Abstract

   Ammonium toxicity in plants is considered a global phenomenon, but the
   primary mechanisms remain poorly characterized. Here, we show that
   although the addition of potassium or nitrate partially alleviated the
   inhibition of rice seedling root growth caused by ammonium toxicity,
   the combination of potassium and nitrate clearly improved the
   alleviation, probably via some synergistic mechanisms. The combined
   treatment with potassium and nitrate led to significantly improved
   alleviation effects on root biomass, root length, and embryonic crown
   root number. The aberrant cell morphology and the rhizosphere
   acidification level caused by ammonium toxicity, recovered only by the
   combined treatment. RNA sequencing analysis and weighted gene
   correlation network analysis (WGCNA) revealed that the transcriptional
   response generated from the combined treatment involved cellulose
   synthesis, auxin, and gibberellin metabolism. Our results point out
   that potassium and nitrate combined treatment effectively promotes cell
   wall formation in rice, and thus, effectively alleviates ammonium
   toxicity.

Introduction

   Compared with nitrate, ammonium is the preferred nitrogen source for
   paddy rice and many other plant species. While a low supply of ammonium
   promotes plant growth, a high supply of ammonium causes toxicity,
   especially when applied as an exclusive nitrogen source [[32]1]. At low
   concentrations (< 3 mM), ammonium is typically the nitrogen source
   preferred by plants, but above a certain threshold, ammonium becomes
   toxic [[33]2]. Ammonium toxicity can lead to plant growth retardation,
   biomass reduction, structural changes, the formation of short and dense
   root systems, and even death [[34]3]. In addition, ammonium toxicity
   inhibits root cell elongation and affects root gravitropism by
   disturbing auxin distribution in root tips [[35]4–[36]6].

   Cells that absorb excess ammonium actively excrete ammonium. These
   transport processes utilize many ion channels, resulting in a large
   waste of energy [[37]7]. To maintain the charge balance, plant cells
   need to pump out protons to counteract the excess ammonium in the
   cytoplasm, which causes rhizosphere acidification [[38]8]. Ammonium can
   enter cells through a variety of nonselective cation channels and
   potassium-specific channels; transport through the latter competitively
   inhibits uptake of potassium [[39]9, [40]10]. Excessive ammonium
   assimilation in the cytoplasm consumes a large amount of α-ketoglutaric
   acid, which decreases the amount of carbon available for metabolism
   [[41]11]. The auxin signal in Arabidopsis thaliana roots is obviously
   reduced under ammonium treatment but exogenous application of auxin is
   unable to completely alleviate the toxicity phenotype in roots
   [[42]12]. Ammonium toxicity can result in higher lipid peroxidation
   levels, a higher ratio of the oxidized state of glutathione and
   ascorbic acid, and higher activity of ROS-scavenging enzymes [[43]13].

   Plant ammonium toxicity can be alleviated, for instance, by increasing
   the activity of the ammonium assimilation system, relieving rhizosphere
   acidification, inhibiting futile transmembrane ammonium cycling, and
   increasing the supply of other nutrients [[44]1–[45]3]. Potassium has a
   significant alleviation effect on ammonium toxicity [[46]14]. A
   possible reason for this effect is that potassium can competitively
   suppress the transport of ammonium, thereby reducing futile
   transmembrane ammonium cycling [[47]15]. Some researchers have reported
   that the presence of nitrate in the growth medium ameliorates the
   ammonium toxicity symptoms, even at a very low nitrate concentration
   [[48]9]. Since nitrate is an essential signaling molecule, the
   alleviation effect of nitrate on ammonium toxicity could be complex
   [[49]16]. Nitrate, as an anionic and oxidized nitrogen source, can
   mitigate the imbalance of ions and redox states caused by cations and
   reduce ammonium ions when applied together [[50]17]. The rhizosphere
   becomes alkalized when plants absorb nitrate, which counteracts the
   acidification caused by ammonium absorption [[51]2]. Nitrate alleviates
   ammonium toxicity without lessening ammonium accumulation, organic acid
   depletion, and inorganic cation depletion in A. thaliana shoots
   [[52]18].

   In this study, we compared the alleviation effects of potassium and
   nitrate when given alone (as KCl and NaNO[3]) or combined (as KNO[3])
   on ammonium toxicity and found that the combined application of
   potassium and nitrate not only improved the alleviation of ammonium
   toxicity compared with the application of individual ions but also
   resulted in some novel beneficial mechanisms, probably due to an
   undiscovered synergism. We provide morphological and physiological
   evidence of the existence of this synergism, and we performed a
   transcriptome analysis to reveal a regulatory network in cell wall
   construction related to this synergism.

Materials and methods

Plant materials and growth conditions

   The plant materials used in this research included ZH11 (Oryza sativa
   japonica) and DR5:GUS insertion lines (ZH11 background) [[53]19].

   Rice seeds were germinated in distilled water for 2 days in a dark
   environment at 30°C and then transferred into hydroponic culture media.
   The growth density of the seedlings was strictly controlled at 24
   plants per 400 mL medium and grown in a growth chamber (30°C-14
   hour-light/22°C-8 hour-dark cycles, 100 μmol m^−2 s^−1 illumination and
   80% humidity).

