Knockout of the Hmt1p Arginine Methyltransferase in Saccharomyces cerevisiae Leads to the Dysregulation of Phosphate-associated Genes and Processes*

Arginine methylation of proteins in the eukaryotic cell is predominantly catalyzed by one conserved enzyme; PRMT1 in mammals or Hmt1p in yeast. Knockout in mammals is embryonic lethal; however, Hmt1p in yeast is non-essential. The systems-level effects of hmt1 knockout in yeast were investigated. Unexpected but significant dysregulation in phosphate homeostasis was seen upon hmt1 knockout. Transcription factor-driven processes may explain these observations, or regulatory processes may link the sensing of S-adenosylmethionine to intracellular phosphate or polyphosphate. Graphical Abstract Highlights Knockout of arginine methyltransferase Hmt1p in S. cerevisiae was investigated. RNA-seq and SILAC MS/MS found downregulation of phosphate-associated processes. Phosphate homeostasis and extracellular levels of acid phosphatases were perturbed. Pho4p was an in vitro Hmt1p substrate, but this was not confirmed in vivo. Hmt1p is the predominant arginine methyltransferase in Saccharomyces cerevisiae. Its substrate proteins are involved in transcription, transcriptional regulation, nucleocytoplasmic transport and RNA splicing. Hmt1p-catalyzed methylation can also modulate protein-protein interactions. Hmt1p is conserved from unicellular eukaryotes through to mammals where its ortholog, PRMT1, is lethal upon knockout. In yeast, however, the effect of knockout on the transcriptome and proteome has not been described. Transcriptome analysis revealed downregulation of phosphate-responsive genes in hmt1Δ, including acid phosphatases PHO5, PHO11, and PHO12, phosphate transporters PHO84 and PHO89 and the vacuolar transporter chaperone VTC3. Analysis of the hmt1Δ proteome revealed decreased abundance of phosphate-associated proteins including phosphate transporter Pho84p, vacuolar alkaline phosphatase Pho8p, acid phosphatase Pho3p and subunits of the vacuolar transporter chaperone complex Vtc1p, Vtc3p and Vtc4p. Consistent with this, phosphate homeostasis was dysregulated in hmt1Δ cells, showing decreased extracellular phosphatase levels and decreased total Pi in phosphate-depleted medium. In vitro, we showed that transcription factor Pho4p can be methylated at Arg-241, which could explain phosphate dysregulation in hmt1Δ if interplay exists with phosphorylation at Ser-242 or Ser-243, or if Arg-241 methylation affects the capacity of Pho4p to homodimerize or interact with Pho2p. However, the Arg-241 methylation site was not validated in vivo and the localization of a Pho4p-GFP fusion in hmt1Δ was not different from wild type. To our knowledge, this is the first study to reveal an association between Hmt1p and phosphate homeostasis and one which suggests a regulatory link between S-adenosyl methionine and intracellular phosphate.


In Brief
Arginine methylation of proteins in the eukaryotic cell is predominantly catalyzed by one conserved enzyme; PRMT1 in mammals or Hmt1p in yeast. Knockout in mammals is embryonic lethal; however, Hmt1p in yeast is non-essential. The systems-level effects of hmt1 knockout in yeast were investigated. Unexpected but significant dysregulation in phosphate homeostasis was seen upon hmt1 knockout. Transcription factor-driven processes may explain these observations, or regulatory processes may link the sensing of S-adenosylmethionine to intracellular phosphate or polyphosphate.

Graphical Abstract
Hmt1p is the predominant arginine methyltransferase in yeast (1). It has an abundance of ϳ37,600 copies per cell and is responsible for 66% of arginine monomethylation and 89% of asymmetric dimethylation of intracellular proteins (2,3). Its substrates, identified through a series of proteome-scale and targeted approaches (4 -7), include both histone and nonhistone proteins. In the case of histone proteins, Hmt1p asymmetrically dimethylates histone H4 on Arg-3 (H4R3me2a) and contributes to the histone code through gene silencing (8,9). On non-histone proteins, Hmt1p-mediated mono-and asymmetric di-methylation of arginine is implicated in transcriptional regulation, nucleocytoplasmic shuttling of proteins and mRNA, and to a lesser extent RNA processing and translational regulation (reviewed in (10)). In total, Hmt1p has been shown to have at least 25 substrate proteins, 10 of which have been validated in vitro (10,11).
Hmt1p is conserved across model eukaryotes (such as Arabidopsis thaliana, Drosophila melanogaster, Caenorhabditis elegans) (12). In mammals, the predominant mammalian arginine methyltransferase, PRMT1, is the ortholog of Hmt1p and is involved in functions including genome integrity, cell proliferation and response to DNA damage (13). Interestingly, the dysregulation of PRMT1 is associated with human diseases such as cardiovascular disease and cancer (14) and in mice, knockout of PRMT1 is embryonic lethal (15). Despite arginine methylation being implicated in a plethora of biological processes, HMT1 in S. cerevisiae is not essential. The knockout has not been documented to show a clear metabolic phenotype (16), but has been reported to show increased transcription from silent chromatin regions because of loss of H4R3me2a (17) and higher tRNA abundance compared with the wild-type (18).
