The ClustalW method with the MEGA 6.06 software was used to align the AtNIP1;2 and rice NIP sequences. A test neighbor-joining phylogenetic tree was built with the same software based on the alignment.

Oryza sativa L. ssp. indica “Minghui86” (“MH86”) served as the transformation host and the wild-type (WT) control. Rice seeds were sterilized for 15–20 min with 20% bleach, then germinated at 30°C for 5 days in the dark. The seedlings with uniform growth were selected and transferred to the control (-Al) growth solution (pH 4.2) overnight to adapt to the low pH condition. Then 10 seedlings for each treatment were transferred to hydroponic growth solutions containing different concentrations of AlCl₃ (10, 20, 40, 60, 80, 120, 160 μM) with the light/temperature condition of 14 h day/10 h night at 30/25°C (day/night). The growth solution contained the macronutrients (mM): KCl, 1.0; NH₄NO₄, 1.5; CaCl₂, 1.0; KH₂PO₄, 0.045; MgSO₄, 0.2; MgNO₃, 0.5; MgCl₂, 0.155; and the micronutrients (μM): MnCl₄, 11.8; H₃BO₃, 33.0; CuSO₄, 0.8; ZnSO₄, 3.06; Na₂MoO₄, 1.07; Fe-HEDTA,77.0. The pH of the solutions was maintained at pH 4.2 by 2 mM Homo-PIPES (homopiperazine-N, N’-bis-2-ethanesulfonic acid). Representative rice seedlings were chosen for root growth phenotyping and image taking after 7-days growth in a hydroponic solution (pH 4.2) containing 0 or 60 μM AlCl₃. Root growth was determined by measuring the root length of individual seedlings prior to and after the 7-day Al treatment (50 μM). Relative root growth (RRG%) was expressed as the root growth of individual plants under Al treatment (+Al) normalized to the average root growth under the control condition (-Al).

For testing rice plants’ responses to other metal toxicity, WT (cv. MH86, Indica) and rice plants were treated for 24 h in growth solutions (pH 4.2) containing (in μM) CdCl₂, 20; ZnSO₄, 100; AlCl₃, 50; or LaCl₃, 5 (Huang et al., 2012). Then, the root growth of the primary roots of individual lines was determined.

The CRISPR-P program was used to predict the two target sites with the PAM (protospacer adjacent motifs) sequences in the exon I and II of OsNIP1;2 (Lei et al., 2014). Subsequently, single guide RNA (sgRNA) sequences targeted the selected OsNIP1;2 sequences were cloned into the vector VK005-01, which contains a maize ubiquitin promoter for Cas9 expression and a rice U6 promoter for sgRNA expression (Viewsolid Biotech, Beijing, China). The primer sequences for the two OsNIP1-2 targets are sgRNA1F: 5′-CAGTGGTCCAAGGAGGCCGTCGT-3′ and sgRNA1R: 5′-A CACGACGGCCTCCTTGGACCA-3′; sgRNA2F: 5′-CAGACC CTGCCTGCTGAAGAACG-3′ and sgRNA2R: 5′-AAC CGTT CTTCAGCAGGCAGGGT-3′. The constructs were transformed into Agrobacterium tumefaciens (EHA105) and then introduced into rice calli of “MH86” with the previously reported methods (Hiei et al., 1994).

Leaf genomic DNAs were extracted with CTAB (cetyltrimethylammonium bromide) (Stewart, 1993). The Cas9/sgRNA T-DNA insertion in the transgenic plants was examined by PCR amplification with the following primer pairs, Cas9F: 5′-GGGAGATCCAGCTAGAGGTC-3′ and Cas9R: 5′-GGAAGGAGGAAGACAAGG-3′. The sgRNA target region of OsNIP1;2 was amplified with the following target-specific primer pairs (OsNIP1-2T1F/R and OaNIP1-2T2F/R), OsNIP1-2T1F: 5′-TGCCGAAGCCTGCTGCTTTC-3′ and OsNIP1-2T 1R: 5′-CGCAAGTTTGGCAAACCACTTG-3′, OsNIP1-2T2F: 5′-GATGGCGGTGGTGGTCGAC-3′, and OsNIP1-2T2R: 5′-G ACTCAATCAGAACACGGTTG-3′, respectively. The PCR products were sequenced. The resulting sequences were decoded with a decoding tool (Liu et al., 2015), and the mutation types and frequency were analyzed. DNA sequences were aligned with the MEGA 6.06 software.

