Plant homeodomain leucine zipper IV (HD‐Zip IV) transcription factors (TFs) contain an evolutionarily conserved steroidogenic acute regulatory protein (StAR)‐related lipid transfer (START) domain. While the START domain is required for TF activity, its presumed role as a lipid sensor is not clear. Here we used tandem affinity purification from We additionally found that PDF2 acts as a transcriptional regulator of phospholipid‐ and phosphate (Pi) starvation‐related genes and binds to a palindromic octamer with consensus to a Pi response element. Phospholipid homeostasis and elongation growth were altered in We propose that the START domain serves as a molecular sensor for membrane phospholipid status in the epidermis. Our data provide insights toward understanding how the lipid metabolome integrates Pi availability with gene expression.
Interactions between proteins and lipids are dynamic in living organisms, yet the full extent and biological significance of such interactions is underexplored, especially in plants. In Arabidopsis thaliana (hereafter Arabidopsis), 16 homeodomain leucine zipper transcription factors of the class IV family (HD-Zip IV TFs) contain a lipid sensor called the START domain (Schrick et al., 2004). Steroidogenic acute regulatory protein (StAR)-related lipid transfer (START) domains were first characterized in mammalian proteins involved in lipid transfer, metabolism and sensing (Ponting & Aravind, 1999;Alpy & Tomasetto, 2005). In humans, the START domain is found in 15 proteins, several of which are known to bind specific sterols, bile acids, phospholipids, sphingolipids, or steroid hormones (Alpy et al., 2009;Letourneau et al., 2012Letourneau et al., , 2015;;Clark, 2020). Homology modeling of START domains from Arabidopsis HD-Zip IV TFs suggests that plant proteins contain a similar ligand-binding pocket (Schrick et al., 2014). In accordance, deletion of this domain from HD-Zip IV member GLABRA2 (GL2) results in loss-of-function phenotypes that are partially complemented by the START domain from mammalian STARD1/StAR (Schrick et al., 2014). The observed complementation is abolished by a binding-site mutation, implying importance of lipid binding for GL2 function. Moreover, START domains from HD-Zip IV TFs PROTODERMAL FACTOR2 (PDF2) and Arabidopsis thaliana MERISTEM LAYER1 (ATML), promote transcriptional activity of a chimeric TF in yeast (Schrick et al., 2014).
PDF2 and its paralog ATML1 are thought to be functionally redundant and play a critical role in maintenance of epidermal (L1) identity of the vegetative, floral and inflorescence shoot apical meristem (Abe et al., 2003). Double knockout mutants of PDF2 and ATML1 result in severe defects in shoot epidermal cell differentiation, leading to embryonic lethality (Ogawa et al., 2015), while overexpression of ATML1 is sufficient to induce epidermal identity in internal cell layers (Takada et al., 2013). Moreover, double mutants of PDF2 with other family members (HDG1, HDG2, HDG5, HDG12), result in floral organ defects (Kamata et al., 2013a). ATML1 and PDF2 TFs bind to the L1 box, a promoter element specific to L1 genes such as those coding for extracellular proline-rich protein PROTODERMAL FACTOR1 (PDF1) (Abe et al., 2003), GDSL lipase LIP1 (Rombola-Caldentey et al., 2014) and ketoacyl-CoA synthase (KCS20), the latter of which catalyzes very long chain fatty acid (VLCFA) biosynthesis (Rombola-Caldentey et al., 2014). VLCFA produced in the epidermis are thought to function as signals affecting proliferation of internal tissues via inhibition of cytokinin synthesis, thus modulating plant growth (Nobusawa et al., 2013). PDF2 and ATML1 are reported to interact with DELLA proteins in regulation of cell expansion (Rombola-Caldentey et al., 2014). Upon gibberellin accumulation, DELLAs are subjected to proteolysis, releasing PDF2 and ATML1 to activate expression of L1 genes (Rombola-Caldentey et al., 2014).
Considering the key role of PDF2 and ATML1 in epidermal development, we investigated additional layers of regulation, whereby TF activity is controlled by a small molecule ligand. Based on the presence of an evolutionary conserved lipid-binding domain and their role as developmental regulators, plant HD-Zip START-containing TFs were suggested to constitute a link between lipid metabolism and plant development (Ponting & Aravind, 1999;Schrick et al., 2004). We hypothesized that START, by binding lipids, controls gene expression analogously to nuclear receptors from animals. An advantage of such a mechanism is that the metabolic states could be linked to cell growth and differentiation. However, the identities of small molecule ligands that bind the START domain in planta have remained elusive. To address this gap in knowledge, we applied a tandem affinity purification protocol using Arabidopsis cell cultures for analysis of small molecule partners of PDF2, a representative HD-Zip IV TF. We then performed an in vitro assay that indicates direct binding of START to lysophosphatidylcholines. Genetic analysis suggests that PDF2 TF function is critical to regulate the expression of several phospholipid and phosphate starvation response genes, and its levels are important for lipid composition. We propose a role for PDF2 in sensing membrane phospholipid status via its START domain.
Plant cell cultures, plants and growth conditions PSB Arabidopsis thaliana cell cultures (Van Leene et al., 2011) were grown in MSMO medium with 3% sucrose, 0.05 mg l À1 kinetin and 0.5 mg l À1 1-naphthaleneacetic acid on a shaker. Cells were passaged weekly to fresh medium and harvested during logarithmic growth using rapid filtration and liquid nitrogen snap freezing. Transformation with TAP constructs was as described previously (Van Leene et al., 2011). Arabidopsis thaliana plants were of the Columbia (Col-0) ecotype. Seeds for pdf2-1 and atml1-1 (Abe et al., 2003) were provided by Taku Takahashi. Both atml1-3 (SALK_033408) and atml1-4 (SALK_128172) are T-DNA insertion alleles (Roeder et al., 2012) provided by Adrienne Roeder. gl2-5 is a En-1 insertion allele of GL2 (Ohashi et al., 2003;Khosla et al., 2014). pdf2-2 (SALK_109425) and pdf2-4 (SAIL_70G06) T-DNA insertion lines (Peterson et al., 2013;Kamata et al., 2013b;San-Bento et al., 2014) were from ABRC. Genotyping primers are listed in Supporting Information Table S1. HA: PDF2 and HA:GL2 were transformed into Col-0 plants. The proGL2:EYFP:GL2, proGL2:EYFP:PDF2 and proGL2:EYFP:PDF2 constructs (and mutant variants) were transformed into gl2-5, while proGL2:EYFP:PDF2 was additionally transformed into ATML1/atml1-1;pdf2-1 and Col-0. Agrobacterium strain GV3101 (MP90) was used for transformation and construction of transgenics by floral dip (Clough & Bent, 1998), followed by selection on 20 lg ml À1 hygromycin B. Segregation patterns of 3 : 1 for EYFP expression were observed among T2 progeny from at least 20 independent transformants, and representative homozygous T3 lines were selected for analysis. Plants were grown at 23°C under continuous light on soil comprised of Metro-Mix 380, vermiculite and perlite (4 : 3 : 2) (Hummert International). For RNA or lipid extraction, seeds were sterilized by chlorine gas treatment and sown onto 0.8% agar (Micropropagation Type II; Caisson Labs, Smithfield, UT, USA) containing Murashige & Skoog (MS) basal salts (Sigma-Aldrich) (Murashige & Skoog, 1962), 1% Suc, and 0.05% MES buffer at pH 5.8. Seeds were transferred to 23°C and grown under continuous light for 12 or 14 d. A razor blade was used to remove roots, and shoots were processed for RNA or lipid extraction. For growth under Pi limitation, vapor-sterilized seeds were germinated on Pi sufficient (P+) media (20.6 mM NH 4 NO 3 , 2.26 mM CaCl 2 dihydrate, 0.759 mM MgSO 4 heptahydrate, 18.8 mM KNO 3 and 1.25 mM KH 2 PO 4 monobasic, MS micronutrient solution (M529, PhytoTech Labs, Lenexa, KS, USA), 1% Suc, 0.05% MES, pH 5.7, 0.8% agar) or Pi limiting (PÀ) media (lacking KH 2 PO 4 monobasic).