Phosphate is a critical macronutrient that is required for energy generation via ATP and GTP synthesis and DNA, RNA and phospholipid biosynthesis. It is also important for cellular signaling processes via phosphor transfer reactions that affect protein function and activity. In yeast, the phosphate (PHO) 1 pathway is modulated by phosphate availability in the environment and the phosphate requirements within the cell. The PHO pathway is regulated by the transcription factor Pho4p and its target genes involve those that encode for membrane-embedded phosphate transporters PHO84, PHO89, PHO87, PHO90, PHO91, acid phosphatases PHO5, PHO11 and PHO12, and polyphosphate synthetases and transporters VTC1, VTC2, VTC3, and VTC4 (19,20). Under phosphate limitation Pho4p becomes activated and transcribes extracellular phosphatases and transporters that sequester extracellular inorganic phosphate (P i ) from phosphate containing organic compounds and imports this P i into the cell (19). Within the cell, excess cytosolic P i is usually synthesized into polymers of P i known as polyphosphates (polyP) and transported through the vacuole membrane for storage by the multi-subunit vacuolar transporter chaperone (VTC) complex (19,21,22). polyP acts as an internal P i buffer and is mobilized from the vacuole and converted into single P i under phosphate limitation (23). polyP has been shown to serve other functions in the cell, with implications in cell cycle progression, metal chelation, pathogenicity, cellular stress and survival (24 -26).
To discover novel functions of Hmt1p, here we first investigated the changes in the transcriptome and proteome in the HMT1 deletion mutant, which revealed an apparent dysregulation in phosphate homeostasis. In hmt1⌬, many components of the PHO pathway, including repressible acid phosphatases, phosphate transporters and vacuolar transporter chaperone proteins showed downregulation and/or lower abundance in hmt1⌬ as compared with wild-type yeast during mid-log growth. These results were associated with a decrease in phosphatase production, a decrease in total P i in phosphate depleted medium, and the dysregulation of polyP homeostasis within the hmt1⌬ cell. We further showed that the transcription factor Pho4p can be methylated, at Arg-241, by in vitro incubation with Hmt1p. Our study is the first to establish an association between Hmt1p-mediated arginine methylation, the regulation of the PHO pathway and of phosphate metabolism.

EXPERIMENTAL PROCEDURES
Yeast Strains and Growth Conditions-Saccharomyces cerevisiae BY4741 haploid strain (Open Biosystems, Huntsville, AL; MATa his3⌬1 leu2⌬0 met15⌬0 ura3⌬0) was used as the wild-type in all experiments in this study unless otherwise specified. HMT1 knockout yeast (Open Biosystems; hmt1:: KANMX4 in BY4741) was used for studies related to deletion of arginine methyltransferase Hmt1p unless otherwise specified. Cells were maintained, and selection performed according to previous methods (27).
Gene Expression Analysis and Raw Reads Filtering-Gene expression analysis was performed on three biological replicates each of wild-type and hmt1⌬, sampled at mid-logarithmic growth phase (O.D. 600 ϭ 0.8 -1.0). Total RNA was extracted using TRIzol (Invitrogen, Carlsbad, CA) according to manufacturer's instructions. Ribosomal RNA depletion was performed prior to library generation (Ribominus Eukaryote kit, Life Technologies). The cDNA libraries for HiSeq 2000 sequencing were constructed from 10 g of total RNA using the TruSeq SBS Kit v3-HS (Illumina, San Diego, CA) according to the manufacturer's instructions, generating 101 bp paired-end reads from a 160 bp insert library. RNA-Seq sequencing was performed using a HiSeq 2000 (Illumina), in the Ramaciotti Centre for Genomics, the University of New South Wales. Initial quality assessment for Illumina HiSeq sequence data was based on FastQC (version 0.11.2) (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Pairend raw reads were trimmed with the BWA trimming mode at a threshold of Q13 (p ϭ 0.05) as implemented by SolexaQA version 1.11 (30). Low quality 3Ј-ends of each read were filtered. Reads that were less than 25 bp in length were discarded.
Mapping of RNA-Seq Reads and Differential Expression Analysis-Filtered reads from six samples were aligned to the S. cerevisiae S288C reference genome (version R64 -1-1) (31) with TopHat 2.0.4 (32) and Bowtie 2-2.0.0-beta7 (33) using default parameters. Count files of the aligned sequencing reads were generated by the HTSeqcount script from the Python package HTSeq (34) with intersectionnonempty mode, using the GFF annotation file downloaded from the Saccharomyces Genome Database (35). Differential gene expression analysis was performed on the count files using the DESeq package (36) from Bioconductor (37), following standard normalization procedures. Genes with less than ten read counts in each replicate in both wild-type and hmt1⌬ were removed from further analysis. Only genes differentially expressed at a false discovery rate Յ 0.1 and adjusted p value Ͻ 0.05 were considered as significantly differentially expressed genes in the analysis.
