Volume 223, Issue 1 p. 261-276
Full paper
Free Access

Diacylglycerol kinase and associated lipid mediators modulate rice root architecture

Shu Yuan

Shu Yuan

National Key Laboratory of Crop Genetic Improvement, Huazhong Agricultural University, Wuhan, 430070 China

College of Life Science and Technology, Huazhong Agricultural University, Wuhan, 430070 China

Donald Danforth Plant Science Center, St Louis, MO, 63132 USA

Department of Biology, University of Missouri, St Louis, MO, 63121 USA

Search for more papers by this author
Sang-Chul Kim

Sang-Chul Kim

Donald Danforth Plant Science Center, St Louis, MO, 63132 USA

Department of Biology, University of Missouri, St Louis, MO, 63121 USA

Search for more papers by this author
Xianjun Deng

Xianjun Deng

National Key Laboratory of Crop Genetic Improvement, Huazhong Agricultural University, Wuhan, 430070 China

College of Life Science and Technology, Huazhong Agricultural University, Wuhan, 430070 China

Search for more papers by this author
Yueyun Hong

Corresponding Author

Yueyun Hong

National Key Laboratory of Crop Genetic Improvement, Huazhong Agricultural University, Wuhan, 430070 China

College of Life Science and Technology, Huazhong Agricultural University, Wuhan, 430070 China

Authors for correspondence:

Xuemin Wang

Tel: +1 314 587 1419

Email: [email protected]

Yueyun Hong

Tel: +86 027 87280545

Email: [email protected]

Search for more papers by this author
Xuemin Wang

Corresponding Author

Xuemin Wang

Donald Danforth Plant Science Center, St Louis, MO, 63132 USA

Department of Biology, University of Missouri, St Louis, MO, 63121 USA

Authors for correspondence:

Xuemin Wang

Tel: +1 314 587 1419

Email: [email protected]

Yueyun Hong

Tel: +86 027 87280545

Email: [email protected]

Search for more papers by this author
First published: 18 March 2019
Citations: 23

Summary

  • Diacylglycerol kinase (DGK) phosphorylates diacylglycerol (DAG) to generate phosphatidic acid (PA), and both DAG and PA are lipid mediators in the cell. Here we show that DGK1 in rice (Oryza sativa) plays important roles in root growth and development.
  • Two independent OsDGK1-knockout (dgk1) lines exhibited a higher density of lateral roots (LRs) and thinner seminal roots (SRs), whereas OsDGK1-overexpressing plants displayed a lower LR density and thicker SRs than wild-type (WT) plants.
  • Overexpression of OsDGK1 led to a decline in the DGK substrate DAG whereas specific PA species decreased in dgk1 roots. Supplementation of DAG to OsDGK1-overexpressing seedlings restored the LR density and SR thickness whereas application of PA to dgk1 seedlings restored the LR density and SR thickness to those of the WT. In addition, treatment of rice seedlings with the DGK inhibitor R59022 increased the level of DAG and decreased PA, which also restored the root phenotype of OsDGK1-overexpressing seedlings close to that of the WT.
  • Together, these results indicate that DGK1 and associated lipid mediators modulate rice root architecture; DAG promotes LR formation and suppresses SR growth whereas PA suppresses LR number and promotes SR thickness.

Introduction

The development of a robust root system is critical for crop productivity because well-developed roots are essential for efficient nutrient and water uptake for plant growth and reproduction (Malamy, 2005; Aloni et al., 2006; Benkova & Bielach, 2010; Zolla et al., 2010). Rice (Oryza sativa) is a major global food crop that supports more than half of the world's population (Khush, 2005). Rice plant growth is supported by a fibrous root system that comprises seminal roots (SRs) and adventitious roots (ARs) (Liu et al., 2009). SRs consist of the primary SR that develops from the embryonic radicle and secondary SRs that develop from the mesocotyl. ARs include post-embryonic crown roots (CRs) developed from the stem base and nodal roots originated from the stem node (Marcon et al., 2013; Zhou et al., 2017). Both SRs and ARs contain many lateral roots (LRs) that constitute a major component of the fibrous system. Root growth and development are tightly regulated by intrinsic cues and plant growth environments (Malamy, 2005; Aloni et al., 2006; Zolla et al., 2010). A number of cereal mutants with deficient root formation have been identified, and plant hormones, particularly auxin, play important roles in the growth and development of specific root types, affecting the number, length and diameter of SRs, ARs and LRs in the cereals (Inukai et al., 2001, 2005; Liu et al., 2005; Hetz et al., 2010; Yu et al., 2016). However, the signaling network underpinning the regulation of root development is poorly understood (Fukaki & Tasaka, 2009; Vanneste & Friml, 2009).

Recent studies indicate that lipid mediators, such as phosphatidic acid (PA) and diacylglycerol (DAG), play a role in root growth and development in Arabidopsis (Wang et al., 2006; Hong et al., 2016). Signaling PA is produced primarily by the activation of phospholipase D (PLD) that hydrolyzes membrane lipids to generate PA and the activity of DAG kinases (DGKs) that produce PA by phosphorylating DAG. Signaling DAG can be produced by two distinctively different families of phospholipase Cs (PLCs), phosphoinositide-specific PLC (PI-PLC) and nonspecific PLC (NPC). Suppression of specific PLDs and PA formation in Arabidopsis led to short roots, such as PLDζs under phosphate deficiency (Li et al., 2006), PLDɛ under nitrogen deprivation (Hong et al., 2009), and PLDα1 and δ under high salinity and drought (Wang et al., 2019). By contrast, suppression of NPC5 and DAG formation led to fewer LRs than in the wild-type (WT) under NaCl stress, and LR primordia number in npc5-1 was restored to that of WT by adding DAG, but not PA, suggesting that NPC5-produced DAG, but not DAG-derived PA, promotes LR formation (Peters et al., 2010, 2014). These results have begun to suggest that PA and DAG have different effects on root growth and development in Arabidopsis, but the role of these lipid mediators in the root development of other plants, particularly cereal crops, is unknown.

DGK phosphorylates DAG to PA and occupies potentially a key position in modulating the signaling function of DAG and PA. DGK-produced PA has been reported in Arabidopsis undergoing dehydration and subject to high salinity and in nod factor-induced root hair deformation in a leguminous plant (Den-Hartog et al., 2001; Darwish et al., 2009; Galvan-Ampudia & Testerink, 2011). Inhibition of DGK by the DGK inhibitor (R59022; [6-{2-{4-[(4-fluorophenyl) phenylmethylene]-1-piperidinyl}ethyl}- 7-methyl-5H-thiazolo(3,2-) pyrimidine-5-one]) that decreased DGK-derived PA also reduced root elongation in Arabidopsis (Gomez-Merino et al., 2005). Recently, DGK2, DGK3 and DGK5 in Arabidopsis were reported to modulate PA and triacylglycerol production in response to cold stress (Tan et al., 2018). The rice genome has eight DGKs that are grouped into three clusters: I for OsDGK (4,5,8), II for OsDGK (3,6) and III for OsDGK (1,2,7) (Zhang et al., 2008; Ge et al., 2012). Cluster I DGKs contain a conserved kinase domain, a transmembrane domain and two C1 domains whereas other DGKs have a much simpler structure, lacking the C1 domain (Ge et al., 2012). The function of DGKs in rice is largely unknown. An early study indicated that overexpression of OsDGK1 (Os04g54200) in tobacco plants enhanced resistance against infection by tobacco mosaic virus and oomycete pathogens (Zhang et al., 2008). A later study used transient RNA silencing of OsDGKs in rice protoplasts, followed by monitoring gene expression, implies that OsDGKs play a role in abiotic and biotic stress (Ge et al., 2012). This study was undertaken to investigate the function of DGK in rice and its associated lipid mdediators DAG and PA in rice root growth and development.

