toxicity. Expression pattern analysis of GmAMT2.1 (b and c) and
   GmAMT2.2 (d and e). f CRISPR/Cas9 mediated gene editing of GmAMT2.1 and
   GmAMT2.2 in soybean. g The growth performance of Mu, WT and OX lines
   under Cd toxicity. The root fresh weight (h), shoot fresh weight (i),
   root N content (j), shoot N content (k), root NH[4]^+-N content (l),
   shoot NH[4]^+-N content (m), and Cd content (n) of Mu, WT and OX lines
   under CK and Cd conditions. o translocation factor of Mu, WT and OX
   lines under Cd toxicity. Different letters represent significant
   differences (one-way ANOVA, n = 6, P < 0.05). The error bars on the
   columns represent the standard errors (SE). Mu: GmAMT2.1/2.2 double
   knockout lines; WT: wild type; OX: GmAMT2.2 overexpression lines; Cd:
   cadmium toxicity.

   When exposed to identical Cd concentrations, the knockout lines
   consistently displayed a more pronounced Cd sensitivity compared to the
   wild-type (WT) lines in terms of plant growth. In contrast, the
   overexpressed lines did not show significant differences compared to
   the WT (Fig. [74]1g). Cd treatment resulted in a significant decrease
   in the fresh weight of all four lines. Furthermore, an increase in N
   content, especially NH[4]^+-N, was observed in response to Cd toxicity.
   This upward trend in N content progression was also observed upon
   overexpression of GmAMT2.2 as well, whereas an opposite relationship
   was observed in the mutants (Fig. [75]1h–o). Taken together, these
   findings suggest that NH[4]^+-N may play a role in alleviating Cd
   toxicity and that the GmAMT2.1/2.2 genes may inhibit Cd transport and
   accumulation by increasing NH[4]^+-N levels, thereby mitigating the
   detrimental effects of Cd toxicity on soybean plants.

Modification of rhizosphere microbiota by GmAMT2.1/2.2 under Cd toxicity

   From 2,236,825 high-quality bacterial 16 S rRNA reads and 2,078,640
   fungal ITS reads, we identified a total of 11,980 bacterial OTUs and
   1,702 fungal OTUs. Cd toxicity had no significant effect (P > 0.05) on
   the alpha diversity of rhizobacterial communities, but increased the
   alpha diversity of fungal communities in both WT and OX soybean
   genotypes. Under no-Cd toxicity conditions, the Mu genotype had a
   higher Shannon index than the other two genotypes, while Cd toxicity
   did not result in a significant difference between the three genotypes
   (Fig. [76]2a, d). In terms of beta diversity, Cd toxicity and the
   interaction of Cd toxicity and genotype significantly altered the
   structure of microbial communities (Fig. [77]2b, e; Supplementary Table
   [78]1). Furthermore, WT, OX, and Mu genotypes generally harbored
   significantly different microbial communities under both Cd and no-Cd
   toxicity conditions (PERMANOVA, pairwise comparison, n = 6, P < 0.05)
   (Fig. [79]2c, d; Supplementary Table [80]1), suggesting that the
   GmAMT2.1/2.2 gene activity affects the assembly of the soybean
   rhizosphere microbiome.

Fig. 2. Effect of Cd toxicity on different genotypes of soybean rhizosphere
microorganisms.

   [81]Fig. 2
   [82]Open in a new tab

   Shannon diversity index of (a) bacterial and (d) fungal communities
   (one-way ANOVA, n = 6, P < 0.05). The Bray-Curtis’s similarity-based
   Principal Coordinate Analysis (PCoA) showed dissimilarities in the
   composition of (b) bacterial and (e) fungal communities (PERMANOVA,
   n = 6, P < 0.05). The relative abundances of the (c) bacterial and (f)
   fungal phyla. Mu: GmAMT2.1/2.2 double knockout lines; WT: wild type;
   OX: GmAMT2.2 overexpression lines; Cd: cadmium toxicity.

   The rhizosphere soil of the three genotypes exhibited comparable
   community composition under Cd and no-Cd toxicity conditions (Fig.
   [83]2e, f). Briefly, the most prevalent bacterial phyla were
   Acidobacteria, Proteobacteria, Actinobacteria and Chloroflexi with the
   relative abundances ranging from 21.56% to 22.65%, 7.42% to 7.98%,
   5.09% to 5.55% and 4.98% to 5.40%, respectively. The fungal communities
   were mainly Ascomycota, Basidiomycota and Chytridiomycota, with the
   relative abundances varying from 62.08% to 70.27%, 12.71% to 25.92% and
   0.73% to 4.26%, respectively. In general, five bacterial phyla (i.e.,
   Proteobacteria, Actinobacteria, Myxococcota, Firmicutes and
   Gemmatimonadetes) and five fungal phyla (i.e., Ascomycota,
   Basidiomycota, Chytridiomycota, Mortierellomycota and Kickxellomycota)
   were significantly affected by Cd toxicity. One bacterial phylum (i.e.,
   Verrucomicrobiota) and one fungal phylum (i.e., Basidiomycot) were
   influenced by genotypes. Moreover, three bacterial (i.e.,
   Acidobacteriota, Actinobacteriota and Verrucomicrobiota) and three
   fungal phyla (i.e., Ascomycota, Basidiomycot and Glomeromycota) were
   altered by the interactions between genotype and Cd toxicity
   (Supplementary Table [84]3).