Microarray Analysis-Total RNA was extracted from four wild-type and three hmt1⌬ biological replicates using the Qiagen (Hilden, Germany) RNeasy isolation kit following mechanical disruption with 0.5 mm glass beads. Microarray analysis was performed in the Ramaciotti Centre for Genomics (University of New South Wales) using the Affymetrix GeneChip® Yeast Genome 2.0 Array. Microarray data were analyzed using R/Bioconductor (37), the '.CEL' data files imported using the affy package (1.52.0) (38). Scripts from Gillespie et al. (39) were used to mask the S. pombe probes, to extract the data for the S. cerevisiae probes. The intensity values from the S. cerevisiae probes were normalized using the Robust Multi-array Average (RMA) (40) function from the affy package. The quality of the normalized data were assessed using arrayQualityMetrics (v3.30.0) (39). Scripts from Gillespie et al. (39) were used to compute the average normalized values from multiple probes of the same gene. The differential expression of genes between the hmt1⌬ and wild-type control yeasts was analyzed using limma (v3.30.4) (41). The Remove Unwanted Variation (RUV4) tool (v0.9.6) (42) was used to remove unwanted variation from the microarray data, with 500 of the least differentially expressed genes identified from an initial limma analysis used as empirical negative control genes. Two unwanted factors identified using RUV4 were removed using the linear model and a subsequent limma analysis was used to identify a final set of differentially expressed genes. The p values were adjusted using the Benjamini-Hochberg procedure.
Protein Extraction, 1-D Gel Electrophoresis, and Immunoblotting-Cells were harvested, lysed, and proteins electrophoresed according to previous methods (43). Immunoblotting was performed according to Low et al. (4). Primary and secondary antibodies used in this study, and their conditions of use, are detailed in supplemental Table S1.
SILAC Mass Spectrometry-For SILAC experiments, lysates were mixed prior to 1-D gel electrophoresis. Resulting gel lanes were excised into 13 slices according to protein mass; these were then reduced and alkylated prior to in-gel tryptic digestion according to established methods (44). The peptide digest pool was vacuum-dried (Savant SPD1010, Thermofisher Scientific, Waltham, MA) before resuspending in 20 l of 0.1% (v/v) formic acid for mass spectrometric analysis. LC-MS/MS analysis of extracted peptides were performed twice on an LTQ Orbitrap Velos Pro (Thermo Fisher Scientific) using an UltiMate 3000 HPLC and autosampler system (Dionex, Sunnyvale, CA) according to previous methods (6).
Sequence Database Search and SILAC Data Analysis-For SILAC samples, raw files were processed using MaxQuant software (version 1.3.0.5) (45) using the Andromeda search engine, searching against the UniProtKB yeast database (July 2013, 6729 entries) (46). Search parameters were as follows: digestion with trypsin, variable modifi-cations of methionine oxidation, serine-threonine-tyrosine phosphorylation and cysteine carbamidomethylation, peptides of minimum six or more amino acids, maximum of two missed cleavages, minimum two razor peptides for quantitation, and peptide and protein false discovery rate of 0.01. No fixed modifications were used. For search tolerances, MS1 was set to Ͻ 5 ppm and MS2 was set to 0.40 Da.
Functional Analysis of Differentially Expressed Genes and Proteins-Gene ontology (GO) terms for all differentially abundant proteins or differentially expressed genes, from proteomics and RNA-Seq experiments respectively, were extracted and overrepresented functional categories were determined by use of GOMiner (47). All unique identified proteins and genes served as the background list, and level 4 GO terms categories were examined, to find terms with a p value of Ͻ 0.05.
Experimental Design and Statistical Rationale-To determine the proteome-level consequences of the deletion of HMT1, four parallel cultures (two of each of the background strain and hmt1⌬ in heavy and light media) were used in SILAC experiments, producing two biological replicates. After the normalization of protein ratios in Max-Quant (45), the resultant data set of ϳ1500 proteins with quantitative information was then subjected to filtering, prior to any statistical analysis. The following rules were applied: a) If a protein was identified with three or more peptides, any peptides with ratios greater than two standard deviations away from the mean for that protein were considered outliers and excluded. b) Proteins identified with less than two peptides were excluded. c) Proteins identified in only one dataset (A or B) were excluded. d) All contaminants (e.g. keratins, trypsin) were excluded. Post-filtering, mean protein ratios were then recalculated for the remaining 1375 proteins, and re-normalized to 1.0 using the limma package in R/Bioconductor (37). Empirical Bayes-moderated p values for each protein were finally calculated in limma; this approach does not assume any fixed distribution of protein ratios.
For RNA-Seq, biological triplicates of wild-type and hmt1⌬ yeast were grown, extracted and then analyzed by HiSeq 2000 next-generation sequencing. For microarray analysis, biological quadruplicates of wild-type and biological triplicates of hmt1⌬ yeast were grown, RNA was extracted and then analyzed. Analysis of gene expression data was as detailed above.
Acid Phosphatase Assay-Extracellular acid phosphatase activity was quantified by the hydrolysis of p-nitrophenyl phosphate (pNPP) to p-nitrophenyl as per Lev et al. (48) or Orkwis et al. (49), with modifications. Cells from overnight cultures in YEPD were pelleted and washed twice in sterile water. To induce a response to high or low phosphate, cells were then resuspended in 4 ml MM-KH 2 PO 4 or MM-KCl media at OD 600 ϭ 1.0 and incubated at 30°C for 2 h. Next, cells were pelleted and resuspended in 3.2 ml of pNPP solution (2.5 mM pNPP (Sigma-Aldrich, St. Louis, MO) in 50 mM NaOAc, pH 5.2) and incubated at 37°C. Every 30 min until 180 min, a 400 l cell and pNPP aliquot was taken and mixed with 800 l 1 M Na 2 CO 3 , cells were pelleted and p-nitrophenyl measured at OD 420 . Two biological and two technical replicates were performed and data analyzed with Student's unpaired t test.