Methods

Isolation of homozygous OsDGK1 mutants and generation of OsDGK1-OE and COM

The rice (Oryza sativa L.) mutants dgk1-1 and dgk1-2 were generated by T-DNA and transposon insertions, respectively. Mutant dgk1-1 was in the Dongjin variety background from the lab of Dr Gynheung An (Kyung Hee University) (Jeon et al., 2000; Jeong et al., 2006), and dgk1-2 from the Zhonghua11 background was from the National Center of Plant Gene Research (Wuhan, China). The dgk1-1 mutant was verified by reverse transcriptase (RT-PCR) detection of the DGK1 transcript using one-pair specific primers, 5′-ATGGCACTAATGGTTCCTGCT-3′ (forward) and 5′-CAACTTGCTGTGCCATCACC-3′ (reverse), and the expression level of dgk1-1 was normalized to that of β-actin (Os01g16414), using primers 5′-TGCTATGTACGTCGCCATCCAG-3′ (forward), 5′-AATGAGTAACCACGCTCCGTCA-3′ (reverse) (Fig. 1d). The homozygous dgk1-2 mutant was identified using PCR screening with DGK1-specific primers (5′-GACAGGCTACCTCGAGGATG-3′ and 5′-ACCCAACCACTGACCAAGAG-3′) and a transposon border primer (5′-GAAGGGGGGTGTTAAATATATATAC-3′). The two DGK1-specific primers amplified an expected DGK1 genomic DNA fragment, but if a homozygous mutant with transposon inserted in the region, no such DNA band was detected (Fig. 1e). Pairing of a DGK1-gene-specific primer with the transposon primer amplified a band, indicating the presence of the transposon insert.

Details are in the caption following the image
Expression pattern of OsDGKs and genetic manipulations of OsDGK1. (a) Expression of eight OsDGKs in flower buds, flowers, leaves after and before flowering, mature seeds, milk grains, and roots after and before flowering based on transcriptomic data from the National Center of Biotechnology Information database (https://www.ncbi.nlm.nih.gov/). (b) Tissue and developmental expression of OsDGK1 at seedling, tillering, florescence and mature stages as quantified by real-time quantitative PCR (RT-qPCR). The relative level of OsDGK1 gene expression was normalized to the transcript level of β-actin. Values are means ± SD (n = 3). (c) T-DNA insertion site for dgk1-1 in Dongjin (DJ) cultivar and transposon-insertion site in dgk1-2 in Zhonghua11 cultivar (ZH). Solid bars represent exons, thin lines represent introns, and left and right white bars indicate the untranslated region (UTR) at both 5′ and 3′ ends, respectively. (d) Reverse transcription-PCR verification with specific primers cover the T-DNA insertion site for the loss of DGK1 transcript in dgk1-1. The level of dgk1-1 transcript was normalized to that of β-actin. WT-DJ denotes wild-type (WT)-Dongjin. (e) The dgk1-2 was isolated by PCR with two pairs of specific primers. One pair of specific primers covers the transposon-insertion site (Upper); the other pair contains one DGK1 genomic gene primer and a transposon border primer (Lower). WT-ZH denotes wild-type Zhonghua11. (f) Expression of DGK1 in dgk1-1 in line COM29 by reverse transcription-PCR using one pair of specific primers that cover the T-DNA insertion site. The level of ubiquitin expression was used as control. (g) Verification of OsDGK1-overexpression (OE) lines, OE2 and OE5. Upper panel: PCR detection of introduced OsDGK1-OE DNA. Lower panel, immunoblotting detection of OsDGK1-Flag protein in OE lines using anti-Flag antibodies. (h) RT-qPCR measurements of OsDGK1 transcript in WT-DJ, dgk1-1, COM29, OE2 and OE5. Expression levels were normalized in comparison to β-actin. Values are means ± SE (n = 3).

To construct the OsDGK1-OEs and OsDGK1-COM, the coding sequence of OsDGK1 was amplified by PCR using primers 5′-ATGGATGGACATACAAATGGCACT-3′ and 5′-GCTAAGGTGAGCAACATCAACCTCA-3′ and cloned into the KpnI and BamHI sites of the binary pU1301D1 vector containing a maize ubiquitin promotor, Flag tag, and terminator. The DNA constructs were introduced into Agrobacterium tumefaciens strain EHA105, which was used to transform rice callus derived from Dongjin for overexpression (OE) or dgk1-1 for COM. Positive transgenic plants were regenerated and selected based on the procedure described previously (Lin & Zhang, 2005). The positive OsDGK1-OE lines were screened by PCR with one-pair primers to the cloning vector covering OsDGK1, 5′-CAAAACAAACGAATCTCAAGCAATC-3′ (forward) and 5′-CTGGTGATTTTTGCGGACTCT-3′ (reverse). The expression level of OsDGK1-COM was normalized to that of ubiquitin (Os05g06770), using primers 5′-ACGTGAAGGCCAAGATCCAG-3′ (forward) and 5′-TCGAAGTGGTTGGCCATGAA-3′ (reverse). One-pair specific primers, 5′-ATGGCACTAATGGTTCCTGCT-3′ (forward) and 5′-CAACTTGCTGTGCCATCACC-3′ (reverse), were used to confirm the expression of OsDGK1-COM (Fig. 1f).

RNA extraction and real-time quantitative PCR (RT-qPCR)

Tissue samples were taken from different organs at different developmental stages and frozen in liquid nitrogen. Total RNA was extracted using a Trizol reagent (TransGene Biotech, Beijing, China) following the manufacturer's instructions, and contaminating DNA was removed using RNase-free DNase I. The first-strand cDNA was synthesized from the RNA samples using the TranScipt cDNA Synthesis SuperMix kit (TransGene Biotech). cDNA samples with the same concentration were used to evaluate the expression pattern of OsDGK1. OsDGK1 expression was measured by the MyIQ Real-Time PCR system (Bio-Rad) with Top Green PCR SuperMix Kit (TransGene Biotech) according to the manufacturer's instruction. The RT-qPCR program was as follows: one cycle of 95°C for 3 min; 40 cycles of 95°C for 10 s and 55°C for 30 s; and one cycle of extension at 95°C for 1 min and 55°C for 1 min; followed by melting curve analysis. The relative level of OsDGK1 gene expression was normalized to the transcript level of β-actin, using primers 5′-GAGGAAGAGAAATCCCGAGAAG-3′ (forward) and 5′-GAGGATGGGTGAGTTAGTGAAG-3′ (reverse) for OsDGK1. The same approach was used to confirm gene expression of selected genes identified in RNA-sequencing data, including IAA24, PIN1C, YUC5, GA20ox2, GXMT1 and DRP-RG3 with gene-specific primers (Supporting Information Table S1).