Seedlings of the WT (indica) and Osnip1;2-1, Osnip1;2-2 mutants were treated with various Al concentrations (0–120 μM) for the indicated duration or different metals. Root tips (0–3 cm) and other tissues (root, stem, leaf, panicle, or seed) were collected in liquid N2 and stored at −80^oC.

Total RNAs were extracted with an RNeasy Mini Kit [QIAGEN China (Shanghai) Co., Ltd., Shanghai, China] and used to synthesize first-strand cDNAs with the SuperScript III First-Strand Synthesis System [Invitrogen China (Shanghai) Co., Ltd., Shanghai, China].

OsNIP1;2 expression was investigated by RT-qPCR with a 7500 Fast Real-Time PCR System (Applied Biosystems). The expression of target genes was calibrated with an endogenous 18S rRNA. The RT-qPCR primer sequences for OsNIP1;2 are 5′-CTCCTTCTTCCTCATGTTCG-3′ and 5′-CCTGCGAA GAGCACGTTCA-3′.

The coding sequence of OsNIP1;2 without the stop codon was amplified from a cDNA plasmid using the primers 5′-TCGCGGATCCAAA ATGGCGGTGGTGGTCGAC-3′ and 5′ -ATGGCTCGAG ACTCCTACGCGAGCTCCTC-3′ (the underlined sequences represent the BamHI and XhoI cutting sites, respectively). The PCR sequences were then cloned into the pGPTV.GFP.Bar vector in front of the GFP coding sequence. The resultant pGPTV-OsNIP1;2-GFP construct was introduced into A. tumefaciens strain GV3101 and then transiently expressed in tobacco leaf epidermal cells with an infiltration method. The plasma membrane marker 35S:PIP2;1-RFP (pm-rk-CD3-1007) was described previously (Nelson et al., 2007). The fluorescent images were obtained with a Leica SP5 confocal scanning laser microscope. The nuclei of the tobacco leaf cells were stained with DAPI (4′,6′-diamidino-2-phenylindole) and visualized with the confocal microscope.

For quantifying Al and other elements, 5-days-old seedlings of the WT (indica), Osnip1;2-1, and Osnip1;2-2 were grown in a growth medium (pH 4.2) containing 50 or 100 μM Al for 8 h. Then, the seedlings were rinsed with a 0.5 mM CaCl₂ solution three times and ddH₂O twice. Shoot and root samples were collected, dried, and digested with pure HNO₃. ICP-AES (inductively coupled plasma-atomic emission spectrometry, Thermo Fisher, iCAP 7500 Series) was used to determine the mineral elements of the samples. Each treatment contained three biological replicates.

Seedlings (5-days-old) were grown in an LPM medium (pH 4.2) containing 50 or 100 μM Al for 8 h. Then, the root apices (0–3 cm) were cut and washed three times with 0.5 mM CaCl₂ and twice with ddH₂O. Next, the apoplastic solution of the root segments was removed via centrifuging at 3,000 g and 4°C for 10 min with the Ultra free-MC Centrifugal filter unit (Millipore) (Xia et al., 2010). Then the samples retained in the filter units were sored at -80°C. To separate the root cell sap from the root cell wall, the frozen samples with the filter units were thawed at 37°C and then centrifuged at 20,600 × g for 10 min. The cell-sap samples were collected from the centrifuge tubes while the cell-wall samples remained in the filter units. The root-cell-wall samples were washed three times with 70% ethanol to remove membrane fractions and then digested with 1 mL of 2 N HCl with gentle shaking for 24 h. Al concentrations were measured by ICP-AES.

Five seedlings (5-days-old) of the Osnip1;2-1 and Osnip1;2-2 mutants and the WT (indica) were pre-treated in an LPM medium (pH 4.2) containing 50 or 100 μM Al for 8 h. Then, plants’ culms were removed, and the xylem-sap exudates were collected in a high humidity environment, with the first droplets excluded to avoid contamination. A micropipette was used to measure the volumes of the collected xylem-sap samples. The unit volume’s K and Al concentrations of the samples were determined by ICP-MS.