   All genotypes were observed to be significantly affected by Cd toxicity
   in terms of the co-occurrence network structure, but the responses
   varied (Fig. [85]3; Supplementary Table [86]3). Specifically, Cd
   toxicity increased network complexity, as evidenced by an increase in
   both the clustering coefficient and the number of edges. Mu genotype
   showed the highest number of edges and clustering coefficient under Cd
   toxicity, followed by WT and OX genotypes. These results underline the
   impact of Cd toxicity on microbial community dynamics and the
   genotype-specific mechanisms that shape community composition under Cd
   toxicity.

Fig. 3. Effect of Cd toxicity on the network structure of rhizosphere
microorganisms in different genotypes of soybean.

   [87]Fig. 3
   [88]Open in a new tab

   The structure of co-occurrence network of rhizosphere microbial
   community for (a) Mu, (b) WT, (c) OX, (d) CdMu, (e) CdWT and (f) CdOX.
   The nodes are colored based on the bacterial and fungal phyla. The
   edges between nodes indicate significant positive (red) or negative
   (green) correlation (Spearman’s r > 0.8 or r < -0.8, P < 0.01). Mu:
   GmAMT2.1/2.2 double knockout lines; WT: wild type; OX: GmAMT2.2
   overexpression lines; Cd: cadmium toxicity.

Enriched potential beneficial microorganisms by GmAMT2.1/2.2 under Cd
toxicity

   We investigated the effect of Cd toxicity on the relative abundance of
   microbes in three soybean genotypes (Figs. [89]4, [90]5). Cd toxicity
   increased the number of microbial OTUs in Mu, WT and OX by 306, 256,
   and 280 (bacteria), and 77, 71, and 94 (fungi), respectively (Figs.
   [91]4b, [92]5b). The enriched bacterial OTUs were mainly belonged to
   Proteobacteria, Firmicutes, Chloroflexi, and Verrucomicrobiota, while
   the enriched fungal OTUs were mainly associated with Ascomycota and
   Rozellomycota. Venn analysis revealed that WT and OX genotypes carrying
   the AMT genes showed a specific enrichment of 274 bacterial and 96
   fungal OTUs under Cd toxicity. Based on the information of the
   isolation and identification of culturable microbes, six bacterial OTUs
   including OTU759 (Tumebacillus), OTU215 (Ralstonia), OTU1135
   (Alicyclobacillus), OTU57 (Burkholderia), OTU419 (Paenibacillus),
   OTU5695 (Methylophilus), and six fungal OTUs including OTU76
   (Aspergillus), OTU99 (Talaromyces), OTU144 (Penicillium) and OTU168
   (Cladosporium) were identified and selected to show their relative
   abundance in different treatments (Figs. [93]4c, [94]5c).

Fig. 4. Effects of Cd toxicity on rhizosphere bacterial OTUs of different
soybean genotypes.

   [95]Fig. 4
   [96]Open in a new tab

   Volcano plots depicting enrichment and depletion bacterial OTUs in (a)
   Mu, WT and OX plants under Cd toxicity compared to no Cd toxicity
   conditions. b Venn analysis showing the unique and shared enriched OTUs
   between genotypes and Cd toxicity. c Relative abundance of six
   bacterial OTUs in different treatments that significant enriched in the
   rhizosphere. Mu: GmAMT2.1/2.2 double knockout lines; WT: wild type; OX:
   GmAMT2.2 overexpression lines; Cd: cadmium toxicity.

Fig. 5. Effects of Cd toxicity on rhizosphere fungal OTUs of different
soybean genotypes.

   [97]Fig. 5
   [98]Open in a new tab

   Volcano plots depicting enrichment and depletion bacterial OTUs in (a)
   Mu, WT and OX plants under Cd toxicity compared to no Cd toxicity
   conditions. b Venn analysis showing the unique and shared enriched OTUs
   between genotypes and Cd toxicity. c Relative abundance of four fungal
   OTUs in different treatments that significant enriched in the
   rhizosphere. Mu: GmAMT2.1/2.2 double knockout lines; WT: wild type; OX:
   GmAMT2.2 overexpression lines; Cd: cadmium toxicity.