Treating plants with DAG, PA and a DGK inhibitor

DAG, PA and R59022 were dried under a stream of nitrogen gas, suspended in 5% DMSO with sonication, and finally diluted in a plant nutrient medium containing 2.5 mM KH2PO4, 5 mM KNO3, 2 mM Ca(NO3)2, 2 mM MgSO4, 51 μM Fe-EDTA, micronutrient solution and 0.5% sucrose with pH adjusted to 5.7 (Haughn & Somerville, 1986). Rice seeds were soaked in water at room temperature and then transferred onto the soil after shoot emergence. Three days later, roots of WT and DGK1-altered seedlings were dipped into 25 μM di8:0 DAG, 25 μM PA mixture (di8:0 PA: egg PA = 9 : 1; Avanti Polar Lipids, Alabaster, AL), or different concentrations (0, 0.8, 2 and 5 μM) of R59022 (EMD Millipore Corporation, Darmstadt, Germany), or DMSO as a control, all of which were diluted in a plant nutrient medium (Peters et al., 2014). Seedlings were incubated in the lipid and inhibitor soultions for 2 h every 2 d. Root parameters, including lateral root density and seminal root thickness, were measured after the seventh day or eighth day of the initial transfer day and images were analyzed using ImageJ 1.50i.

Plant growth and root trait measurements

Rice seeds were surface-sterilized by soaking in 70% (v/v) ethanol for 12 min and then in 20% (v/v) Clorox bleach solution for 12 min, followed by washing in sterilized distilled water three times. Seeds were stratified in water for 2 d at 4°C in the dark and then were planted onto half-strength Murashige & Skoog (½MS) agar medium supplemented with 0.5% sucrose and 0.9% agar (pH 5.7) under 12 h light at 24°C and 12 h dark at 22°C. For some experiments, seeds were germinated on Turface (Turface Athletics MVP, from Profile Products LLC, IL, USA) under 14 h light at 28°C and 10 h dark at 24°C, fertilized daily with 15-16-17 (compound fertilizer), 200 ppm N in a glasshouse.

Rice seedlings 7 or 8 d old grown in soil were unearthed carefully, and residual soil was washed out. The total number of visible LRs was counted along the entire SR. The thickest part of the SR was used for root thickness measurement, and the thickness of the SR at 2 cm from the root tip was also measured. SR diameter was measured with ImageJ 1.50i. To visualize LR primordia, rice SR of soil-grown seedlings was stained with methylene blue (Johnson et al., 1996). SR was fixed with FAA solution (formaldehyde solution : acetic acid : alcohol = 2 : 1 : 17) at 4°C for 24 h, rinsed with double-distilled water for 10 min and stained with 0.01% (w/v) methylene blue solution for about 2.5 h. Roots were then rinsed with double-distilled water for 10 min. A dissecting microscope was used to observe and quantify the LR primordia.

Microscopic observation of root cells

Three-day-old SR of WT-DJ, dgk1-1, OE2 and OE5 seedlings grown in ½MS medium were cut at 1 cm from the tip. The tip roots were fixed with 2% (w/v) formaldehyde in Pipes buffer (piperazine-N,N′bis(2-ethanesulfonic acid)) for 1 h under vacuum at room temperature. After washing twice with 0.1 M Pipes buffer, the root tissues were cleared with ClearSee solution (10% (w/v) xylitol powder, 15% (w/v) sodium deoxycholate and 25% (w/v) urea) at room temperature for 4 wk or longer. For post-staining, the tissues were incubated with 100 μg ml–1 Calcofluor White in a ClearSee solution for 1 h to stain cleared tissues, and then washed in ClearSee solution for 1 h (Kurihara et al., 2015). The transition zone of stained root tips was observed under a Leica SP-8 confocal microscope.

Protein extraction and immunoblotting

Total proteins were extracted from 4-wk-old WT-DJ, OsDGK1-OE2 and OE5 leaves using an extraction buffer (50 mM Tris-HCl, pH 7.5; 150 mM NaCl; 0.5% TritonX-100; and 1% protease inhibitor). After centrifugation at 13 000 g for 10 min, equal amounts of supernatant proteins were separated by 10% (v/v) SDS-PAGE and then transferred onto a polyvinylidene difluoride (PVDF) membrane for immunoblotting. The membrane was preblotted with 5% nonfat milk, followed by incubation with anti-Flag antibody (1 : 1000) that was raised in mouse and then a secondary antibody (1 : 2500) before the final color development (Bio-Rad) at room temperature. The membrane was washed four times for 15 min with PBST buffer (10 mM Tris-HCl, pH 8; 140 mM NaCl; and 0.1% Tween 20).

Lipid extraction and analysis

Rice root samples were collected from 7-d-old soil-grown seedlings of WT-DJ and OsDGK1-altered lines treated with or without R59022. Lipids were extracted from rice roots as previously described (Welti et al., 2002). Briefly, fresh rice root tissue was immersed in 75°C isopropanol with 0.01% butylated hydroxytoluene (BHT) for 15 min to suppress PLD activity. Chloroform : methanol (2 : 1) with 0.01% BHT was used to extract lipids at least three times. The lipid extracts were washed twice with 1 M KCl and once with water, and the upper aqueous phase was discarded. The remaining tissue was dried in an oven overnight at 105°C and weighed. Extracted lipids were dissolved in 1 ml chloroform. Phospholipids and galactolipids were analyzed using an electrospray ionization triple quadrupole mass spectrometry (ESI-MS/MS)-based approach as described previously (Welti et al., 2002). DAG species were determined by a series of neutral loss (NL) scans that detected DAG species as [M + NH4]+ ions (ammonium adducts) as described previously (Peters et al., 2010). Isotopically corrected signals corresponding to each DAG species were combined and quantified according to an internal standard (15:0/15:0 DAG) (Peters et al., 2010). Because of varied ionization and fragmentation efficiencies among different acyl groups (Han & Gross, 2001), the level of DAG species was expressed as mass spectral signal relative to 15:0/15:0 DAG. The lipid MS data were analyzed and processed using Analyst 1.5.F.