There are 10 NIP members in the rice genome (Oryza sativa) (Bansal and Sankararamakrishnan, 2007). Phylogenetic analysis (Supplementary Figure 1) and sequence alignment (Supplementary Figure 2) indicated that AtNIP1;2 and its closest homolog, OsNIP1;2, share 63.7% amino-acid sequence identity.

Aquaporins (AQPs) share highly conserved structural features (Hove and Bhave, 2011), containing two highly conserved constrictions in the pore region thought to specialize AQPs’ functions (de Groot and Grubmuller, 2001; Forrest and Bhave, 2007). The first constriction consists of two highly conserved NPA (asparagine-proline-alanine) motifs in the inter-helical loop B (LB) and loop E (LE) (Supplementary Figure 2). The second constriction is the so-call ar/R (aromatic/arginine) region, which contains four residues located in helix 2 (H2), H5, LE1, and LE2 (Supplementary Figure 2; de Groot and Grubmuller, 2001; Wallace and Roberts, 2004; Forrest and Bhave, 2007).

The plant’s unique NIP members could be further classified into NIP-I and NIP-II subgroups (Wallace and Roberts, 2004; Wallace et al., 2006). Members of the NIP-I subgroup have a conserved ar/R tetrad sequence of Trp (W), Val/Ile (V/I), Ala (A), and Arg (R), and an invariable NPA triad sequence for the NPA1motif. However, for the NPA2 motifs, the triad sequence of a NIP-I member could be variable as NPA, NPG, or NPV (Mitani et al., 2008). OsNIP1;2 and AtNIP1;2 are NIP-I members and share a conserved ar/R tetrad sequence of WVAR (Supplementary Figure 2). However, they have different triad sequences in the NIP2 motif, i.e., NPG in AtNIP1;1 and NPA in OsNIP1;2 (Supplementary Figure 2).

To evaluate the biological function of OsNIP1;2 in rice, a CRISPR/Cas9 system was used to generate Osnip1;2 mutants. Through Agrobacterium-mediated genetic transformation, 27 tissue-cultured plantlets were obtained, among which 25 plantlets showed edited OsNIP1;2 sequences, indicating an editing efficiency of 92.6%.

Four types of plantlets could be identified by sequence decoding in the T₀ generation, including those with homozygous, monoallelic heterozygous, biallelic heterozygous, and non-editing. First, an online CRISPR-P tool was used to screen five potential off-target sites carrying three to five mismatched bases¹ (Supplementary Table 1). No off-target effects were identified in selected putative loci against sgRNA1 and sgRNA2 (Supplementary Table 1). Then, two homozygous mutant lines, with one nucleotide deletion in target 1 or one nucleotide insertion in target 2, which led to the shift of open reading frame and mutated OsNIP1;2, were designated as Osnip1;2-1 and Osnip1;2-2 (Supplementary Figures 3, 4) and chosen for further characterization.

Root growth of the wild-type (WT), Osnip1;2-1, and Osnip1;2-2 plants was comparable under the -Al (control) condition (Figure 1A). However, root growth of Osnip1;2-1 and Osnip1;2-2 showed more potent root-growth inhibition by 500 μM Al than the WT (Figure 1A). Moreover, such a root-growth inhibition in the two Osnip1;2 mutant lines were Al-dose dependent (Figures 1A,B), indicating that OsNIP1;2 plays a crucial role in Al tolerance in rice.

The responses of the Osnip1;2 mutants to toxic levels of other metal ions, including La³⁺, Zn²⁺, and Cd²⁺, were examined to investigate the sensitive specificity to Al stress. The results indicated that the WT and mutants showed no difference in growth inhibition by 20 μM CdCl₂, 5 μM LaCl₃, or 60 μM ZnSO₄ (Figure 1C). Therefore, the Osnip1;2 mutants are specifically sensitive to Al toxicity.

