Synergy between a shallow root system with a DRO1 homologue and localized P application improves P uptake of lowland rice

Improved phosphorus (P) use efficiency for crop production is needed, given the depletion of phosphorus ore deposits, and increasing ecological concerns about its excessive use. Root system architecture (RSA) is important in efficiently capturing immobile P in soils, while agronomically, localized P application near the roots is a potential approach to address this issue. However, the interaction between genetic traits of RSA and localized P application has been little understood. Near-isogenic lines (NILs) and their parent of rice (qsor1-NIL, Dro1-NIL, and IR64, with shallow, deep, and intermediate root growth angles (RGA), respectively) were grown in flooded pots after placing P near the roots at transplanting (P-dipping). The experiment identified that the P-dipping created an available P hotspot at the plant base of the soil surface layer where the qsor1-NIL had the greatest root biomass and root surface area despite no genotyipic differences in total values, whereby the qsor1-NIL had significantly greater biomass and P uptake than the other genotypes in the P-dipping. The superior surface root development of qsor1-NIL could have facilitated P uptakes from the P hotspot, implying that P-use efficiency in crop production can be further increased by combining genetic traits of RSA and localized P application.

Synergy between a shallow root system with a DRO1 homologue and localized P application improves P uptake of lowland rice Aung Zaw Oo 1,3 , Yasuhiro Tsujimoto 1,3* , Mana Mukai 1 , Tomohiro Nishigaki 1 , Toshiyuki Takai 1 & Yusaku Uga 2 Improved phosphorus (P) use efficiency for crop production is needed, given the depletion of phosphorus ore deposits, and increasing ecological concerns about its excessive use. Root system architecture (RSA) is important in efficiently capturing immobile P in soils, while agronomically, localized P application near the roots is a potential approach to address this issue. However, the interaction between genetic traits of RSA and localized P application has been little understood. Near-isogenic lines (NILs) and their parent of rice (qsor1-NIL, Dro1-NIL, and IR64, with shallow, deep, and intermediate root growth angles (RGA), respectively) were grown in flooded pots after placing P near the roots at transplanting (P-dipping). The experiment identified that the P-dipping created an available P hotspot at the plant base of the soil surface layer where the qsor1-NIL had the greatest root biomass and root surface area despite no genotyipic differences in total values, whereby the qsor1-NIL had significantly greater biomass and P uptake than the other genotypes in the P-dipping. The superior surface root development of qsor1-NIL could have facilitated P uptakes from the P hotspot, implying that P-use efficiency in crop production can be further increased by combining genetic traits of RSA and localized P application.
Phosphorous deficiency restricts crop growth, particularly in the tropics, due to the inherently low P content of soils and the high P-fixing capacity of other minerals such as active Al-and Fe-oxides 1 . Large amounts of mineral P fertilizer have been continuously applied to overcome low P-use efficiency and achieve high grain yields. Given the finite nature of the P fertilizer resource and increasing ecological concerns about the excess use of P in agricultural systems [2][3][4] , it is vital to investigate sustainable crop production strategies that facilitate the efficient utilization of applied and available P in soils. Such strategies are also critical for the food security of resource-poor farmers with low fertilizer inputs in developing countries 5 .
Roots play a pivotal role in exploring immobile P in the soil. An increased root surface area with minimal carbon costs is one strategy, through the formation of finer roots, aerenchyma, and root hairs [6][7][8] . Changes in root system architecture (RSA) such as the development of surface roots is another root function to adapt to P deficiency, that is called 'topsoil foraging' , because P is most available in surface soil layers 9 . This topsoil foraging can be enhanced by a shallower growth angle of axial roots 9 , adventitious root abundance 10 , and many/short lateral root branching 11 . Field-based studies have demonstrated the yield advantages of genotypes with these architectural traits for several crops under P-deficiency 8 . Therefore, identification of key root traits and their genetic mechanisms and conferring genes or quantitative trait loci (QTL) should offer avenues for improving P acquisition efficiency in crop breeding 12 .