RNA sequencing

Total RNA samples were extracted from roots of 7-d-old soil-grown WT-DJ, dgk1-1 and OsDGK1-OE2 seedlings with three biological repeats for each line. RNA levels and integrity were tested with agarose gel electrophoresis, a nanophotometer spectrophotometer, Qubit2.0 fluorometer precise, and Agilent 2100 bioanalyzer. RNA was sequenced using an Illumina HiSeq 1000 at Novogene in Beijing (https://en.novogene.com). Raw data were analyzed using Illumina Casava 1.8, and three types of sequence reads, namely reads with adapter, no exact base information or low-quality reads, were filtered and removed. The reference genome used was Oryza sativa (ftp://ftp.ensemblgenomes.org/pub/plants/release37/fasta/oryza_sativa/dna/Oryza_sativa.IRGSP-1.0.dna.toplevel.fa.gz). The processed data with sequence reads and alignments were generated based on Hisat2 software.

Data availability

Accession numbers: OsDGK1, Os04g54200; OsDGK2, Os08g08110; OsDGK3, Os02g54650; OsDGK4, Os12g38780; OsDGK5, Os03g31180; OsDGK6, Os08g15090; OsDGK7, Os01g57420; OsDGK8, Os12g12260; AtDGK1, At5g07920; AtDGK2, At5g63770; AtDGK3, At2g18730; AtDGK4, At5g57690; AtDGK5, At2g20900; AtDGK6, At4g28130; AtDGK7, At4g30340; dgk1-1, PFG_3A-51430.L; dgk1-2, RMD_04Z11HI70; and β-actin, Os01g16414. RNA-seq data in this paper have been deposited in NCBI's Gene Expression Omnibus with accession number GSE118455 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE118455).

Results

Knockout and overexpression of OsDGK1 that is highly expressed in roots

Of eight DGKs in rice, the transcript level of OsDGK1, OsDGK8 and OsDGK7 was high in roots before flowering compared to that of other OsDGKs and tissues (Fig. 1a). The expression level of OsDGK1 remained high throughout developmental stages and became higher than that of OsDGK8 and OsDGK7 after flowering. To verify the expression pattern, we used RT-qPCR to measure the transcript level of OsDGK1 in roots, leaves, sheath, panicles and stems at four developmental phases: seedling, tillering, flowering and mature. Among the various tissues, OsDGK1 was most highly expressed in roots at tillering, florescence and mature stages (Fig. 1b). At the tillering stage, the transcript level of OsDGK1 in roots was more than 25-fold higher than that in leaves (Fig. 1b).

To investigate the functions of DGKs in rice, we screened available mutants from different sources and identified two mutants with DNA insertion in the coding region of OsDGK1 (Fig. 1c). OsDGK1 belongs to the cluster III DGKs and contains a conserved catalytic domain with an ATP binding site and an accessory domain (Gomez-Merino et al., 2004; Arisz et al., 2009). The mutant dgk1-1 was disrupted by a T-DNA insertion in the first exon whereas dgk1-2 contained a transposon insertion in the fifth exon (Fig. 1c). The absence of OsDGK1 mRNA transcript in the dgk1-1 mutant was confirmed by reverse transcription-PCR (Fig. 1d). The dgk1-2 mutant was isolated from the Zhonghua11 background and the transposon insertion was verified by PCR using DGK1-specific and transposon-border primers (Fig. 1e).

To generate OsDGK1-OE transgenic rice, we placed the OsDGK1-coding DNA sequence fused with a Flag tag under the control of a maize ubiquitin promoter (Ubiq:DGK1) and introduced it into the WT. Presence of the OsDGK1 transgene and the production of OsDGK1 protein in OE plants were verified by PCR and immunoblotting using an anti-Flag antibody, respectively (Fig 1g). In addition, Ubiq:DGK1 was introduced into mutant dgk1-1 plants. The OsDGK1 transcript was undetectable in the knockout (KO) dgk1-1 but present in the dgk1-1 harboring the Ubiq:DGK1 line (designated as COM29) (Fig. 1f). The transcript level of DGK1 in these lines was quantified using RT-qPCR. The OsDGK1 transcript level in OsDGK1-OE lines was five times more than in the WT (Fig. 1h).

OsDGK1 suppresses LR number but promotes SR and CR thickness

The high level of OsDGK1 expression in roots led us to scrutinize the effect of OsDGK1-alterations on root growth and development in rice. Because dgk1-1, OE2 and OE5 were in Dongjin whereas dgk1-2 was in Zhonghua11 variety background, we compared the OsDGK1-altered plants with their corresponding WT for various root phenotypic traits, including SR number, SR length, LR density and SR thickness in young seedlings (Fig. 2; Fig. S1). Seedlings of dgk1-1 and dgk1-2 exhibited more LRs from SR than WT, and the LR density of dgk1-1 and dgk1-2 was c. 20% higher than that of WT (Fig. 2a, b). In addition, the dgk1 seedlings had thinner SRs (Fig. 2c). Introducing DGK1 into dgk1-1 decreased LR density and increased SR thickness (Fig 2a–c), indicating that the loss of DGK1 is responsible for the altered root traits. DGK1-OE seedlings displayed opposite changes to those of dgk1 mutants in LR density and SR thickness. DGK1-OE2 and OE5 seedlings had c. 15% fewer LRs and c. 15% thicker SRs than WT (Fig. 2b,c). DGK1-KO and OE had no significant effect on SR number or SR length (Fig. S1). The same results were obtained in additional plant lines DGK1-COM19, OE36, OE59 and OE61 (Fig. S2).

Details are in the caption following the image
Opposite effects of OsDGK1-KO and overexpression (OE) on lateral root (LR) density and seminal root (SR) thickness. (a) SR phenotype of 7-d-old seedlings of WT-DJ, dgk1-1, COM29, OE2, OE5, WT-ZH and dgk1-2. Scale bars = 1 cm. (b) Lateral root (LR) density and (c) SR thickness of WT-DJ, dgk1-1, COM29, OE2, OE5, WT-ZH and dgk1-2. Values are means ± SE (n = 20 seedlings). (d) Crown roots (CRs) from 1-month-old (upper) and 2-month-old (lower) WT-DJ, dgk1-1, OE2 and OE5 plants grown in the field. Scale bars = 2 cm. (e) Two-month-old WT-DJ, dgk1-1, OE2 and OE5 plants grown in the field. Scale bars = 10 cm. (f) LR density and (g) CR thickness of 1-month-old WT-DJ, dgk1-1, OE2 and OE5 plants. (h) Plant height of 2-month-old WT-DJ, dgk1-1, OE2 and OE5 lines, which were measured from stem base to top of panicle. Values from f, g and h are means ± SE (n = 15 plants). Asterisks mark significant change from wild-type based on Student's t test: *, 0.01 < P < 0.05; **, P < 0.01. CR, crown root.