RT-qPCR analyses showed OsNIP1;2 transcript levels were significantly higher in the root, stem, and leaf than in the seed and panicle (Figure 2A). Time-course RT-qPCR analyses indicated that OsNIP1;2 transcripts were rapidly up-regulated in the root, peaked at 4 h after Al treatment, and remained high after 24 h (Figure 2B). Furthermore, increased Al concentrations in the treatment solutions were associated with enhanced OsNIP1;2 expression in the root (Figure 2C). In addition, OsNIP1;2 expression was induced by Al³⁺ ions but not responsive to Zn²⁺, Cd²⁺, and La³⁺ (Figure 2D).

The subcellular localization of OsNIP1;2-GFP was evaluated by transient co-expression of OsNIP1;2-GFP and the red fluorescence protein (RFP)-PIP2;1, a PM marker protein, in tobacco (Nicotiana benthamiana) leaf epidermal cells (Nelson et al., 2007). The RFP-PIP2;1 signal was localized in the plasma membrane of N. benthamiana cells, overlapping with OsNIP1;2:GFP (Figure 3A). In addition, the DAPI-stained nucleus was enclosed by the OsNIP1;2-GFP fluorescence in the cytoplasm (Figure 3B). These results demonstrated OsNIP1;2 as a PM-localized protein.

Al accumulation in the root cell wall can be visualized by hematoxylin staining. When the roots of WT, Osnip1;2-1, and Osnip1;2-2 seedlings were stained with hematoxylin after Al treatment, the Osnip1;2-1 and Osnip1;2-2 plants showed much stronger hematoxylin staining in the root apex than the WT plants (Figure 4A). These results suggest that OsNIP1;2 facilitates Al removal from the root cell wall.

Furthermore, compared with the WT plants, the Al-treated Osnip1;2 mutants had significantly lower Al concentrations in the root cell sap (Figure 4B). In contrast, the root-cell-wall Al concentrations were considerably higher in the Osnip1;2 mutants than in WT plants (Figure 4C). These results indicate that OsNIP1;2 participates in removing the Al in the root cell wall to the cytosol.

We measured Al concentrations of the root and shoot after Al treatment to further evaluate the role of OsNIP1;2 in the root-to-shoot Al distributions. Compared with the WT, the Osnip1;2-1 and Osnip1;2-2 mutants possessed remarkably higher and lower Al concentrations in the root (Figure 5A) and shoot, respectively (Figure 5B). These results indicated that OsNIP1;2 plays a vital role in Al translocation to the shoot in rice.

The role of OsNIP1;2 in loading Al to the xylem was examined by evaluating Al concentrations in the root xylem exudates. The results indicated that Al concentrations of the root xylem exudates were considerably lower in the Osnip1;2-1 and Osnip1;2-2 mutants than in the WT (Figure 5C), while the K concentrations were comparable among the WT and the Osnip1;2 mutants (Figure 5D). Furthermore, compared to the WT, 43 and 68% decreases in Al concentrations were observed in the root cell sap (Figure 4A) and the root xylem exudate (Figure 5C), respectively, in the Osnip1;2 mutants. The extra percentage decreases in Al concentrations in the xylem sap of the Osnip1;2 mutants could be attributed to the impaired function of OsNIP1;2 in loading Al to the xylem of the root cells.

Taken together, our results suggested that OsNIP1;2 involves removing Al from the root cell wall and promotes Al translocation to the shoot by facilitating Al loading to the root xylem.

We transferred the pYES2-GFP (control) and pYES2-OsNIP1;2-GFP (OsNIP1;2) constructs to the wild-type (BY4741) yeast cells. When expressed in the yeast cell, the OsNIP1;2-GFP fusion protein was detected at the plasma membrane under a confocal microscope (Figure 6A). This result was consistent with the OsNIP1;2 subcellular localization observed in Nicotiana benthamiana epidermal cells (Figure 3).

We compared the Al sensitivity of the transformed yeast lines to test if OsNIP1;2 could contribute to Al resistance in yeast. Without Al treatment, the OsNIP1;2-expressing line and the control (empty vector) line displayed similar cell-growth rates (Figure 6B). In contrast, expressing OsNIP1;2 in yeast led to remarkably more sensitivity of the yeast line to Al stress (Figure 6B).