Response of soil metabolites to GmAMT2.1/2.2 under Cd toxicity

   To explore the complex metabolic alterations in the rhizosphere soil
   under different treatments and genotypes, we performed LC-MS analysis
   on soil metabolites of Mu, WT, and OX with and without Cd exposure. A
   total of 798 distinct peaks with defined identities were detected
   across the three genotypes. Orthogonal partial least squares
   discriminant analysis (OPLS-DA) illustrated a clear separation between
   the treatments (Supplementary Fig. [99]4; Supplementary Table [100]4).
   Using a combination of filtering procedures and a threshold of variable
   importance in projection (VIP) > 1.0 and P < 0.05, we identified 35
   Cd-induced metabolites that increased and 30 that decreased in OX
   and/or WT, respectively (Supplementary Table [101]5). KEGG pathway
   enrichment analysis revealed differential enrichment and depletion of
   metabolites in secondary metabolism biosynthesis, including purine
   metabolism, arginine biosynthesis, galactose metabolism, stilbenoid,
   flavonoid biosynthesis, tryptophan metabolism, isoflavonoid
   biosynthesis, and phenylpropanoid biosynthesis (Fig. [102]6a). Notably,
   the levels of genistein, chrysin, piceatannol, glycitein, daidzein,
   daidzin, and coumestrol increased, while xanthurenic acid, sinapoyl
   alcohol, L-glutamic acid, guanine, 2-deoxyguanosine and sucrose
   decreased in WT and/or OX plants under Cd exposure (Fig. [103]6a).
   Correlation analysis between metabolites and microbial species revealed
   a positive association between sinapyl Alcohol, daidzein, and glycitein
   with Alicyclobacillus, Tumebacillus, Ralstonia, and Methylophilus in Mu
   and WT genotypes. However, a significant increase in the levels of
   these metabolites in the OX genotypewas positively correlated with the
   abundance of different microbial species (Fig. [104]6b). These findings
   suggest that GmAMT2.1/2.2 may play a role in regulating metabolic
   pathways associated with microbial community assembly in the
   rhizosphere.

Fig. 6. Effect of Cd toxicity on metabolites and metabolite-microbe
correlations.

   [105]Fig. 6
   [106]Open in a new tab

   a The process involves screening for metabolic pathway maps related to
   crucial differentially expressed metabolites. The red color indicates
   the metabolites that were increased in plants with GmAMT2.1/2.2 gene
   under Cd toxicity, while the green color indicates the metabolites that
   were decreased. b Effect of Cd toxicity on metabolite-microbe
   correlations in the rhizosphere soil of Mu, WT and OX. Mu: GmAMT2.1/2.2
   double knockout lines; WT: wild type; OX: GmAMT2.2 overexpression
   lines; Cd: cadmium toxicity.

Effect of different SynComs on Cd tolerance in soybean

   Using a high-throughput cultivation method, we sequenced and identified
   403 bacterial and 268 fungal colonies from the rhizosphere compartment
   across different treatments. Among them, six bacterial and eight fungal
   strains, identical to the microorganisms enriched in the soybean
   genotypes carrying the AMT genes based on the sequencing results, were
   used to construct three SynComs (bacterial, fungal, and cross-kingdom
   SynComs) (Supplementary Table [107]6). The efficacy of these SynComs in
   enhancing soybean Cd toxicity tolerance was evaluated (Fig. [108]7).
   The results showed that SynComs-treated plants had significantly higher
   shoot and root fresh weight, NH[4]^+-N content, and translocation
   factor, with reduced Cd content in shoot and root compared to
   heat-killed control plants (Fig. [109]7a–d). Additionally,
   cross-kingdom and fungal SynComs had more pronounced impact on the
   indicators mentioned above compared to bacterial SynCom.

Fig. 7. Effect of three SynComs on soybean gene transcription and their
regulatory effects.

   [110]Fig. 7
   [111]Open in a new tab

   Effect of different SynComs (SynComs1: bacterial SynComs; SynComs2:
   fungal SynComs; SynComs3: cross-kingdom SynComs) on (a) soybean plant
   growth (b) fresh weight, (c) Cd content in soybean root, (d) NH[4]^+-N
   content and (e) translocation factor under Cd toxicity. Soybean plants
   were grown in sterile soil and inoculated with SynComs or heat-killed
   SynComs. Soybean plants were grown in sterile soil, inoculated with
   SynComs or heat-killed SynComs. f Principal component analysis (PCA)
   showed distinctions in gene expression of Mu genotype plants between
   the three SynComs and heat-killed control treatments. g Heatmap with
   hierarchical clustering analysis showing the differential DEGs between
   the three SynComs and heat-killed control treatments. KEGG analysis of
   DEGs in response to (h) bacterial, (i) fungal, and (j) cross-kingdom
   SynComs inoculation under Cd toxicity. Dot size and color indicate the
   number of DEGs and p-value, respectively. The error bars on the columns
   represent the standard errors (SE).