Tandem affinity purification constructs were generated by Gateway technology using pKCTAP and pKNGSTAP as described previously (Van Leene et al., 2011). PDF2 was amplified from an Arabidopsis cDNA library using PCR primers listed in Table S1. The pdf2 DSTART constructs were generated by PCR amplification of PDF2 binary N 0 and C 0 TAP constructs using PCR primers flanking the START domain (Table S1) followed by ligation. The SR54 binary vector for expression of GL2 under its native promoter (proGL2:EYFP:GL2) was previously described (Schrick et al., 2014). To construct binary vectors expressing PDF2 and ATML1, cDNA sequences were PCR amplified using Q5 High Fidelity Polymerase (New England Biolabs, Ipswich, MA, USA) and cloned into SR54 proGL2:EYFP cleaved with SalI and KpnI New Phytologist (2024) www.newphytologist.com Ó 2024 The Authors New Phytologist Ó 2024 New Phytologist Foundation using NEBuilder HiFi DNA Assembly Master Mix (New England Biolabs) with gene-specific primers (Table S1). The ∆START (∆ST) and K107E mutations in PDF2 were generated using the Q5 Site-Directed Mutagenesis Kit (New England Biolabs) (Mukherjee et al., 2022). The L480P mutation in GL2 was generated by one-step PCR-based site-directed mutagenesis (Scott et al., 2002) using PfuUltra II Fusion HS DNA polymerase (Agilent Technologies) with primers listed in Table S1. HA: PDF2 and HA:GL2 were constructed by transferring the respective cDNAs from pENTR/D-TOPO plasmids into pEarleyGate 201 (Earley et al., 2006) using Gateway LR Clonase II (Invitrogen).
Affinity purification was performed as previously described (Luzarowski et al., 2017(Luzarowski et al., , 2018)). Whole cell native protein lysates (inputs) were harvested from Arabidopsis cell cultures expressing 35S:TAP:PDF2, 35S:PDF2:TAP, 35S:TAP:pdf2 DSTART , 35S: pdf2 DSTART :TAP, or empty vector. A soluble (membrane depleted) fraction was obtained by centrifugation of the lysate for 10 min at 14 000 rcf at 4°C, followed by ultracentrifugation for 1 h at 35 000 rcf at 4°C, and subsequent incubation with IgG Sepharose. After stringent washes, bait proteins were released from the beads by TEV protease cleavage. Samples were extracted as previously described (Giavalisco et al., 2011), using a methyltert-butyl ether (MTBE)/methanol/water solvent system to separate proteins, lipids, and polar compounds into pellet, organic, and aqueous phases, respectively. Following extraction, organic and aqueous phases were dried and stored at À20°C until LC/MS analysis.
Ultra-performance liquid chromatography (Waters Acquity UPLC System, Milford, MA, USA) coupled to an Exactive mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) in positive and negative ionization mode was used to analyze the samples as described (Giavalisco et al., 2011). UPLC separation of the polar fraction was performed using an HSS T3 C18 reversed-phase column (100 mm 9 2.1 mm 9 1.8 lm particles; Waters). The mobile phases were 0.1% formic acid in water (Buffer A, ULC/MS; Biosolve, Dracut, MA, USA) and 0.1% formic acid in acetonitrile (Buffer B, ULC/MS; Biosolve). A 2 ll sample (the dried-down aqueous fraction was resuspended in 200 ll of UPLC-grade water) was loaded per injection. UPLC separation of the lipid fraction was performed using a C8 reversed-phase column (100 lm 9 2.1 lm 9 1.7 lm particles; Waters, Milford, MA, USA). Mobile phases were H 2 O (ULC/MS; Biosolve) with 1% 1 mM NH 4 Ac, 0.1% acetic acid (Buffer A) and acetonitrile : isopropanol (7 : 3, ULC/MS; Biosolve) containing 1% 1 mM NH 4 Ac, 0.1% acetic acid (Buffer B). A 2 ll sample (of the dried-down organic fraction resuspended in 200 ll of acetonitrile : isopropanol (7 : 3)) was loaded per injection. Processing of chromatograms, peak detection, and integration were performed using REFINER MS 12.0 (Genedata, Basel, Switzerland). Processing of mass spectrometry data included removal of isotopic peaks and of chemical noise, retention time alignment and adduct detection. Metabolic features (m/z at a given retention time) were queried against an in-house reference compound library (allowing 10 ppm error and up to 0.2 min deviation from the retention time). Lipid annotation was based on a previously generated library of polar and lipophilic metabolites (Giavalisco et al., 2011). See Dataset S1 for lipidomics mass spectrometry details.
The START domain coding region of PDF2 was PCR amplified using gene-specific primers having ligation independent cloning (LIC) compatible extensions (Table S1). Gel-purified PCR product and SspI-digested pET-His6-MBP-TEV-LIC vector (Addgene, Watertown, MA, USA) were treated with T4 DNA polymerase with 25 mM dCTP and dGTP for 30 min at 22°C followed by heat inactivation. A 6 ll mixture of PCR product and vector was incubated at 22°C for 30 min, followed by addition of 1 ll 25 mM EDTA and E. coli transformation. Primers used to generate pdf2 L467P via site-directed mutagenesis are listed in Table S1. E. coli BL21 Rosetta 2 (DE3) (Novagen, Madison, WI, USA) cells carrying pET-His6-MBP-TEV-PDF2 (START) and pdf2(START) L467P were grown overnight in 5 ml LB with 40 lg ml À1 kanamycin at 37°C. The next day, 0.5 l freshly prepared media was inoculated with 1 ml of culture and growth was continued at 28°C. At OD 600 0.6, expression was induced with 0.5 mM IPTG (Sigma-Aldrich), followed by incubation at 16°C for 16 h. Cells were harvested by centrifugation at 4000 rcf, 10 min at 4°C, the pellet was frozen in liquid nitrogen and stored at À20°C for 1 h. The cells were resuspended in 20 ml of ice-cold lysis buffer containing 50 mM sodium phosphate pH 7.4, 500 mM NaCl, 1 mM imidazole, 0.5 mM TCEP, 1 mM PMSF (Sigma-Aldrich), 10% glycerol, 0.1% (w/v) lysozyme (AppliChem, Darmstadt, Germany) and cOmplete Protease Inhibitor Cocktail, EDTA free (Sigma-Aldrich). Bacterial slurry was sonicated in an ice-cold ultrasonic bath (RK 31; Bandelin) for 10 min, followed by centrifugation at 13 000 rcf for 10 min at 4°C. Supernatant was mixed with 2 ml of Ni-NTA agarose (Qiagen) on a rotary shaker for 1 h at 4°C. Ni-NTA beads with bound MBP-PDF2(START) protein were washed with 12 ml of ice-cold NaCl solutions. Protein was released from the beads using a step elution gradient (100-500 mM imidazole). Each step included 3 min incubations with 0.5 ml elution buffer containing 50 mM sodium phosphate pH 7.4, 500 mM NaCl, 0.5 mM TCEP, 1 mM PMSF, 10% glycerol, and increasing imidazole concentrations (100-500 mM). Concentration and purity of MBP-PDF2(START) in elution fractions was estimated by SDS-PAGE. Imidazole was removed and proteins were concentrated using Amicon Ultra 15 ml centrifugal filters with 10 kDa cutoff. Protein folding was assessed using nano differential scanning fluorimetry (nanoDSF). Aliquots of purified protein were stored at À20°C in 50 mM sodium phosphate buffer (pH 7.4) supplemented with 500 mM NaCl.