Short-term (up to 8 h) Al uptake by OsNIP1;2 were tested in the presence of Al³⁺ (Figure 6C) or Al³⁺ complexed with different cellular ligands, including citrate (Cit), malate (Mal), oxalate (Oxa), succinate (Suc), fumarate (Fum), aconite (Aco), cysteine (Cys), histidine (His), glutathione (GSH), phytochelatin (PC), and metallothionein (MT) (Supplementary Figure 5). In a short-term (up to 8 h) time-course assay, significant OsNIP1;2-mediated Al uptake activities were observed in the OsNIP1;2-expressing yeast cells after incubation in the uptake solution containing 50 μM AlCl₃ for 4 h (Figure 6C). This result suggested that OsNIP1;2 facilitated Al uptake in yeast.

To test the effects of different cellular ligands on OsNIP1;2-mediated Al uptake, we performed a short-term (4 h) Al uptake assay for the yeast lines (BY4741) carrying the pYES2-GFP or pYES2-OsNIP1;2-GFP construct in the presence of Al³⁺ or Al³⁺ conjugated with different cellular ligands mentioned above. The pYES2-OsNIP1;2-GFP line showed significantly enhanced Al uptake activities in the presence of Al³⁺ and the Al-Cys conjugated complex in the uptake solution but not with other Al-ligands (Supplementary Figure 5). This result suggests that the Al-Cys complex could be a transport substrate for OsNIP1;2.

Aquaporin members can transport various small molecules, including B, As, glycerol, Al-malate complexes, and Zn complexes (Wang et al., 2017, 2022; Singh et al., 2020). OsNIP1;2 is one of the 10 members of the NIP subfamily in rice (Bansal and Sankararamakrishnan, 2007). Expressing OsNIP1;2 caused the yeast cells to be more susceptible to Al toxicity (Figure 6B). In addition, the yeast cells expressing OsNIP1;2 displayed enhanced Al uptake activities when grown in the liquid medium supplemented with AlCl₃ or the Al-Cys complex (Figure 6C and Supplementary Figure 5). As the OsNIP1;2 is localized to the PM in the yeast cell (Figure 6A), it is reasonable to assume that OsNIP1;2 can facilitate Al transport and uptake into yeast cells.

Our previous results suggested that the aluminum-malate (Al-Mal) complex could be a transport substrate for AtNIP1;2 (Wang et al., 2017). As the AtALMT1-facilitated malate release system accounts for most of Al resistance in Arabidopsis (Hoekenga et al., 2006; Liu et al., 2009), AtNIP1;2 plays a critical complementary role in removing the Al-Mal complex from the root cell wall for Arabidopsis plants to achieve higher overall Al resistance (Wang et al., 2018, 2020).

OsNIP1;2 is the closest sequence homolog of AtNIP1;2 (Supplementary Figure 2) and is involved in Al transport, translocation, and resistance in rice (Figures 1, 4, 5). However, OsNIP1;2 appeared not to facilitate Al-Mal transport in yeast (Supplementary Figure 5). Instead, an Al-Cys transport activity was observed for the OsNIP1;2-expressing yeast line (Supplementary Figure 5). Although AtNIP1;2 and OsNIP1;2 share a high degree of sequence homology in overall sequences, especially in the ar/R region and NPA motifs (Supplementary Figure 5), a single amino acid difference in the NPA2 constriction (Supplementary Figure 5) might change the transport substrate preferences for these two transporters.

Aquaporin transporters are believed to facilitate transporting polar but non-charged small molecules. However, a significant but minor Al transport activity was also observed for the yeast line carrying OsNIP1;2 in the presence of Al³⁺ ions (Figure 6 and Supplementary Figure 5). As yeast cells secrete cellular ligands into the growth media (Kutralam-Muniasamy et al., 2015), the external ligands and Al³⁺ could form Al-ligand complexes in the uptake medium, which could be taken up by OsNIP1;2. Therefore, the Al taken up by OsNIP1;2 in the Al³⁺ condition (Figure 6 and Supplementary Figure 5) could be in the form of Al-ligands but not as Al³⁺ ions. Further studies are required to distinguish these possibilities and to identify the identity of the ligand.