   There were 6 groups of hydroponic culture media used in this research:
   control (pure water); ammonium toxicity treatment (4 mM NH[4]Cl
   solution); potassium alleviation treatment (4 mM NH[4]Cl and 4 mM KCl
   solution); nitrate alleviation treatment (4 mM NH[4]Cl and 4 mM NaNO[3]
   solution); potassium and nitrate alleviation treatment (4 mM NH[4]Cl
   and 4 mM KNO[3] solution); sodium and chlorine alleviation treatment (4
   mM NH[4]Cl and 4 mM NaCl solution). No other nutrients were added to
   the hydroponic culture media except for the above mentioned ions.

   If not mentioned, 7-day-old seedlings were harvested for experiments.

Analysis of root morphology

   Seedlings were removed from the boxes, and the roots were immersed in
   water contained in petri dishes. The petri dishes were placed on a
   black cloth, and the whole root area per plant was photographed using a
   D7000 camera (Nikon, Japan). All the root morphology parameters were
   measured with SmartRoot software V4.2 [[54]20] according to the user
   guide.

   The root tips were cut and fixed with FAA fixative (1.9% formaldehyde,
   4.9% acetic acid, 63% ethanol, wt/vol) for 2 days and then embedded in
   paraffin. Then, 6-μm-thick slices were obtained with a microtome. After
   dewaxing with xylene, the sections were placed on glass slides and
   photographed using a microscope (DM4000B equipped with a DFC490 camera,
   Leica, Germany).

Culture pH measurement

   At selected times after germination, the pH of the hydroponic culture
   media was measured using a pH meter (PB-10, Sartorius, Germany).

DR5:GUS-based auxin assay

   The DR5:GUS insertion line grown as above was harvested and rinsed in
   distilled water for 30 s. After fixing in 90% (wt/wt) acetone in vacuo,
   the roots were rinsed with GUS staining buffer (50 mM Na[2]HPO[4], 50
   mM NaH[2]PO[4], 0.1% Triton X-100, 2 mM K[4][Fe(CN)[6]], 10 mM EDTA) 3
   times and then stained with 5 mL of GUS staining buffer containing 5 μg
   of X-gluc (Sigma-Aldrich, USA) for 2h, then they were immersed in the
   GUS staining buffer for observation. Images were photographed under a
   microscope (DM4000B equipped with a DFC490 camera, Leica, Germany).

Ion content, enzyme activity, and soluble protein content determination in
roots

   Seedling roots were harvested and rinsed in distilled water for 30 s to
   remove extracellular ions. About 2 g roots from each treatment were
   homogenized (1 min for each sample) under liquid N[2] using a mortar
   and pestle.

   For tissue ammonium content determination, 0.2 g of homogenized tissue
   was mixed with 2 mL of 10 mM precooled formic acid before
   centrifugation (4°C, 10000 ×g, 10 minutes). The supernatants were
   analyzed by the o-phthalaldehyde (OPA) method to determine the ammonium
   content using spectrophotometry as described in [[55]21].

   For tissue nitrate content determination, 0.1 g of homogenized tissue
   was mixed with 1 mL of H[2]O before centrifugation (4°C, 10000 ×g, 10
   minutes). Then, 30 μL of supernatant was mixed with 100 μL of 5%
   (wt/vol) salicylic acid-H[2]SO[4]. After reacting for 20 minutes at
   room temperature, 2 mL of 2 M NaOH was added to the mixture. The sample
   absorbance was measured at 410 nm, and the nitrate content was computed
   according to a standard curve.

   The potassium content was determined by the atomic absorption
   spectroscopy (AAS) method using a contraAA700 (Analytik Jena,
   Thuringia, Germany). After determining the dry root weights, the
   samples were digested with 5 mL of concentrated nitric acid and 1 mL of
   H[2]O[2] at 150°C until the mixtures were transparent. The solutions
   were then diluted to 50 mL with 0.5% nitric acid. The potassium content
   was determined at 404.4 nm.

   Homogenized tissue was lysed with Western and IP lysis buffer (Beyotime
   Biotechnology, Shanghai, China). The soluble protein content was
   measured with a BCA protein assay kit (Beyotime Biotechnology,
   Shanghai, China).

   Glutamine synthetase (GS) activity was measured using a GS activity
   assay kit (Comin Biotechnology, Suzhou, China) as described in
   [[56]22].

   Glutamate synthase (GOGAT) activity was measured using a GOGAT activity
   assay kit (Comin Biotechnology, Suzhou, China) as described in
   [[57]23].

Net ^15NH[4] flux measurement

   Rice seedlings were germinated and grown on 4 mM ^14NH[4]Cl for 7 days.
   After harvesting, their roots were rinsed with 0.1 mM CaSO[4]
   immediately for 1 minute. Then, the samples were evenly separated into
   4 groups and immersed in 4 mM ^15NH[4]Cl, 4 mM ^15NH[4]Cl + 4 mM KCl, 4
   mM ^15NH[4]Cl + 4 mM Na^14NO[3], or 4 mM ^15NH[4]Cl + 4 mM K^14NO[3]
   for 1 hour. After rinsing again with 0.1 mM CaSO[4] for 1 minute, the
   roots from the 4 groups were dried in a 70°C oven for 48 hours. Then,
   they were homogenized using liquid N[2]. The proportions of ^15N in the
   total nitrogen content (wt/wt) in the homogenized roots, which
   represents the root surface ^15NH[4] influx in 1 hour, were measured by
   isotope mass spectrometry (isoprime-100, Vario ISOTOPE cube CN).

