Plastid-localized amino acid metabolism coordinates rice ammonium tolerance and nitrogen use e�ciency

Ammonium toxicity affecting plant metabolism and development is a worldwide problem impeding crop production. Remarkably, rice (Oryza sativa L) favors ammonium as its major nitrogen source in paddy �elds. We set up a forward-genetic screen to decipher the molecular mechanisms conferring rice ammonium tolerance and identi�ed the rohan mutant showing root hyper-sensitivity to ammonium due to a missense mutation in an arginine-succinate lyase-encoding gene. ROHAN localizes to plastids while its expression is induced by ammonium. ROHAN alleviates ammonium-inhibited root elongation by converting the excessive glutamine into arginine. Consequently, arginine leads to auxin accumulation in the root meristem thereby stimulating root elongation under high ammonium. Furthermore, we identi�ed natural variation in the ROHAN allele between Japonica and Indica subspecies explaining their different root sensitivity towards ammonium. Finally, we show that ROHAN expression positively correlates with root ammonium tolerance and that nitrogen use e�ciency and yield can be improved through a gain-of-function approach.


Introduction
Nitrogen (N) is one of the essential nutrients for plant growth and crop yield because its metabolism in vivo leads to the production of diverse organic compounds, such as proteins, nucleic acids, chlorophyll, and plant hormones. Nitrate (NO 3 -) and ammonium (NH 4 + ) are two primary inorganic N sources for plants. Nitrate is reduced into NH 4 + by ferredoxin-dependent nitrite reductases, consuming NADH or NADPH as reductants, while NH 4 + can be directly incorporated into amino acids via the GS-GOGAT pathway 1 . Thus NH 4 + is considered a more e cient and economical N fertilizer than NO 3 to promote crop yield 2 . However, excessive NH 4 + causes severe toxic effects on plant shoot and root growth 3 .
Because the intensive use of NH 4 + -based fertilizers in the past decades has resulted in dangerous accumulation of NH 4 + in agricultural soils 4 , improving plant tolerance to NH 4 + is key to the enhanced utilization of NH 4 + and improve N use e ciency (NUE).
The mechanism underlying plant NH 4 + tolerance have been extensively investigated in Arabidopsis through genetic approaches, and NH 4 + toxicity is suggested to be a consequence of repressed metabolic processes 3 , protein glycosylation 5 , chloroplast development 6 , hormone metabolism [6][7][8] , glucosinolate metabolism and Fe homeostasis 9 . Recent studies have also revealed that NH 4 + toxicity is coupled with NH 4 + uptake and glutamine (Gln) accumulation by reducing root apoplast pH and provoking severe acidic stress to Arabidopsis 10,11 . In Arabidopsis mutants showing reduced NH 4 + uptake (for instance, the ammonium transporters AMTs quadruple mutant) or Gln synthesis (e.g. gln2-1), the toxic effects of NH 4 + on seedling growth are vastly relieved. The NH 4 + toxicity can also be alleviated when the NRT1.1mediated nitrate in ux is reduced or the SLAH3-mediated nitrate e ux is enhanced, which both increase external nitrate levels to buffer the rhizosphere pH 12,13 . Together, these results suggest that root acidi cation is the primary cause of NH 4 + toxicity in Arabidopsis.
Rice is considered as a NH 4 + tolerant species since NH 4 + is the primary N source for rice in the paddy eld, where the nitri cation process is low 14 . AMT-mediated NH 4 + uptake plays a predominant role in promoting rice shoot growth and yield 15 but negatively regulates root system architecture, in particular by reducing rice root elongation and gravitropism 16,17 . Several studies suggested that the inhibitory effect of NH 4 + on rice root elongation is probably due to its impact on the biosynthesis of plant hormones such as auxin, ethylene, and brassinosteroids, which are critical for sustaining root development 11,18,19 . Our recent work also showed that root acidi cation resulting from NH 4 + uptake triggered the asymmetric auxin distribution in the root cap, leading to the loss of root gravitropism under high NH 4 +11 . However, root elongation inhibited by NH 4 + could not be fully mitigated by buffering the rhizosphere pH or by supplying external nitrate supply, thus suggesting a different impact of NH 4 + in root development between Arabidopsis and rice.
In order to decipher the molecular machinery allowing rice to circumvent NH 4 + toxicity on root growth, we employed a forward genetic approach and identi ed ROOT HYPERSENSITIVE TO AMMONIUM NITROGEN (ROHAN) as a key regulator of NH 4 + tolerance in rice. We further show that ROHAN encodes a plastid-localized argininosuccinate lyase which metabolizes excessive Gln into arginine (Arg), by doing so, alleviating NH 4 + -inhibition of root elongation. Additionally, we show that a substitution in the ROHAN coding sequence present of a particular rice cultivar can confer root tolerance to high NH 4 + . Finally, we demonstrate that ROHAN not only help to cope with NH4 + toxicity but also promotes rice yield and NUE under low and high N supplies, which represents an important target for rice molecular breeding of NH 4 + tolerance and NUE.