Lipids (Avanti Polar Lipids, Alabaster, AL, USA) were dissolved in chloroform. A total of 5 mg lipid for each liposome batch was dried in a glass tube under N 2 at 60°C. Residual chloroform was removed under vacuum overnight. Dried lipid cakes were rehydrated in 500 mM NaCl and 50 mM sodium phosphate (pH 7.4) at room temperature. Small unilamellar vesicles (SUVs) were formed by sonication using an ultrasonic bath (RK 31, Bandelin) for 15 min or by extrusion through two layers of polycarbonate membranes with 50 nm pore size (Nuclepore hydrophilic membrane; Whatman, Maidstone, UK) in a handheld extruder (Avanti Polar Lipids) or by sonication. Hydrodynamic radii of liposomes were determined by dynamic light scattering (DLS) to validate successful SUV formation.
Microscale thermophoresis (MST) measurements were performed using a Monolith NT.115 (NanoTemper,Munich,Germany). Capillaries were loaded into the instrument assets in 16-point ligand titrations. MBP-PDF2(START), MBP-PDF2 (START) L467P were labeled in 50 mM sodium phosphate buffer (pH 7.4) supplemented with 500 mM NaCl using Monolith Protein Labeling kit RED-MALEIMIDE (NanoTemper) according to manufacturer's instructions. To remove the MBP tag, labeled proteins were incubated with Ni-NTA agarose (Qiagen) on a rotary shaker for 1 h at RT. Ni-NTA beads were washed with 50 mM sodium phosphate buffer (pH 7.4) supplemented with 500 mM NaCl before release with two rounds of TEV protease digestion, each with 30 U of TEV for 1 h at RT. Binding was performed in 50 mM sodium phosphate buffer (pH 7.4) supplemented with 500 mM NaCl using standard capillaries. MO. Affinity Analysis software (NanoTemper) was used to analyze binding affinities from changes in fluorescence. SDS-Test was performed according to the NanoTemper MST manual to exclude that observed changes in fluorescence were due to ligand induced changes in protein aggregation.
Plant samples of c. 50 mg were frozen in liquid nitrogen and stored at À80°C before RNA extraction with RNeasy Plant Mini Kit and on-column RNase-Free DNase Set (Qiagen). Total RNA (0.5 lg) was used as a template for cDNA synthesis with GoScript Reverse Transcriptase (Promega). Quantitative real-time polymerase chain reaction was performed using iTaq SYBR Green Supermix with the CFX96 Touch Real-Time PCR Detection System (Bio-Rad) with gene-specific primers (Table S1). Reactions contained 10 ll SYBR Green Supermix, 1 ll forward and reverse 10 lM primers, and 5 ll cDNA (diluted fivefold) in 20 ll. Standard curves were generated from 10-fold dilutions of amplicons for each primer pair. ACT7 served as the reference gene. Data represent at least 3-4 biological samples of seedling shoots with three technical replicates for each biological sample.
Plant tissues were transferred to hot isopropanol (70°C) with 0.01% BHT (butylated hydroxytoluene; Sigma) for 15 min followed by cooling to room temperature, and storage at À80°C before processing. Lipid extraction was done with chloroform: (isopropanol + methanol) : water (30 : 65 : 3.5). Samples were incubated overnight on a shakerat room temperature followed by solvent evaporation. Extracted lipids were transferred to 2 ml glass vials and dried under N 2 . Based on lipid dry weight and formula weight of c. 800 Da, lipids were eluted at 100 mM with chloroform. 100 ll of a 100 lM lipid mixture was dried under N 2 and stored at À80°C before analysis. Dried lipid fractions were resuspended in 200 ll UPLC-grade acetonitrile:isopropanol (7 : 3). A 2 ll sample was loaded per injection. LC/MS analysis was performed as described above. Raw intensities were normalized to the median of chromatogram intensity.
Seedlings, trichome phenotypes, and EYFP expression were imaged with a Leica M125 fluorescence stereo microscope fitted with a GFP2 filter set, and a Leica DFC295 digital camera with Leica Application Suite 4.1. Trichome quantification was performed as previously described (Schrick et al., 2014). Root lengths were measured using IMAGEJ software analysis of seedling images from Bio-Rad Gel Doc XR+ Imaging System. Mature plants were imaged with a Canon PowerShot ELPH 350 HS digital camera.
In vitro transcription and translation and electrophoretic mobility shift assay (EMSA) Wild-type (WT) and mutant PDF2 cDNAs were cloned from pENTR/SD/D-TOPO vectors (ABRC) into pIX-HALO (ABRC) using Gateway LR Clonase II Enzyme mix (Thermo Fisher Scientific). Halo fusion proteins were produced from 1.5 lg plasmid DNA in a 15 ll reaction using TNT SP6 High-Yield Wheat Germ Protein Expression System (Promega). Protein expression was confirmed by western blot with Anti-HaloTag monoclonal Ab (1 : 2000) (Promega) as the primary Ab and Goat Anti-Mouse IgG [HRP] (1 : 3000; GenScript A00160) as the secondary Ab. Cy3-or fluorescein (FAM)-labeled and unlabeled dsDNA probes were generated with oligonucleotides listed in Table S1. Annealing was performed with 25 lM oligonucleotides in 100 mM Tris-Cl (pH 7.5), 1 M NaCl, 10 mM EDTA at 95°C for 2 min, followed by 57°C for 5 min, 37°C for 90 min and 37°C for 2 min. EMSA reactions were prepared as previously described (Mukherjee et al., 2022). Binding reactions were performed with 200 nM (agarose gel) or 800 nM (or 9.2 lM for competition) of oligonucleotide probe (polyacrylamide gel). Electrophoresis of the protein-DNA complexes was at 4°C in a 0.6% agarose gel (1X TBE, pH 8.3) at 150 V for 1 h, or in a 7.5% polyacrylamide gel (456-1025, Bio-Rad) at 150 V for 4.5 h. The gels were analyzed with a Typhoon Trio Imager (GE Healthcare, Chicago, IL, USA) using high sensitivity New Phytologist (2024) www.newphytologist.com Ó 2024 The Authors New Phytologist Ó 2024 New Phytologist Foundation and the 532-nm green laser and 580-nm emission filter for Cy3 (600 PMT voltage) or the 488-nm blue laser and 520-nm emission filter for FAM (595 PMT voltage).
At 5-6 d after germination on P+ or PÀ agar media, 20-30 seedlings per sample were transferred to liquid P+ or PÀ media and growth was continued for 16 h at 23°C under continuous light. Cycloheximide (Sigma-Aldrich) (400 lM final concentration) or DMSO was added at 0 h, and harvesting occurred at 0, 2, 5, 10 or 24 h. For proteasome inhibition experiments, cycloheximide was added together with MG132 (50 lM) (Sigma-Aldrich 474787) or DMSO control at 0 h. Seedling samples were frozen in liquid nitrogen and stored at À80°C before protein extraction. Tissue was homogenized in liquid nitrogen and hot SDS buffer (8 M urea, 2% SDS, 0.1 M DTT, 20% glycerol, 0.1 M Tris pH 6.8, 0.004% bromophenol blue) was added before SDS-PAGE and western blotting. Anti-HA (1 : 10 000; Pierce, Rockford, IL, USA) or Anti-GFP (1 : 2000; Roche) served as primary Abs, followed by Goat Anti-Mouse IgG [HRP] (1 : 3000; GenScript A00160) as the secondary. Proteins were detected with SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific) using an Azure 300 chemiluminescence imager (Azure Biosystems, Dublin, CA, USA), and blots were stained with Bio-Safe Coomassie Blue G-250 (Bio-Rad) to monitor protein loading. Band intensities were quantified with IMAGEJ.
To investigate binding partners of the START domain from PDF2 we used tandem affinity purification (TAP) adapted for parallel analysis of protein and metabolite interactors of the bait protein of choice (Luzarowski et al., 2017(Luzarowski et al., , 2018) ) (Fig. 1a,b). We generated Arabidopsis cell lines expressing either full-length PDF2 or mutants lacking the START domain (pdf2 DSTART ) under control of the constitutive CaMV 35S promoter, with a TAP tag fused to either the amino-or carboxyl end. Whole cell native protein lysates (referred to as input) from cultures expressing PDF2, pdf2 DSTART or empty vector were ultracentrifuged to deplete cellular membranes. TAP-tagged proteins were immunoisolated from soluble fractions and, following stringent washes, bait proteins together with interactors were released. The eluate was extracted yielding protein pellets, polar and nonpolar (lipid) metabolite fractions.