Polyphosphate Purification and Phosphate Quantification-The gene encoding the exopolyphosphatase of S. cerevisiae (ScPpx1p), with a 6ϫ histidine tag at the C terminus, was overexpressed using the BG1805 overexpression plasmid in the yeast strain BY4741. Overexpression of proteins was induced overnight and 300 ml of culture was pelleted, resuspended in binding buffer (50 mM Tris-HCl buffer pH 8.0, 50 mM NaCl, 40 mM imidazole, 20% (v/v) glycerol, 0.25% (v/v) Triton X-100, supplemented with EDTA-free protease inhibitor (Roche, Basel, Switzerland)), lysed and clarified as described previously (43). Ppx1p was His-purified via 1 ml Ni NTA cartridges (Qiagen) precharged with Ni 2ϩ -Sepharose for affinity purification ac-cording to manufacturer's instructions. The final eluate was concentrated using the Amicon Ultra-4 10K centrifugal filter (Merck Millipore, Burlington, MA) at 25°C and buffer exchanged with 50 mM Tris-HCl (pH 7.4), 200 mM NaCl to reduce NaCl to at least 200 mM and remove imidazole. Glycerol was added to a final concentration of 50% (v/v) before storage at Ϫ80°C.
To monitor the cellular concentration of polyphosphates (polyP) in the wild-type and hmt1⌬ mutant during growth, overnight cultures of cells were subcultured into fresh YEPD at O.D. 600 ϭ 0.2 and MM-KCl and MM-KH 2 PO 4 at O.D. 600 ϭ 1.0. polyP was extracted at the lag (3 h from subculture), log (7 h) and stationary phases (10 h) of growth in YEPD as well as at 25 h after subculture. polyP in cells grown in MM-KCl and MM-KH 2 PO 4 were extracted at the same harvesting times as for YEPD. Equal numbers of cells were adjusted by cell density for polyP extraction and purification and quantification were performed as described previously by Canadell et al. (50) with previously purified ScPpx1p.
To quantify total phosphate levels, cells were grown to O.D. 600 ϭ 1.0 in YEPD. Total phosphate levels were also quantified from MM-KCl and MM-KH 2 PO 4 after 3 h of subculture from YEPD at a starting O.D. 600 ϭ 1.0. Equivalent numbers of cells were washed with sterile Milli Q water, boiled in 1 M H 2 SO 4 for 20 min and assessed using P i ColorLock™ Gold (Innova Biosciences, Cambridge, UK) according to manufacturer's instructions. Both polyphosphate and total phosphate quantification assays were performed with two technical replicates in at least three biological replicates.
Expression and His-purification of Pho2p and Pho4p-PHO2 and PHO4 were cloned from BG1805 plasmids (Thermo Fisher Scientific) into pRsfT25MCS1 (51) (see supplemental Table S2 for plasmids) using PhoLinkF and PhoR primers specific for PHO2 and PHO4 respectively (see supplemental Table S3 for primers). The vector pRsfT25MCS1 was amplified using DuetPstF and DuetHisRLink primers. PHO2 and PHO4 were cloned separately into the vector using Gibson assembly cloning kit (New England Biolabs, Ipswich, MA), according to manufacturer's instruction. Assembled plasmids were transformed into Alpha-select Gold efficiency competent cells (Bioline, London, UK). Transformants were plated onto LB plates with 50 g/ml kanamycin for selection. Successful assembly was screened with PCR using T7 promoter and DuetDOWN1 primers (supplemental Table S3) and confirmed with Sanger sequencing.
Plasmids were grown, extracted and then transformed into BL21 Rosetta (DE3) E. coli cells for protein expression. Plasmids carrying HMT1 and NPL3, a known substrate of Hmt1p (52), were also transformed into the BL21 Rosetta (DE3) expression cells. All proteins were overexpressed by growing in LB at 37°C with orbital shaking at 200 rpm and inducing expression of proteins by 1 mM IPTG. Protein induction was left for 6 h after which cells were pelleted and then lysed by cell sonication in binding buffer (50 mM Na-phosphate buffer pH 8.0, 0.5 M NaCl, 40 mM imidazole, 20% (v/v) glycerol, 0.25% (v/v) Triton X-100, 10 mM ␤-mercaptoethanol), supplemented with EDTAfree protease inhibitor (Roche). The lysates were clarified by centrifugation (16,000 ϫ g, 20 min, 4°C) and put onto HisTrap HP 1 ml columns (GE Healthcare, Chicago, IL). The final eluate was concentrated using centricon Plus-20 centrifugal column (Merck Millipore) at 4°C, subjected to buffer exchange with 50 mM sodium-phosphate buffer pH 7.4, 0.2 M NaCl, 20% (v/v) glycerol and stored as detailed previously.