Results
Rice rohan mutants display root growth hypersensitivity to ammonium To discover the genetic control of rice root responses to NH 4 + , an EMS-mutagenized population (10,000 mutant lines) was generated starting from a local elite Japonica cultivar Wuyungeng 7 (hereafter designated as the wild-type). We then screened for mutants showing altered root length when grown in hydroponic solutions supplied with 2.5 mM NH 4 + , a concentration that is known to cause inhibition of seminal root (SR) elongation 11 . Because NH 4 + uptake results in proton release 11 , MES was supplemented in order to exclude root phenotypes caused by lowering pH. We identi ed one mutant displaying severely inhibited SR elongation when compared with the wild-type under high NH 4 + supply (73.9% shorter than the wild-type SR length). In absence of N however, the mutant SR length was slightly shorter than the wild-type (35.9% decrease) ( Fig. 1a and b). We further observed that the highest sensitivity of SR elongation to NH 4 + in the mutant occurred at 2.5 mM NH 4 + , and that SR elongation was less sensitive to varying NO 3 concentrations (Extended Data Fig. 1a). A time-course experiment moreover showed that the inhibitory effect of NH 4 + on SR elongation in the mutant was more pronounced after three days of treatment than one day (Extended Data Fig. 1b). Furthermore, the root phenotype of this mutant was consistently observed in the paddy eld supplied with low N (75 kg/ha urea) and high N supply (350 kg/ha urea) ( Fig. 1e and f). Because of the striking inhibition of root growth under ammonium supply, we named the mutant root hypersensitive to ammonium nitrogen (rohan).
Root elongation is determined by the meristem cell division activity and the subsequent cell elongation.
Thus, we investigated the root meristem development in rohan under different N conditions. Compared to the N-free control condition, NH 4 + treatment caused a signi cant reduction of both meristem length and cortical cell number in wild-type, and this inhibitory effect on meristem division was enhanced in rohan ( Fig. 1c and d). In contrast, the cortical cell elongation was not affected in rohan neither for N-free or NH 4 + supply (Extended Data Fig. 1c and d). We further determined the root meristem activity of the mutant rohan by employing an EdU-based S-phase (DNA synthesis phase) assay 20 . NH 4 + treatment led to weaker EdU signal in the root tips of both the wild-type and rohan than under N-free condition. Nevertheless, root meristems of rohan showed less EdU signal than wild-type under high NH 4 + supply ( Fig. 1c and d). These results indicated that root phenotype of rohan resulted from reduced meristem activity under high NH 4 + supply.
Next, we investigated whether the hypersensitive rohan root response to NH 4 + toxicity was caused by NH 4 + accumulation and the herewith associated in vivo acidic stress. Transcription analysis showed that the respective expression levels of AMT1.1, AMT1.2, AMT1.3, and AMT2.1 in the roots were similar between wild-type and rohan (Extended Data Fig. 2a). Consistently, the rohan mutation had a minor effect on NH 4 + uptake of roots and it also showed a lower root proton in ux than the wild-type (Fig. 1g, and Extended Data Fig. 2b). Additionally, rohan exhibited an equal reduction in root elongation under low pH treatment as wild-type (Extended Data Fig. 2c and d). These results therefore suggested that the root response of rohan to NH 4 + was most likely not a consequence of NH 4 + uptake and acidic stress.
ROHAN encodes an N-responsive and plasmid-localized argininosuccinate lyase To identify the mutation causing the root hypersensitivity of rohan to ammonium, we performed a MutMap analysis using an F 2 segregating population of rohan backcrossed with the parental line. We found that the rohan mutation was inherited as a single recessive mutation by observing the root phenotype of F 1 progeny and self-pollinated F 2 population. The seedlings of the F 1 progeny showed a parent-like root sensitivity to NH 4 + treatment, and F 2 progeny phenotypically segregated in nearly 3:1 ratio for a respective NH 4 + sensitive and hypersensitive root phenotype (χ2 = 0.58; P < 0.05) (Extended Data Fig. 3a and b). Detailed bulked genomic sequencing and MutMap analysis on the F 2 progeny identi ed a non-synonymous single base change (C to T) at position 1706-bp downstream of the start codon of Os03g05500 ( Fig. 2a and Extended Data Fig. 3c). The gene encodes an argininosuccinate lyase (ASL), which is known to catalyzes the conversion of argininosuccinate into arginine and fumarate 21 . ROHAN/ASL gene is highly conserved within the plant kingdom (Extended Data Fig. 4a).