The presence of the bait protein was confirmed using mass spectrometry-based proteomics (Fig. 1c). We used an LC-MS lipidomics platform to identify lipids that co-purified with PDF2, and calculated the enrichment of specific lipids in the eluate in relation to the input. Comparison of PDF2 vs pdf2 DSTART cell lines (using input normalized data) identified 12 lipid species that were at least fourfold more abundant (t-test, P < 0.05; n = 6) in PDF2 vs pdf2 DSTART lines (Fig. 1d; Dataset S2). Of the 12 differential lipid species, six were also at least fourfold more abundant (t-test, P < 0.05; n = 6) in PDF2 vs empty vector lines, constituting a list of high confidence lipid binders (Fig. 1d). The highest enrichment was for lysophosphatidylcholines (LysoPC 18:1 and LysoPC 18 : 2), with 14-and 22-fold enrichment over the empty vector and 7-and 12-fold enrichment over the pdf2 DSTART mutant (Fig. 1e). No differential lipid accumulation was found between the empty vector control and pdf2 DSTART lines (Dataset S2). Overall, the TAP experiments indicated that the START domain of PDF2 is associated with lipids, preferentially LysoPCs in Arabidopsis cell cultures.
We next tested direct binding of PDF2 to LysoPCs in a reconstituted system. The c. 26 kDa START domain from PDF2 was produced in E. coli (Fig. S1a). Like mammalian STARD1/StAR (Sluchanko et al., 2016), the PDF2 START domain, designated hereafter as PDF2(START), is highly insoluble when expressed in E. coli. To enhance solubility, the maltose binding protein (MBP) was fused to its amino terminus. The MBP tag was removed by TEV protease cleavage before binding analysis.
Alongside WT PDF2(START), we tested binding specificity using the mutant pdf2(START) L467P having a missense mutation in the C-terminal a-helix that forms the lid of the binding pocket. L467 is a predicted lipid contact site (Roderick et al., 2002), and the L467P mutation likely introduces a structural kink in the ahelix. Thus, this mutation could affect ligand binding by altering a ligand-binding site and/or by interfering with binding pocket closure. The C-terminal a-helix is conserved in START proteins from humans and plant HD-Zip IV TFs (Fig. 2a). Analogous L to P mutations in human StAR result in congenital lipoid adrenal hyperplasia, consistent with StAR loss-of-function (Bose et al., 1996;Fluck et al., 2005). Homology modeling (Roy et al., 2010;Yang & Zhang, 2015) reveals structural similarity to START from PDF2 and GL2 (Fig. 2b). In GL2, the analogous L480P mutation leads to a loss-of-function phenotype in trichome cell differentiation (Fig. 2c,d) (Mukherjee et al., 2022).
To examine binding of PDF(START) and pdf2(START) L467P to lysophospholipids, we used microscale thermophoresis (MST) in conjunction with small unilamellar liposomes. Since LysoPC does not form liposomes on its own, we tested a 1 : 1 mixture of LysoPC and DOPC (36:2 PC; 1,2-dioleoyl-sn-glycero-3phosphocholine). We used liposomes carrying DOPC alone, and liposomes carrying another phospholipid, PG 34:2 (1-palmitoyl-2-linoleoyl-sn-glycero-3-phosphoglycerol) as controls (Figs 2e,. The presence of LysoPC 18:1 favored interaction with WT PDF2(START) over the mutant. Specifically, the binding affinity to DOPC/LysoPC 18:1 liposomes was c. 12-fold greater for PDF2 (START) (K d = 17 lM) in comparison to pdf2 (START) L467P (K d = 200 lM) (Fig. 2e). By contrast, the binding data show that WT PDF(START) and mutant pdf2 (START) L467P bind DOPC and PG 34:2 liposomes with comparable affinities. These in vitro binding data indicate that PDF2 associates with and directly binds LysoPCs through its START domain, consistent with our TAP results (Fig. 1).
Ó 2024 The Authors New Phytologist Ó 2024 New Phytologist Foundation New Phytologist (2024) www.newphytologist.com
We investigated a possible regulatory connection between PDF2 and phospholipids. Previously reported genome-wide DNA affinity purification sequencing (DAP-seq) peak data (O'Malley et al., 2016) for PDF2 revealed the palindrome GAATATTC as the main DNA-binding motif (Fig. 3a). This octamer displays consensus to the previously identified P1BS element (GNA-TATNC) (Rubio et al., 2001). Under Pi limitation, P1BS is the (Pant et al., 2015).
To validate that PDF2 binds the P1BS palindrome, we performed electrophoretic mobility shift assays (EMSA) with in vitro translated proteins and a fluorescently labeled oligonucleotide containing GAATATTC. The WT PDF2 and the mutant pdf2 DST proteins caused a shift in mobility of the fluorescent probe, in contrast to missense mutant pdf2 K107E in which a conserved arginine in the HD DNA-binding domain is replaced with glutamic acid (Fig. 3b,c). Our EMSA competition experiments indicated binding with an unlabeled oligonucleotide containing the WT but not a mutant P1BS element (Fig. 3c). The results additionally show that, while the HD is required for binding to the P1BS palindrome, the START domain of PDF2 is dispensable for DNA binding, as previously reported for PDF2 binding to the L1 box (Mukherjee et al., 2022).
The binding of PDF2 to the P1BS palindrome may affect the gene expression of nearby genes, and such genes are candidates to be transcriptional targets of PDF2. We mined the available PDF2 DAP-seq data (O'Malley et al., 2016) for candidate transcriptional targets using the PANTHER (Mi et al., 2019) overrepresentation test. The top gene ontology (GO) terms were 'phospholipid catabolic process' and 'cellular response to phosphate starvation' and these displayed enrichment of c. 9.8-fold and c. 6.3-fold, respectively (Fig. 3d; Dataset S3).
We scanned the genomic regions from candidate gene targets for overlap between DAP-seq peaks and P1BS palindromes (Figs 3e, S2a). Matches of 100% were found in the promoters or 5 0 -UTR regions of several Pi starvation-induced genes, including those encoding a glycerophosphodiester phosphodiesterase (GDPD1) (Cheng et al., 2011), an SPX domain-containing nuclear protein (SPX1) (Puga et al., 2014), a nonspecific phospholipase C (NPC4) (Nakamura et al., 2005) (Fig. 3e), and other phospholipid-or Pi starvation-related candidates (Fig. S2a). These observations suggested that PDF2 transcriptionally regulates genes that are involved in phospholipid metabolism and/or Pi sensing.
To test whether PDF2 controls the mRNA expression of the candidate transcriptional targets, we applied reverse transcription quantitative polymerase chain reaction in conjunction with mutant analysis. Since PDF2 is expressed in the epidermis and the DAP-seq experiment utilized genomic DNA from young leaves (O'Malley et al., 2016), we extracted mRNA from seedling shoots. This material contains epidermis as well as other tissues that do not express PDF2. Therefore, we considered small differences from WT, if statistically significant, to be indicative of altered gene expression in mutants.
