A random mutation library of ptxD was constructed by error-prone PCR. The PCR mixture (50 μl) was consisted of Taq buffer containing 5 mM MgCl₂, 20 ng of template plasmid pGEX-6p-ptxD, 1 mM dNTPs, 0.2 mM dTTP and dCTP, 0.6 mM MnCl₂, 2.5 units of Taq DNA polymerase, and 0.4 μM primers (ptxD-F and ptxD-R). After PCR, the amplified products were purified and double digested with BamHI and XhoI. They were then ligated into the expression vector pGEX-6p-1 using T4 DNA ligase. Finally, the ligation product was transformed into E. coli DH5α and plated onto LB-ampicillin medium to generate the random mutation library.

To screen for the activity-improved variants in the above mutation library, protein induction coupled with T7 phage-mediated cell lysis was performed. Briefly, single colonies from the mutant library were picked and grown in a 96-deep well plate (Qiagen, Germany) containing 600 μl of LB medium with ampicillin (100 mg/ml). After overnight incubation at 37°C, 200 μl of fresh LB culture containing 0.1 mM IPTG and T7 phage was added to each well. This was incubated at 22°C for 6 h at 200 rpm. Subsequently, aliquots for cell lysis (10 μl) were transferred to another 96-well plate with each well containing 150 μl of reaction buffer [50 mM phosphite, 5 mM NAD⁺, and Nitro blue Tetrazolium Chloride (NBT; 0.5 mg/ml)]. After 30 min of incubation at 37°C, the amount of NBT reduction in each well, which corresponded to the activity of the variant, was determined by measuring the absorbance at 550 nm using a Multiskan Spectrum spectrophotometer (Thermo Fisher Scientific, Vantaa, Finland).

For the most improved variant from the random mutation library, site saturation mutagenesis was performed at position 139Y by overlapping extension PCR (Heckman and Pease, 2007). For each mutagenesis, specific primer pairs containing desired nucleotide mutations (as listed in Supplementary Table S1) were used. The PCR mixture (50 μl) consisted of the PCR buffer, 10 ng of template pGEX-6p-ptxD, 0.2 μM pair of flanking primers, and a pair of corresponding specific primers, 0.3 μM dNTPs, and 1 unit of pfu DNA polymerase (TransGen, China). After the PCR, the full-length genes containing the desired mutation were ligated into the expression vector pGEX-6p-1. The recombinant vector was transformed into E. coli DH5α, and the introduction of the desired mutations was confirmed by sequencing the DNA.

The codon-optimized gene encoding PtxD was chemically synthesized by GenScript Corporation (Naing, China) and inserted into E. coli pGEX-6p-1 expression vector, resulting in a recombinant pGEX-6p-PtxD vector. For protein expression, E. coli BL21 (DE3) cells with the recombinant plasmid were grown overnight at 37°C in LB broth containing ampicillin (100 μg/ml). Subsequently, the culture was inoculated into fresh LB medium and grown at 37°C till OD600 was 0.6; IPTG (isopropyl-D-thiogalactopyranoside) was then added to induce protein expression at a final concentration of 0.1 mM. After an overnight induction at 18°C, cells were harvested, centrifuged, and washed twice before being resuspended in ice-cold phosphate-buffered solution (140.0 mM NaCl, 2.7 mM KCl, 10.0 mM Na₂HPO₄, and 1.8 mM KH₂PO4; pH 7.0). The suspended solids were disrupted by a French cell press and the lysates were centrifuged at 12,000 g for 30 min. The supernatants were collected to purify the recombinant enzyme by using the GST fusion protein purification kit according to the instructions of the manufacturer. The purity of the recombinant PtxD and its variant enzymes were analyzed using 12% SDS-PAGE (sodium dodecyl sulfate–polyacrylamide gel electrophoresis). The protein concentrations were measured by the Bradford method using the bovine serum albumin standard.

The kinetic parameters for the wild-type (WT) PtxD and its variants with phosphite were assayed as described previously (Hirota et al., 2012). The enzymatic reaction consisted of 100 mM Tris-HCl buffer containing 5 mM NAD⁺ (saturating concentration), sodium phosphite at varying concentrations (ranging from 0.02 to 2 mM), and the purified enzyme. The reaction was performed at 30°C, and the initial rates were determined by measuring the absorbance of NADH at 340 nm. The kinetic constants were analyzed by the Michaelis–Menten equation using the initial reaction parameters and phosphite concentration in GraphPad Prism 5.0. The effects of pH, temperature, and common ions at the final concentration of 1 mM on the activity of PtxD were assayed via routine methods described elsewhere (Li et al., 2020).