In rice, ROHAN/ASL contains seven exons, and the mutation in rohan results in a Pro211Leu (P211L) substitution located at the third exon (Fig. 2a). Pro211 is highly conserved among plant species (Extended Data Fig. 4c). By structural modelling the ROHAN/ASL protein based on the ASL crystal structure in Bacillus coli, we found the Pro211 at the hinge region linking protein domains ( Fig. 2f and Extended Data Fig. 4c). To con rm that the P211L mutation is responsible for ROHAN function and root response to NH 4 + , we further performed genetic complementation by transforming rohan mutants with VENUS-fused ROHAN/ASL wild-type protein driven by its native promoter (proROHAN:ROHAN-VENUS rohan). Over 24 individual lines were obtained, and lines with ROHAN-VENUS expression in the root showed identical root sensitivity to NH 4 + as the wild-type ( Fig. 2b and c). Additionally, knock-out lines of ROHAN/ASL1, generated by CRISPR-cas9, mimicked the root phenotype of rohan under high NH 4 + supply ( Fig. 2b and c). Thus, ASL is responsible for rohan root phenotype while its function is dependent on the Pro211 residue.
By generating a transcriptional reporter line of ROHAN in the wild-type background, we observed a strong expression of proROHAN:GUS in the root meristem and the stele of the elongation zone. proROHAN:GUS was also found to be expressed in lateral root primordia and emerged lateral roots (Extended Data Fig.  4b), indicating a critical role of ROHAN in regulating rice root development. In the complemented lines, the ROHAN-VENUS signal was also highly detected at the root meristem and localized to plastids at the cellular level (Fig. 2d). The Pro211Leu (P211L) did not affect the subcellular localization of ROHAN ( Fig.  2e) and suggested the Pro211 residue is key to its protein function in plastids.
We next investigated whether ROHAN expression was regulated by external NH 4 + supply. qRT-PCR analysis showed that ROHAN expression in roots gradually increased by NH 4 + application over 24h, while it was less induced by NO 3 treatment (Fig. 2g). This was further con rmed by the expression analysis of proROHAN:GUS, showing a strong induction of ROHAN expression in the root meristem by NH 4 + after 24h ( Fig. 2h). All together, these data demonstrate that ROHAN is a NH 4 + responsive gene that encodes a plasmid-localized protein regulating root development.
Metabolic defect of NH 4 + /Gln conversion to Arg in rohan causes its root hypersensitivity to ammonium The ROHAN encoded ASL has previously been shown to play a role in amino acid (AA) metabolism by catalyzing the last step of Arg biosynthesis. Furthermore, Arg produced by ASL is believed to be required for normal root elongation in rice 22 suggesting a possible link between ROHAN function in plastids and root tolerance to NH 4 + . To address the latter, we measured the content of 20 free AAs in roots of rohan and wild-type under N-free condition and high NH 4 + supply. NH 4 + treatment increased the overall content of AAs, especially Asn and Gln both in wild-type and rohan roots (Extended Data Fig. 5a). Remarkably, upon high NH 4 + supply, more Gln and free NH 4 + accumulated in roots of the rohan and rohan CR 4 mutants than in the wild-type. The opposite trend was observed for the Arg content, being more reduced in the mutants than in the wild-type roots (Fig. 3a). NH 4 + is assimilated to Gln by Glutamine synthetase (GS), and Gln is subsequently converted to Arg by the Urea cycle ( Fig. 3b) 23,24 . Thus, the increased Gln and NH 4 + in rohan roots might be due to the perturbation of Arg synthesis. In the complementing line of rohan, the contents of Gln, NH 4 + and Arg in the root were all recovered to the wild-type levels ( Fig. 3a and Extended Data Fig. 5a), con rming that ROHAN acts on the conversion of NH 4 + /Gln to Arg.
Next, we questioned if the observed metabolic defect was associated with rohan root sensitivity to NH 4 + by testing the effects of Gln and Arg application on root growth. We observed a substantial inhibition of SR meristem cell division and elongation at 0.3 mM Gln treatment in the wild-type. In rohan however, Gln concentration as low as 0.03 mM Gln already caused signi cant inhibition of SR meristem cell division and elongation (Fig. 3c, d, h, i and Extended Data Fig. 5g). This suggested that root elongation of rohan is hypersensitive to Gln, mimicking its response to NH 4 + . Interestingly, the addition of methionine sulfoximine (MSO), a potent inhibitor of GS 25 , restored SR elongation in both the wild-type and rohan under high NH 4 + supply ( Fig. 3c and d). We further used CRISPR-Cas9 to generate quadruple mutants of AMTs (qko) 26 and a gs1;1 single mutant, which are defective in Gln synthesis (Extended Data Fig. 5b and c) 27 . Both qko and gs1;1 mutants showed resistance to NH 4 + treatment towards SR elongation ( Fig. 3e and f) indicating that Gln accumulation is a major factor responsible for NH 4 + toxicity to root elongation in rice.
Intriguingly, the arrested SR meristem cell division and elongation in rohan under high NH 4 + could be rescued by the supply of Arg, with the optimum concentration at 0.3 mM, while it had a minor effect on SR meristem cell division and elongation in the wild-type (Fig. 3d, h, i, and Extended Data Fig. 5d).
However, external Arg supplies could not rescue the NH 4 + -and Gln-inhibited SR elongation in the wild-type (Extended Data Fig. 5e and f), suggesting that endogenous Arg, produced by ROHAN, is required for root elongation under high NH 4 + . Thus, we concluded that the biological function of ROHAN in converting excessive NH 4 + /Gln to Arg in vivo, enables the alleviation of NH 4 + inhibited rice root elongation.

