Five rice (O. sativa) lines, the wild type Nipponbare, OsPT2 overexpression line (OsPT2-ov), OsPT4 overexpression line (OsPT4-ov), OsPT4 RNA interference line (OsPT4-Ri), and OsPT4 CRISPR (Clustered regularly interspaced short palindromic repeats) line (OsPT4-cr), were used in this study. The generation of OsPT4-cr is described below. OsPT2-ov, OsPT4-ov, and OsPT4-Ri were characterized previously (Liu et al., 2010; Ye et al., 2015).

To generate the construct of OsPT4-CRISPR vector, we designed the DNA spacer in NEB cutter¹. The amplified PCR product including U3 promoter, spacer of OsPT4 and sgRNA (small guide RNA) were cloned into vector PJE 45/pH-Ubi-cas9-7 (Miao et al., 2013). The primers were OsPT4-spcer-F:5′-AGCCGGGGCTCTTGGACGCCGTTTTAGAGCTATGCTGAAA-3′, spacer-sgRNA-R: 5′-AAAAAGCAGGCTTAAAAAAAAAGCACCGACTCG- 3′, OsPT4-spcer-F: 5′- GGCGTCCAAGAGCCCCGGCTTGCCACGGATCATCTGCAC-3′ and Pu3-spacer-F: 5′- AGAAAGCTGGGTAAAGGGATCTTTAAACATAC GAAC-3′. The construct was transformed into Nipponbare via Agrobacterium tumefaciens-mediated transformation (Wu et al., 2003).

Standard rice culture solution was used in hydroponic experiments. The composition of the culture solution follows: 1.44 mM NH₄NO₃, 0.5 mM K₂SO₄, 1.0 mM CaCl₂, 1.6 mM MgSO₄, 0.17 mM Na₂SiO₃, 0.3 mM NaH₂PO₄, 50 μM Fe-EDTA, 0.06 μM (NH₄)₆Mo₇O₂₄, 15 μM H₃BO₃, 8 μM MnCl₂, 0.12 μM CuSO₄, 0.12 μM ZnSO₄, 29 μM FeCl₃, and 40.5 μM citric acid at pH 5.5 (Yoshida et al., 1976). The transgenic lines were grown in solution containing different concentrations of As using Na₃AsO₄. The solution was renewed every 5 days. The Nipponare and transgenic lines were grown in a greenhouse under 16 h/8 h, 30/22°C, day/night conditions after germination, with c. 60% relative humidity.

A soil experiment was performed in the experimental field in Huazhong Agricultural University, Wuhan, China. The experimental field was divided into two parts with or without Pi fertilizer. The Pi concentrations of these two fields were 15 mg kg^-1 soil P (-P) and 30 mg kg^-1 soil P (+P). Seedlings of Nipponbare and OsPT4-overexpressing plants (20-day-old) were transplanted into the soil and grown to maturity. Each treatment had 10 replicates.

Plant tissue samples (50–100 mg) were cut and ground with a mortar and pestle to a fine powder in liquid nitrogen. Afterward, total RNA was extracted using TRizol regent (Invitrogen, Carlsbad, CA, United States). Then, the resulted total RNAs were checked by gel electrophoresis (Supplementary Figure S1A). According to the manufacturer’s instructions, 3 μg of total RNA was used to synthesize the first-strand cDNA in 20 μL of reaction mixture using M-MLV reverse transcriptase (Invitrogen). Real-time PCR was performed using the SYBR Premix Ex TaqTM (TaKaRa, Shiga, Japan) with the following gene-specific primers (Table 1). The amplification reaction was performed on an Applied Biosystems (Foster City, CA, United States) 7500 PCR instrument. The rice Ubiquitin 5 gene was used as the internal control.

Rice seeds were soaked in deionized water overnight and germinated at 37°C in darkness for 3 days. Seedlings were transferred to 0.5 mM CaCl₂ solution containing a gradient of As(V) concentrations: 0, 25, and 50 μM. Each treatment was replicated with 10 seedlings. Seedlings were grown in a controlled-environment room at 25°C constant temperature and 12-h day length. After 7 days, we photographed the growth phenotype and measured root length and shoot height.

Shoots and roots were harvested separately and roots washed with distilled water before sampling. After drying at 80°C for 3 days, all samples were digested in 65% nitric acid in a MARS6 microwave (CEM) at a temperature gradient of 120–180°C for 45 min, and then diluted in deionized water. The As content of samples was determined with inductively coupled plasma-mass spectrometry (Agilent 7700 series, CA, United States).