The agronomic approach for improving P-use efficiency includes localized fertilization, which refers to the placement of small amounts of fertilizers nearby the root zone. Several field experiments have demonstrated the positive impacts of localized P fertilization on grain yields and/or fertilizer use efficiencies for crop production (e.g., Vandamme et al. 13 ). Our recent study identified that applied P-use efficiency can be substantially improved by dipping seedling roots in P-enriched slurry at transplanting (P-dipping) in severely P-deficient rice fields in Madagascar 14 . The P-dipping transfers P with the slurry attached to seedling roots, creating a soluble P hotspot www.nature.com/scientificreports/ nearby the transplanted roots and facilitating plant P uptake, even under the high P-fixing soils of the tropics 15 . The use of P-dipping is currently being tested by hundreds of smallholder farmers in Madagascar. Despite a range of studies in both genetic and agronomic approaches, none have examined how the combination of RSA traits and localized fertilization would affect plant P-use and acquisition efficiencies. In the present study, we aimed to identify the combination effect by using near-isogenic lines (NILs) of DRO1, and its homologue (qSOR1), the major QTLs of rice controlling root growth angle (RGA). The parent variety, 'IR64' , is a high-yielding, modern variety with a relatively shallow RGA with the combination of the nonfunctional allele of DRO1, and the functional allele of qSOR1. The Dro1-NIL, developed by Uga et al. 16 , has a relatively deep RGA with the combination of functional alleles of both DRO1 and qSOR1. The qsor1-NIL, developed by Kitomi et al. 17 , has a shallower RGA than IR64, with the combination of nonfunctional alleles of both DRO1 and qSOR1. We hypothesize that P-dipping, creating a P hotspot at the soil surface, will have a positive interaction with the shallow root system in rice. By understanding the interaction, further research can be expected to improve applied P-use efficiencies by designing RSA traits for localized fertilizer application techniques.

Results
Shoot growth and P uptake. Localized P application via P-dipping (P dip ) achieved equivalent biomass and P uptakes at one fifth of the application rate of uniform P incorporation (P inco ) (Fig. 1). The ANOVA detected consistent and significant interactions between genotype and P treatment for shoot biomass and P uptakes at and shoot P uptake (B) of rice genotypes as affected by different P application methods (P incorporation (P inco ) of 500 mg P 2 O 5 box −1 vs. P-dipping (P dip ) of 90 mg P 2 O 5 box −1 ) at 21 days after transplanting (DAT), and 42 DAT. Different letters and ns within each treatment indicate significant and nonsignificant differences, respectively, among genotypes at 5% using Tukey's HSD test. Error bars represent the standard error of replications. The *, **, and *** indicate that the individual effects of and interaction between genotype (G) and P application method (P) are significant at P < 0.05, P < 0.01, and P < 0.001, respectively. www.nature.com/scientificreports/ both 21 days after transplanting (DAT) and 42 DAT. In the P dip treatment, qsor1-NIL consistently had greater shoot biomass and P uptake than Dro1-NIL. In contrast, in P inco , Dro1-NIL tended to have greater shoot biomass and significantly greater P uptakes than the other genotypes at 42 DAT. Applied P-use efficiency (calculated as the ratio of shoot P uptake at 42 DAT to the amount of P applied) increased from 3.4 to 16.2% for IR64 by changing the P application methods from P incorporation to P-dipping and further increased to 20.0% by using qsor1-NIL (data not shown).