ROHAN promotes nitrogen metabolism and photosynthesis under high ammonium supply
We next aimed to unveil the mechanism underlying ROHAN-mediated root tolerance to NH 4 + by performing RNA-seq analysis of root meristems in rohan and wild-type under N-free condition and NH 4 + supply. In agreement with the hypersensitive root phenotypes of rohan mutants under high NH 4 + , a more diverged transcriptional pattern between rohan mutants and the wild-type under high NH 4 + compared to N-free was observed (Extended Data Fig. 6a).
To specify the molecular pathways downstream of ROHAN, we analyzed differentially expressed genes in the wild-type and rohan under NH 4 + and N-free conditions. NH 4 + treatment led to a higher number of differentially expressed genes in rohan than wild-type after 1d treatment, and the number of differentially expressed genes in rohan was increased mainly after 3 days of treatment (Extended Data Fig. 6b). Gene set enrichment analysis of these differentially expressed genes based on GO and KEGG suggested that the N metabolism pathway was signi cantly affected in rohan (Extended Data Fig. 6c), consistent with its altered amino acid metabolism. Interestingly, genes associated with ammonium assimilation and Gln synthesis, including three GS1 members and GDH1, were all found to be signi cantly down-regulated in rohan under NH 4 + treatment (Extended Data Fig. 6d and e), which might result from a feedback inhibition by the accumulated Gln in rohan.
In addition, photosynthesis pathway genes were also signi cantly enriched in rohan under high NH 4 + (Extended Data Fig. 6b). Increasing N supply can promote photosynthesis in rice 28,29 . Consistently, high NH 4 + supply signi cantly induced photosynthetic rate, transpiration rate, stomatal conductance and PSII in wild-type seedlings, severely repressed in rohan, correlating with retarded shoot growth in rohan mutants under NH 4 + supply (Extended Data Fig.7b). The photosynthesis and shoot growth de ciency in rohan was recovered in the complementation line or by external Arg supply (Extended Data Fig.7a). These results show the importance of ROHAN for e cient AA metabolism in photosynthesis.
The rohan root response to NH 4 + is mediated by auxin metabolism Interestingly, gene set enrichment analysis of differentially expressed genes revealed that hormone metabolism was also affected in rohan under high NH 4 + (Extended Data Fig. 8a). Among plant hormones, auxin plays a crucial role in regulating root elongation and responses to N sources and photosynthesis 30,31 . The expression of several genes involved in auxin conjugation (i.e., GH3s) and auxin signaling (i.e., ARF2, IAA6, IAA14, and small auxin-up RNAs, SAURs) was found to be down-regulated in rohan under high NH 4 + (Fig. 4a). GH3s and SAURs are auxin-inducible genes 32 , thus the changes in their expression indicated altered endogenous auxin levels in rohan. By measuring endogenous IAA content, we indeed detected a signi cant reduction of IAA content in the root tip of rohan as compared to that of in the wild-type under high NH 4 + supply, while it was only slightly reduced in rohan under N de ciency (Extended Data Fig. 8b). In the complementation line of rohan, IAA content in the root tip was fully recovered (Extended Data Fig. 8b), showing the importance of ROHAN for endogenous auxin metabolism in the root tip.
To reveal spatial changes in auxin levels in root tips, we introduced DR5rev:3xVENUS, a sensitive auxin reporter 33 , into rohan mutants by crossing. The DR5 signal was strong in the root epidermis and stele of both the wild-type and rohan in N-free condition. Under high NH 4 + treatment, wild-type roots had a lower DR5 signal in the stele but an increased signal in the epidermis, whereas it almost disappeared from the root stele and epidermis of rohan. Interestingly, exogenous Arg elevated DR5 expression in the root stele and epidermis of rohan mutants (Fig. 4b, c and Extended Data Fig. 8c), corresponding to its effect in restoring SR elongation in rohan mutants. These results indicate that the changes of auxin levels in the root tip are correlated with the ROHAN-mediated SR elongation in response to NH 4 + .
Regulation of the root meristem activity and root elongation in response to N sources is mediated by auxin e ux carrier PIN-dependent polar auxin transport (PAT) 34 . Accordingly, we also found that several PIN encoding genes were transcriptionally repressed in rohan mutants under high NH 4 + (Fig. 4a). We further validated the role of PINs in mediating rohan root response to NH 4 + by using two wellcharacterized potent polar auxin transport inhibitors, 1-N-naphthylphthalamic acid (NPA) and BUM 35,36 , in order to bypass the functional redundancy of PIN proteins. Under high NH 4 + supply, rohan mutants showed higher sensitivity to NPA and BUM with respect to SR elongation, as this was completely arrested at low concentration of NPA and BUM ( Fig. 4d and e). Following inhibition of SR elongation by NPA, NPA treatment resulted in a higher accumulation of DR5 expression in the root epidermis of the wild-type upon NH 4 + treatment (N-free), while this induction was blocked in rohan. Moreover, exogenous IAA failed to trigger DR5 expression and root elongation in rohan mutants, suggesting a complete suppression of PIN activity in rohan mutants ( Fig. 4b and Extended Data Fig. 8c). We further generated a CRISPR-Cas9 knock-out mutant of PIN1a and PIN1b, whose expression abundances were signi cantly repressed in rohan mutants under high NH 4 + . The pin1apin1b mutants showed shorter root meristems and SRs, mimicking the rohan mutant root phenotype ( Fig. 4g and i), con rming the requirement of PIN-mediated polar auxin transport for the ROHAN-dependent effect on SR elongation. Altogether, our data demonstrated a critical role of auxin metabolism and polar auxin transport underlying ROHAN-controlled root tolerance to ammonium stress.
ROHAN allele confers rice root tolerance to NH 4 + toxicity Because of the signi cant role of ROHAN in NH 4 + root tolerance, we further investigated its genetic variation in rice germplasm. By nucleotide diversity (π) and Fst analysis, we found evidence for decreased nucleotide diversity of ROHAN in the japonica and indica subpopulations as compared to the wild rice population (Oryza ru pogon), while regions around the ROHAN locus showed signi cant differentiation between subspecies japonica and indica, indicative for a possible result of selective pressure during domestication ( Fig. 5a and b). Detailed analysis showed 37 SNP variants, including 35 introns, one missense, and one synonymous within the ROHAN gene. We further focused on the missense SNP (Chr3:10847318, c.3972A>G), which showed apparent Indica (94.91%)-Japonica (96.76%) differentiation and led to a Lys470Arg substitution ( Fig. 5d and e). The allele frequency of the SNP from japonica, indica I, indica II, indica III, aus, and aromas subpopulations were calculated and shown on the geographic map (Fig. 5c).
To validate the function of the SNP, we randomly selected in total 100 SNP A -variant and SNP G -variant accessions to evaluate their root sensitivity to high NH 4 + . Our results revealed a higher root sensitivity to NH 4 + in SNP G -variant accessions than SNP A -variant accessions (Fig. 5f). Furthermore, ROHAN expression in SNP G -variant accessions was less induced than SNP A -variant accessions by high NH 4 + (Fig. 5g) . 6a and b). In contrast, rohan displayed lower expression of ROHAN in roots and shorter SRs than wild-type (Extended Data Fig. 9b). These results underline once again the importance of ROHAN expression for root tolerance to NH 4 + .
We further planted the mutant and overexpression lines of ROHAN in the paddy eld supplied with low (75 kg/ha), moderate (150 kg/ha), and high nitrogen (350 kg/ha). Compared to the wild-type, ROHAN overexpression led to a signi cant increase in tiller number, grain yield, grain proteins and nitrogen use e ciency under all N conditions, whereas rohan mutants showed lower tiller number, yield, grain proteins and NUE ( Fig. 6c and d). Collectively, these results show that ROHAN expression positively correlates with root tolerance to high NH 4 + and NUE.