   The method for root surface ^15NH[4]^+ efflux (in 1 hour) measurement
   is the same as above, except for the following part. Rice seedlings
   were germinated and grown on 4 mM ^15NH[4]Cl solution for 7 days, and
   they were immersed in 4 mM ^14NH[4]Cl, 4 mM ^14NH[4]Cl + 4 mM KCl, 4 mM
   ^14NH[4]Cl + 4 mM Na^14NO[3], or 4 mM ^14NH[4]Cl + 4 mM K^14NO[3] for 1
   hour. Finally, the proportions of ^15N in the total nitrogen content
   (wt/wt) in the homogenized roots were measured, which represents the
   root surface ^15NH[4]^+ efflux in 1 hour.

   ^15NH[4]Cl was obtained from Shanghai Engineering Research Center of
   Stable Isotope (Shanghai, China), and 99% of the nitrogen atoms were
   ^15N.

RNA sequencing and data analysis

   Total RNA was extracted from 15 samples of ZH11 roots under 5 different
   treatments (3 biological replications for each treatment, 40 plants for
   each biological replication). The cDNA libraries were constructed by
   Personal Biotechnology (Shanghai, China) using the TruSeq RNA Sample
   Prep Kit (Illumina, USA). The mRNA paired-end sequencing was performed
   using the NextSeq 500 platform (Illumina, USA) by the same company. The
   raw RNA sequencing reads have been deposited in the NCBI Sequence Read
   Archive (SRA) under accession No. PRJNA693667.

   After removing the adapter sequences from raw reads, clean reads were
   obtained by Trimmomatic [[58]24]. The clean reads were mapped to the
   reference genome of the Oryza sativa japonica variety
   ([59]ftp://ftp.ensemblgenomes.org/pub/plants/release-36/fasta/oryza_sat
   iva/dna/Oryza_sativa.IRGSP-1.0.dna.toplevel.fa.gz) using STAR [[60]25],
   and gene structure information was obtained from EnsemblPlants
   annotation
   ([61]ftp://ftp.ensemblgenomes.org/pub/plants/release-36/gtf/oryza_sativ
   a/Oryza_sativa.IRGSP-1.0.36.chr.gtf.gz). The normalized gene expression
   levels (TMM normalized FPKM) were calculated using Trinity [[62]26].
   DESeq2 [[63]27] was used to identify the differentially expressed
   genes. Those with an adjusted p-value (padj) < 0.05 and |log[2]FC| > 1
   genes between every pair of groups were considered significantly
   differentially expressed genes.

Weighted gene correlation network analysis (WGCNA)

   A matrix containing the TMM values of all genes was used as an input
   file, and 40% of the genes were first filtered out using the varFilter
   command (var.func = IQR) in the R package genefilter. Then, WGCNA was
   performed using the R package WGCNA [[64]28]. A suitable beta value of
   18 was calculated, and the TOMType was selected as "signed" to
   construct a coexpression matrix, which divided genes into 15 modules,
   including 14 valid modules and 1 “gray” module. The correlations
   between the modules and traits were calculated to select the modules
   for further analysis. Hub genes were selected using a module membership
   > 0.9 and a gene significance > 0.8.

Results

Growth, morphology, and auxin distribution

   We found that ammonium toxicity in the roots of Oryza sativa japonica
   ZH11 seedlings was alleviated to a much greater extent when potassium
   and nitrate were supplied together than when they were supplied
   separately ([65]Fig 1A and 1B). We tried various concentrations (from
   0.5 to 12 mM) of NH[4]Cl, and found that the 4 mM NH[4]Cl is a
   propriate concentration to generate enough ammonium toxic effect on
   rice seedling roots while facilitate sampling ([66]S1A Fig). We also
   tried different concentration ratios of alleviative ions versus toxic
   ion in the 12 mM NH[4]Cl ([67]S1B Fig), and found that the differences
   among 5 treatments are consistent with that in [68]Fig 1C. Therefore,
   we speculated that there might be a synergistic mechanism by which
   potassium and nitrate participate in ammonium toxicity alleviation. To
   test our hypothesis, we first determined that neither sodium nor
   chlorine had any effect on ammonium toxicity and its alleviation
   ([69]Fig 1B–1D). The combined application of potassium and nitrate had
   significantly stronger alleviation effects on embryonic crown root
   (ECR) length than individual applications, while the effects on radical
   length, crown root number and root diameter seemed additive ([70]Fig
   2). The morphologies and auxin distributions of the radicles and the
   embryonic crown roots in each treatment were consistent. Under ammonium
   toxicity treatment, the root cap was enlarged and grew abnormally,
   which was ameliorated only by KNO[3] ([71]Fig 3B). We can see cells
   which are beginning to elongate (a sign indicating the beginning of the
   elongation zone) in the H[2]O and KNO[3] alleviation treatments, but we
   cannot see this kind of cell in the other 3 treatments ([72]Fig 3A).
   The cells in the cortex of the control and KNO[3] alleviation treatment
   were similar and rectangular, while cells in the cortex of the ammonium
   toxicity and KCl or NaNO[3] alleviation treatments were elliptical and
   irregular ([73]Fig 3A).