Genetic root traits. The RSA traits among genotypes were consistent under P dip : the RGA was the shallowest in the order of qsor1-NIL > IR64 > Dro1-NIL at both 21 DAT and 42 DAT (Fig. 2). As a result of the RGA differences, qsor1-NIL developed a large proportion of root biomass and root surface area in the 0-3 cm layer and little in the 14-28 cm layer. In contrast, Dro1-NIL distributed a relatively large proportion of root mass in the 14-28 cm layer. For instance, at 21 DAT, qsor1-NIL developed 50.3% of the root mass in the 0-3 cm layer and only 2.0% in the 14-28 cm layer while these proportions were 32.7% and 10.3%, respectively, for Dro1-NIL. The root distribution pattern of IR64 was intermediate between qsor1-NIL and Dro1-NIL. The trend in RSA among genotypes was the same in P inco while IR64 and Dro1-NIL tended to have deeper RGAs than those in P dip (Fig. 3). The RGAs of qsor1-NIL, IR64, and Dro1-NIL at 21 DAT were 7.1°, 23.6°, and 33.3° in P dip and 5.0°, 39.8°, and 52.2° in P inco , respectively. By reflecting on the differences in RGA, qsor1-NIL had a greater root biomass, greater root surface area, and longer lateral root length than Dro1-NIL at both the center position (0 to 3 cm from the plant base on both sides), and side position (3 cm from the plant base to the edges of the box) in the 0-3 cm soil depth layer at 21 DAT (Figs. 4;5). On the other hand, Dro1-NIL had greater root surface area and longer lateral and nodal root length than qsor1-NIL at the side position in the 14-28 cm layer despite its significantly lower values in total for these parameters at 21 DAT (the mean values were also higher in the center position, but the differences were not statistically significant) (Fig. 4). IR64 was intermediate for these parameters in both the 0-3 cm and 14-28 cm layers. At 42 DAT, genotypic differences in root distribution patterns vertically and horizontally became less significant as root development and root growth angle were increasingly constrained by the size of the root box (Fig. S1). Yet, qsor1-NIL had consistently greater root mass and greater root surface area than Dro1-NIL at the side position in the 0-3 cm layer and vice versa in both center and side positions in the 14-28 cm layer.
Spatio-temporal dynamics in soluble P concentrations. Soluble P concentrations in soils were averaged across genotypes because there were no significant genotype differences in any sampling times or sampling layers. The P dip , in which high P solution was applied by spot at a depth of 1.5 cm from the soil surface, had a substantially large soluble P concentration at a depth of 3 cm (Fig. 6). The maximum P concentration at a depth of 3 cm for P dip was > 100 times greater than the other depths for both P treatments throughout the growing period. In P dip , soluble P concentrations were greater at a depth of 7 cm than at 21 cm in the latter growth stages, but apparently the vertical P diffusion from the 3 cm hotspot was relatively small. In contrast, the soluble P concentrations were significantly higher at a depth of 21 cm than at 7 cm in P inco after 28 DAT.

Discussion
The results support the hypothesis that the shallow root system of qsor1-NIL has a positive interaction with localized P application via P-dipping and that the combination additively improves applied P-use efficiency for initial rice growth. The other genotypes also reduced the RGA by 16-19° in response to the P hotspot (Figs. 2, 3), yet the synergy with P-dipping was greater in qsor1-NIL. This implies that breeding efforts to design the RGA in localized P spots can be more beneficial than relying on the intrinsic root plasticity of each genotype. It is also considered that root phenotypic adaptations to growing environments (as found in RGA changes of the other genotypes) may have a certain metabolic cost 18 .
Superior P uptake of qsor1-NIL with P-dipping is attributable to the greater root biomass, greater root surface area, and longer lateral root length, especially at the center position in the 0-3 cm soil layer where high soluble P is available throughout the growing period. This is most likely the same mechanism as topsoil foraging, prioritizing the root development in the P-rich domains to efficiently capture immobile P in soils. Spatio-temporal P variations in the P-dipping indicate that applied P mobility is highly restricted despite a general understanding that P becomes less immobile under flooded conditions 19 , emphasizing the importance of RSA for the localized P acquisition, even under flooded soil culture. The relative immobility of applied P even in irrigated lowlands was also reported by Akahane et al. 20 , in which they detected high amounts of P retained up to harvest within a small area of 2-3 cm (vertical) by 4-5 cm (horizontal) centered on the application spot. This area of P distribution surrounding the application spot under the flooded soil culture corresponds to the center position of the 0-3 cm soil layer surrounding the P-applied spot of the P dip treatment in the present study.