Discussion
Ammonium nitrogen can cause severe toxic effects on various plant physiological and developmental processes. In particular, NH 4 + represses growth of the root system, which in turn leads to reduced nutrient and water uptake and therefore dramatically decreases crop yield. Improving plant tolerance to NH 4 + thus represents a promising strategy to improve crop nitrogen use e ciency and yield. Although previous studies in rice have shown that reduced ammonium uptake or Gln synthesis could considerably alleviate the NH 4 + -inhibition of root growth, they also reported on retarded shoot growth, indicating an antagonistic interplay between NH 4 + tolerance and NUE 26,37 . Here, we identi ed ROHAN, a gene encoding a plastidlocalized ASL, essential in determining the NH 4 + tolerance of rice roots. Our study revealed that NH 4 + toxicity impeding rice root growth results from the accumulation of NH 4 + /Gln in roots. ROHAN/ASL is transcriptionally induced in the root meristem by NH 4 + and facilitates the conversion of over-accumulated NH 4 + /Gln into Arg, which is essential for auxin homeostasis and root elongation under high NH 4 + (Fig.   6e). Signi cantly, induction of ROHAN expression not only promotes rice root tolerance to NH 4 + but also increases the NUE and yield under low and high N supplies in the paddy eld. Therefore, our results suggest that the ROHAN-mediated Gln to Arg conversion coordinates root tolerance to NH 4 + toxicity and NUE.
Amino acids are a primary storage form of N and exert diverse functions in plant development and defense 38 . NH 4 + is assimilated via the GS-GOOAT pathway into Gln and further converted into other amino acids. The ROHAN encoded ASL is one of the rate-limiting enzymes of the urea cycle and is utilized in the production of arginine 21 . ROHAN shares the same plastidial localization with GS1;1 10 ( Fig.   2), a critical enzyme for Gln synthesis. By analyzing the amino acid content, ROHAN was found to act speci cally in the conversion pathway of Gln to Arg (Fig. 3), indicating a critical role of the urea cycle in Gln metabolism in plastids. In Arabidopsis, the elevated Gln content and root apoplast acidi cation by high NH 4 + supply has been suggested as the primary cause of NH 4 + toxicity 10 . Consistently, Gln had a similar inhibitory effect on root growth in rice, and rohan mutant roots show hypersensitive to external Gln supply, which can be suppressed when Gln synthesis is attenuated (Fig. 3). These results demonstrate that NH 4 + -induced accumulation of endogenous Gln is a common mechanism underlying NH 4 + toxicity in rice root development. However, rohan mutant roots exhibited normal sensitivity to low pH (Extended Data Fig. 2), arguing that ROHAN-mediated root tolerance to NH 4 + is unlikely linked to root acidi cation.
The regulation of N in plant root development is mediated by phytohormones such as auxin, ethylene, and brassinosteroid 11,39,40 . Auxin is a prominent regulator of root meristem activity and root elongation, and also interacts with other hormones to regulate root development 41 . Our transcriptome data revealed a robust transcriptional regulation of ROHAN in auxin signaling pathways (Fig. 4). We demonstrated that ROHAN plays a role in maintaining high auxin levels and spatial distribution at the root tip, required for root elongation under high NH 4 + . Local auxin accumulation is mediated by auxin e ux carrier PINdependent polar auxin transport. We further identi ed PIN1 as a downstream signaling component of ROHAN. PIN1 activity is required for rice root meristem activity and SR elongation. PIN1 is transcriptionally activated by ROHAN to facilitate the tissue-speci c auxin accumulation at the root tip to stimulate root elongation. In Arabidopsis, high ammonium was also found to affect PIN activity.to adjust root growth 34 . Therefore, our study suggests a molecular link between the N metabolic pathway and the phytohormone signal in regulating rice root tolerance to ammonium. Further investigations will be needed to uncover the mechanism underlying the transcriptional regulation of ROHAN and NH 4 + in auxin homeostasis.
ROHAN is highly conserved in various plant species. In this study, we further identi ed a natural variant of ROHAN (SNP A ) in rice that increases root tolerance to high external NH 4 + . This tolerance allele SNP A is mainly distributed in Japonica cultivars but is absent from Indica cultivars, correlating with the higher root tolerance of Japonica cultivars as compared to Indica cultivars (Fig. 5). As ROHAN expression is highly elevated by high NH 4 + in Japonica cultivars in contrast to Indica cultivars, the elite allele of ROHAN might determine root tolerance to NH 4 + toxicity through regulating its expression in response to external NH 4 + supply.
In summary, we identi ed ROHAN as a critical regulator of NH 4 + tolerance and NUE in rice. Our study reveals a molecular link between amino acid metabolism and root response to external N supply and suggests that polar auxin transport-mediated local auxin accumulation in the root meristem is required for ROHAN function. As rice varieties harboring the elite variant of ROHAN display a higher root tolerance to NH 4 + without penalty in plant growth and yield, ROHAN may represent a potential candidate target for genetic editing or marker-assisted breeding to develop crop varieties tolerant to ammonium toxicity and help to increase rice NUE and yield under the large quantities of urea-based fertilizer. The identi cation of ROHAN therefore represents an important element to set up strategies to cope with the excessive use of N fertilizer that has resulted in several environmental issues, such as surface water eutrophication, groundwater pollution, greenhouse gas emission and soil acidi cation. to the nalization of the manuscript.