   RNA-Seq analysis was conducted to elucidate the molecular mechanisms
   underlying the SynCom-mediated enhancement of soybean growth under Cd
   toxicity. PCA showed distinct separation of the four treatments
   (P < 0.05) (Fig. [112]7f). SynComs application induced differential
   expression of numerous genes in soybean root tissues (Fig. [113]7g).
   Notably, all three SynComs significantly upregulated the expression of
   genes involved in cytochrome P450 (Glyma.08G350800), endopeptidase
   inhibitor activity (Glyma.16G211700), metabolic process related
   (Glyma.07G007000, Glyma.01G205900 and Glyma.02G104600), zinc finger
   protein (Glyma.04G044900), messenger RNA biogenesis (Glyma.12G093100),
   ethylene response factor (Glyma.12G117000) and BURP domain family
   member (Glyma.12G217300). Bacterial SynCom specifically upregulated
   genes related to cell redox homeostasis and casparian strip membrane
   domain proteins (CASP) family member, while fungal SynCom exclusively
   upregulated genes involved in protein phosphorylation. KEGG enrichment
   analysis was further performed to explore the distinct metabolic
   pathways modulated by the three SynComs in alleviating Cd toxicity
   (Fig. [114]7h–j). The results showed that the three SynComs mainly
   influenced metabolic processes associated with amino acid metabolism,
   biosynthesis of other secondary metabolites, carbohydrate metabolism
   and signal transduction. Among these, phenylpropanoid biosynthesis
   exhibited the highest enrichment in both bacterial and fungal SynComs,
   while plant hormone signal transduction showed the highest enrichment
   in cross-kingdom SynCom. However, certain pathways demonstrated
   specific enrichment under different SynCom treatments. For instance,
   the toll-like receptor signaling pathway was specially enriched in
   bacterial SynCom treatment and linked to plant disease resistance and
   immune response. Flavonoid biosynthesis, ascorbate and aldarate
   metabolism and glucosinolate biosynthesis were exclusively associated
   with fungal SynCom, while plant-pathogen interaction and mineral
   absorption were only associated with cross-kingdom SynCom.

Discussion

   Plant functional genes can actively shape the microbial community
   structure of the rhizosphere by regulating root metabolites, which
   improving the nutrient uptake efficiency and environmental stress
   resistance of crops^[115]14,[116]15,[117]29. AMTs regulate the uptake
   and transport of NH[4]^+-N, thereby promoting plant uptake of N
   sources^[118]24. NH[4]^+-N produces various N-containing
   macromolecules, which regulate Cd uptake and translocation by plants,
   and alter the soybean root exudate and rhizosphere
   environment^[119]32,[120]33. In this study, we used three soybean
   genotypes with normal expression (WT), overexpression (OX) and knockout
   (Mu) of the GmAMT2.1/2.2 genes, respectively, to demonstrate the key
   role of the GmAMT2.1/2.2 genes in influencing NH[4]^+-N uptake, Cd
   levels, and soil metabolites, resulting in significant shifts in
   rhizosphere microbial community structure (Fig. [121]8).

Fig. 8. A schematic model summarizing the hypothesized mechanisms of how
GmAMT2.1/2.2 confers soybean Cd resistance and plant growth.

   [122]Fig. 8
   [123]Open in a new tab

   Soybean under Cd toxicity induces the GmAMT2.1/2.2, leading to the
   recruitment of beneficial microorganisms, such as Tumebacillus,
   Alicyclobacillus, Methylophilus, Aspergillus, Talaromyces, and
   Penicillium, by altering metabolites such as sinapyl alcohol,
   genistein, coumestrol, and piceatannol. SynComs composed of these
   specific microorganisms enhance soybean resistance to Cd toxicity by
   collectively influencing various molecular mechanisms, including
   ascorbate and aldarate metabolism, glucosinolate biosynthesis,
   plant-pathogen interactions, and mineral absorption pathways, with
   cross-kingdom and fungal SynComs demonstrating significant synergistic
   effects.

   Plant materials carrying the GmAMT2.1/2.2 genes have been found to
   attract a diverse array of potentially beneficial microorganisms,
   including bacterial species such as Tumebacillus, Alicyclobacillus,
   Paenibacillus, and Methylphilus, and fungal species such as
   Aspergillus, Penicillium, Cladosporium, and Talaromyces (Figs. [124]4,
   [125]5). The enriched microbial species with relevant function may play
   an essential role in enhancing plant resilience to Cd toxicity. For
   instance, Tumebacillus and Alicyclobacillus are soil organic matter
   (SOM) decomposers that can mobilize scarce nutrients, thereby improving
   plant adaptation to Cd toxicity^[126]34–[127]36. Paenibacillus and
   Methylphilus have been demonstrated to promote plant N
   uptake^[128]37,[129]38. Talaromyces sp., a plant growth-promoting
   fungus, releases terpenoid-like volatiles that stimulate plant
   growth^[130]39. Certain Ralstonia species, such as Ralstonia eutropha
   Q2-8 and Ralstonia Bcul-1, are considered heavy metal-resistant strains
   and can enhance plant Cd tolerance through different mechanisms,
   including modulating heavy metal resistance gene expression,
   strengthening cell wall components, or sequestering Cd into root cell
   vacuoles^[131]40,[132]41. Aspergillus sp. activated genes encoding
   glutathione (GSH) under Cd toxicity, alleviating the detrimental
   effects of Cd^[133]42. Moreover, Paenibacillus, Penicillium and
   Cladosporium are known to directly absorb and immobilize soil Cd by
   exopolysaccharide production, formation of stable phosphate
   precipitates and oxidation of Mn (II) to biogenic manganese oxides
   (BMOs), respectively^[134]43–[135]45. Thus, these rhizosphere
   microorganisms with diverse functionalities can synergistically enhance
   plant Cd tolerance through various mechanisms, including nutrient
   provision, gene expression regulation, chemical transformation, and
   direct Cd absorption/immobilization. They formed a complex functional
   network of rhizosphere microorganisms that promote healthy plant growth
   under Cd toxicity.