For this analysis we utilized previously characterized T-DNA insertion mutants of PDF2 and other HD-Zip IV genes ATML1 and GL2 (Fig. 3f). Our quantitative real-time polymerase chain reaction analysis revealed differences in PDF2 mRNA levels for each of the pdf2 alleles (Fig. 3g). The null mutant allele, pdf2-4, in which the T-DNA insertion occurs before the HD (Kamata et al., 2013b), exhibited a c. threefold increase in mRNA. By contrast, the pdf2-1 and pdf2-2 alleles, in which the T-DNA insertion disrupts the START domain, result in lower levels of the transcript (Fig. 3g).The pdf2-1 partial function mutant showed approximately WT levels of PDF2 transcript, as previously reported (Kamata et al., 2013a). By contrast, the pdf2-2 allele resulted in c. twofold lower levels of PDF2 expression (Fig. 3g). Strikingly, the EYFP:pdf2 DST transgenic line in which the START domain is deleted exhibited a similar reduction in mRNA levels as the pdf2-2 allele (Fig. 3h). The difference in endogenous PDF2 mRNA expression for WT EYFP:PDF2 vs mutant EYFP:pdf2 DST (Fig. 3h) cannot be attributed to differences in transgene expression because the transgenes exhibit similar levels of mRNA (Fig. 3i).
We performed quantitative real-time polymerase chain reaction with several candidate transcriptional targets from the DAPseq data (Figs 3e, S2b). The quantitative real-time polymerase chain reaction data show that, in comparison to WT, the Ó 2024 The Authors New Phytologist Ó 2024 New Phytologist Foundation New Phytologist (2024) www.newphytologist.com (a) (c) (d) (e) (b) * * Fig. 2 Conserved ligand contact site is required for activity and PDF2 START domain binds LysoPC in vitro. (a) Alignment of C-terminal a-helix of START from human (Hs) StAR and PCTP, and HD-Zip IV TFs from Arabidopsis thaliana (At) and Physcomitrium patens (Pp). Conserved amino acids (bold, yellow); conserved Leu/Met (green). Ligand contact sites as determined from PCTP-PC co-crystal (Roderick et al., 2002) (red). (b) Structural homology models of PDF2 and GL2 START domains generated in I-TASSER (Roy et al., 2010; Yang & Zhang, 2015) reveal conserved Leu (green) in C-terminal a-helix. (c) First leaves expressing proGL2:EYFP:GL2 vs proGL2:EYFP:gl2 L480P in gl2-5 background in comparison to wild-type (WT) and gl2-5. Normal trichomes on leaves of WT (arrows) but not mutant gl2. Bar, 1 mm. (d) Quantification of leaf trichomes: gl2 L480P mutants exhibit trichome defects similar to gl2-5. Error bars indicate SD for n ≥ 20 plants. Significant differences for gl2 L480P vs WT (unpaired t-test): *, P < 1.0EÀ10. (e) PDF2 START domain binds to LysoPC 18:1 in vitro. Binding of purified WT PDF2(START) or mutant pdf2(START) L467P to liposomes prepared using indicated lipids and lipid mixtures and measured by microscale thermophoresis (MST). Mean AE SD is shown for n = 3 independent titrations, and individual data points are shown for n = 2-3 independent titrations. Dose response curves were used to calculate binding affinities expressed as dissociation constants K d . See also Supporting Information Fig. S1. New Phytologist (2024) www.newphytologist.com Ó 2024 The Authors New Phytologist Ó 2024 New Phytologist Foundation
phospholipid catabolic gene GDPD1 was upregulated in pdf2 mutants, but not in atml1 or gl2 mutants (Fig. 3j). SPX1, which encodes a transcriptional regulator associated with the Pi starvation response (Puga et al., 2014), showed remarkably similar regulation (Fig. 3j). The NPC4 gene also showed upregulation in pdf2-1 mutants, but the expression profiles were not as consistent. We observed inconsistencies in gene regulation of other predicted transcriptional targets, some of which had DAP-seq peaks in their promoter or 5 0 -UTR (Fig. S2a). For these gene target candidates, complex regulatory mechanisms may only partially or transiently involve PDF2 function.
To examine the expression of the selected candidate transcriptional target genes under Pi sufficiency and starvation, we tested pdf2-4 null mutants alongside WT and pdf2-1 mutants. Consistent with previous studies (Nakamura et al., 2005;Cheng et al., 2011;Puga et al., 2014), WT showed upregulation of GDPD1, SPX1 and NPC4 under Pi limitation (Fig. 3k). The pdf2 mutants exhibited further upregulation of these three genes under PI limitation (Fig. 3k), suggesting that PDF2 plays a role in fine-tuning gene expression in response to Pi availability. The robust upregulation of GDPD1 in all of the pdf2 mutant alleles that we tested is consistent with a role for PDF2 in transcriptional repression of that gene. Overall, the data indicate that PDF2 activity is required for maintaining normal transcript levels of GDPD1 and a few other phospholipid-and Pi response-related genes (SPX1, NPC4) at the seedling stage and Pi conditions that we tested.
We asked whether ectopic expression of PDF2 can drive repression of the selected transcriptional targets by comparing transgenic lines expressing EYFP:PDF2 with the mutant EYFP: pdf2 DST which lacks the START domain (Fig. 3l). The WT EYFP:PDF2 expression, but not EYFP:pdf2 DST , resulted in downregulation of GDPD1, SPX1 and NPC4 (Fig. 3l). This expression pattern was not observed under Pi limitation, likely due to low levels of mRNA expression of the EYFP-tagged transgenes relative to the endogenous PDF2 gene (Fig. 3i,j). Other candidate gene targets failed to show consistent PDF2-dependent regulation in our experiments (Fig. S2). Nonetheless, our PDF2 ectopic expression studies together with our mutant analysis (Fig. 3j,k) provides evidence that PDF2 acts a transcriptional repressor of at least three transcriptional target genes that have functions related to phosphate response (GDPD1, SPX1, and NPC4). Moreover, comparison of the WT EYFP:PDF2 and mutant EYFP:pdf2 DST transgenic lines indicates that the START domain is required for transcriptional repression.
We next focused on the link between PDF2 and phospholipid metabolism. Gene expression changes in the phospholipid catabolic genes GDPD1 and NPC4 are expected to result in altered phospholipid profiles and abnormal membrane lipid remodeling.
We performed a lipidomic analysis from the same shoot tissues as those used for quantitative real-time polymerase chain reaction. Our LC-MS platform targeted > 240 lipid species including phospholipids (LysoPC, PC, PE, PG, PI, PS), sphingolipids (ceramides (Cer) and glucosylceramides (GlcCer)), glycolipids (DGDG, MGDG, SQDG), diacyl-and triacylglycerols (DAG, TAG), and fatty acids (FA). Representative lipids from each major class were quantified in WT and mutants for PDF2, ATML1 and GL2. (Dataset S4; Figs S3, S4). Double mutants for atml1-1 and pdf2-1 display morphological defects at the seedling stage (Fig. S4a) (Abe et al., 2003), and we detected significant differences from WT in > 100 lipid species (Fig. S3b; Dataset S4; Table S2). Notably, the phospholipids LysoPC, PE, PI, and PS were generally increased in atml1;pdf2. Other lipids that showed increases included DAGs, TAGs, FA, and Cer, while GlcCer, DGDG, MGDG, and SQDG were decreased (Dataset S4; Fig. S3c). Since the atml1;pdf2 double mutants display developmental defects (Abe et al., 2003), the associated lipid changes may be attributed to their abnormal physiology. The other HD-Zip mutants, which display normal growth patterns, showed phospholipid defects to a lesser extent. For example, the pdf2-1 single mutants exhibited increases in several PC and PS lipids. Alterations in DAG, TAG, galactolipids and FA were additionally observed, as expected from membrane lipid remodeling (Dataset S4; Fig. S4).
In a second lipidomics experiment we monitored lipid composition under Pi sufficiency and limitation for pdf2, atml1 and gl2 single mutants in comparison to WT (Dataset S5; Figs 4a, S5, S6). To address the role of the START domain in lipid homeostasis, we included transgenic lines that were either WT (PDF2 and GL2) or mutant for the START domain (pdf2 DST and gl2 L480P ). These transgenes were expressed as EYFP-tagged proteins in the gl2-5 background under the GL2 promoter, which drives expression in specialized epidermal cell types including trichomes (Khosla et al., 2014). We included three pdf2 alleles (pdf2-1, pdf2-2 and pdf2-4). Based on the position of their T-DNA insertion (Fig. 3f), pdf2-4 represents a null allele (Kamata et al., 2013b), while the pdf2-1 and pdf2-2 alleles are both expected to affect the function of the START domain. Likewise, atml1-3 represents a null allele (Fig. 3f), whereas atml1-4 affects START but not the HD.