With growth, SRs degenerate gradually in rice, while CRs and LRs become the main components of the rice root system. We therefore examined the effect of OsDGK1-KO and OE on root growth and development at later stages (Fig. 2d–h). Like SRs, dgk1-1 plants had a higher LR density whereas OsDGK1-OE had a lower LR density on CRs of 1-month- or 2-month-old plants when compared with WT (Fig. 2d). The LR density on dgk1-1 CRs was two-fold higher whereas that of OE plants was 20% lower than that of WT (Fig. 2f). Conversely, the CRs of dgk1-1 plants were about 40% thinner whereas those of OsDGK1-OE plants were about 10% thicker than CRs of WT (Fig. 2g). In addition, the above-ground plant height of dgk1-1 seedlings was significantly shorter while that of OsDGK1-OE plants was higher than that of WT at 1 month (Fig. 2e,h). These results indicate that OsDGK1 suppresses LR number of SRs and CRs but increases SR or CR thickness.

OsDGK1 inhibits LR primordia formation and increases SR cell width

To investigate how OsDGK1 affected LR density and number, we stained for LR primordia, the pericycle cell layers before emerging into a new LR (Fig. 3a,b). The LR primordia density of 3-d-old dgk1-1 seedlings was about 20% higher while both OsDGK1-OE2 and OsDGK1-OE5 were c. 15% lower than that of WT (Fig. 3b). The magnitude of changes in LR primordia density was similar to that of LR density (Fig. 2b), suggesting that OsDGK1 decreased LR number by suppressing LR primordial formation, rather than LR elongation and growth.

Details are in the caption following the image
OsDGK1-KO and overexpression (OE) effects on lateral root (LR) primordia and seminal root (SR) cell size. (a) SRs of 3-d-old seedlings of WT-DJ, dgk1-1, OE2 and OE5 plants were stained with 0.01% methylene blue. Arrows mark LR primordia. Scale bars = 1 mm. (b) LR primordia density of WT-DJ and OsDGK1-altered lines. Values are means ± SE (n = 10 seedlings). (c) Transverse section of root tips in the transition zone of 3-d-old seedlings of WT-DJ, dgk1-1, OE2 and OE plants5. Scale bars = 20 μm. (d) SR radius of WT-DJ and OsDGK1-altered lines measured on root transverse sections. (e) Cell or tissue width measured in different root cell types, including epidermis (ep), exodermis (ex), sclerenchyma (sc), cortex (co-1 to co-5, outermost cortex cell to innermost cortex cell), endodermis (en), radius of stele (st) and secondary metaxylem (sm). Values from d and e are means ± SE. Asterisks mark significant change from WT-DJ based on Student's t test: *, 0.01 < P < 0.05; **, P < 0.01.

To determine the effect of OsDGK1 on SR thickness, we compared cell size and shape in SRs of 3-d-old seedlings. Images of the transition zone above columella and the meristematic zone were measured for cell or tissue width in different root cell layers, including epidermis, exodermis, sclerenchyma, cortex, endodermis, stele and secondary metaxylem (Fig. 3c). The radius of the SR transverse section of OsDGK1-OE2 and OsDGK1-OE5 was c. 10% greater than that of WT (Fig. 3d). The width of root cells, including the epidermis, the fourth and fifth cortex cells, endodermis, stele and secondary metaxylem, were smaller in dgk1-1 than in WT, whereas the width of the first and second cortex cells of dgk1-1 roots were greater than WT (Fig. 3e). OsDGK1-OEs had a larger width of the exodermis, sclerenchyma and the first to fourth cortex cells than WT (Fig. 3e). These data indicate that OsDGK1 increases SR cell width, leading to increased SR diameter and thickness.

DAG and PA restore root growth of OsDGK1-altered plants

Because DGK1 phosphorylates DAG to PA, we tested the effect of DAG and PA on root phenotypes. WT and OsDGK1-altered seedlings were grown in media supplemented with 25 μM DAG or 25 μM PA. We used lipids with a C8:0 acyl chain at both sn-1 and sn-2 positions because the short chain lipids are relatively more water-soluble and readily absorbed by roots. Treatments of DAG and PA to OsDGK1-OE and dgk1-1, respectively, restored the LR density of KO and OE lines to that of WT (Fig. 4a). Without added PA, dgk1-1 seedlings had a higher LR density than WT, but the addition of PA decreased the LR density of dgk1-1 plants to that comparable of WT in the absence of PA (Fig. 4a). Without added DAG, OsDGK1-OE seedlings had a lower LR density than WT, but the addition of DAG increased the LR density of OsDGK1-OE more than WT so the LR density was comparable between OsDGK1-OE and WT (Fig. 4a). In addition, the DAG and PA supplementations restored the SR thickness of OsDGK1-OEs and dgk1, respectively, to that of WT (Fig. 4b). Without added PA, SRs of dgk1-1 seedlings were thinner than WT, but the addition of PA increased the SR thickness of dgk1-1 so that the SR thickness became similar between dgk1-1 and WT (Fig. 4b). Without added DAG, OsDGK1-OE seedlings had a thicker SR than WT, but with added DAG, the SR thickness was similar between OsDGK1-OE and WT (Fig. 4b).

Details are in the caption following the image
Diacylglycerol (DAG), phosphatidic acid (PA) and R59022 effects on lateral root (LR) density and seminal root (SR) thickness. (a) LR density and (b) SR thickness of 8-d-old WT-DJ and OsDGK1-altered rice seedlings treated with 25 μM DAG or 25 μM PA. (c) LR density and (d) SR thickness of 7-d-old seedlings of WT-DJ, dgk1-1, OE2 and OE5 plants treated with 0–5 μM R59022 (DGK inhibitor). Values are means ± SE (= 16 seedlings). Different letters indicate difference at P < 0.05 based on one-way ANOVA.

To test further the effect of DGK1 on roots, we treated WT and OsDGK1-altered plants with the DGK inhibitor R59022 and measured LR density and SR thickness. The DGK inhibitor R59022 was reported previously to reduce DGK activity and PA formation in Arabidopsis (Gomez-Merino et al., 2005). R59022 increased LR density and decreased SR thickness of OsDGK1-OE roots in a dose-dependent manner (Fig. 4c,d). R59022 at 5 μM almost restored the LR density of OsDGK1-OEs to the level of WT (Fig. 4c), and the SR thickness of OsDGK1-OEs decreased nearly to that of WT without the inhibitor treatment (Fig. 4d). By contrast, the LR density and SR thickness of dgk1 plants showed no significant response to R59022 (Fig. 4c,d). For WT plants, at 5 μM R59022 the SR thickness was decreased but the LR density was not changed compared with WT plants without the inhibitor treatment (Fig. 4c,d). These pharmacological results support the genetic manipulation results that OsDGK1 plays a negative role in LR formation but positive role in SR thickness. This lack of response to R59022 in dgk1 could mean that among the eight members of OsDGKs, OsDGK1 is primarily responsible for modulating LR formation and SR thickness.