   Microbial community composition and metabolite profiles exhibited
   significant variations among soybean varieties. In particular, the
   abundances of Paenibacillus, Tumebacillus, Methylophilus, Aspergillus,
   Penicillium and Talaromyces were positively correlated with sinapyl
   alcohol and coumestrol levels. Sinapyl alcohol serves as a precursor of
   coumestrol, a type of coumarin. Coumarin exudation, regulated by MYB72,
   have been shown to shape the root microbiome, promoting the health of
   Arabidopsis thaliana^[136]18. Therefore, our findings suggest that
   GmAMT2.1/2.2 may enhance soybean adaptation to Cd toxicity by
   modulating the root microbiome through coumarin synthesis. GmAMT2.1/2.2
   could influence soil N content and alter the synthesis of flavonoids,
   including genistein, daidzin and daidzein (Fig. [137]6). These
   flavonoids may influence the symbiotic relationship between rhizobia
   and legumes, enriching beneficial microorganisms such as Cladosporium
   and Aspergillus, and further modulate plant defense
   responses^[138]46,[139]47. In addition, low concentrations of the
   flavonoid chrysin can promote spore germination and root colonization
   by arbuscular mycorrhizal fungi, enhancing the rhizobia-plant symbiosis
   and potentially increasing soybean tolerance to Cd^[140]48. Moreover,
   Paenibacillus, Ralstonia and Penicillium were significantly correlated
   with piceatannol, a compound known to enhance plant resistance to
   biotic and abiotic stresses. This indicates that piceatannol may also
   selectively recruit beneficial microbes. Although a direct link between
   piceatannol and specific microbes has not yet been established, this
   could be a direction for future research.

   All three SynComs alleviated Cd toxicity, with fungal SynCom and
   cross-kindom SynCom exhibiting greater efficacy than the bacterial
   SynCom (Fig. [141]7). This suggests that GmAMT2.1/2.2 genes-regulated
   microbiome plays a crucial role in enhancing soybean resistance Cd
   toxicity, and fungi may hold greater potential than bacteria in
   assisting soybeans to adapt to Cd toxicity. Fungi mitigation of Cd
   toxicity involves diverse mechanisms, including metal biosorption onto
   the cell wall, intracellular accumulation and sequestration, and
   deposition of metal compounds within and around the mycelium^[142]49.
   Additionally, unlike bacteria, fungi can form extensive mycelial
   networks, facilitating bidirectional nutrient exchange with
   plants^[143]50. Additionally, the fungal SynCom may promote soybean
   adaptation Cd toxicity through the induction of genes involved in
   flavonoid and glucosinolate biosynthesis and ascorbic acid metabolism
   in the soybean root, which was not observed with the bacterial SynCom.
   These substances contribute to Cd toxicity mitigation by enriching
   certain beneficial microorganisms, increasing sulfur availability and
   scavenging excess H[2]O[2] from plant cells^[144]51–[145]53. Of note, a
   recent study by Xie et al. ^[146]54 showed that microorganisms might
   regulate the immune system responsible for Cd resistance in host
   plants^[147]54. This aligns with our finding that the bacterial SynCom
   specifically enriched for toll-like receptor signalling pathway, which
   served as a sensor for pathogens and enhances the immune response of
   plants^[148]55,[149]56.

   Furthermore, DEGs significantly upregulated by SynComs were involved in
   plant defense signalling, abiotic stress resistance and plant hormone
   signalling. It has been reported that lipoxygenases 1 (LOX1) and ZAT10
   were directly related to Cd toxicity response. When Arabidopsis
   thaliana was exposed to Cd toxicity, LOX1 was involved in Cd-induced
   signalling transduction and increased jasmonic acid (JA)
   biosynthesis^[150]57,[151]58. Similarly, ZAT10, a member of the
   C[2]H[2]zinc finger gene family, negatively regulated Cd uptake in
   Arabidopsis thaliana and enhanced Cd detoxification, and positively
   regulating heavy metal detoxification-related genes such as NAS1, IRT2
   and MTP3^[152]59. RD22, containing the plant-specific BURP domain, was
   also highly up-regulated under SynCom application and was involved in
   several abiotic resistances, including drought, salt and osmotic
   stresses^[153]60,[154]61. CYP93D1, belonging to the cytochrome P450
   monooxygenases family, is involved in JA metabolism and can be
   activated by ZmCLA4 to affect lignin biosynthesis^[155]62. Cd toxicity
   stimulates lignin biosynthesis, and the upregulation of CYP93D1
   increases lignin content, which enhances Cd adsorption and promotes
   plant growth^[156]63,[157]64. Two other significantly up-regulated DEGs
   under SynComs application are related to plant defense mechanisms.
   UGT73B3 is involved in regulating redox status and ROS reactivity,
   while ERF9 serves as a negative regulator of DREB AND EAR MOTIF PROTEIN
   1 (DEAR1)-dependent ethylene/JA-mediated plant defense
   mechanisms^[158]65,[159]66. Furthermore, phenylpropanoid biosynthesis
   was the key enriched pathway under SynComs treatments, indicating that
   SynComs activate the phenylpropanoid biosynthetic pathway, synthesizing
   lignin from the accumulated phenolic compounds to overcome heavy
   metal-induced stress^[160]67. In summary, SymComs application enables
   the coordinated regulation of LOX1, ZAT10, RD22, CYP93D1, UGT73B3 and
   ERF9 expression, triggering an interconnected network of defense
   responses, plant hormone signalling pathways, lignin biosynthesis and
   heavy metal detoxification mechanisms to enhance soybean’s ability to
   mitigate Cd toxicity.