Pi limitation resulted in lower levels of phospholipids in WT, as previously reported (Li et al., 2006), and we observed this trend in all the lines (Dataset S5; Figs 4a, S6). In comparison to WT, the pdf2-2 seedlings exhibited a notably altered phospholipid profile: LysoPCs were significantly increased, and others (PC, PG, PS) were increased or decreased under Pi sufficiency, whereas >30 phospholipids (LysoPC, PC, PE, PG, PS) exhibited enhanced accumulations FC ≥ 2 under Pi limitation (Fig. 4a). The pdf2-1 mutants also exhibited altered levels of several PCs, as well as other abnormal lipid accumulations, especially TAGs and FAs, similar to pdf2-2, and these defects were more pronounced
Ó 2024 The Authors New Phytologist Ó 2024 New Phytologist Foundation New Phytologist (2024) www.newphytologist.com S3, S4). The lipid defects in pdf2-2 were more severe than those in the pdf2-1 mutants, possibly due to enhanced stability of the corresponding truncated protein resulting in more DNA-binding activity from pdf2-2 allele.
We compared lipid profiles of seedlings expressing WT EYFP:PDF2 to the EYFP:pdf2 DST mutant. Strikingly, pdf2 DST exhibited FC ≥ 2 increases in LysoPCs (16:0, 18:2, 18:3) and several other phospholipids under Pi limitation (Figs 4c, S6; Table S5). Similarly, when we compared WT EYFP:GL2 to
(a) (d) (e) (f) (g) (h) (i) (j) (k) (l) (b) (c) * New Phytologist (2024) www.newphytologist.com Ó 2024 The Authors New Phytologist Ó 2024 New Phytologist Foundation
START domain mutant EYFP:gl2 L480P we also noted lipid changes that varied with Pi status: LysoPC 18:2 and several FAs were elevated FC ≥ 2 in gl2 L480P under Pi limitation (Fig. S6; Dataset S5). We compared lipid changes in pdf2-2 and pdf2 DST which both affect START domain function (but not HD DNA binding) and both exhibit reduced levels of endogenous PDF transcript (Fig. 3h,i). Under Pi limitation, pdf2-2 and pdf2 DST shared numerous phospholipid increases in comparison to controls (Fig. 4c; Tables S4, S5). We also observed increases in numerous FA species in both START domain mutants (Fig. 4d). Overall, the results suggest the strongest imbalances in phospholipid and FA levels in the pdf2 START domain mutants, notably under Pi starvation.
To examine the role of HD-Zip IV genes in regulating growth according to Pi status, we assayed WT seedlings alongside pdf2, atml1 and gl2 mutants for vertical root elongation in Pi sufficient and limiting media (Fig. 5a). For this experiment we included the pdf2-4 null mutant in comparison to the pdf2-2 START mutant that had resulted in a severe lipid phenotype. Both the pdf2-2 and pdf2-4 mutants exhibited decreased elongation growth under Pi sufficiency, while the atml1-3 and gl2-5 mutants appeared indistinguishable from WT (Fig. 5b). Under Pi limitation, the pdf2-2 mutant had a more pronounced defect in root elongation and the pdf2-4 mutant showed a slight decrease (Fig. 5b). By contrast, root elongation appeared mildly increased for the atml1-3 and gl2-5 mutants under Pi limitation.
We next examined seedlings expressing EYFP:PDF2 under the epidermal specific GL2 promoter in the gl2-5 background. The GL2 promoter drives expression in trichomes and in nonroot hair cells (Khosla et al., 2014), which undergo extensive elongation in the seedling. Expression of WT EYFP:PDF2, but not HD mutant EYFP:pdf2 K107E or START mutant EYFP:pdf2 DST , partially rescued the trichome defect of gl2-5 (Fig. 5c,d). To further test whether PDF2 is critical for elongation growth we measured root lengths in EYFP:PDF2 vs EYFP:pdf2 DST seedlings (Fig. 5e, f). The data indicate that ectopic expression of WT PDF2 under both Pi sufficiency and limitation results in increased elongation, whereas elongation in pdf2 K107E or pdf2 DST was indistinguishable from the control. At later stages, we observed growth defects and aberrant leaf morphologies in the PDF2 expressing lines, but not in pdf2 K107E or pdf2 DST lines (Fig. S7). EYFP-tagged PDF2 protein exhibited nuclear localization under both Pi sufficiency and limitation (Fig. 5g), and mutant pdf2 proteins were similarly expressed in nuclei (Fig. 5h). The data indicate that ectopic epidermal PDF2 expression drives elongation growth in the primary root, and the observed growth phenotype is dependent on both the HD and START domains. Our results corroborate our finding that both the pdf2-4 null mutant (which abolishes HD function) and the pdf2-2 mutant (which abolishes START function) are defective in elongation growth.
We further tested whether the START domain is required to control elongation growth by comparing gl2-5 seedlings stably expressing proGL2:EYFP:GL2 or proGL2:EYFP:gl2 L480P . The START domain mutation L480P leads to trichome defects (Figs 2d, 6a) (Mukherjee et al., 2022). Additionally, the gl2 L480P seedlings displayed slightly decreased root elongation under Pi sufficiency, and increased elongation in comparison to WT under Pi limitation (Fig. 6b). Our quantitative real-time polymerase (j-l). Significant differences between genotypes determined by one-way ANOVA, Tukey's test, and indicated by letters: P < 0.05. See also Fig. S2(c-e).
Ó 2024 The Authors New Phytologist Ó 2024 New Phytologist Foundation New Phytologist (2024) www.newphytologist.com chain reaction analysis showed that both the EYFP:GL2 and EYFP:gl2 L480P transgenes were expressed (Fig. 6c), arguing against the possibility that the mutant phenotype is due to the lack of mRNA expression.
The WT EYFP:GL2 showed transcriptional repression of a previously identified phospholipase target gene of GL2, namely PLDf1 (Ohashi et al., 2003), whereas the mutant EYFP:gl2 L480P failed to show this transcriptional repression (Fig. 6d). Consistent
(a) (c) (d) (b) * * * * * * * * New Phytologist (2024) www.newphytologist.com Ó 2024 The Authors New Phytologist Ó 2024 New Phytologist Foundation
with our finding that the EYFP:GL2 and EYFP:gl2 L480P transgenes are expressed at the mRNA level (Fig. 6c), both WT and mutant EYFP-tagged proteins were visible and both exhibited nuclear localization (Fig. 6e), despite loss of transcriptional repressor activity of the gl2 L480P mutant protein. Therefore, the differential root elongation phenotype of mutant gl2 L480P vs WT GL2 as well as our quantitative real-time polymerase chain reaction analysis with the phospholipase target gene PLDf1 suggests that a functional START domain is critical for normal elongation growth and target gene repression of PLDf1 in response to Pi availability.
Time-course microarray profiles of Arabidopsis thaliana seedlings previously indicated that PDF2, ATML1 and GL2 transcripts are not significantly up-or downregulated in the initial response to Pi starvation (Lin et al., 2011). However, our quantitative realtime polymerase chain reaction data show that prolonged Pi limitation results in downregulation of both PDF2 and GL2 mRNA in seedlings (Figs 3i,6c), possibly due to feedback mechanisms affecting TF function. We next examined whether the PDF2 and GL2 proteins are posttranslationally regulated by Pi availability. Cycloheximide assays with seedlings expressing hemagglutinin (HA)-or EYFP-tagged TFs were performed to track the stability of the proteins in the presence or absence of Pi. We noted that EYFP:PDF2 displayed increased stability in comparison to HA: PDF2 (Fig. 7a), similar to the higher stability of EYFP:GL2 in comparison to HA:GL2 (Fig. 7b) that was previously reported (Subedi & Schrick, 2022). The reason for the increased stability due to the presence of the EYFP tag is not known, but it was speculated that this tag may interfere with degradation by some unknown mechanism (Subedi & Schrick, 2022).