Immunoprecipitation of Chromosomally GFP Tagged Pho4p-The Pho4p-GFP strain was cultured in MM-KCl or MM-KH 2 PO 4 to induce a high and low phosphate response, as above, before the Pho4p-GFP fusion protein was immunoprecipitated using GFP-Trap®_MA (ChromoTek, Munich, Germany). Cells were lysed in lysis buffer (10 mM Tris-Cl pH 7.5, 150 mM NaCl, 0.5 mM EDTAm 0.5% Triton X-100, 2.5 mM MgCl 2 , 20 U/ml DNase (New England Biolabs), cOmplete™ mini EDTA-free protease inhibitor (Roche), and PhoSTOP™ phosphatase inhibitor (Roche)) by bead-beating 3 times for 30 s with 3 min incubations on ice between steps. Lysates were then clarified (21,000 ϫ g for 40 min at 4°C) and filtered through a 0.45 m filter. 25 l of GFP-Trap®_MA bead slurry was washed three times with 500 l chilled wash buffer (10 mM Tris-Cl pH 7.5, 150 mM NaCl, 0.5 mM EDTA) before 2.5 mg of lysate was added to beads and incubated for 1 h at 4°C with end-over-end mixing. Beads were washed three more times before boiling in 50 l 2ϫ SDS-PAGE sample buffer for 10 min. Eluates were separated by 1-D electrophoresis and the band corresponding to Pho4p-GFP was analyzed by mass spectrometry as described below.
In Vitro and In Vivo Methylation and ETD-MS/MS Analysis-All assays were performed essentially as described in (27), using recombinantly generated Hmt1p (53). Methylation reactions were incubated at 30°C for 2 h. Negative controls for each reaction were performed with the omission of SAM (S-adenosyl methionine) in a replicate reaction, substituted with water. Electron transfer dissociation (ETD) analysis of in vitro protein methylation was conducted on an LTQ Orbitrap Velos Pro ETD (Thermo Fisher Scientific) as described previously (53). MS-Digest (version 5.19.1, University of California, San Francisco) outputs were used to calculate theoretical m/z values associated with doubly and triply charged arginine methylated peptides; these m/z values were incorporated into LC-MS/MS inclusion lists for use in mixed targeted and data dependent acquisition LC-MS/MS experiments. Theoretical Pho2p or Pho4p peptide masses were generated in MS-Digest using the following parameters: trypsin digest (up to two missed cleavages); variable modifications of carbamidomethyl (C), Methyl (R), Dimethyl (Uncleaved R), Oxidation (M), Phospho (STY); peptide masses 700 to 8000; and a minimum peptide length of 5. No fixed modifications were used.
For analysis of in vitro protein methylation data, files were submitted to the database search program Mascot (version 2.3, Matrix Science). Searches and verification of methylpeptides by manual examination for neutral losses were performed according to (5): instrument type was set as ETD-TRAP; precursor and MS/MS tolerances were Ϯ4 ppm and Ϯ0.4 Da, respectively; variable modifications of acrylamide (C), carbamidomethyl (C), oxidation (M), methylation (R) and demethylation (R) was used; digestion with trypsin was specified with two or four allowed missed cleavages; and the SwissProt database was searched. No fixed modifications were used. For the validation of in vitro methylated peptide of Pho4p, a peptide of sequence R met SSGALVDDDKR was chemically synthesized (ChinaPeptides, Shanghai, China) and analyzed as above. Analysis of in vivo methylation of Pho4p involved the overexpressing the His-tagged protein on a BG1805 plasmid in wild-type BY4741 yeast. This and the affinity purification of Pho4p was done according to (54). The resulting protein, as well as Pho4p-GFP immunoprecipitated as described above, were prepared for mass spectrometric analysis as described previously (53). Peptide samples were analyzed using an UltiMate 3000 HPLC and autosampler system (Dionex) coupled to a Fusion Lumos Tribrid (Thermo Fisher Scientific). Nano-LC and nano-ESI were performed following experimental procedures described previously (53). Survey scans m/z 350 -1750 were acquired in the Orbitrap (resolution ϭ 120,000 FWHM) with an AGC target value of 4 ϫ 10 5 charges (maximum ion injection time ϭ 50 ms). Peptide ions (Ͼ5000 counts) with charge states of 2-8 were sequentially isolated and fragmented via ETD performed with a 100 ms reaction time, supplemental activation employed (ETciD at 10% collision energy) and a fluoranthene anion target of 6 ϫ 10 5 . Fragment ions were mass analyzed in the linear ion trap. Dynamic exclusion was applied to ions subjected to MS/MS using the following parameters: exclude after n ϭ 1, exclusion duration ϭ 25 s and mass tolerance ϭ 10 ppm. A mixed targeted and data dependent acquisition approach was employed using inclusion lists generated following the methods described above.