Fresh samples were milled in liquid nitrogen and kept at 4°C until samples thawed. The milled samples were homogenized in 10% (w/v) perchloric acid:5% (w/v) perchloric acid (1:9) and placed on ice for 30 min. Following centrifugation at 10 000 g for 10 min at 4°C, the supernatant was used for Pi measurement by molybdenum blue method. The working fluid was a 6:1 ratio 0.4% (w/v) ammonium molybdate dissolved in 0.5M H₂SO₄ mixed with 10% ascorbic acid. Of the working fluid, 2 mL was added to 1 mL of sample solution, incubated in a water bath at 42°C for 20 min, and then cooled on ice. Sample absorbance was measured at 820 nm, and Pi concentration was calculated by normalization to fresh-weight values.

Data were examined using one-way ANOVA, followed by comparisons of means using the LSD test (Fisher’s Least Significant Difference).

As(V) is a chemical analog of Pi and can be taken up by Pi transporters, so we assayed expression levels of the Pht1 family members except for OsPT11 and OsPT13, which are induced specifically during mycorrhizal symbiosis. The transcript levels of OsPT1, OsPT2, OsPT4, and OsPT8 were significantly higher than that of other Pht1 family members (Figure 1). When rice had been grown in hydroponic conditions with 5 μM As(V) for 7 days, the expression levels of OsPT1, OsPT4, and OsPT8 were significantly increased by 1.7, 2.7, and 5.0 times, respectively. Considering that OsPT1 and OsPT8 are involved in As(V) uptake and transport in rice, OsPT2 and OsPT4 may also have roles in As(V) uptake and transport. The OsPT2- and OsPT4-overexpressing plants were obtained via A. tumefaciens-mediated transformation, and expression levels of OsPT2 and OsPT4 were measured using real-time PCR (Supplementary Figures S1B,C). When transgenic plants had been grown in culture solution with 5 μM As(V) for 3 days, differences in As concentration in both shoots and roots were observed in OsPT4-overexpressing but not OsPT2-overexpressing plants (Supplementary Figures S1D,E). Thus, we focused on the role of OsPT4 in As(V) uptake and transport.

To examine the hypothesis that OsPT4 was involved in As(V) absorption in rice plants, a phytotoxicity experiment was performed to observe growth of OsPT4-overexpressing rice. The Nipponbare and OsPT4-overexpressing lines were seeded in hydroponic conditions with 25 or 50 μM As(V). When exposed to the As(V) condition for 7 days, rice showed an As-toxicity phenotype in which growth of aboveground parts was delayed and the roots were stunted (Figures 2A–C). The OsPT4-overexpressing plants were more sensitive to As(V) and their root length and shoot height decreased 50 and 30% compared with wild-type, respectively (Figures 2D,E). Furthermore, As(V) uptake by OsPT4-overexpressing lines and wild-type roots were determined. At both 25 and 50 μM As(V), the As concentration in roots of OsPT4-overexpressing rice increased 10 and 33% compared with Nipponbare (Figure 2F).

A time-course experiment was used to investigate the ability of OsPT4 to take up As(V). The As accumulation was determined when plants were exposed to hydroponic culture with 5 μM As(V) for 2 h, 1 and 7 days. The concentration of As in rice roots and shoots increased significantly with the As(V) treatment time (Figure 3). When OsPT4-overexpressing lines grew in culture with 5 μM As(V) for 7 days, the As accumulation increased by twice in both roots and shoots.

Certainly, the overexpression lines of OsPT4 accumulated more As both in shoots and roots. It is important to determine the As uptake rate of Nipponbare and OsPT4-overexpressing rice. Wild type and overexpression lines were cultured in the hydroponic solution with 1–50 μM As(V) concentration under +P (100 μM) and -P (0 μM). According to the results (Figure 4), the As(V) uptake rate of OsPT4 overexpression lines was higher compared with wild type under the condition with or without Pi. In the absence of Pi, the As(V) uptake kinetics could be described by a Michaelis–Menten equation (Table 2). The V_max (maximum influx velocity) of As(V) uptake in OsPT4-overexpressing plants increased 23–45% compared with Nipponbare. Additionally, the K_m values of As(V) influx in overexpression lines were 8–28% higher than that in wild type. Under the +P condition, the As(V) uptake rates of rice were significantly lower compared with plants grown in the -P condition. Moreover, As(V) uptake rate was linear over the range of As(V) concentrations tested in the solution with Pi and the slopes of OsPT4 overexpressing plants were 1.4 to 2 times greater compared with Nipponbare. The data suggested that OsPT4 contributes to the As(V) uptake in rice root.