The effect of topsoil foraging itself has been reported in several upland crops 11,21-23 , but not in rice. Previous studies detected no significant effects of root distribution patterns or RGA for rice P acquisition under P deficiency (e.g., Mori et al. 24 ), which may be due to the materials differing not only in root system architecture but in other traits or in more complex screening environments. The present study had an advantage in using NILs that differed in RGAs (but were otherwise equivalent phenotypes 17 ), under non-water-stressed, and greatly uneven P availability because of P-dipping.
In addition, the present study detected a positive effect of Dro1-NIL on P uptake under uniform, P-sufficient conditions. The reason for this positive interaction should be further explored but can be related to consistent P acquisition from the P-rich subsoil layers after the depletion of available P in topsoil layers (Fig. 6). Another potential reason is the more efficient acquisition of other nutrients, such as N, which are vertically more mobile than P. Deep rooting has been reported as a positive trait for N acquisition in upland crops 8   www.nature.com/scientificreports/ It should be noted that crop production environments are complex with multiple abiotic stresses, particularly on smallholder farms in developing countries where stress-resilient and nutrient-efficient technologies are most needed. In this respect, field-based experiments to maturity are further required to confirm the effect of the combination of genetic RSA traits and P fertilizer management practices. The combination of shallow roots and localized P application can never be a silver bullet. A careful selection of field environments where P deficiency is the primary limiting factor is needed to effectively apply this combination, ideally together with the development of bimodal root phenotypes (shallow and deep), or high RSA plasticity 27 against complex growing environments.  www.nature.com/scientificreports/ In rice, qSOR1 and DRO1 can be promising genetic resources for the development of such bimodal root phenotypes, without increasing costs of root elongation but by controlling the root growth angles 28,29 . A combination of shallow root system and RSA plasticity may be another possible trait ensuring both efficient P uptake from localized spots, and flexible responses to unpredictable changes in growing environments.

Conclusion
The study provides significant evidence that a shallow root system has a positive interaction with localized P application nearby the root at transplanting by using NILs differing in their RGA. The combination substantially improves applied-P use efficiency for initial rice growth. This finding should encourage relevant research focusing not only on physiological root traits or agronomic management approaches, but on their combination to address to the global issue of increasing crop production while minimizing the environmental impacts.

Materials and methods
Experimental design and treatments. The experiment was conducted in a greenhouse with an automatic ventilation system at the Japan International Research Center for Agricultural Science (JIRCAS), Tsukuba, Japan. The average daytime and nighttime temperatures during the experiment ranged from 26.2 to 35.8 °C and 24.7 to 28.7 °C, respectively (Thermo Recorder TR-50U2, T&D Corporation, Japan). The soil for the experiment was collected from a subsoil layer (40-50 cm in depth) at the JIRCAS Tropical Agricultural Research Front, Okinawa, Japan. Physicochemical properties of experimental soil are summarized in Table 1. The soil was silty clay with a pH (H 2 O) of 4.86 and low available P content, and high P retention capacity with abundant active Al and Fe oxides. The soil was air-dried and passed through an 8 mm sieve prior to the experiment.
Two different P treatments, sufficient P incorporation (P inco ) and localized P application via P-dipping (P dip ), were factorially combined with three rice genotypes in a randomized complete block design with seven replications. For both treatments, NH 4 NO 3 and K 2 SO 4 were mixed with soils and puddled in a bucket at a rate of 220 mg N box −1 , and 220 mg K 2 O box −1 to develop uniform and N-and K-sufficient conditions. For the P inco , triple super phosphate (TSP) was added at puddling. The mixed soils were filled into a root box at a rate of 500 mg P 2 O 5 box -1 to develop a uniform and P-sufficient condition. The P application rate of the P inco treatment was determined based on Oo et al. 15 to expect similar levels of plant P uptakes and biomass production with the P dip treatment for the comparison of root development. The root box was made of transparent acrylic sheets with a size of 30 cm height × 30 cm length × 3 cm width. The soil was added to the box to a depth of 28 cm.