   It is noteworthy that the soil metabolites in this study are not the
   same as those from root exudates. Although the study used a single soil
   type, theoretically, the majority of the differences in soil
   metabolites are likely due to soybean genotypes. However, we cannot
   ignore the potential for soil microorganisms to be influenced by root
   exudates under GmAMT2.1/2.2 genetic regulation, which could lead to the
   up-regulation or down-regulation of extracellular metabolites. Several
   metabolites from root exudates have been shown to alter soil microbial
   community composition^[161]68, suggesting that changes in the soil
   metabolite profile may partly due to extracellular compounds readily
   released by microorganisms. Thus, future studies should dedicate
   greater effort to distinguishing the relative importance of specific
   root- or soil-derived metabolites in shaping rhizosphere microbial
   structures^[162]20,[163]69. Moreover, advanced techniques such as
   microfluidic-based cultivation and diversification, should be applied
   to isolate additional uncultivated microbes, including fungi and
   bacteria^[164]28. The efficacy of SynComs in promoting Cd tolerance in
   other crops and under field agricultural conditions requires further
   validation.

   In summary, our study reveals that Cd toxicity induces the
   GmAMT2.1/2.2, leading to the recruitment of beneficial microorganisms,
   such as Tumebacillus, Alicyclobacillus, Methylophilus, Aspergillus,
   Talaromyces, and Penicillium, by altering metabolites such as sinapyl
   alcohol, genistein, coumestrol, and piceatannol. SynComs composed of
   these microorganisms collectively affecting various molecular
   mechanisms, thereby enhancing plant resistance to Cd toxicity. These
   mechanisms include ascorbate and aldarate metabolism and glucosinolate
   biosynthesis pathways associated with fungal SynCom, and plant-pathogen
   interaction and mineral absorption pathways associated with
   cross-kingdom SynCom. Notably, cross-kingdom SynCom and fungal SynCom
   showed greater potential for synergistic effects. Overall, our study
   results draw attention to the crucial role of microorganisms,
   especially fungi within SynComs, in the molecular mechanism underlying
   Cd toxicity resistance in soybean. The advanced insights provided by
   our metabarcoding data, and subsequent validation by microbial culture
   profiling, substantially enhance our understanding of plant functional
   gene regulation and synergistic resistance mechanisms within the root
   microbiome. This study provides critical data and guidance for the
   identification of agriculturally important microbial communities as
   potential breeding targets, as well as the development of microbial
   formulations to enhance resistance to Cd toxicity in sustainable
   agriculture.

Materials and Methods

Soil and soybean materials

   Experimental soils were collected at the surface layer at a depth of
   10 cm from acidic soils in China in the summer of 2020. The sampling
   site was Guangzhou (113°35′N, 23°15′E), Guangdong Province, which is
   classified as Udic Agrosol, according to USDA soil taxonomy. The
   characteristics of the soil were: pH 5.3, porosity 40.2, Cd 0.2 mg/kg,
   organic matter 5.2 g/kg, NH[4]^+-N 5.3 mg/kg, available phosphorus (P)
   15.5 mg/kg, NO[3]^--N 24 mg/kg, and available potassium (K) 85.2 mg/kg.

   To construct the CaMV35S-drived GmAMT2.2, the full-length CDS of
   GmAMT2.2 was cloned into the pTF101-eGFP vector, and the resultant
   construct was introduced into Agrobacterium tumefaciens EHA101. To
   produce the GmAMT2.1 and GmAMT2.2 loss-of-function mutants, the
   gene-editing tool CRISPR-Cas9 was used to knockout GmAMT2.1 and
   GmAMT2.2. CRISPR/Cas9-mediated gene editing was performed as previously
   reported^[165]70. Briefly, the two sgRNAs targeting the first exon of
   GmAMT2.1 and GmAMT2.2 were designed on CRISPR-P server
   ([166]http://crispr.hzau.edu.cn/CRISPR/)^[167]71. The sgRNAs was
   sub-cloned into pGES201 plasmid, and the resultant construct was
   further introduced into Agrobacterium tumefaciens EHA105.
   Agrobacterium-mediated transformation as previously
   reported^[168]70,[169]72, and the soybean variety Young was used as the
   transgene receptor.