The cell walls in the root apex are a significant target of Al toxicity. The reason is that Al toxicity disrupts the root cell walls’ integrity, structure, and function in the root tip, as evidenced by the distorted and swollen cells in the root apex (Ma et al., 2004; Horst et al., 2010; Sivaguru et al., 2013). Therefore, reducing the cell wall’s Al concentrations in the root apices could alleviate Al toxicity and thus improve plants’ Al resistance.

Decreases in Al accumulation in the root cells could be achieved by lowering the binding capacity to Al via modifying cell-wall components, such as reducing the polysaccharide concentrations and/or increasing degrees of pectin methylation (Yang et al., 2008, 2011). Another means is to remove Al from the root cell wall for further detoxification inside the cell (Xia et al., 2010; Wang et al., 2017). The Al could be sequestered into the vacuole of the root cell or moved to xylem parenchyma for uploading to xylem for translocation to the less vulnerable shoot via the xylem stream. Our previous results indicate that AtNIP1;2 functions as a bi-directional channel, facilitating the Al uptake from the root cell wall and subsequent Al uploading to the xylem sap (Wang et al., 2017).

Loss-of-OsNIP1;2 functions impaired Al removal from the root cell wall (Figure 4) and subsequent Al translocation to shoot (Figure 5), which made the mutant plants susceptible to Al toxicity (Figure 1). The results indicate that the PM-localized OsNIP1;2 is a crucial component of the internal detoxification mechanism in rice.

In plants, members of the Nramp (natural resistance-associated macrophage protein) transporters are localized to membranes of different subcellular compartments, facilitating the transport of divalent and trivalent metal ions, including Fe²⁺, Zn²⁺, Cd²⁺, Mn²⁺, As²⁺, and Al³⁺, in monocots and dicots plants (Williams et al., 2000; Mäser et al., 2001; Pittman, 2005; Ding and Cai, 2017). Recently, Nrat1 (Nramp aluminum transporter1) has been suggested to facilitate transporting the root cell walls’ Al³⁺ ions to the cytosol in rice (Xia et al., 2010; Li et al., 2014). Then, the cytosolic Al could be transported to the vacuoles of the root cells by OsALS1 or other unknown transporters.

In contrast, members of the aquaporin superfamily are channel proteins and are believed to transport polar but non-charged small solutes (Forrest and Bhave, 2007; Maurel et al., 2008). However, our recent studies suggest that some members of the aquaporin subfamilies could be involved in transporting divalent metal ions, Zn²⁺, and trivalent valent metal ions, Al³⁺, complexed with cellular ligands such as glutathione (GSH) and malate, respectively (Wang et al., 2017, 2022). The metal-ligand complexes are presumably neutrally charged; thus, they could be transport substrates for aquaporin transporters.

Under adverse conditions, plant roots secrete various metabolites into the rhizosphere. The root exudates could adjust soil pH to solubilize mineral nutrients and make them more accessible to plants; chelate toxic compounds; facilitate the formation of beneficial microbiota communities, or function as toxic compounds for pathogens (Bertin et al., 2003; Bais et al., 2006; Badri and Vivanco, 2009; Baetz and Martinoia, 2014; Zhao et al., 2019; Vives-Peris et al., 2020; Upadhyay et al., 2022). For instance, under Al toxicity, plant roots secrete various organic acids (OAs), i.e., malate, citrate, and oxalate, into the rhizosphere (Sasaki et al., 2004; Magalhaes et al., 2007; Liu et al., 2009, 2012; Matonyei et al., 2014; Kochian et al., 2015; Qiu et al., 2019). The OAs and Al³⁺ form OA-Al complexes in the rhizosphere, which are inaccessible to root cells and thus non-toxic to plants. However, our studies indicated that the OA-Al complexes retained in the root cell wall could be toxic, and they need to be removed from the root cell wall to reach a higher resistance level for plants (Wang et al., 2017, 2020). Although root OA exudation could not explain Al resistance in rice (Famoso et al., 2011), the Al complexed with other cellular ligands in the root cell wall could be harmful to the root cells and needs to be removed.

Thus, although structurally distinct, the putative Al³⁺ transporter, Nrat1, and Al-ligand transporter, NIP1;2, are functionally overlapped but complementary in the same biochemical process to clean out the Al retained in the root cell wall under the acid soil condition.