To further understand the role of OsPT4 in As concentration and distribution in rice, the Nipponbare and OsPT4-overexpressing plants were grown to heading stage in hydroponic culture with 25 μM As(V). In this study, As accumulated mainly in the roots and to a lesser degree in aboveground organs (Figure 5A). In shoots, As was mainly in the nodes, which is the most important storage location for various metallic elements. The sum of As content in nodes accounted for 60% of total As in aboveground parts (Figure 5B). The As content in OsPT4-overexpressing lines increased in both roots and shoots by 22 and 47%, respectively. In addition, the As content of all organs in aboveground parts increased significantly, especially in nodes. The As content of the first node in OsPT4-overexpressing lines increased by twice that for the wild-type, while the second and third node only increased by 55 and 39% compared with wild-type, respectively. Furthermore, the As accumulation in flag leaf and flag sheath of OsPT4-overexpressing lines increased by 40 and 32%. The total As distribution of Nipponbare and OsPT4-overexpressing lines grown in the field soil was similar to that in the hydroponic experiment (Supplementary Figure S2A). This result was consistent with the expression pattern in different organs of rice, in which OsPT4 was mainly expressed in root and flag leaf (Supplementary Figure S2B).

Previous studies suggested that OsPT1 and OsPT8 were involved in As(V) uptake. The transcript levels of OsPT1, OsPT4, and OsPT8 in rice grown in 5 μM As(V) for 2 h, 1 and 7 days were measured to determine any interaction between these genes. In roots, OsPT8 expression rapidly increased by 10 times, then decreased and remained at a high level; however, OsPT1 and OsPT4 expression increased gradually and maintained a constant level until 7 days (Figure 6A). In the shoot, Real-time PCR analysis showed that expressions of these three genes were enhanced by As(V) (Figure 6B). OsPT1 and OsPT8 were significantly induced by 30 and 8 times, respectively, within a short period and then quickly returned to their original state. The transcript level of OsPT4 increased gradually with treatment time, and finally increased 11 times at 7-days treatment. The induction of OsPT1, OsPT4, and OsPT8 by As(V) raised the question of whether there was functional redundancy across the three genes. To determine this, the relative expression levels of OsPT1 and OsPT8 were evaluated in Nipponbare and OsPT4-overexpressing rice grown in hydroponic culture with or without As(V) (Figures 6C,D). Interestingly, expressions of OsPT1 and OsPT8 had no change in OsPT4-overexpressing rice cultured under normal conditions. The expression levels of OsPT1 both in roots and shoots of OsPT4-overexpressing rice significantly increased in the As(V) condition. However, expression of OsPT8 in OsPT4-overexpressing rice decreased over twofold in roots and was almost unchanged in shoots. The results suggest a lack of functional redundancy among OsPT1, OsPT4 and OsPT8.

A long-term experiment was used to investigate transgenic and wild-type plants grown to maturity in flooded soil conditions with two levels of P: -P (15 mg P kg^-1 soil) and +P (30 mg P kg^-1 soil). Overexpression of OsPT4 significantly increased total Pi concentrations in both grain and straw both in -P field and +P field (Figures 7A,B). In -P field, OsPT4 overexpression significantly enhanced grain and straw As accumulation by 22–30% and 40–66%, respectively, but no significant difference was shown in +P field (Figures 7C,D).

In order to further study the role of OsPT4 in rice As(V) uptake and transport, we studied the phenotype of OsPT4-Ri plants in solution with 0, 25, and 50 μM arsenate. The growth of OsPT4-Ri plants was similar to the wild type. No obvious difference in As concentration between NIP and OsPT4-Ri plants was observed (Supplementary Figure S3). Meanwhile, we obtained two different OsPT4 CRISPR lines that were treated with different As concentrations. As we expected, the OsPT4-cr lines showed stronger resistance to As(V) compared to wild type (Figure 8). The root length and shoot height of OsPT4-cr plants were significantly longer than wild type. Furthermore, the As accumulation of roots in OsPT4-cr plants decreased 17–30% compared with Nipponbare. The results obtained from OsPT4-cr plants confirmed that OsPT4 is involved in As(V) uptake and transport.

As(V) is a toxic analog of Pi, so it can be absorbed and transported via Pi transport in plant (Xu et al., 2007). In Arabidopsis, previous studies suggested that AtPht1;1 and AtPht1;4 mediated a significant proportion of the As(V) uptake and a AtPHF1 mutant was more resistant to As(V) than wild-type (Shin et al., 2004; Gonzalez et al., 2005). In rice, research has shown that Pht1 family genes participate in As(V) uptake – OsPT1 was involved in As(V) transport from soil to apoplast and OsPT8 functioned in As(V) uptake and resulted in a high affinity for As (Wu et al., 2011; Kamiya et al., 2013).