To generate plants overexpressing ptxD _Q , the full-length coding sequence of ptxD _Q was inserted into the pTF101 vector driven by the Ubi promoter at BamHI restriction sites. The full-length coding sequence of ptxD _Q was also inserted into the pCAMBIA1300 vector driven by the CaMV 35S promoter at BamHI restriction sites. The two vectors were constructed using GB clonart Seamless Cloning Kit (GBI, Suzhou, China) according to the instructions of the manufacturer. The sequences of the ptxD _Q primers were listed in Supplementary Table S2.

The pTF101-ptxD _Q was transformed into Agrobacterium tumefaciens strain GV3101 and then transformed into Arabidopsis (Arabidopsis thaliana) using the floral dip method. All WT and transgenic Arabidopsis plants used in this study were in the Columbia background. The pTF101-ptxD _Q was recombined into Agrobacterium tumefaciens strain EHA105, and transgenic rice was generated by Agrobacterium-mediated transformation as described in previous studies (Lu et al., 2016). The resultant transgenic plants were confirmed by testing their resistance to Basta and RT-PCR analysis. Total RNA was extracted from the tissues of rice plants using TRIzol reagent (Thermo Fisher Scientific). Total RNA (2 μg) from each sample was used to synthesize complementary DNA (cDNA) using a High-Capacity cDNA reverse transcription kit (Thermo Fisher Scientific); 30 cycles in RT-PCR were used. The ACTIN genes of Arabidopsis and rice were used as the internal references, respectively. The sequences of the primers were listed in Supplementary Table S2.

Arabidopsis seeds were surface-sterilized with 75% ethanol for 2 min and then dipped in 1% (w/v) NaClO₄ for 10 min. They were washed with sterilized water six times. Finally, the seeds were kept dipped in the proper volume of sterilized water and incubated in dark at 4°C for 2 days. All Arabidopsis seedlings germinated initially on Murashige and Skoog (MS) medium with 1% sucrose, 3.5 mM 2-N-Morpholino ethanesulfonic acid, and 1.2% Agar at pH 5.6 (adjusted with 1 M KOH). Seeds were sown in a horizontal line on plates that were vertically disposed in the growing chamber (16-h light/8-h dark photoperiod) with a light intensity of 200 μmol m⁻² s⁻¹ and 60% relative humidity, at 22°C. Three different Pi level treatments included LP (10 μM), HP (high Pi, 1.25 mM), and Phi (replace Pi with Phi, 1.25 mM). Fifteen-day-old seedlings were photographed. There was a little difference in the vermiculite-hydroponics culture medium. Ten-day-old seedlings germinated on MS medium in the plate were transformed and grown in pots with an equal weight of vermiculite. Four different Pi level MS hydroponics were prepared, which included LP (0.01 mM), HP (1.25 mM), Phi (1.25 mM), and ½Pi + ½Phi (625 μM Pi + 625 μM Phi). An equal volume of the nutrient solution was provided every 3 days. About 35-day-old plants were photographed and their shoot dry weights were measured. The plants continued to grow till the flowering period; about 60-day-old plants were photographed. Until the end of the plant growth period, we collected all the seeds from every single plant. Plant growth conditions were the same as described above.

Seeds of rice (Oryza sativa) cv. ZhongHua 11 were surface-sterilized with 1% nitric acid for 14 h, in dark, and then rinsed with deionized water five times. The seeds germinated at 37°C in the dark for 2 days. Uniformly germinated seeds were cultured hydroponically using Yoshida solution, pH 5.5 (adjusted using 1 M HCl). The solution contained 1.425 mM NH₄NO₃, 0.513 mM K₂SO₄, 0.998 mM CaCl₂, 1.643 mM MgSO₄, 0.25 mM Na₂SiO₃, 0.009 mM MnCl₂, 0.075 μM (NH₄)₆Mo₇O₂₄, 0.019 μM H₃BO₃, 0.155 μM CuSO₄, 0.152 μM ZnSO₄, and 0.125 mM EDTA-Fe (II). The 10-day-old seedlings were transferred to the nutrient solutions with different Pi sources: LP (10 μM), HP (300 μM), Phi (300 μM), LP + Phi (10 μM Pi + 300 μM Phi), and HP + Phi (300 μM Pi + 300 μM Phi). The seedlings were grown in a greenhouse for 20 days, at a 12-h light/12-h dark photoperiod, 32/24°C, and 60% relative humidity, with a photon flux density of 200 µmol m⁻²s⁻¹.