Fig 1. Phenotypes of rice seedling roots under.

   [74]Fig 1
   [75]Open in a new tab

   (A) Photograph of 10-day-old rice seedlings. (B) Photograph of
   7-day-old rice seedlings. (C and D) Fresh weight (FW) and dry weight
   (DW) of roots of 7-day-old seedlings. Scale bars = 5 cm. Data are mean
   ± SD; t-tests were used to identify significant differences, and
   different letters represent significant differences among different
   treatments (p < 0.05).

Fig 2. Root architecture.

   [76]Fig 2
   [77]Open in a new tab

   (A) Photograph of roots grown for 7 days. Scale bar = 2 cm. (B-F)
   Measured parameters. ECR = embryonic crown root. Data are mean ± SD, n
   = 10; t-tests were used to identify significant differences, and
   different letters represent significant differences among different
   treatments (p < 0.05).

Fig 3. Morphology, auxin responsiveness, and rhizosphere acidification.

   [78]Fig 3
   [79]Open in a new tab

   (A) Micrographs of paraffin sections of radicles from 7-day-old
   seedlings. The approximate boundary of the elongation zone is marked by
   white arrowheads. (B) Micrographs of 7-day-old roots expressing
   DR-5::GUS, stained for GUS activity. In A and B, roots shown are the
   roots representative, selected from 10 seedlings in each treatment.
   Scale bars = 50 μm. (C) The pH of the culture solutions was measured
   every 2 days from the 3^rd day after seed germination. Data are the
   mean ± SD; t-tests were used to identify significant differences; ** (p
   < 0.01) represent significant differences between each alleviation
   treatment and ammonium toxicity treatment.

   DR5 is a synthetic auxin response element which reports on auxin
   responsiveness in a DR5:GUS transgenic plant. Under ammonium toxicity,
   the auxin signal in the rice seedling root tips was markedly weakened
   compared with that in the control ([80]Fig 3B). Comparing to the
   control, KCl treatment partly recovered the distribution of the GUS
   signal in the root cap, and slightly reduced the GUS signal in the
   meristem zone. NaNO[3] treatment could not restore the GUS signal in
   the root cap and the meristem zone ([81]Fig 3B). In the case of KNO[3],
   the distribution of the GUS signal was essentially the same as that in
   the control ([82]Fig 3B), indicating that KNO[3] had a better
   alleviating effect on auxin responsiveness. Rhizosphere acidification
   caused by ammonium toxicity was lessened by only KNO[3] and not by KCl
   or NaNO[3] ([83]Fig 3E).

   Therefore, in terms of the biomass, root tip cell morphology, and
   rhizosphere acidification, it appears that potassium and nitrate act
   synergistically to alleviate ammonium toxicity in rice roots.

Ammonium uptake and assimilation

   To analyze how potassium and nitrate alleviated rice ammonium toxicity,
   we measured the ammonium content in roots. The ammonium content under
   the ammonium toxicity treatment was the highest among all treatments
   ([84]Fig 4A). Both KCl and KNO[3] reduced the ammonium content to a
   level comparable to that of the control, while the alleviation effect
   of NaNO[3] was somewhat weaker ([85]Fig 4A). These results suggested
   that both potassium and nitrate alleviated ammonium toxicity by
   reducing the net influx of ammonium. Furthermore, we compared the
   effects of potassium and nitrate on root ammonium influx and efflux
   using ^15N-labeled NH[4]Cl. Potassium reduced ammonium influx and
   efflux, while nitrate reduced only ammonium efflux ([86]Fig 4B and 4C).
   However, consistent with the ammonium content, the impacts of KCl and
   KNO[3] on root ammonium influx and efflux were largely the same,
   suggesting that the differences in ammonium toxicity alleviation
   between KCl and KNO[3] could not be attributed only to the regulation
   of ammonium content. In addition, the GS and GOGAT enzyme activities
   were similar to ammonium content ([87]Fig 4D). In contrast, protein
   levels were elevated in all of the treatments although somewhat less so
   by KNO[3] ([88]Fig 4F).

Fig 4. Composition, transport, and enzyme activity related to nitrogen
status.

   [89]Fig 4
   [90]Open in a new tab

   (A) The ammonium concentrations. (B and C) For the ^15NH[4]^+ transport
   assay, rice seedlings were grown for 7 days in ^14NH[4]Cl (for B) or
   ^15NH[4]Cl (for C) solution containing no other substance and then
   transferred to solutions containing ^15NH[4]Cl (for B) or ^14NH[4]Cl
   (for C) with the 3 alleviation treatments for 1 hour (see [91]Materials
   and methods) before the root ^15N content was measured. (D-H) Data are
   mean ± SD; t-tests were used to identify significant differences, and
   different letters represent significant differences among different
   treatments (p < 0.05).

   We also measured the potassium and nitrate levels in roots ([92]Fig 4G
   and 4H). As expected, ammonium treatment reduced concentrations of both
   potassium and nitrate (by about 50%); also as expected, provision of
   either ion raised the level in the water control. However, the level of
   each ion in the root was increased even more in the combined treatment
   than in the single treatment, suggesting metabolism of the ions is
   linked.