RNA extraction and quantitative real-time PCR (qRT-PCR)

   Total RNA was extracted from soybean or Arabidopsis thaliana with a
   TRNzol Universal Kit (DP424, TIANGEN, Beijing, China). The cDNA was
   synthesized from total RNA using a PrimeScript RT Reagent Kit with gDNA
   Eraser (RR047A, Takara Bio, Japan) according to the manufacturer’s
   instructions. DNA fragment amplification was performed using KOD FX neo
   (TOYOBO (SHANGHAI) BIOTECH CO., Shanghai, China). qRT-PCR was conducted
   using TB Green Premix Ex Taq II (RR820, Takara Bio, Japan) with a CFX96
   Real-Time System (Bio-Rad, Hercules, CA, USA). Data were normalized to
   the reference genes GmActin3. All analyses were performed with three
   biological replicates and three technical replicates. The results were
   analyzed using the 2^–ΔΔCt method. Student’s t-test implemented in
   Excel software (Excel 2016) was used to evaluate the statistical
   significance of the data. The primers for the markers were listed in
   Supplementary Table [170]7.

Experimental design

   A pot experiment was conducted at the College of Agriculture, South
   China Agricultural University, located in Guangzhou, China. A
   randomized complete block design was used for the experiment, with
   three soybean genotypes (with or without Cd), for a total of six
   treatments (3 soybean genotypes × 2 Cd treatments = 6 treatments).
   Soybean seeds of the three genotypes were sterilized with alcohol and
   then planted in separate pots. The pots were of the same size and
   contained approximately 2.5 kg of soil. In each pot, eight seeds of
   uniform size were initially sown and subsequently thinned to two on the
   10th day. The soil moisture content was maintained at 80% of the field
   capacity throughout the experiment. For each treatment, there were six
   replicates (plots). The soybeans were grown under controlled greenhouse
   conditions (day temperature 28 ~ 32 °C, night temperature 16 ~ 20 °C).

Sample collection and soil chemical analysis

   Twenty days after soybean sowing, 36 samples were collected and
   subjected to amplicon sequencing. Briley, the roots were shaken gently
   to remove the soil adhering to the roots. The roots and attached soil,
   which consider the rhizosphere soil, were then transferred to 1x
   phosphate-buffered saline. Ten grams of rhizosphere soil were obtained,
   of which two grams were stored at -80°C for the microbial experiment
   and LC-MS analysis. The remaining samples were stored at 4 °C for
   subsequent microbial isolation and soil chemical characterization
   measurements.

Amplicon sequencing and data analysis

   Utilizing the Fast DNA SPIN Kit for Soil (MP Biomedicals, Santa Ana,
   CA), microbial DNA was extracted, and the bacterial 16 S rRNA gene was
   specifically targeted by amplifying the V4 region using primers
   515 F/806R^[171]73. Meanwhile, amplification of the fungal ITS region
   was achieved by targeting the ITS1 region using primers ITS5/1737 F.
   Briefly, each PCR reaction contained 4 μL buffer, 2 μL dNTPs (2 mM),
   1 μL of forward/reverse primer (10 μM), 10 ng of DNA and 10 μL ddH[2]O.
   The PCR was then performed on an ABI 7900 system according to the
   following program 94 °C for 45 s; 35 cycles of 95 °C for 15 s, 55 °C
   for 10 s, 72 °C for 10 s; and 50 for 15 min. The amplicons that were
   pooled in equimolar amounts were subjected to paired-end sequencing on
   the Illumina MiSeq platform (Shanghai Majorbio Bio-pharm Technology
   Co., Ltd). The bacterial 16 S rRNA gene and fungal ITS sequences were
   processed with QIIME (v1.9.1) and VSEARCH, starting from the raw FASTQ
   files. After quality filtering, the primers and low-quality sequences
   with scores < 20 and lengths < 200 bp were removed. The paired-end
   reads that correspond to bacterial 16 S rRNA genes and fungal ITS
   regions were merged into single files, respectively. The operational
   taxonomic units (OTUs) were classified at a 97% sequence similarity
   cutoff using CD-HIT^[172]74. Taxonomic assignment was performed using
   the SILVA (v138) and UNITE (v8.0) databases for bacteria and fungi,
   respectively. Mothur and QIIME (v1.91) were used, respectively, to
   calculate the alpha and beta diversity of microbial communities.
   Rarefaction of bacterial and fungal OTU tables resulted in 60,024 and
   60,049 reads, respectively, for alpha diversity estimation.

Metabolite measurement

   Fifty mg of the rhizosphere soil sample were added to a solvent of
   acetonitrile, methanol, and water (volume ratio 2:2:1). The sample was
   then vortexed and sonicated three times in an ice-water bath. The
   supernatant was centrifuged and subjected to UHPLC-Q Exactive (QE)
   Orbitrap MS analysis. Quality control samples were generated by pooling
   the supernatant from all samples and detecting soil metabolites via
   Liquid Chromatography-Tandem Mass Spectrometry (LC-MS/MS) methodology
   (Shanghai Majorbio Bio-pharm Technology Co., Ltd). The OPLS model was
   used to identify metabolites with the greatest group-specific
   differences based on their variable importance prediction (VIP) scores,
   with a threshold of 1 (OPLS, VIP > 1).