Con ict of interest statement
All authors state no con ict of interest concerning this manuscript.

Plant materials
Rice (Orzya sativa cv. Wuyungeng7) seeds were mutagenized by treating them with 1.0% ethyl methylsulfonate (EMS) for 12 h as previously described 42 . M 2 seeds obtained from self-pollinated M 1 plants were used for the screening of NH 4 + -sensitive mutants based on root elongation at 2.5 mM ammonium. In brief, seeds from 10,000 M 2 lines were bulked, and 100 seeds were sown on a mesh in a 500ml volume cup. Young seedlings were subjected to ammonium treatment 3 days after germination with a hydroponic culture containing 2.5 mM ammonium. The root elongation of M 3 progeny was nally assessed in the same manner to con rm the NH 4 + -sensitive phenotype of the candidate mutant rohan.

Plant growth conditions
Wild-type, mutants and transgenic rice seeds were surface-sterilized with 70% (v/v) ethanol for 2 min, followed by 30% (v/v) bleach containing 0.01% Tween 80 for 30 min. After 5 times washing with sterilized water, rice seeds were germinated at 37 ℃ for 3 days. Germinated rice seedlings were rst grown in water for another 3 days in a growth chamber under a photoperiod of 14 h light (~ 200 μmol•m -2 •sec -1 light density and ~ 60% humidity) and 10 h dark at 28 ℃. Subsequently, the rice seedlings with around 2 cm seminal roots were transferred to a hydroponic medium with modi ed Kimura B solution (500 ml volume for each cup with 10 seedlings, pH 5.5, 0.25g MES) for different treatments according to the previous study 11 . For N-free treatment, nitrogen sources (NH 4 ) 2 SO 4 and KNO 3 were replaced with K 2 SO 4 at a concentration of 1.25 mM. For NO 3 treatment, (NH 4 ) 2 SO 4 was replaced with 2.5 mM KNO 3 .
The rice seedlings were treated for 6 days unless otherwise noted, and the hydroponic culture was refreshed every 2 days.