Growth parameters were measured for the WT and transgenic plants under different P conditions. The fresh weights of shoots and roots in the culture medium of LP, HP, and Phi were calculated. The fresh weights and lengths of individual plants in HP, LP + Phi, and HP + Phi culture media were counted. The freshly obtained samples were incubated in a 65°C drying chest for 5 days.

The Pi and Phi concentration was determined as follows: The fully dried sample was ground into powder; 8 ml of deionized water was added to about 50 mg of the dried samples and incubated at 250 rpm for 1 h at 25°C; the sample was placed in a metal bath at 70°C for 2 h and was shaken every 30 min; the sample was centrifuged at 10,000 rpm for 10 min, and 2 ml of the supernatant was filtered using a 0.2-μm filter membrane; the filtrate was measured with an ion chromatograph (Dionex ICS-1100) using appropriate concentrations of NaH₂PO₄ and H₃PO₃ as standards.

SPSS Statistics Base software (version 22) was used for statistical analysis. Significant differences were evaluated using one-way ANOVA and Turkey’s test.

The ptxD gene sequence from Ralstonia sp. strain 4,506 was optimized on the basis of the codon bias of E. coli and chemically synthesized. For protein PtxD expression in E. coli, the synthesized ptxD _R4506 gene was cloned into the pGEX-6p-1 expression plasmid. Crude cell extracts were prepared from IPTG-induced strains carrying the ptxD overexpression and empty vectors. A band of overexpressed 62-kDa protein was detected in the cell extract from the ptxD overexpression sample using SDS-PAGE (Figure 1A). A single protein of approximately 36 kDa was obtained using GST affinity columns, which was consistent with the predicted PtxD protein size from Ralstonia sp. strain 4,506 (Figure 1A). The purified PtxD_R4506 showed maximum Phi oxidation activity at 45°C, and relatively high activity (>90%) was maintained at pH range from 5.5 to 7.5 (Figures 1B,C). The optimal pH of 7.0 rendered PtxD good fittness in plant cell microenvironment. In addition, we tested influence of common ions existing in plant cell including NH₄ ⁺, Fe³⁺, Mg²⁺, Mn²⁺, K⁺, Na⁺, and Ca²⁺ on enzyme activity. It was observed that Fe³⁺ could dramatically stimulated the PtxD activity, whereas other ions showed weak enhancing effect, suggesting that these ions commonly existing in plant cell might imposed no inhibition effects on PtxD activity (Supplementary Figure S1).

To improve the activity of WT PtxD_R4506, a mutant library was created by introducing random mutations in the ptxD gene through error-prone PCR and then screened by 96-well plate assay (Supplementary Figure S2). Using NBT colorimetry, the ptxD-M20 mutant that showed the highest PtxD activity was selected from about 10,000 clones. Sequencing analysis showed that the Tyr¹³⁹ of PtxD_R4506 was substituted by Trp in the improved mutant. To further confirm whether the Tyr¹³⁹ mutation increased the activity of PtxD_R4506, we purified the mutated PtxD protein and determined its enzymatic kinetic constants. As shown in Table 1, Trp to Tyr mutation at site 139 increased the affinity and turnover number of PtxD protein by approximately 1.5-fold and 2.9-fold higher than WT PtxD, respectively. This indicated that the 139th amino acid residue of PtxD_R4506 could be used as candidate target for optimizing its catalytic efficiency.

To further enhance the enzymatic efficiency of PtxD, saturation mutagenesis was performed by introducing 18 different amino acids excluding Trp and Tyr, at the 139th site of PtxD using overlap extension PCR. The activity of these 18 mutants was screened by using the NBT assay, which was employed for random mutation libraries screening. As compared to the WT, six mutants retained comparable or even higher enzymatic activity; whereas the other 13 mutants showed no obvious improvement in activity (Supplementary Figure S3). The six PtxD variants were purified to high homogeneity by GST affinity chromatography. Enzyme kinetic constants were measured and compared with the parent enzyme. We found that four variants (Y139F, Y139Q, Y139M, and Y139L) obtained from saturation mutagenesis exhibited significantly enhanced PtxD activity, whereas one variant (Y139R) had a slightly reduced PtxD activity (Table 1). Notably, among all the tested PtxD proteins, the variant Y139Q showed the lowest K_m and highest k _cat with phosphite, with an enhanced 5.8-fold catalytic efficiency as compared to the WT PtxD (Table 1).