High-throughput cultivation of microbes and construction of synthetic
communities

   The rhizosphere bacteria and fungi were isolated using high-throughput
   culture according to Zhang et al., (2021)^[173]20 and Zhou et al.,
   (2022)^[174]28 with minor modifications. Briefly, the rhizosphere
   suspensions were diluted to an optimal dilution such that 30% of the
   wells showed bacterial growth. Then, 100 μL of the diluted suspension
   was plated on the 96-well plates containing different culture media.
   For the bacteria, the media included LB medium, beef extract peptone
   and Trytone yeast extract glucose medium. For the fungi, the media
   included 1/4 strength RBM, 1/4 strength RBM, 1/10 strength PDA and 1/4
   strength MEA^[175]28. In addition, 0.01 M CdCl[2] was added to the
   media to obtain bacteria and fungi more resistant to Cd. After 7 days
   of inoculation, microbial identification was used to identify each
   well. Bacterial 16 S rDNA and fungal 18 S rDNA were amplified and
   sequenced at Sangon Biotech Co., Ltd (Shanghai, China). Sequence
   alignment was performed on the NCBI website, and 1 mL of 30% glycerol
   (v/v) at -80°C was used to preserve each microbial species.

   We identified bacteria and fungi that were significantly enriched in
   the rhizosphere of the soybean carrying GmAMT2.1/2.2 genes by analyzing
   the composition of the rhizosphere microbial community in different
   treatments. These strains were then considered as potential candidates
   for building SynComs. The OD[600] of each bacterial fermentation broth
   was adjusted to 0.02 (approximately 10^7 cells/ml). Fungal strains were
   propagated using shake flask fermentation in 1/10th strength PDB
   medium, and diluted to 10^6 conidia/ml^[176]28. To identify how
   root-derived microbes enhance plant adaptation to Cd toxicity, we
   conducted an experiment consisting of three synthetic SynComs,
   including bacterial, fungal, and bacterial and fungal cross-kingdom
   SynComs. Each SynCom consisted of an equal proportion of the respective
   bacterial or fungal strains. Different SynComs were added to the soil
   10 days after soybean sowing. Plant fresh weight, root Cd^2+ and
   NH[4]^+-N content and translocation factor were determined 15 days
   after the addition of the SynComs.

RNA‐seq for soybean treat with SynComs

   RNA-seq was used to investigate the effect of SynComs on the expression
   of related genes in soybean. Total RNA was extracted from soybean root
   samples using RNAiso Plus reagent (Takara Bio). The constructed
   libraries underwent a thorough quality assessment before sequencing on
   the Illumina platform (PE 150). Then, the raw data files were converted
   into raw reads via base calling analysis. RSEM and STAR were used to
   perform the sequence the alignment and quantification of the detected
   genes^[177]75. Using the “stats” package in R 4.1.2, PCA (Principal
   components analysis) was conducted^[178]76 to show the effect of three
   SynComs on the gene expression of the soybean root. DESeq2 with the
   FDR < 0.05 was used to identify the differentially expressed genes
   (DEGs)^[179]77. A hierarchical clustering heatmap was employed to
   identify DEGs, which were subsequently annotated and analyzed for
   enriched pathways using the “stats” package’s hclust function^[180]78.
   The identified genes were annotated according to SoyBase
   ([181]https://www.soybase.org) and then KEGG
   ([182]http://www.genome.jp/kegg/) analysis was performed using the
   software package “clusterProfiler” to find significantly enriched
   metabolic or signalling pathways with a threshold of
   FDR < 0.05^[183]79,[184]80.

Statistical analysis

   Principal coordinate analysis (PCoA) was conducted in R 4.1.2 using the
   “vegan” package, based on Bray-Curtis dissimilarities. The PERMANOVA
   (Permutation Multivariate Analysis of Variance) and the Mantel test
   were also used to evaluate the significance of the PCoA results. A
   generalized linear model was run in R using the “edgeR” package to
   analyze the microbial and metabolite differences between each two
   treatments. The results of the generalized linear model were visualized
   in volcano plots^[185]81. Using the “vcd” package, the Kruskal-Wallis
   test was performed in R to calculate the enrichment of microbes, and
   was then visualized in ternary plots using “ggplot2” package^[186]82.
   Differences in chemical properties of soil and microbial relative
   abundance on the phylum level were evaluated using Genstat (version
   13.0) with the two-way analysis of variance (ANOVA). Analysis of the
   microbiome and metabolome data integration was performed using
   M^2IA^[187]83. Microbial co-occurrence networks with an average OTU
   abundance greater than 0.1% across samples were constructed and
   analyzed to determine network connectivity in the genotypes. With the
   ’Hmisc’ and ’igraph’ packages in the R environment, spearman
   coefficients between OTUs were determined, and correlations with
   r > 0.8 and P < 0.05 were included in the network^[188]84. The networks
   were explored and visualized using Gephi (v 0.8.2). The codes used for
   the study are accessible on GitHub
   ([189]https://github.com/liantengxiang1988/GmAMT2.1-2.2-shape-rhizosphe
   re-microbiome-mitigate-cadmium-toxicity).

Reporting summary

   Further information on research design is available in the [190]Nature
   Research Reporting Summary linked to this article.

Supplementary information

   [191]Supplementary Information^ (1.2MB, pdf)
   [192]Reporting summary^ (2.9MB, pdf)

Acknowledgements