Root phenotype analysis
The rice seedling roots were imaged with a 400dpi resolution by an EPSON Expression 11000XL scanner.
Seedling roots were scanned before and after treatments. We then measured the seminal roots (SR) length using Fiji image analysis software (http:// ji.sc/). The SR elongation was nally calculated as the difference between the SR lengths after and before treatments.

Genetic mapping of the rohan mutation
To identify the causal mutation, a F 2 population was generated by backcrossing rohan mutant with the parent Wuyungeng 7 then used for genetic mapping through a modi ed MutMap analysis 43  RNA-seq datasets were then analyzed following a custom protocol previously published 48 . Raw data were also cleaned using Fastp v0.20 44 . Cleaned reads were aligned to the rice reference genome using STAR v2.7.8a with a splicing-aware method and two-pass mode 49 . The aligned reads were separately assembled into transcripts for each sample with the reference annotation-based transcript (RABT) assembly algorithm and generated an updated transcript annotation with GTF-formatted le using StringTie v2.1.3 50 . Finally, the expression level of genes was quanti ed and normalized with the aboveupdated GTF le using HTSeq, respectively 51 . Only the genes with an FPKM (fragments per kilobase of transcript per million fragments mapped) >1 in at least six samples were used for downstream gene expression analysis. DESeq2 was used to perform pairwise comparisons between conditional samples to identify differentially expressed (DEG) genes with the updated GTF le 52 . In our study, genes were considered as differentially expressed according to the following criteria: Log2 (Fold change) ≥ 1 and the adjusted p-value < 0.05. Raw RNA-Seq datasets were also available on NCBI BioProject under PRJNA808101.
Venn diagrams and heatmaps were plotted from the differentially expressed genes using custom R scripts. Enrichment analysis based on GO and KEGG databases was carried out using PlantGSAD 53 . The Benjamini-Yekutieli method was used for P-value adjustment. Besides, clusterPro ler v3.14, a gene set enrichment analysis (GSEA) method 54 , was further employed to determine if a prior de ned set of genes such as nitrogen metabolism and plant hormone signal transduction shows concordant differences in response to NH 4 + treatment between the mutant and the wild-type.

Allelic variation and local adaption analysis
For allelic variation analysis of ROHAN, SNP datasets of 446 accessions of common wild rice (Oryza ru pogon) and 3K Rice Genomes Project were obtained from a previous study 55,56 . The nucleotide diversity (Pi) and Fst statistics were calculated and tted with the sibling window method (window size:2 kb; step size:1 kb) for wild rice and rice subpopulations using VCFtools v.0.1.17 57 . Functional effects of variants from ROHAN and its surrounding regions (with 2 Kb anking sequence on both sides) were predicted by SNPeff v4.3 58 . Allele frequency of missense SNP (Chr3:10847318) from six rice subgroups (japonica, indica I, indica II, indica III, aromatic and aus) of the 3,000 varieties were calculated and integrated into a geographical map according to their subpopulation and origin information from the Rice SNP-Seek Database (https://snp-seek.irri.org). Moreover, SR length and expression abundance corresponding to different alleles were also measured from a selection of 50 Japonica and 50 Indica varieties.

Plastid construction and plant transformation
The Gateway system® (Invitrogen, Carlsbad, CA, USA) was employed to generate most genetic constructs. For transcriptional fusions, rst, ~2k promoter fragments upstream of the start codon were ampli ed by PCR from genomic DNA and cloned into pdonrp41R by Bp reactions. Second, the ROHAN open reading frame with or without stop codon was ampli ed from genomic DNA and was cloned into pDONR221 or pDONR2F3R, respectively. Subsequently, the vectors were introduced into different expression vectors by LR reactions. The expression vectors pmk7snfm14GW, pHb7m34GW and pHb7m24GW were respectively used for the promoter-GUS fusion, the complementation of the mutant and generating the overexpression lines.
Histochemical analysis and microscopy GUS assays were done as previously described 60 . Roots were imaged with a Leica DM2500 microscope (Leica Microsystems, Wetzlar, Germany). For the anatomical sections, GUS-stained samples were xed overnight and embedded following a published protocol 61 .
DR5rev:VENUS-N7 expression in the root tip of rice was observed with a Leica SP8 laser-scanning microscope equipped with a white laser and hybrid laser detectors. The root tips of rice DR5rev:3xVENUS-N7 transgenic rice seedlings grown under different treatments were cleaned with a modi ed ClearSee method preceding confocal imaging as previously described 11 . Rice root meristems were imaged using a modi ed mPS-PI staining to clear and visualize the cell organization of rice SR tips 11 .