Regardless of the epitope tag, both the PDF2 and GL2 proteins exhibited increased turnover under Pi limitation (Figs 7a-c, S8). In comparison to WT EYFP:PDF2, the mutant EYFP: pdf2 DST protein tended to be less stable under Pi sufficiency, and this instability was enhanced to under Pi limitation (Fig. 7c). Similarly, in comparison to WT EYFP:GL2, which is relatively stable at 24 h under Pi sufficiency, the EYFP:gl2 L480P mutant protein exhibited a decrease in stability and half-life of c. 10 h (Figs 7d, S8). The half-life of EYFP:gl2 L480P was further reduced to c. 2 h under Pi limitation (Fig. 7d), indicating that START is critical for protein stability under both conditions. Coincubation of seedlings under Pi limitation with cycloheximide and proteasome inhibitor MG132 resulted in increased stability of EYFP: pdf2 DST and EYFP:gl2 L480P (Fig. 7c,d), consistent with the possibility that the START mutant proteins are degraded via the proteasome. Alternatively or additionally, it is possible that unknown components that normally destabilize START mutant proteins, are degraded via the proteasome. Taken together, these experiments reveal that the HD-Zip TFs PDF2 and GL2 are destabilized under Pi limitation, and that the START domain is critical for protein stability under both Pi sufficiency and limitation.
The main finding herein is that PDF2, via its START domain, directly interacts with lysophosphatidylcholines (LysoPCs). Our initial strategy was to identify in vivo binding partners of this representative HD-Zip IV TF by performing TAP experiments with Arabidopsis cell lines. We followed up on lysophospholipids as candidate ligands using in vitro binding validation. Our data are consistent with a previous study in which START domains of PDF2, ATML1, and GL2 were expressed in yeast and subjected to immunoisolation (Schrick et al., 2014). Subsequent lipidomic analysis revealed enrichment of LysoPCs and other phospholipids (PC and PS) in START domain pull-down samples (Schrick et al., 2014). Although it is possible that the epidermal cells in which these TFs are predominantly expressed contain additional ligands, lysophospholipids now emerge as important PDF2 interactors.
LysoPC arises from partial hydrolysis of PC to remove one of the fatty acid groups. Since Pi starvation induces breakdown of PC in plants, LysoPCs serve as intermediates of the plastidic lipid biogenesis pathway. It was proposed c. 20 yr ago that LysoPC is exported from ER to chloroplast as a precursor for galactolipid synthesis (Mongrand et al., 2000). Lysophospholipids are additionally thought to serve as messengers in plants. In arbuscular mycorrhizal symbiosis, roots use LysoPC as a signaling molecule to induce expression of endogenous Pi transporter genes (Drissner et al., 2007). LysoPC is known to stimulate plasma S4). Parentheses next to lipid species indicate alternative combinations of fatty acid chains corresponding to nomenclature for numbers of carbons and double bonds. Minimum and maximum values were normalized to 0.0 and < 1.0, respectively, for visualization purposes. See Ó 2024 The Authors New Phytologist Ó 2024 New Phytologist Foundation New Phytologist (2024) www.newphytologist.com
New Phytologist membrane H + -ATPase activity (Palmgren et al., 1988;Palmgren & Sommarin, 1989), which in turn provides the driving force for uptake of Pi in the root epidermis under Pi limitation (Yan et al., 2002;Yuan et al., 2017). Interestingly, genetic analysis in Medicago truncatula indicates that H + -ATPase is required for Pi transport during arbuscular mycorrhizal symbiosis (Krajinski
100 Pm 100 Pm 1 mm gl2-5 GL2 WT 'ST in gl2-5 W T g l 2 -5 P D F 2 T 9 P D F 2 T 1 1 p d f 2 K 1 0 7 E T 4 p d f 2 K 1 0 7 E T 1 6 p d f 2 S T T 1 7 p d f 2 S T T 2 9 W T g l 2 -5 P D F 2 T 9 P D F 2 T 1 1 p d f 2 K 1 0 7 E T 4 p d f 2 K 1 0 7 E T 1 6 T 1 7 T 2 9 0 1 2 3 4 5 6 7 8 Root length (cm) P+ P-5.0 5.8 5.4 6.5 1.7 5.1 5.4 1.4 5.5 2.0 5.7 1.4 1.5 2.0 1.6 1.4 * ** ** ns ns ** p < 0.04 p < 1.0E-7 ** * p d f 2 S T p d f 2 S T 0 10 20 30 40 50 60 70 80 90 EYFP:pdf2 START T29 EYFP:pdf2 START T17 EYFP:pdf2 K107E T16 EYFP:pdf2 K107E T4 EYFP:PDF2 T10 EYFP:PDF2 T9 EYFP:GL2 gl2-5 WT Trichomes on first leaves c a a b c c c c b p < 0.004 P-P+ P-1 cm 1 cm P+ 1 cm P-1 cm P+ P-1 mm 1 mm WT gl2-5 EYFP:pdf2 'ST EYFP:pdf2 K107E EYFP:PDF2 P+ EYFP:PDF2 100 Pm 100 Pm 100 Pm EYFP:pdf2 'ST 100 Pm EYFP:pdf2 K107E 100 Pm EYFP:PDF2 100 Pm EYFP:GL2 PDF2 pdf2 pdf2 K107E 'ST 1 mm 1 mm ns ns (b) (a) (c) (e) (f) (h) (g) (d) P-WT P+ P-1 cm W T p d f 2 -4 a t m l 1 -3 g l 2 -5 Root length (cm) * * * ** p < 0.03 p < 1.0E-4 ** * pdf2-2 pdf2-4 gl2-5 atml1-3 0 1 2 3 4 5 6 1.43 1.36 6 5 . 1 5 5 . 1 1.15 3.46 2.88 3.49 3.56 3.24 p d f 2 -4 a t m l 1 -3 g l 2 -5 p d f 2 -2 p d f 2 -2 W T ** * P+ P-P-EYFP:PDF2 New Phytologist (2024) www.newphytologist.com Ó 2024 The Authors New Phytologist Ó 2024 New Phytologist Foundation et al., 2014). In Arabidopsis, extracellular LysoPC is taken up by an ATPase flippase in the root epidermis, and it is speculated that the detection of LysoPC signals the availability of Pi in the decaying humus in soils (Poulsen et al., 2015). Our study raises the possibility that the internalization of LysoPCs activates not only plasma membrane H + -ATPase, but also elicits a transcriptional response by binding to the START domain of PDF2.
PDF2 START domain binding to LysoPCs in Arabidopsis cells (Fig. 1), in yeast (Schrick et al., 2014), and in vitro (Fig. 2f) builds on mounting evidence that links HD-Zip IV TFs with phospholipid sensing. In 2003, GL2 was identified as a negative regulator of phospholipase D (PLDf1) in root hair patterning (Ohashi et al., 2003). Further insights came from studies with mammalian STARD2/PC transfer protein Ó 2024 The Authors New Phytologist Ó 2024 New Phytologist Foundation
New Phytologist (2024) www.newphytologist.com (PCTP), which binds PC and is expressed during embryonic development in the mouse. STARD2/PCTP interacts with and enhances TF activity of Pax3, a mammalian HD protein (Kanno et al., 2007). The START domain from human PCTP, similarly to the PDF2 START domain, also recruits LysoPCs in pull-down experiments in yeast (Schrick et al., 2014) Dual role of PDF2 as a metabolic sensor and transcriptional regulator of phospholipid catabolism and Pi-induced starvation Here we identify PDF2 as a negative regulator of two phospholipid catabolism genes, GDPD1 and NPC4. Until now, PDF2 was viewed primarily as an activator that functions redundantly with ATML1 to upregulate L1 genes. Surprisingly, the DAP-seq data identified a P1BS palindrome (GAATATTC) as the main DNAbinding motif for PDF2 (Fig. 3a), as opposed to the L1 box (TAAATCTA), which was reported as the DNA-binding motif for both ATML1 and PDF2 (Rombola-Caldentey et al., 2014).