Site-directed Mutagenesis of PHO4 and HMT1-The Arg-241 (R241) of Pho4p was mutated to lysine in pRsfT25MCS1-Pho4 (supplemental Table S2). This was performed by use of Site-directed Ligase Independent Mutagenesis (SLIM) developed by Chiu et al. (55) with primers listed in supplemental Table S3. Plasmids were mixed and subjected to DpnI restriction enzyme digestion (New England Biolabs) to eliminate template DNA. Ten microliters from each PCR reaction were combined and hybridized. A hybridization program consisting of a denaturing step at 99°C for 5 min, 3 cycles at 65°C for 5 min and 30°C for 45 min was used. The hybridization products were transformed into Alpha-select Gold efficiency competent cells (Bioline) at 5 l of hybridization product per 50 l of competent cells. Verification of the SLIM-modified plasmids was performed via colony PCR screening followed by Sanger sequencing. Next, mutated PHO4 gene was amplified from pRsfT25MCS1 using Pho4_Fwd145 and Pho4_Rev173 primers, whereas G68R HMT1 was amplified from DuetHmt1G68R-Npl3T18 (51) using Hmt1_Fwd145 and Hmt1_Rev173 primers (supplemental Table S3). Separately, the vector pRS426, containing URA3 marker, was amplified using p426_Fwd173 and p426_Rev145 primers. The fragments were assembled using Gibson assembly cloning kit (New England Biolabs), subjected to DpnI digestion, plasmid hybridization as above, and then screened for successful incorporation of mutated PHO4 or HMT1 into pRS426 using p426Ura primers. Subsequently, mutated PHO4 or HMT1 was amplified alongside URA3, using Fwd_Pho4/Rev_Ura3 or Fwd_Hmt1/Rev_Ura3_Hmt1 primers, respectively, containing flanking regions upstream and downstream of transformation target site as described in (28). The PCR products were transformed into wild-type S. cerevisiae as described in (56) and plated onto SC-URA plates (0.192% (w/v) yeast synthetic drop-out without uracil (Sigma-Aldrich), 0.17% (w/v) yeast nitrogen base without amino acids and ammonium sulfate (BD Biosciences), 0.5% (w/v) ammonium sulfate, 2% (w/v) D-glucose, 2% (w/v) agar). DNA was extracted from successful transformants, screened with Pho4_A/Pho4_D or Hmt1_Check_ A/Hmt1_Check_D primers, and finally confirmed with Sanger sequencing.
Fluorescence Microscopy-Low phosphate YEPD (1% (w/v) yeast extract, 2% (w/v) bacteriological peptone, 2% (w/v) D-glucose and 0.246% (w/v) MgSO 4 (pH 6.5)) was prepared by chelating phosphates with 8 ml of concentrated NH 4 OH. After 30 min at room temperature, the solution was filtered twice through Whatmann paper before filter sterilization. Wild-type, Tef1p-GFP, Pho2p-GFP, Pho4p-GFP and hmt1⌬ Pho2p-GFP and hmt1⌬ Pho4p-GFP cells were grown overnight in YEPD, washed twice with water and resuspended in a 1:10 dilution of low phosphate YEPD with and without 10 mM KH 2 PO 4 . The cultures were incubated for 5-6 h at 30°C with shaking. The localization of Pho4p was viewed in the specified growth media at room temperature under an inverted fluorescence microscope (Olympus IX71), fitted with a CoolSNAP HQ2/ICX285 camera (Photometrics, Tucson, AZ). For each strain, 20 images of representative cells were taken to quantify the nucleolar localization of Pho4p. Pho2p-GFP and Tef1p-GFP, which are proteins localized in the nucleus and cytoplasm, respectively, were included in fluorescence microscopy for comparison. Nuclei were stained by the addition of 1 g/ml DAPI in the media for 20 -30 min prior to viewing. Statistical analysis of Pho4p-GFP nuclear localization was performed using the student's unpaired t test, where p Ͻ 0.05 was considered significant.
Images were taken under an Olympus 100x OIL/1.40 objective. Image acquisition and deconvolution were performed with the Resolve3D softWoRx-Acquire (version 6.5.2) imaging software. Images were taken after 1 s of exposure, with a FITC filter for GFP fluorescence (475/28 nm excitation and 523/36 nm emission) and DAPI filter for nuclear fluorescence staining (390/18 nm excitation and 435/48 nm emission).

Transcriptome Analysis of hmt1⌬ Yeast Reveals Downregulation of Genes Associated with Phosphate Homeostasis-To
investigate the effects of Hmt1 gene deletion on the transcriptome, we first performed RNA-Seq analysis. Global gene expression in wild-type and hmt1⌬ yeast at O.D. 600 ϭ 0.8 -1.0 was directly compared, with an average of 21.9 and 24.8 million reads generated for the biological triplicates of wildtype and hmt1⌬, respectively. Filtered RNA-Seq reads were aligned to the yeast genome and differential gene expression analysis performed using DESeq (36). A total of only 17 genes were found to be significantly differentially expressed between wild-type and hmt1⌬ (Table I). Four genes showed up-regulation, whereas 13 genes including HMT1 showed downregulation in hmt1⌬. HMT1 showed more than a 1000fold difference in read counts between wild-type and hmt1⌬ yeast, confirming its knockout.
We investigated any functional relationships between differentially expressed genes using GOMiner (47). Six of the 17 genes were of unknown function. Analysis of the 11 remaining genes revealed that acid phosphatase activity and phosphate metabolic process were significantly downregulated ( Fig. 1A and supplemental Table S4 for genes in each category). Genes involved were the acid phosphatases PHO5, PHO11, PHO12, the phosphate permease PHO89 and cyclin-dependent kinase inhibitor SPL2. Interestingly, all are PHO (phosphate-responsive signaling) regulated and known to be controlled by transcription factor Pho4p (19). The VTC3 member of the vacuolar transporter chaperone, involved in intracellular polyphosphate generation (19,59,60), was also significantly downregulated.