RNA extraction and quantitative real-time PCR analysis
Total RNA was isolated using the plant RNA puri cation reagent (Invitrogen). cDNA was synthesized from Olympus MXV10 microscope with GFP channel was used to capture the images.

Amino acid analysis
Wild-type and rohan plants were grown under different treatments for 4 days. Roots were collected for amino acid extraction as previously described 63 . Roots were collected in liquid nitrogen, then grinded and mixed manually. Samples were then transferred to 5 ml tubes, frozen with liquid nitrogen, and transferred to the lyophilizer (BILON, FD-2C, Shanghai, China) for 3 days. Lyophilized root tissues (~ 3 mg) were used for extraction with 80% methanol followed by incubation at 70℃ for 15 mins and shaking (600 rpm). Following centrifugation at 10,000 g for 15 min at room temperature, the supernatants were collected, the pellet was re-extracted with 20% methanol as described above. All supernatants were collected in a new pellet and dried in a Termovap Sample Concentrator MD200-1(Allsheng, China). The dried pellets were dissolved in 500 µL ultrapure water. 500 µL 4% 5-Sulfosalicylic acid dihydrate was added to re-dissolve followed by centrifugation at 12,000g for 15min at room temperature. The supernatants were collected and ltered by 0.22 µm water lter. The samples were analyzed with a LA8080 automatic amino acid analyzer (Hitachi, Tokyo, Japan). All steps were performed accordingly to the manufacturer's instructions.

Photosynthesis parameter and seeds protein analysis
Simultaneous measurements of photosynthesis parameters were performed on light-adapted leaves with a LI-6,800 infrared gas analysis system. Prior to the measurements, the leaves were placed in a greenhouse at a photosynthetic photon ux density of 1,500 μmol·m -2 s -1 with an ambient CO 2 concentration of 400 μmol/mol. After equilibration to a steady state, 0.8 s saturating pulses of light (~8,000 μmol·m -2 s -1 ) were applied to measure the maximum (F m ') and steady-state uorescence (F s ); gas exchange parameters were recorded. The Φ PSII was calculated as: Φ PSII = (F m '-F s )/ F m '. Gas exchange parameters include net photosynthesis rate, stomatal conductance and transpiration rate. The seeds crude protein content was detected with a Near Infrared Spectroscope (NIRS) (DA7250, China).

IAA content analysis
Germinated rice seedlings (3-day-old) were transferred to nutrient solution containing NH 4 + or not. After 4 days of treatment, the roots were harvested and immediately frozen in liquid nitrogen. Auxin extraction procedure was conducted as described in 64

Statistical analysis
The experiments performed in this study were repeated at least three times, and all the results were presented as the mean ± SD. SPSS software was used for statistical analysis. The signi cant difference between the two sets of data was determined by Student's t-test, whereas the difference among more than two sets of data was analyzed with one-way ANOVA followed by Duncan's multiple comparisons.   and cortical cell number and of the wild-type and pin1apin1b mutant seedlings shown in h. Data represent the means ± s.d. (n ≥ 10 seedlings), and the asterisks indicate signi cant differences relative to the wild-type (student's t-test, *P<0.05 and **P < 0.01).

Figure 5
Genetic variation of ROHAN is associated with its regulatory role in root response to NH 4 + .
a, Nucleotide diversity(π) of ROHAN gene and anking regions between different rice subspecies. b, Mean Fst value of ROHAN gene and anking regions between different rice subspecies. c, Geographic distribution of of the SNP(Chr3:10847318)'s allele frequency in different rice varieties based on the 3,000 Rice Genome database. d, Distribution of the SNP(Chr3:10847318) in Japonica and Indica rice subspecies based on the 3,000 Rice Genome Project database. e, Amino acid change caused by the SNP mutant (A<-G). f, Comparison of relative seminal roots (SR) length (+N/-N) between Japonica and Indica ssb. seedlings grown under 2.5 mM NH 4 + (+N) or N-free (-N) condition for 6 days. Data represent the means ± s.d. (n = 50 cultivars for either Japonica or Indica subspecies, ≥ 10 seedlings per cultivar), and the asterisk indicates signi cant difference (student's t-test, ***P < 0.001). g, Comparison of relative gene expression (+N/-N) of ROHAN in roots between Japonica (n = 8 cultivars) and Indica subspecies (n = 8 cultivars) grown under 2.5 mM NH 4 + (+N) or N-free (-N) conditions. Data represent the means ± s.e. of three biological replicates, and the asterisk indicates signi cant difference between two subspecies (student's t-test, **P < 0.01). h and i, Root phenotype h and the seminal root (SR) length i of the wild-type (Orzya sativa cv. Wuyungeng7) and of ve independent complementary lines of rohan mutant carrying K470R substitution, treated with 2.5 mM NH 4 + for 6 days. The white dotted line indicates the position of the root tip when the seedlings were transferred to media supplied with 2.5 mM. Scale bars, 1 cm. Data represent the means ± s.d. (n ≥ 10 seedlings), and the asterisk indicates signi cant difference between two subspecies (student's t-test, ***P < 0.001).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. SupplementaryInformationtosubmit.docx