Analysis of differentially expressed genes

   Whole-transcriptome analysis revealed that the number of differentially
   expressed genes (|log2FC| > 1 and adjusted p value < 0.05) in each
   treatment compared with the control was related inversely to the same
   comparison for root biomass (Figs [93]1C, 1D and [94]5A). From these
   differentially expressed genes, we arbitrarily selected 24 genes that
   appeared in all 4 groups to detect their expression level by
   quantitative real-time polymerase chain reaction (qRT-PCR), and all 24
   gene expression differences among the 5 treatments were consistent with
   the corresponding fragments per kilobase per million (FPKM) values
   generated from the transcriptome analysis, verifying the validity of
   the RNA-Seq analysis results ([95]S2 Fig).

Fig 5. Differentially expressed genes (DEGs).

   [96]Fig 5
   [97]Open in a new tab

   (A) Numbers of differentially expressed genes between given treatments
   and the control (H[2]O). (B) Venn diagram showing alleviation-related
   gene numbers in each alleviation treatment. The expression of these
   genes was differentially regulated by ammonium toxicity and restored to
   the control level by the 3 alleviation treatments.

   We defined 4365 differentially expressed genes between the NH[4]Cl
   treatment and the control as toxicity-related genes ([98]Fig 5A). To
   further investigate the unique role of KNO[3] in the alleviation of
   ammonium toxicity, we compared the number of genes among the
   toxicity-related genes with expression that returned to levels
   comparable to that in the control (|log2FC| < 1 and adjusted p value <
   0.05) under different alleviation treatments. There were 1716, 1335,
   and 2492 genes that met these requirements in the differentially
   expressed gene groups of the KCl, NaNO[3], and KNO[3] treatments,
   respectively ([99]Fig 5B). These genes were treated as
   alleviation-related genes. There were 812 genes with expression that
   could only be recovered under KNO[3] alleviation ([100]Fig 5B, [101]S1
   Table, [102]S3 Fig).

   We found 3 proton exchanger/antiporter genes (OS05G0113300,
   OS03G0828600, OS07G0666900) among the KNO[3]-specific
   alleviation-related genes (812 genes), and these genes might play a
   role in the specific mechanisms that alleviate rhizosphere
   acidification. To gain further insight into the possible processes
   involved in ammonium toxicity and alleviation, we examined the
   distributions of toxicity-related genes and alleviation-related genes
   based on pathway enrichment (KEGG) analysis ([103]S4A Fig) and gene
   ontology (GO) classification ([104]S4B Fig). The KEGG and GO results
   suggest that the combined application of potassium and nitrate
   generated alleviation via changes in gene expression related to stress
   responses, redox states, and metabolism.

Weighted gene correlation network analysis

   To characterize the gene expression changes that occurred under
   different treatments, we clustered the expression patterns by the WGCNA
   method, leading to the identification of 15 distinct WGCNA modules
   ([105]Fig 6A and 6C). Analysis of the module-trait relationships
   revealed that the ‘turquoise’ module was highly positively correlated
   with root biomass and all root traits except for root diameter
   ([106]Fig 6C). The ‘turquoise’ module was also highly negatively
   correlated with GOGAT enzyme activity and protein concentration
   ([107]Fig 6C). Therefore, the genes of the ‘turquoise’ module were
   considered to play an important role in root growth. The ‘yellow’
   module was highly positively correlated with the GOGAT enzyme activity,
   protein concentration, and ammonium concentration ([108]Fig 6C). The
   genes of the ‘yellow’ module might be negatively associated with root
   growth.

Fig 6. The weighted gene correlation network analysis.

   [109]Fig 6
   [110]Open in a new tab

   (A) Hierarchical cluster tree showing coexpression modules identified
   by WGCNA. Each leaf in the tree represents 1 gene. The major tree
   branch comprises 15 modules labeled by different colors. (B) Venn
   diagram showing the interactions of genes in root-trait coupled modules
   and differentially expressed genes with expression that was recovered
   only by KNO[3] under ammonium toxicity. (C) Module-trait association.
   Each row corresponds to a module. Each column corresponds to a trait.
   The color of each cell at the row-column intersection indicates the
   correlation coefficient from -1 (green) to 1 (red).

   There were 507 genes in the ‘turquoise’ module and 115 genes in the
   ‘yellow’ module with differentially regulated expression in response to
   ammonium toxicity that was specifically recovered by KNO[3] ([111]Fig
   6B). These genes with expression that was highly correlated with traits
   might be responsible for the KNO[3]-specific alleviation effect on root
   growth. We used a module membership (MM) > 0.9 and a gene significance
   (GS) > 0.8 as parameters to identify hub genes among the 507 genes, and
   we found 62 hub genes ([112]S2 Table). The coexpression network was
   visualized by Cytoscape with a WGCNA edge weight > 0.35 ([113]S4C Fig).
   Several genes involved in cell wall formation were identified in the
   network, such as genes associated with cellulose synthesis (OsCesA4,
   OsCesA7, OsCesA9), lignin catabolic process (Os01g0850700), cell wall
   modification (OsEXPB5), and cell wall proteins (OsFLA11, OsFLA27)
   ([114]S2 Table). These results suggested that under ammonium toxicity,
   KNO[3] specifically restores the expression of genes related to cell
   wall formation and alleviates the effects of ammonium toxicity on cell
   morphology and root growth by restoring cell wall formation.