It was reported that the expression of OsPT4 significantly increased in root of BRRT51 under As stress (Begum et al., 2016). In our study, the transcript levels of Pht1 family genes in Nipponbare grown in normal and As(V) conditions were measured. The expression level of OsPT4 significantly increased in hydroponic culture containing As(V). The up-regulated expression of OsPT4 hinted at a role in As(V) uptake. Furthermore, the ability of OsPT4 to absorb As was clearly demonstrated in the As phytotoxicity, hydroponic and field experiments. Root length and shoot height of OsPT4-overexpressing lines decreased 50 and 30%, respectively, compared with Nipponbare, and the As accumulation in roots of transgenic lines increased 33%. Meanwhile, the OsPT4-cr plants produced the opposite phenotype. Furthermore, the As(V) uptake rates of OsPT4-overexpressing plants were significantly higher than that in wild type under the growth condition with or without Pi. Differences in As concentration were also observed in grain and straw of OsPT4-overexpressing plants compared with wild-type in the flooded soil. All these results suggested that OsPT4 was a functional transporter in As(V) uptake.

Previous studies showed that OsPT4, a Pi-influx transporter involved in Pi acquisition and mobilization in rice, facilitates embryo development (Ye et al., 2015; Zhang et al., 2015). OsPT4 was highly expressed in roots, specifically in the exodermis cells and cortex. Although the sclerenchymatous cells at the exodermis in rice is the first apoplastic barrier to entry of toxic As, it does not completely stop entry of As because it shares transporters with essential Pi. The strong expression of OsPT4 in the root exodermis cells and cortex (Ye et al., 2015) may explain the high As accumulation in roots and the heightened effect of As(V) in the As phytotoxicity experiment. Overall, the results suggested that OsPT4 was a Pi transporter sharing an ion channel with As(V) and playing an important role in As(V) uptake.

In rice, OsPT4 is constitutively expressed in roots and shoots (Ye et al., 2015), with the highest expression in flag leaves (Supplementary Figure S2). However, the As concentration in roots was much higher than that in shoots (Figure 5). The reason for this discrepancy is likely to be due to the mitigation strategies in rice As uptake and transport. The As(V) was absorbed by OsPT4 and the other transporters, and then quickly reduced into As(III). OsHAC1;1 and OsHAC1;2 have been reported to act as arsenate reductases in rice (Shi et al., 2016). Most of As(III) was mainly fastened in root cortex and stele, forming complex with thiol (Su et al., 2010). The uncomplexed As(III) is transported to shoots and even to grain. The formation of As(III)-thiol complex in rice roots helps to explain the reason why the As concentration in roots is four times higher than that in shoots.

The flag leaf is an essential tissue for the growth and development of rice panicles, and plays a key role in remobilization of many mineral elements from leaves to developing grain. In addition, Zhang and his team reported that OsPT4 was involved in Pi mobilization that facilitates embryo development in rice (Zhang et al., 2015). In our work, overexpressing OsPT4 resulted in higher As concentration than wild-type in various organs of rice shoots, such as nodes, flag leaves and panicles Thus, we deduced that OsPT4 was probably involved in As mobilization from flag leaf to panicles and immobilization in grain.

Many researches have reported that nodes are critical hubs in controlling the distribution of mineral elements including As. Nodes have a markedly larger concentration of As than the other tissues of rice shoots (Moore et al., 2014). This was confirmed in present study, showing that the As concentration of nodes represents 60% of the total As in shoot. The possible explanation is that a large portion of the node tissues are vascular bundles and that As accumulates strongly in the phloem (Chen et al., 2015). The nodes that produce or are near crown roots may accumulate higher concentrations of As accumulation. This may explain the observation that OsPT4-overexpressing rice accumulated more As in node II and node III where generates crown roots. In the last few years, many mineral element transporters have been reported to function in rice nodes. A member of the rice C-type ATP-binding cassette (ABC) transporter family, OsABCC1, was reported as a As(III)-phytochelatin transporter. Knockout of OsABCC1 in rice resulted in less As accumulation in the nodes and more As accumulation in the grain (Song et al., 2014). A strategy to prevent As accumulating in the grain is accumulate As in the nodes, especially those close to crown roots.