To overexpress the ptxD _Q in Arabidopsis, the ptxD _Q sequence was optimized for the codon bias of plants and chemically synthesized (Supplementary Figure S4). The synthetic ptxD _Q gene was cloned into the plant transformation vector pTF101-Ubi. By Agrobacterium-mediated transformation, 12 independent ptxD _Q overexpression Arabidopsis lines were generated (Supplementary Figure S5). The homozygous T3 seeds germinated and were grown on MS medium with LP, HP, and Phi. The ptxD _Q overexpression lines showed similar growth phenotypes as the WT under both LP and HP conditions. Interestingly, although the non-transgenic Arabidopsis did not grow on medium with phosphite as the sole source of P, the ptxD _Q overexpression lines could grow normally under Phi treatment condition (Figure 2A). To further examine the ability of the ptxD _Q overexpression lines to utilize Phi as the sole source of P, 10-day-old seedlings were transformed and grown in pots with an equal weight of vermiculite. Nutrient solutions containing LP, HP, and Phi were used to water the plants regularly. The growth of seedlings with ptxD _Q overexpression was similar to Col-0 under both Pi-deficient and Pi-sufficient conditions (Figures 2B,C). Although the growth of Col-0 seedlings was completely inhibited by Phi treatment, the ptxD _Q overexpression grew well under Phi treatment condition (Figures 2B,C).

We further examined the effects of different P nutrient conditions on the agronomic traits in Arabidopsis. As expected, LP treatment delayed flowering time and produced fewer seeds as compared to the HP treatment condition. The flowering time and seed production were similar between WT plants and ptxD _Q overexpression lines under both Pi-deficient and Pi-sufficient conditions (Figure 3). With phosphite as the only P source, flowering was inhibited and no seeds were produced in WT plants (Figure 3). In contrast, the flowering time and seed production of ptxD _Q overexpression plants were not affected by Phi treatment. This indicated that the ptxD _Q overexpression plants could use phosphite as the only P source to complete their life cycle (Figure 3). We further treated the plants with a combination of phosphate and phosphite in the ratio of 1:1. Interestingly, the WT plants showed a delayed flowering time and reduced seed production similar to the Pi-deficient condition; the ptxD _Q overexpression plants showed a similar flowering time and seed production as in Pi-sufficient conditions (Figure 3).

In addition to Arabidopsis, the ptxD _Q gene was also transformed into the rice, and 11 independent overexpression lines were generated (Supplementary Figure S6). To examine whether the ptxD _Q overexpression rice could utilize Phi as the sole source of P, we performed hydroponic experiments with T₂ generation transgenic lines. The transgenic plants showed similar growth as WT plants under both Pi-deficient (LP) and Pi-sufficient (HP) conditions. When Phi was provided as the sole source of P in the nutrient solution, the WT plants were smaller than those under LP conditions. Under a prolonged Phi treatment, the WT seedlings died (data not shown). In contrast, when Phi was the sole source of P, the ptxD _Q overexpression plants grew normally as in the Pi-sufficient conditions (Figure 4). We compared the effects of Phi on the growth of transgenic rice under LP and HP conditions. Although Phi treatment inhibited the growth of non-transgenic plants under both LP and HP conditions, the ptxD _Q overexpression rice had similar plant heights as in the HP conditions under Phi treatment. Moreover, the fresh weight of ptxD _Q overexpression rice treated with Phi increased by approximately 14.9% and 30.7%, under LP and HP conditions, respectively, as compared to the plants grown under HP conditions (Figure 5). Together, these results indicated that overexpression of ptxD _Q conferred the ability to use Phi as a P resource in rice plants, which could further stimulate the growth of transgenic plants under both LP and HP conditions. We further tested the effect of spaying Phi on the growth of rice under Pi-sufficient condition. Spaying high concentration of Phi significantly inhibited the growth of WT plants (Supplementary Figure S7). Interesting, ptxD _Q overexpression rice plants are resistance to spaying Phi on the shoots (Supplementary Figure S7).

The Pi and Phi contents under different P treatment conditions were measured in the WT and transgenic plants. The Pi content showed no significant differences between WT and transgenic plants under LP and HP conditions. The addition of Phi in the nutrient solutions with LP and HP decreased the Pi content by 19.3% and 17.8%, respectively, in the shoots of WT plants (Figure 6A). However, Pi content in the shoots of transgenic plants increased by approximately 135% and 21.1% under Phi treatment for LP and HP conditions, respectively (Figure 6A). We could not detect any Phi in the WT and transgenic plants under LP and HP conditions without Phi applicant (Figure 6B). When Phi was added to the nutrient solution, its content in WT and transgenic plants was higher under the LP condition as compared to the HP condition. However, Phi content in WT was very low compared with that in transgenic plants when Phi was added into the nutrient solution under both LP and HP conditions (Figure 6B).