Changes in DAG and PA accumulation in DGK1-KO and OE roots

To test whether the DGK1-KO and OE altered DAG and PA levels, we measured the DAG and PA levels and compositions of rice roots using ESI-MS/MS. The total DAG content of two OsDGK1-OEs was c. 25% lower than that of WT, while dgk1-1 showed no significant difference from WT (Fig. 5a). The lack of difference in total DAG and PA amounts between dgk1-1 and WT-DJ might result from functional redundancy because the expression levels of OsDGK1, OsDGK7 and OsDGK8 were high in seedling roots (Fig. 1a). DAG molecular species significantly decreased in both OE lines were 18:2/16:0, 18:1/16:0, 18:3/18:0, 18:2/18:1 and 18:2/18:0 (Fig. 5c). The total amount of PA was similar among WT and OsDGK1-altered lines (Fig. 5b), but the level of 34:1 PA and 36:3 PA species in dgk1-1 roots was significantly lower than that in WT (Fig. 5d).

Details are in the caption following the image
Effects of OsDGK1 alterations on diacylglycerol (DAG), phosphatidic acid (PA) and other glycerolipids in rice roots. (a) Total DAG content in roots treated with 5 μM R59022 or not, (b) total PA content in roots treated with 5 μM R59022 or not, (c) DAG molecular species without R59022, (d) PA molecular species without R59022, (e) DAG content in roots treated with a gradient of R59022 concentrations, (f) PA content in roots treated with a gradient of R59022 concentrations, (g) glycerolipid (PC, PE, PI, PS, PG, lysoPC, MGDG and DGDG) contents, and (h) selected glycerolipid molecular species. Lipids were extracted from roots of 7-d-old seedlings treated with a gradient of R59022 concentrations. Molecular species were determined using ESI-MS/MS. DAG molecular species were expressed as the mass spectral signal, while others were expressed as mol% of total glycerolipids quantified. Glycerolipid molecular species of four genotypes without R59022 treatment were normalized with a minimum–maximum normalization method from 0 to 100, divided into three ranges including major, medium and minor levels. Lipids in the medium band are shown because this shows the most differences among wild-type (WT), knockout (KO) and overexpression (OE) roots. DAG molecular species, acyl carbons of one of the fatty acid chains: acyl carbon double bonds of one of the fatty acid chains/acyl carbons of the other fatty acid chain: acyl carbon double bonds of the other fatty acid chain; other glycerolipid molecular species, total acyl carbons: total acyl carbon double bonds. Different letters indicate difference at P < 0.05 based on one-way ANOVA. Asterisks mark significant change from WT based on Student's t test: *, 0.01 < P < 0.05; **, P < 0.01. Values are means ± SE (= 5 biological repeats). PC, phosphatidylcholine; PE, phosphatidylethanolamine; PI, phosphatidylinositol; PS, phosphatidylserine; PG, phosphatidylglycine; lysoPC, lysophosphatidylcholine; MGDG, monogalactosyldiacylglycerol; DGDG, digalactosyldiacylglycerol.

When seedlings were grown in the presence of 5 μM R59022, the DGK inhibitor treatment increased DAG but decreased PA accumulation in roots in all genotypes in a dose-dependent manner (Fig. 5e,f). The reduction of 34:1 PA and 36:3 PA levels observed in dgk1-1 in the absence of the inhibitor was abolished in the presence of R59022 (Fig. S3). The total DAG amount increased in WT and OsDGK1-altered plants, and no significant difference in the total DAG and DAG species was observed among WT, dgk1-1 and OsDGK1-OEs (Figs 5a, S3).

When the composition of other membrane glycerolipids was compared, dgk1-1 roots had a higher level of phosphatidylcholine (PC) and monogalactosyldiacylglycerol (MGDG), but a lower level of phosphatidylethanolamine (PE) while no significant difference was detected for phosphatidylglycine (PG), phosphatidylinositol (PI), phosphatidylserine (PS), lysophosphatidylcholine (lysoPC) or digalactosyldiacylglycerol (DGDG) when compared with WT (Fig. 5g). By comparison, OsDGK1-OE roots displayed no significant difference from WT in total levels of MGDG, PI, PC or other lipids detected except for a lower PS level (Fig. 5g). However, the concentrartion of some molecular species (mol%), such as 34:3 PC, 36:5 PC and 36:6 PC was higher, whereas that of 34:2 PG and 32:0 PG was lower in OE than WT (Fig. 5h). The treatment of rice seedlings with the DGK inhibitor R59022 changed membrane glycerolipid composition in WT and OsDGK1-altered lines (Fig. 5g). The percentage of PC was increased whereas that of PI was decreased in a dose-dependent manner in all genotypes (Fig. S4). At 5 μM R59022, the percentage of PC was increased whereas that of PE and PI was decreased in WT when compared with no inhibitor treatment. Like WT, dgk1-1 roots showed an increase in PC and a decrease in PE and PI, but unlike WT, dgk1-1 roots had an increase in lysoPC and DGDG but a decrease in PS compared with roots without R59022 treatments (Fig. 5g). OsDGK1-OE roots were similar to WT and dgk1-1 roots in that R59022 treatment increased PC and decreased PI, but unlike WT or dgk1-1 roots, the inhibitor caused no significant decrease in PE (Fig. 5g).

Transcriptomic changes in OsDGK1-altered roots

To gain insights into the effect of OsDGK1 on root growth and development, we profiled the transcriptomes of WT, dgk1-1 and OE2 using RNA isolated from 7-d-old roots. Among 35 000 gene transcripts detected in roots, 470 and 1847 genes displayed altered expression in dgk1-1 and OsDGK1-OE2, respectively, when compared with those in WT. Among the 470 genes with altered expression in dgk1-1 roots, 311 were upregulated whereas 159 were downregulated (Fig. 6a). By comparison, OsDGK1-OE2 roots had 994 upregulated genes and 853 downregulated genes compared with WT (Fig. 6b). Of those genes with altered expression, 178 were altered in both dgk1 and OsDGK1-OE2 (Dataset S1).

Details are in the caption following the image
Transcriptomic profiling of WT-DJ and OsDGK1-altered roots. (a,b) Volcano plots shwoing the distribution of differentially expressed genes in roots from 7-d-old soil-grown seedlings in dgk1-1 vs WT-DJ and OE2 vs WT-DJ, respectively. RNA-seq data were from three biological repeats. (c) Hierarchical clustering of transcript levels of WT-DJ, dgk1-1 and OE2 as profiled by RNA-seq (left) and several genes for which dgk1-1 or OE2 displayed dramatic changes compared with WT-DJ (right). The log10 (FPKM+1) values were normalized, transformed (scale number) and clustered. Red indicates upregulated, while blue indicates downregulated. FPKM, expected number of fragments per kilobase of transcript sequence per million base pairs sequenced. (d) GO enrichment clustering of differentially expressed genes in dgk1-1 vs WT-DJ and OE2 vs WT-DJ. (e) Real-time quantitative PCR measurements of transcripts for IAA24 (auxin-responsive protein IAA24), PIN1C (putative auxin efflux carrier component 1c), YUC5 (indole-3-pyruvate monooxygenase YUCCA5), GA20ox2 (gibberellin 20 oxidase 2-like), GXMT1 (glucuronoxylan 4-O-methyltransferase 1) and DRP-RG3 (putative disease resistance protein RGA3) were compared with those determined by RNA-seq. Values are means ± SE (= 3). Asterisks mark significant change from WT-DJ based on Student's t test: *, 0.01 < P < 0.05; **, P < 0.01.