For the P dip treatment, a P solution was placed in a spot nearby the transplanted root zone to apply the exact amount of P in all boxes. We estimated the amount of P-enriched slurry transferred or attached to seedling roots at transplanting as 90 mg P 2 O 5 box −1 based on our previous study 30 . After the N-and K-added soil was filled in the root box, 90 mg P 2 O 5 as TSP dissolved in 20 ml water was injected into the soil at a depth of 1.5 cm in the center of the root box (Fig. 5). On the same day of these P treatments, one 10-day old seedling was transplanted in the middle of each root box and grown under continuously flooded conditions. Measurement. Soil solution samplers (DIK-8393, Daiki Rika Kogyo Co. Ltd., Japan) were installed in one side of the acrylic board in the middle of the 3 cm, 7 cm, and 21 cm depths for the P dip treatment and at 7 cm and 21 cm for the P inco treatment for four out of the seven replicates (Fig. 5). Based on our previous observation 15 , we assumed that the spatial variation in soluble P concentration were relatively small in the box because we thoroughly mixed P with soil at the time of puddling and thus omitted the measurement at 3 cm in the P inco treatment.
Soil water samples were collected at 3, 7, 14, 21, 28, and 35 DAT. The samples were analyzed for soluble P concentration as an index of P available to plants using a microplate reader spectrophotometer at an absorbance of 630 nm by following the Malachite Green method 31 . Table 1. Physicochemical properties of soil. a NC analyzer, Sumigraph NC-220F (SCAS, Japan); b The proportion of absorbed P after shaking 5 g of soil with 25 ml of 1000 ppm P solution for 24 h; c UV spectrophotometer (UV-1800, Shimadzu); d Inductively coupled plasma mass spectrometer (ICPE-9000, Shimazu, Japan) after oxalate extraction. e Sieving and pipetting method. www.nature.com/scientificreports/ Three and four replicates were harvested at 21 DAT and 42 DAT, respectively. At each harvest time, shoots were cut at ground level and oven-dried at 70 °C for > 48 h to determine shoot biomass. Shoot P concentration was measured with the molybdate blue method 32 after dry-ashing at 550 °C for 2 h and digestion with 0.5 M HCl. Shoot P uptake was calculated by multiplying the P concentration and shoot biomass.
After shoots were removed, root samples were collected using pin-board method as per Kano-Nakata et al 33 . In brief, roots were pinned with a 5 mm mesh net and pinboard after which soils were washed off and digital images were taken. The RGA was determined from the digital image taken by a commercial camera (D7000, Nikon Corp., Japan) as the angle from the soil surface to the shallowest nodal root using ImageJ software (Version 1.52a, NIH, USA). The root system was then divided into 12 compartments or into the center (0 to 3 cm from the plant base) and both sides of the 0-3 cm, 3-7 cm, 7-14 cm, and 14-28 cm soil layers to assess spatial root distributions (Fig. 5). The values of both sides (left and right) were summed as there must be no physiological meaning in the difference between these two. Root length and surface area of each compartment were measured using Epson Pro-selection X980 Scanner and WinRhizo Pro software (Regent Instruments, Quebec, Canada). Roots were classified as lateral roots (< 0.2 mm) as per Sandhu et al. 27 and nodal roots (0.2 to 2 mm) as per Kano-Nakata et al. 34 . Roots of > 2 mm were excluded from the analysis, as they were too large for a single root diameter and most likely occurred as a result of a measurement error. After the morphological analysis, root biomass of each compartment was determined by oven-drying at 70 °C for > 48 h. Statistical analysis. JMP software (v14.0.0, SAS Institute Inc., Japan) was used to perform the statistical analyses. The treatment means were compared at 5% level of probability using Tukey's honestly significant difference (HSD) test after the single and/or interaction effects of genotypes and P treatment were confirmed by a generalized linear model.

The use of plant materials.
Near-isogenic lines (NILs) and their parent of rice (qsor1-NIL, Dro1-NIL, and IR64) that we used in the present experiment was transferred from National Agricultural Research Organization (NARO) to Japan International Research Center for Agricultural Sciences (JIRCAS) by the Joint Research Contract, and the experiment was conducted by compiling with the guideline and regulation of this contract.