Deletion of HMT1 Leads to Downregulation of Proteins Associated with Phosphate Homeostasis-To determine if the gene expression changes in hmt1⌬ affected the proteome, and in particular the PHO pathway, we used SILAC MS/MS. S. cerevisiae strain BY4741 lys2⌬/arg4⌬ was used as background with critical steps in lysine and arginine biosynthesis knocked out to allow complete isotopic labeling. We then deleted the gene for methyltransferase Hmt1p in that back- ground, to produce a lys2⌬/arg4⌬/hmt1⌬ strain. A reciprocal SILAC design integrating an isotope label swap was used. Four parallel cultures ( Fig. 2; each of the background and hmt1⌬ strains in heavy and in light media) were grown to mid-log phase. We had previously determined that there was no discernible difference in growth rates between hmt1⌬ and wild-type (supplemental Fig. S2). After harvest and lysis, protein extracts were combined and subjected to separation by SDS-PAGE. This produced two sets of biological replicates, A and B. We confirmed the completeness of metabolic labeling by analyzing single slices of light-labeled, heavy-labeled, and 1:1 mixed labeled lysates, using LC-MS/MS and MaxQuant. Heavy-labeled lysate showed isotope incorporation of ϳ97% (heavy peak intensity divided by light peak intensity, supplemental Fig. S3). LC-MS/MS analysis was performed on an Orbitrap mass spectrometer with one iteration of exclusion list analysis. After MaxQuant (45) processing, there were ϳ1,900 proteins identified, and quantitative information attained for ϳ1,500 proteins. We subjected the quantitative data set to filtering prior to statistical analysis; the rules used for data filtering were as in the Experimental Design and Statistical Rationale. Postfiltering, mean protein ratios were recalculated for the remaining 1375 proteins, and B series ratios were inverted to match the A series ratios (i.e. a B ratio of 0.5 was converted to 2.0). Protein ratios for both series had distributions (supplemental Fig. S4) typical of SILAC analyses (62,63). Statistically, we found 32 proteins to be of differential abundance between the background and hmt1⌬ strains, where p Ͻ 0.05. Of the 32 proteins, 13 were of increased abundance in hmt1⌬ whereas 19 showed decreased abundance (Table II). Interestingly, VTC3/Vtc3p were the only identical gene/protein pair that both showed significant downregulation. A lack of overlap in the genes/proteins of interest is not unexpected, given that the correlation between protein and mRNA levels can be poor (64). Several proteins of interest were unlikely to be detected by whole cell proteomics; for example, Pho5p, Pho11p and Pho12p are secreted extracellular acid phosphatases (65).
We investigated if there were functional relationships between the 32 differentially abundant proteins in hmt1⌬ yeast (Table II). The GOMiner tool (47) was used to analyze these proteins, relative to a background of all identified proteins postfiltering. We found a significant enrichment of four functionally related categories. The category of phosphate metabolic process was enriched among the differentially abundant proteins in hmt1⌬ (Fig. 1B), including the vacuolar transporter chaperone complex proteins (Vtc1p, Vtc3p, Vtc4p), the acid phosphatase Pho3p and the high-affinity phosphate transporter Pho84p (19,59,60). The category of transmembrane transporter activity also contained proteins associated with phosphate metabolism; Pho8p is a phosphatase (66) whereas Mir1p is a mitochondrial phosphate transporter (67). We noted that the vacuolar transporter chaperone complex proteins mapped to multiple categories, additionally being present in  Table S5.
Phosphate regulation and metabolism is controlled by the PHO (phosphate-responsive signaling) pathway in yeast (68). The vacuolar transporter chaperone (VTC) proteins are known to be induced under low-phosphate conditions, as are the Pho84p phosphate transporter and the Pho8p vacuolar phosphatase (19). All are under the control of the Pho4p transcriptional activator. The fact that these PHO-regulated proteins uniformly displayed a decreased abundance in hmt1⌬ suggests PHO pathway repression, as was also seen in the gene expression analysis. It is notable that three of the four proteins of the VTC complex (Vtc1p, Vtc3p, Vtc4p) demonstrated decreased abundance in hmt1⌬. The VTC complex plays a crucial role in phosphate homeostasis as part of the PHO pathway (19,69,70). The VTC complex is present on the endoplasmic reticulum, at vacuoles, and the cell periphery, but is enriched at the vacuolar membrane (71). The Emp70p and Pho8p proteins, significantly downregulated as per Vtc1p, Vtc3p, Vtc4p, also localize to this cellular component (72,73).