When these differentially expressed genes were clustered based on GO enrichment analysis, the cellular processes that altered the most were extracellular and wall component metabolism, defense response, and terpenoid and hormone metabolism (Fig. 6d). These include laccase series genes (LAC8, LAC19 and LAC24), chitinase, and expansin-like proteins (expansin-like A3 and B1), XIP2L (xylanase inhibitor protein 2-like), GXMT1 (glucuronoxylan 4-O-methyltransferase 1), DRP-RP13 (disease resistance protein RPP13) and YR49 (putative disease resistance RPP13-like protein 1) (Dataset S1). The increased expression of defense-related genes is consistent with an early report that OsDGK1 plays a role in disease responses (Zhang et al., 2008). The genes involved in hormone metabolism and response include YUC5 (an indole-3-pyruvate monooxygenase), IAA24 (an auxin response element-binding transcriptional factor) and PIN1C (an auxin transporter) involved in auxin synthesis, response and transport, and GA20ox2 (a gibberellin (GA) 20-oxidase) involved in GA biosynthesis.

To verify the transcript changes, we conducted RT-qPCR of selected genes to quantify the level of transcript in WT, dgk1-1 and OsDGK1-OE2 roots (Fig. 6e; Table S1). The RNA-seq data show that transcripts of three auxin-related genes, IAA24, PIN1C and YUC5, were decreased in OsDGK1-OE2 roots (Fig. 6c), and RT-qPCR measurements indicated a c. 73%, 51% and 63% decrease for IAA24, PIN1C and YUC5, respectively, in OE2 roots compared to WT roots (Fig. 6e). By contrast, both RNA-seq and RT-qPCR data showed that the transcript level of YUC5 of dgk1-1 was higher than WT whereas that of IAA24 and PIN1C in dgk1-1 was comparable to that of WT roots (Fig. 6c,e). The transcript level of GA20ox2 was increased greatly in OsDGK1-OE2, c. 20-fold and 40-fold higher, by RNA-seq and RT-qPCR, respectively, than WT and dgk1-1 (Fig. 6c,e). GXMT1 (glucuronoxylan methyltransferase1) catalyzes 4-O-methylation of the glucuronic acid in glucuronoxylan (Urbanowicz et al., 2012). The expression level of GXMT1 was lower in dgk1-1 and higher in OsDGK1-OE roots as detected by RNA-seq and RT-qPCR (Fig. 6c,e). Compared with WT, the expression level of DRP-RG3 increased by c. 11-fold in OsDGK1-OE roots, but dgk1-1 roots exhibited only a slight increase (Fig. 6c,e). Expression data from RT-qPCR and RNA-seq for these genes were similar, indicating the reliability of RNA-seq.

Discussion

Development of the complex root architecture plays a crucial role in water and nutrient uptake and in communicating with other plants and the environment (Hochholdinger et al., 2004; Coudert et al., 2010; Varney & Canny, 2010). Here we show that OsDGK1 and associated lipid mediators, DAG and PA, play an important role in LR number and SR/CR root growth in rice (Fig. 7). The role of OsDGK1 is supported by several lines of evidence. Genetically, OsDGK1 is highly expressed in rice roots throughout development, and KO of OsDGK1 increased LR number and decreased SR radius whereas overexpression of OsDGK1 reduced the number of LRs with increased radius. Pharmacologically, inhibition of DGK activity by a DGK inhibitor has a similar effect as OsDGK1-KO, and supplementation of the DGK lipid product PA to roots restored the phenotype of OsDGK1-KO to that of WT while application of the DGK lipid substrate DAG had the opposite effect of PA on LR number and SR growth.

Details are in the caption following the image
Proposed model regarding the role of diacylglycerol (DAG) and phosphatidic acid (PA) in regulating rice root architecture. Diacylglycerol kinase 1 (DGK1) phosphorylates DAG to generate PA. DAG promotes lateral root (LR) formation whereas PA promotes seminal root (SR) and crown root (CR) growth. PA has been shown to interact with and regulate the functions of specific proteins involved in metabolism, hormone distribution, signaling and transcriptional regulation, thus altering gene expression, root growth and development. Knockout of DGK1 increases the transcript level of auxin-synthesizing genes, such as YUC5, and promotes LR formation. Overexpression of DGK1 decreases transcript levels of auxin-related genes and increases the transcript level of the GA-related gene GA20ox2, which increases SR and CR thickness and suppresses LR formation. DAG can be produced by C-type phospholipase Cs PLC and NPC whereas PA production by phospholipase D (PLD) is often associated with stress responses. PLC, phospholipase C; NPC, nonspecific phospholipase C; PLD, phospholipase D; ARF, auxin response factor.

DAG and PA have distinctively different roles in regulating root growth and architecture

The effects of the DGK inhibitor, DAG, and PA on roots suggest that the enzymatic activity of OsDGK1 affects rice LR development and SR/CR growth. DAG is a potent lipid messenger in animal cells but its signaling function in plants remains elusive. Recent studies indicate that DAG functions as a signaling molecule in plants, promoting stomatal opening and LR development (Peters et al., 2010, 2014; Wimalasekera et al., 2010). A previous study suggested that DAG activated an ion pump in the plasma membrane of guard cells and inhibited stomatal closure (Lee & Assmann, 1991). We later showed that under well-watered conditions, NPC4-KO plants had lower stomatal conductance and transpiration rates than WT, and these results, together with DAG treatments, indicated that NPC4-produced DAG promoted stomatal opening under well-watered conditions (Peters et al., 2010). KO of NPC3 or NPC5 also resulted in lower LR density (Wimalasekera et al., 2010). The decrease in LR density resulted from a decrease in LR primordia and the LR number in KO was restored to that of WT by adding DAG, but not PA, suggesting that DAG from NPC5 promotes LR development in Arabidopsis (Peters et al., 2014). These results are consistent with those of the present finding that DAG increased LR development and number in rice.