A cell wall regulation network related to this synergism

   Based on these findings, we re-examined the expression of genes
   associated with root cell formation. Cellulose is an important
   component of the cell wall. Analysis of the transcription regulation
   network for cellulose synthesis suggests that OsNAC29 and OsNAC62
   directly regulate OsMYB61, which in turn activates the expression of
   cellulose synthase genes (CESAs), such as OsCESA4, OsCESA7, and OsCESA9
   [[115]29]. The expression of OsMYB61 and OsCESAs is negatively
   regulated by OsIIP4, which does not bind their promoters and interacts
   with OsNAC29 [[116]30]. Our results showed that the expression of these
   genes was repressed under NH[4]Cl treatment, and KNO[3] resulted in
   better recovery of the expression of these genes ([117]Fig 7).
   Furthermore, similar results were observed with OsCSLF6 and OsIRX10
   ([118]Fig 7), which are involved in the hemicellulose synthesis
   pathway. Thus, the cellulose and hemicellulose synthesis pathways
   mentioned herein might be a regulatory mechanism induced by KNO[3].

Fig 7. The cell wall regulation network related to this synergism.

   [119]Fig 7
   [120]Open in a new tab

   (A) A network involved in cell wall formation was constructed according
   to the literature (line) and predicted transcriptional regulation
   (dashed line). Blue boxes indicate genes or proteins; yellow ovals
   indicate chemicals; green boxes indicate biological processes. (B)
   Expression patterns of genes in this network under the treatments.

   Xyloglucan endotransglucosylases/hydrolases (XTHs) play an important
   role in the construction of the primary cell wall by mediating the
   cleavage and rejoining of β-(1–4)-xyloglucans. There were 2 XTH genes
   (OsXTH8 and OsXTH2) among the 507 genes. OsXTH8 is upregulated by
   gibberellin and is highly expressed in vascular bundles of the leaf
   sheath and young roots, where cells are actively undergoing elongation
   and differentiation [[121]31]. OsXTH2 is also upregulated by
   gibberellin [[122]32]. Gibberellin can be converted to nonbiologically
   active gibberellin in plants by 2β-hydroxylation, a reaction catalyzed
   by gibberellin 2-oxidase. OsGA2ox1 encodes a gibberellin 2-osidase,
   which is functional in vitro and in vivo [[123]33]. Our results showed
   that KNO[3] restored or nearly so the expression levels of OsXTH8,
   OsXTH2, and OsGA2ox1 under NH[4]Cl toxicity ([124]Fig 7). This result
   suggests that the gibberellin metabolic pathway mentioned herein was
   disrupted by ammonium toxicity, leading to abnormal root cell
   elongation, which could be effectively recovered by the synergistic
   effects of KNO[3].

   Auxin is known to stimulate cell elongation by increasing cell wall
   extensibility. The OsGH3 protein controls auxin homeostasis by
   catalyzing the conjugation of IAA to amino acids. Overexpression of
   OsGH3-2 reduces the free IAA level [[125]34], and downregulation of
   OsGH3-8 leads to an auxin overproduction phenotype [[126]35]. Our
   results again showed that the expression levels of OsGH3-2 and OsGH3-8
   were upregulated by NH[4]Cl treatment, and high expression was
   maintained under the KCl and NaNO[3] treatments ([127]Fig 7). Only the
   KNO[3] treatment restored the expression of these 2 genes to the
   control level, which was consistent with the auxin distribution result
   indicated by DR5:GUS ([128]Fig 3B). Thus, the synergistic effects might
   also regulate the auxin pathways to result in enhanced beneficial
   effects on rice roots under ammonium treatment.

Discussion

Potassium and nitrate might have a synergistic effect on ammonium toxicity
alleviation in rice roots

   In 4 mM NH[4]Cl solution, the growth of rice seedling roots was
   severely hampered, which is a typical ammonium toxicity. It is widely
   known that supplementation with nitrate or potassium alleviates
   ammonium toxicity in plants [[129]1, [130]5, [131]17]. To exclude the
   interference of other substances on the alleviation effect of potassium
   and nitrate on rice ammonium toxicity, we did not use any culture media
   in this research. In rice, the endosperm can provide sufficient
   nutrition for seedling growth over 1 week after germination. We found
   that the addition of KNO[3] resulted in a significantly enhanced
   promotion of root growth compared with that of KCl or NaNO[3].

   The rhizosphere acidification is the most typical characteristics of
   plant ammonium toxicity [[132]36]. We found that only the combined
   treatment of potassium and nitrate reduced the level of rice
   rhizosphere acidification caused by ammonium toxicity. In the roots of
   KNO[3] treatment, there were more nitrate uptake and less ammonium
   uptake compared to that of the NaNO[3] treatment under ammonium
   toxicity. This might help to alleviate the root rhizosphere
   acidification level in the KNO[3] treatment. Our study showed that the
   ammonium toxicity notably decreased the number of embryonic crown
   roots, while the KNO[3] treatment increased their number to even more
   than that of the control ([133]Fig 2F). In rice in general, the number
   of embryonic crown roots is 5. Considering that the number of embryonic
   seminal roots in maize is variable [[134]37], we speculate that the
   number in rice might be altered by certain conditions. Evidently,
   ammonium toxicity and the combined treatment of potassium and nitrate
   alter embryonic crown root number in rice. Additionally, roots in the
   KNO[3] treatment are shorter and thicker than those of the control,
   though there is no biomass difference between them. We suggest that the
   cause of this phenomenon is related to the nutrient status in these two
   treatments.