There are a reported 13 members of the rice Pht1 family and most share a similar protein structure and the same protein destination (localized to the plasma membrane). Among their encoding genes, OsPT1, OsPT4, and OsPT8 were induced by As(V) stress. Individually overexpressing these three genes led to higher As accumulation than their background. These results raised the question of whether there is a network among these three Pi transporters playing specific and/or overlapping roles in As accumulation in rice.

The sensitivity to As(V) stress of OsPT4-overexpressing plants and OsPT8 mutants, assessed in As phytotoxicity experiments, showed that both OsPT4 and OsPT8 were involved in As(V) uptake from roots (Wang et al., 2016) (Figure 2). Analyses of the GUS reporter gene driven by the promoters of OsPT4 and OsPT8 indicated a partial overlap in their spatial expression patterns, because the strongest expression of OsPT4 and OsPT8 were in the epidermis and cortex (Jia et al., 2011; Ye et al., 2015). Additionally, real-time PCR analysis revealed the same expression pattern of OsPT4 and OsPT8 in BRRI33 and BRRI51 under As stress along with lower regulation in BRRI33 and higher regulation in BRRI51 (Begum et al., 2016). And both OsPT4 and OsPT8 were up-regulated in overexpression lines of OsPAP21b and OsHAD1 under Pi-deficient conditions (Mehra et al., 2017; Pandey et al., 2017). Moreover, the transcript level of OsPT8 in roots was dramatically attenuated in OsPT4-overexpressing plants grown in As(V) conditions (Figure 6D). These results indicated that there may be a functional overlap of OsPT4 with OsPT8. However, contrary to this assumption, ospt8 mutants lost almost half of their As(V) uptake ability when seedlings were exposed to 1–2 μM As(V)(Wang et al., 2016). In our study, OsPT4-cr lines displayed significantly lower As concentrations in roots. The similar phenomenon has been observed in ospt4 mutant (Cao et al., 2017). The attenuation of function of OsPT8 was not compensated for by OsPT4, and so OsPT4 and OsPT8 were non-redundant in As(V) uptake. Furthermore, the expression variance of OsPT4 and OsPT8 in the time-course experiment also differed. In the present study, OsPT8 expression in rice roots quickly increased by 10 times when plants were treated with As(V) for 2 h and then dropped (Figure 6A). Over the same time periods, the transcript level of OsPT4 increased gradually and remained at a high level. In flooded soil, overexpressing OsPT8 did not change the As accumulation in straw and grain (Wu et al., 2011), but the As accumulation in OsPT4-overexpressing plants increased in Pi-replete conditions (Figures 7C,D). These results strongly suggest that OsPT4 and OsPT8 had similar expression patterns but different regulation pathways in As uptake.

In Arabidopsis, there was a possible functional overlap of Pht1;1 with Pht1;4. In the As(V) condition, the double mutant showed a more resistant phenotype compared with background and the mutants for single genes (Shin et al., 2004). In rice, the promoters of OsPT1 and OsPT4, unlike OsPT8, did not contain the P1BS element. The expression level of OsPT1 and OsPT4 was not effect by Pi supply conditions (Wu et al., 2013). Kamiya et al. (2013) reported that OsPT1 was involved in As accumulation in shoots, a conclusion consistent with our result that the expression of OsPT1 in shoots rapidly increased by 30 times (Figure 6B). The As accumulation in shoots of ospt1 mutant was not significantly different to its background in Pi-replete condition (Kamiya et al., 2013), indicating that there may be a functional overlap between OsPT1 and OsPT4. We also found that altered expression of OsPT4 did affect expression of OsPT1. The transcript level of OsPT1 in rice roots and shoots dramatically increased in OsPT4-overexpressing plants grown in normal and As(V) conditions (Figure 6C). Therefore, it was logical to assume that OsPT1 and OsPT4 shared a similar or the same pathway in rice As(V) uptake and translocation. Studies utilizing double mutants of these two genes are needed to test this hypothesis.

Although there were significant increases of As concentration in OsPT4-overexpressing plants grown in hydroponic solution with Pi supplement, differences in As concentration in grain and straw were observed under -P but not +P flooded soil (Figure 7). The latest study showed that knockout of OsPT4 could significantly decrease the inorganic As in rice grain (Cao et al., 2017). Since inorganic As is classified as Class-1 carcinogen, decreasing the concentration of inorganic As in rice grain is an important rice breeding target to protect human health. Under these circumstances, OsPT4 which is involved in Pi and As uptake is a candidate gene to generate a high Pi-efficiency and low As-accumulating rice.