By contrast, the present results indicate that PA suppresses LR number, an effect opposite to DAG. Previous studies in Arabidopsis have indicated that PA promotes root elongation. For example, AtNPC4 KO resulted in shorter roots that can be restored to that of WT (Peters et al., 2010; Kocourková et al., 2011), but the addition of R59022 prevented DAG restoration, indicating that the DAG phosphorylation by DGK to produce PA is responsible for root elongation (Peters et al., 2010). Consistent with the above suggestion is that treatments of Arabidopsis with the DGK inhibitor R59022 inhibited root elongation (Gomez-Merino et al., 2005). In addition, PLD is another major family that produces PA by hydrolyzing phospholipids, and PLDs and PA derived from them have been reported to enhance primary root growth under stress conditions (Hong et al., 2016). These include PLDα1 and δ involved in hyperosmotic stress-induced PA production and root elongation under NaCl and drought (Zhang et al., 2004; Wang et al., 2019), PLDζs in promoting primary root elongation under phosphate deficiency (Li et al., 2006; Su et al., 2018), and PLDε in promoting root length under nitrogen deficiency (Hong et al., 2009). Recent studies have indicated that PA mediates cellular functions by interacting with effector proteins involved in different processes, such as hormone responses, metabolism, gene transcription and DNA replication (Testerink et al., 2004, 2008; Zhang et al., 2004; Kim et al., 2013, 2019; Yao et al., 2013). The total amount of PA remains unchanged between WT and dgk1-1 (Fig. 5b) even though the levels of 34:1 PA and 36:3 PA species in dgk1-1 roots are significantly lower than those in WT (Fig. 5d). It is possible that specific DAG/PA species, rather than total DAG or PA, play a more important role in rice root architecture. Our MS-based DAG/PA profiling indicates that levels of some DAG and PA species change in OsDGK1-OE and KO tissues. The inconsistency observed in the steady-state level between DAG and PA species in the OsDGK1-altered plants implies rapid turnover of the lipid metabolites. As such, the dynamic nature of lipid metabolism makes it difficult to identify specific lipid species responsible for the physiological changes.

DGK1 and associated lipid mediators in hormone metabolism, distribution and signaling

Hormones play important roles in the development of appropriate root architecture and affect the number and length of LRs and other roots. The regulation involves networking among different hormones and, in particular, auxin promotes LR initiation and primordial development whereas ABA negatively regulates lateral root formation (Mockaitis & Estelle, 2008; Fukaki & Tasaka, 2009). However, the signaling mechanism underlying the hormone production, distribution and response are much less well understood than hormone function. Recent results indicate that PLD, NPC and/or PA are mediators in auxin and ABA distribution and responses. PA mimics the effect of ABA, including stomatal movement and suppression of seed germination, and binds the protein phosphatase and NADPH oxidase to mediate the ABA effect (Ritchie & Gilroy, 1998; Zhang et al., 2004, 2009; Mishra et al., 2006; Uraji et al., 2012). The positive effect of PA in ABA signaling is consistent with the hypothesis that PA negatively regulates lateral root growth. By contrast, the regulation of PLD and PA on auxin distribution is also consistent with the hypothesis that PA suppresses LR formation. An earlier study suggested that PA regulates auxin transport and distribution through the cycling of PIN2-containing vesicles (Li & Xue, 2007). A recent study shows that PA interacts with PINOID (PID) kinase to control PIN2 activity and auxin redistribution, and a loss of PLD-mediated PA production impaired auxin redistribution and resulted in markedly reduced primary root growth in Arabidopsis in response to salt stress (Wang et al., 2019).

Data from transcriptome profiling of rice roots showed that KO and OE of DGK1 altered the expression of genes involved in hormone metabolism, distribution and signaling. The expression of YUC5 was decreased in DGK1-OE but increased in KO. This change is consistent with the effect of the auxin-biosynthetic gene that is predominantly expressed in roots and promotes LR initiation and development (Woodward et al., 2005; Lee et al., 2012). Similarly, the decrease of IAA24 and PIN1C in OE is consistent with the role of these genes in root growth and development. IAA24 is a transcription factor that binds to an early auxin response element (Kim et al., 1997; Wen et al., 2016) whereas PIN1C is an auxin transporter and is expressed in early stages of rice lateral root primordia (Bakshi et al., 2017a,b). GA20ox2 was reported to increase root thickness and decrease root number in rice (Shen et al., 2001; Zhang et al., 2013). The transcript level of GA20ox2 was increased greatly in DGK1-OE that displayed increased root thickness. The decrease in expression of genes involved in hormone metabolism, transport and signaling in DGK1-OE plants is consistent with the decreased LR formation in the OE mutant. Thus, the altered metabolism and signaling of hormones, such auxin and GA, may underlie a basis for the effect of DGK1 and associated PA and DAG changes on LR formation and root growth.

In addition, recent results have begun to shed light onto how altered metabolism of DAG and PA impact gene expression. Specifically, PA, DAG and their metabolizing enzymes, including DGK, are present in the nucleus (Arisz et al., 2009; Baldanzi et al., 2016; Poli et al., 2017). PA phosphohydrolase (PAH) that catalyzes the reverse reaction of DGK by dephosphorylating PA to DAG translocates between the endoplasmic reticulum and nucleus (Ren et al., 2010). In addition, PA outside the nuclei can enter the nuclei via vesicular trafficking, and it was reported that PA cycled back and forth from the cellular membrane to nuclei (Henkels et al., 2016). PA has been reported to act in nuclei and interacts with proteins mediating nuclear processes, such as protein nuclear translocation, transcriptional regulation, DNA replication and cell proliferation. In Arabidopsis, PA interacts with MYB transcription factors and modulates the protein nuclear translocation and/or transcription factor–promoter interaction (Yao et al., 2013; Kim et al., 2019). In animal cells, PA alters nuclear receptor binding to DNA in cancer cells (Mahankali et al., 2015; Henkels et al., 2016). Thus, the DAG conversion to PA and the interaction of DAG and PA with effector proteins could be in the plasma membrane and nuclear membranes (Fig. 7). Thus, it is possible that the DGK-mediated changes in DAG and PA may regulate directly the transcription of genes involved in hormone metabolism and signaling.

In summary, these data suggest that the interconversion between DAG and PA by lipid (de)phosphorylation modulates establishment of the rice root system. Based on the present results in rice and previous results in Arabidopsis, we propose that the regulated conversion between PA and DAG is a cellular regulator for root growth, development and architecture (Fig. 7). DAG and PA have opposite effects on LR formation and development: by contrast, DAG promotes LR growth PA suppresses it. On the other hand, PA promotes the growth and expansion of primarily roots and SRs and CRs.

Acknowledgements

We thank Dr Gynheung An lab (Kyung Hee University, Korea) for the mutant dgk1-1, and the National Center of Plant Gene Research (Wuhan, China) for the mutant dgk1-2. We also thank Dr Howard Berg for microscopic analysis and Dr Chuanmei Zhu for suggestions on the manuscript. This work was supported by grants from Agriculture and Food Research Initiative (AFRI) award no. (2016-67013-24429/project accession no. 1007600) from the USDA National Institute of Food and Agriculture, National Science Foundation (MCB-1412901), and the National Science Foundation of China (31470762, 31271514).

    Author contributions

    SY designed and performed all experiments, except for those done by the co-authors, obtained and verified DNA and plant materials, collected and analyzed numerical data, generated and modified visual images, and wrote and revised the manuscript. S-CK helped with lipid analysis, experimental design, and writing. XD aided in rice growth and transformation. YH supervised the study and edited the manuscript, and XW proposed and guided the study and edited the manuscript.