   The nitrogen status regulates cell division and expansion and thus
   shapes architecture [[135]38]. Ammonium inhibits cell elongation rather
   than proliferation in A. thaliana roots [[136]4]. The inhibitory effect
   of ammonium on A. thaliana primary root elongation was mainly achieved
   by reducing the length of the meristem and elongation zones, although
   elemental elongation rate was also inhibited [[137]6]. Our results show
   that ammonium toxicity causes an irregular cell morphology, seemingly
   impaired cell elongation, and loss of auxin signaling in rice seedling
   roots. KNO[3] treatment effectively recovered the root cell morphology,
   resulting in a regular cell arrangement and a recovered auxin
   distribution that was almost the same as that of the control.

Potassium-nitrate synergism might regulate gene expression associated with
root growth and ammonium assimilation

   Ammonium causes changes in gene expression in roots, which in turn
   leads to a series of changes in growth [[138]36]. Our results show that
   potassium and nitrate restore the expression of some genes to levels
   comparable to those in the control. Therefore, the restoration of the
   expression of these genes represents an intrinsic mechanism of ammonium
   toxicity alleviation by these 2 ions. We identified 812 genes with
   expression that was specifically restored by KNO[3] but not by KCl or
   NaNO[3], suggesting that the synergism between potassium and nitrate
   might activates specific transcriptional regulation to alleviate
   ammonium toxicity in rice roots.

   Coexpression network analysis helps investigate thousands of genes with
   identical expression patterns. It is an important method for gene
   interaction analysis and the interpretation of molecular mechanisms
   [[139]39]. In this study, we identified a “turquoise” module that was
   highly positively correlated with almost all root traits (excluding
   root diameter), and a “yellow” module that was highly positively
   correlated with tissue ammonium content and ammonium assimilation. The
   correlation analysis of the traits showed a negative correlation
   between most root traits and ammonium assimilation. Therefore, we
   hypothesized that gene expression in these 2 modules was positively and
   negatively correlated with root growth. Among the 812
   alleviation-related genes with expression that was specifically
   regulated by KNO[3], 507 genes belonged to the ‘turquoise’ module, and
   115 genes belonged to the ‘yellow’ module. Therefore, the specific
   synergistic alleviation effect of KNO[3] on root ammonium toxicity
   might be mainly achieved by regulation of the expression of these
   genes. No significant enrichment of GO terms was found in the 115
   genes. The 507 genes were predominantly enriched in GO terms such as
   oxidation reduction, carbohydrate metabolic process, lignin catabolic
   process, and response to oxidation stress. Furthermore, we identified
   62 hub genes from the 507 genes. MapMan analysis [[140]40] of these hub
   genes showed that they were enriched in 3 categories: genes associated
   with cell wall formation, various enzymes, and transporters. Some genes
   were significantly associated with root growth traits and cell wall
   formation, so changes in their expression could result in different
   morphologies of rice root tip cells under ammonium toxicity or
   different alleviation conditions.

This synergism can alleviate rice root ammonium toxicity by regulating root
cell wall formation

   With changes in external nutrients, the composition of the cell wall
   determines the morphology of root cells and tissues, further
   influencing plant growth [[141]41]. Although there is growing evidence
   that the reprogramming of cell wall genes is critical for plant
   adaptation to nutritional status, an understanding of the molecular
   mechanisms that control these changes is now emerging. In crops,
   nitrogen status affects the mechanical strength and disease resistance
   of stems, which are regulated by the cell wall organization and
   strength, suggesting that the cell wall structure is regulated by
   nitrogen status [[142]41, [143]42]. The expression of genes associated
   with lignin and cellulose synthesis is significantly downregulated when
   rice is grown in a high-nitrogen environment [[144]42, [145]43].
   Consistent with this phenomenon, a high-nitrogen growth environment
   causes a significant decrease in cellulose and lignin in the rice root
   cell wall, while nitrogen deficiency causes an increase in cellulose in
   rice roots [[146]44].