Our gene expression studies show that pdf2, and not atml1 mutants exhibit transcriptional upregulation of GDPD1 and SPX1, suggesting that PDF2 is the main repressor of these genes in the shoot. Moreover, ectopic PDF2 expression that was dependent on the START domain was sufficient to drive repression of three genes, namely GDPD1, SPX1 and NPC4 (Fig. 3l). The other gene targets that we selected for quantitative real-time polymerase chain reaction analysis showed inconsistencies in mRNA expression, suggesting complex gene regulatory networks with respect to PDF2 function at the seedling stage. Thus, it would be interesting to investigate how PDF2 affects genome-wide mRNA expression of the DAP-seq predicted transcriptional targets at multiple developmental stages.
We propose that PDF2 functions as a lipid sensor for phospholipids via its START domain (Fig. 7e). In one possible model, LysoPCs bind to START to stabilize the protein, resulting in transcriptional activity. PDF2 directly binds to the promoters of several phospholipid catabolism genes resulting in repression of phospholipid catabolism. Thereby, PDF2 activity promotes incorporation of phospholipids into membranes, driving elongation growth. Mutant analysis suggests that PDF2 is important for fine-tuning transcript levels of phospholipid catabolic genes (Fig. 3). Under Pi starvation, overall phospholipid levels including LysoPCs decrease, resulting in reduced PDF2 protein levels and reduced cell elongation. However, PDF2 levels are not completely abolished. According to this model, the transcriptional activity of PDF2 is critical to regulate phospholipid catabolic genes to allow measured elongation growth according to lysophospholipid levels. Under Pi limitation, we suggest that a pool of PDF2 protein is bound to a destabilizing ligand or unliganded, resulting in degradation that could in part be regulated by the proteasome and components that remain to be identified.
PDF2 activity is positioned to protect membrane lipid biogenesis in the epidermis when Pi is limiting. Derepression of phospholipid catabolic genes leads to the production of fatty acids, galactolipids, as well as DAG and TAG, and recycling back to phospholipids (Fig. 7f). Our lipidomic profiling of pdf2, atml1, and gl2 mutants uncovered altered levels of several types of phospholipids, as well as products of phospholipid catabolism. While pdf2 mutants exhibited elongation defects in the seedling, we found that ectopic expression of PDF2 drives root elongation. The growth promoting activity in the seedling requires the function of both the START domain and HD (Fig. 5f). By contrast, ectopic expression of EYFP:PDF2 (or EYFP:ATML1) under the GL2 epidermis-specific promoter leads to dwarfism in adult plants (Fig. S7), a phenotype that is abolished by HD or START domain mutation. These observations highlight the importance of PDF2 in maintaining the normal growth pattern. We also noted that in comparison to PDF2, mutations in GL2 had opposite effects on root elongation depending on Pi availability (Fig. 6b), suggesting further complexities underlying the function of HD-Zip IV proteins in growth regulation.
Why should a phospholipid sensing mechanism that transcriptionally controls phospholipid catabolism function in the epidermis? In addition to its myriad protective functions, the epidermis plays a critical role in controlling growth. The brassinosteroid pathway for cell expansion and cell division is required in the L1 layer (Savaldi-Goldstein et al., 2007), and epidermis-localized VLCFA biosynthesis is implicated in growth control (Nobusawa et al., 2013). Here we show that PDF2 negatively regulates GDPD1 (Fig. 3), which encodes an enzyme that hydrolyzes glycerophosphodiesters to glycerol-3-phosphate to control phospholipid homeostasis (Cheng et al., 2011). We identified another PDF2 target gene, NPC4 (Fig. 3), that is important for hydrolysis and breakdown of glycosyl inositol phosphoceramides (GIPC) (Yang et al., 2021). These phosphosphingolipids, along with phosphoglycerolipids, are major constituents of the plasma membrane. It is estimated that in plants, about one-third of cellular Pi is stored in membrane phospholipids. Our study highlights the importance of membrane phospholipids and lipid homeostasis as a regulator of growth in the epidermis.
Whether other START domain-containing HD-Zip TFs besides PDF2 bind lysophospholipids needs to be tested experimentally. Considering that mammalian START proteins differ in their specificity toward various lipids ranging from fatty acids to sterols, a similar diversification is expected in plants. VLCFA-ceramides were proposed to bind the START domain of ATML1 and PDF2 in maintaining epidermal specificity (Nagata et al., 2021;Nagata & Abe, 2023). Fitting with this possibility, our TAP results for PDF2 identified one ceramide species (Cer t18:1/c24:0) that is enriched in WT vs the pdf2 DSTART mutant (Fig. 1e). In vitro studies suggest that a HD-Zip III TF binds PC and other phospholipids through its START domain (Husbands et al., 2023). Aside from HD-Zip III and IV TFs, Arabidopsis contains 14 START proteins whose ligands are unknown (Schrick et al., 2004). The START domain-containing wheat stripe rust resistance protein WKS1 shows specificity toward phosphatidic acid and phosphatidylinositol phosphates in protein lipid overlay assays (Gou et al., 2015). Another START protein from Marchantia was implicated in lipid transfer activity during Pi deprivation (Hirashima et al., 2021).
It is noteworthy that this newly discovered lipid metabolism connection between PDF2 and lysophospholipids relates to sensing of Pi, a nutrient that is crucial for plant growth. Our findings open a new area of research that will further explore how Pi sensing and membrane lipid metabolism are integrated with the developmental program in plants and across multicellular organisms. Intriguingly, a human START protein of the thioesterase family (THEM1/STARD14) that is critical for brown fat metabolism is allosterically regulated via its binding to LysoPC and fatty acids (Tillman et al., 2020). Since both LysoPCs and fatty acids are breakdown products of membrane lipid catabolism in plants, future investigations will explore how START domains evolved to effectively orchestrate gene expression networks according to environmentally guided metabolic inputs.
Arabidopsis thaliana HD-Zip IV TFs: PDF2 (At4g04890), GL2 (At1g79840), ATML1 (At4g21750); Physcomitrium patens HD-Zip IV TF: PpHDZIV (XP_024401280.1), Homo sapiens START domain proteins: StAR/STARD1 (NP_000340.2); Homo sapiens PCTP/STARD2 (NP_067036.2); Arabidopsis thaliana phospholipid catabolism enzymes: GDPD1 (At3g02040), GDPD2 (At5g41080), GDPD3 (At5g43300), NPC2 (At2g26870), NPC4 (At3g03 530), NPC6 (At3g48610), PLA2b (At2g19690); PLDɛ (At1g55 180), PLDf1 (At3g16785), PLDf2 (At3g05630); Other Pi starvation-related proteins: PHO1 (At3g23430), PHO1;H1 (At1g68740), SPX1 (At5g20150). Table S1 Oligonucleotides used in this study.
Table S2 Lipidomic changes in atml1;pdf2-1 vs wild-type Arabidopsis thaliana seedling apices.
Table S3 Lipidomic changes in pdf2-1 vs wild-type Arabidopsis thaliana seedling apices under Pi limitation.
Table S4 Lipidomic changes in pdf2-2 vs wild-type Arabidopsis thaliana seedling apices under Pi limitation.
Table S5 Lipidomic changes in EYFP:pdf2 DSTART vs EYFP: PDF2 Arabidopsis thaliana seedling apices under Pi limitation. Please note: Wiley is not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office. Ó 2024 The Authors New Phytologist Ó 2024 New Phytologist Foundation New Phytologist (2024) www.newphytologist.com
14698137, 0, Downloaded from https://nph.onlinelibrary.wiley.com/doi/10.1111/nph.19917 by Kathrin Schrick -Kansas State University , Wiley Online Library on [02/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License