   Based on differentially expressed gene analysis and WGCNA, we revealed
   a transcriptional regulatory network of genes associated with rice root
   cell wall formation that are affected by nutritional status ([147]Fig
   4, [148]S3 Table). Previously, the importance of the changes in the A.
   thaliana cell wall in response to ammonium stress has been reported
   [[149]45]. These authors found that the ammonium-mediated inhibition of
   growth is related to a more rigid cell wall structure which limits the
   expansion of cells, and identified some genes as important contributors
   to changes in the cell wall of leaves [[150]45]. Additionally, in the
   roots of the legume Medicago truncatula, Royo et al. (2019) showed that
   the higher expression of the enzymes of the phenylpropanoid metabolism
   during ammonium nutrition is accompanied with a higher content of
   lignin, confirming reinforcement of the cell wall as a tolerance
   mechanism of [[151]46]. Among our toxicity related genes, we found
   putative orthologues for all of the cell wall modifying proteins
   mentioned by Podgórska et al. (2017) [[152]45]. And most of these
   orthologous genes, specifically, pectin lyase, polygalacturonase,
   peroxidase, xyloglucan endotransglycosylase genes also appeared in our
   KNO[3]-specific alleviation-related genes ([153]S4 Fig). We even found
   2 XTH genes (OsXTH2 and OsXTH8) in our regulation network of the root
   cell wall formation ([154]Fig 7). Likewise, similar to the findings of
   Royo et al. (2019) [[155]46], we identify as related to ammonium
   toxicity many of the same genes, including genes involved in metabolism
   of phenylpropanoids, amino acids, carbohydrates, flavonoids,
   diterpenoids, antioxidants, and numerous membrane protein genes such as
   integral and intrinsic protein genes, various ion binding protein and
   transporter genes, and many peroxidase genes. The genes involved in
   metabolisms of glyoxylate, dicarboxylate, glycine, serine, threonine,
   glutathione, flavonoid, and the genes of integral and intrinsic to
   membrane were also appeared in our KNO[3]-specific alleviation-related
   genes. Additionally, many cellulose metabolism genes only differently
   expressed in the KNO[3] alleviation treatment (they are not a DEG in
   the other 4 treatments) ([156]S4 Fig). These findings suggest that cell
   wall formation in rice roots is related not only to nitrogen status but
   also to interactions between different nutrients. The separate addition
   of nitrate and potassium into the ammonium growth environment hardly
   alleviated the damage to cell wall formation, while the mixture of
   nitrate and potassium restored the expression of genes related to cell
   wall synthesis, and evidently restored cell wall formation. Moreover,
   these processes also involve hormonal pathways such as the auxin and
   gibberellin pathways. These results suggest that the effects of
   nitrogen on cell wall formation involve not only the morphological
   impact of nitrogen itself (ammonium or nitrate) but also the synergism
   between nitrogen and other nutrients (potassium). This phenomenon is
   also consistent with the knowledge that, in agricultural practices,
   mixed fertilizer application results in better crop growth performance.

   We found 2 transcription factors in the KNO[3]-specific
   alleviation-related genes (Os03g0820300 and Os12g0582600). These genes
   might play a role in regulating nitrogen use-associated genes, thereby
   contributing to nitrogen use efficiency [[157]47]. Rice is a member of
   the Gramineae, a plant that is a model organism for most crops for
   human consumption. An in-depth study of these mechanisms will reveal
   the link between environmental nutrition and crop growth and provide a
   theoretical basis for resolving the conflict between high yield and
   poor disease resistance in crops by improving the fertilizer
   utilization efficiency [[158]38, [159]47].

Supporting information

   S1 Fig. Supplementary data of the root growth.

   (A) 5-day-old roots of rice seedlings under 0/0.5/1/2/4/8/12 mM (from
   left to right) NH[4]Cl. (B) The fresh weight of roots under different
   treatments. The concentration ratio of “alleviative ions”: “toxic ion”
   = 1: 2, 1: 1, 2: 1 in 12 mM concentration.

   (TIF)
   [160]Click here for additional data file.^ (673.2KB, tif)
   S2 Fig. The validity of the RNA-Seq analysis results.

   The relative expression patterns among 5 different treatments of 24
   DEGs (randomly selected from all DEGs that appeared in all 4 DEG
   groups) were consistent with the corresponding relative FPKMs generated
   from transcriptome analysis.

   (TIF)
   [161]Click here for additional data file.^ (918.6KB, tif)
   S3 Fig. Expression heatmap of the clustered genes.

   The gene expression heatmap of 10 groups of clustered KNO[3]-specific
   alleviation-related genes. The color of each cell indicates the
   Log[2]FC of genes comparing to control (H[2]O) from -6 (blue) to 6
   (red).

   (TIF)
   [162]Click here for additional data file.^ (5MB, tif)
   S4 Fig. KEGG, GO and coexpressed hub genes.

   (A) KEGG pathway enrichment analysis of DEGs. (B) GO enrichment
   analysis of DEGs. (C) Cytoscape presentation of coexpressed hub genes
   among 507 genes, with the criterion of a module membership > 0.9 and a
   gene significance > 0.8.

   (TIF)
   [163]Click here for additional data file.^ (1.7MB, tif)
   S1 Table. The Log[2]FC of the KNO[3]-specific alleviation-related
   genes.

   The order of the genes in this table is consistent with that in the
   heatmap ([164]S3 Fig).

   (CSV)
   [165]Click here for additional data file.^ (140.6KB, csv)
   S2 Table. Sixty-two hub genes with expression that could be recovered
   by specific synergism.

   (XLSX)
   [166]Click here for additional data file.^ (24.2KB, xlsx)
   S3 Table. Hub genes in the cell wall construction network that were
   regulated by specific synergism.

   (XLSX)
   [167]Click here for additional data file.^ (10.8KB, xlsx)
   S1 File. Supporting methods.

   (PDF)
   [168]Click here for additional data file.^ (115.8KB, pdf)

Acknowledgments