Some Hmt1p Substrates Display Changed Protein Abundance but Are Unchanged at Transcript Level-Comparatively few proteins (13 out of 32) displayed an increase in abun- dance in hmt1⌬ yeast (Table II). Three of these 13 proteins were of interest as they are either known or putative substrates of the methyltransferase Hmt1p (4), however it is notable that there was no significant increase or decrease in transcript levels of these or any other known or putative Hmt1 substrate. Gar1p has been shown to be methylated by Hmt1p (5,74), whereas Psp2p and Dbp2p have been proposed as substrates of Hmt1p because of the presence of RGG-rich regions (75). To further investigate the increase in abundance of Hmt1p substrates, we subjected yeast strains carrying TAP-tagged Gar1p and Dbp2p to HMT1 knockout. One-dimensional SDS-PAGE and immunoblotting validated an increased abundance of Dbp2p in hmt1⌬ (supplemental Fig.  S5). This suggests the regulation of Dbp2p abundance at the protein level. However, Gar1p in hmt1⌬ yeast was found to migrate anomalously because of the absence of arginine methylation, appearing as a band of lower mass (supplemental Fig. S5).

Network Analysis Reveals No Associations Between Known Hmt1p
Substrates and Phosphate Regulation-It was not immediately apparent how deletion of HMT1 could affect the regulation of phosphate in the yeast cell. No known Hmt1p substrates showed any differences in gene expression, whereas the above-mentioned Dbp2p, Gar1p and Psp2p proteins have no known association with the regulation of phosphate. Given that arginine methylation can modulate proteinprotein or protein-RNA interactions (52,76) we first examined networks of physical and functional interactions in the STRING database (77) for evidence of interactions between known substrates of Hmt1p and the differentially expressed genes/differentially abundant proteins. As Hmt1p is known to methylate many RNA-binding heterogeneous nuclear ribonucleoproteins (hnRNPs), it was not surprising that several of our differentially expressed transcripts have been reported to interact with known Hmt1p substrates Nab2p and Sbp1p (supplemental Fig. S6). However, Nab2p is a general RNAbinding protein, involved in nuclear export of over 2500 mRNAs (78). Similarly, Sbp1p has been reported to interact with Ͼ 1000 different RNAs (79). These interactions are therefore not specifically associated with phosphate regulation. Though not revealed in the STRING analysis, several our differentially expressed genes/differentially abundant proteins have been reported to interact with Hmt1p substrates. Siz1p has been reported to interact physically with Nab2p and Tif4632p, another putative substrate of Hmt1p (81). However, there is no evidence of those interactions being methylationdependent. Interestingly, transcripts YFL067W and YHR214W-A have been reported to bind Hek2p (82). Although Hek2p is not known to be a substrate of Hmt1p, or reported to carry arginine methylation in yeast, its human homolog hnRNP K has been reported to be methylated at five sites by PRMT1, the human equivalent of Hmt1p (83). Notwithstanding these observations, there was no apparent functional association between phosphate regulation and known substrates of Hmt1p that had differentially expressed genes or differentially abundant proteins.
HMT1 Knockout Yeast Shows Decreased Acid Phosphatase Activity and Lower Total P i Levels-During mid-log growth, hmt1⌬ showed a decrease in expression or abundance of genes and proteins associated with phosphate regulation and metabolism (Tables I and II, Fig. 1). These included repressible acid phosphatase family proteins (PHO5, PHO11, and PHO12), high-affinity phosphate transporters (Pho84p and PHO89), subunits of the VTC complex associated with polyP accumulation (Vtc1p, Vtc3p, and Vtc4p), and SPL2 -a gene whose product affects the localization of the Pho87p low affinity phosphate transporter (84). Accordingly, we investigated whether there was any change in extracellular acid phosphatase activity, along with total P i and inorganic polyphosphate (polyP) stores between hmt1⌬ and wild-type yeast.
To assay extracellular acid phosphatases, wild-type and hmt1⌬ cells were first grown in YEPD and then conditioned in phosphate depleted (MM-KCl) or phosphate replete (MM-KH 2 PO 4 ) media. Cells were then washed prior to the pNPP hydrolysis assay. Compared with the wild type, hmt1⌬ showed a ϳ33% reduction (p Ͻ 0.05) in pNPP hydrolysis when cells were first conditioned in either phosphate depleted or phosphate replete media (Fig. 3A). This indicates a significant decrease in extracellular acid phosphatase activity. A longer conditioning also showed a similar trend (supplemental Fig. S7). These observations validated the gene expression results (Table I), in which hmt1⌬ showed a significant downregulation in the genes of secreted acid phosphatases PHO5, PHO11, and PHO12. To determine that the loss of enzymatic activity of Hmt1p was the cause of the decreased acid phosphatase activity, we also assayed a strain where we had engineered the G68R mutation into the chromosomal gene for Hmt1p. This mutation generates an inactive version of the enzyme (52,85). The inactive G68R Hmt1p showed similar decreases in extracellular acid phosphatase activity as the hmt1⌬ strain (Fig. 3A), indicating that loss of Hmt1p activity and not just the loss of the Hmt1p protein itself causes a decreased level of extracellular acid phosphatases.
Total P i was extracted and quantified from wild-type and hmt1⌬ cells grown in phosphate depleted, phosphate replete or YEPD media (Fig. 3B). The hmt1⌬ cells showed a significant 30% reduction (p Ͻ 0.0001) of total P i levels, as compared with wild-type, when both were grown in phosphate depleted media. This likely reflects the lower abundance of the high affinity phosphate transporter Pho84p in hmt1⌬, as seen in the proteomic analysis (Table II). Differences in response to different media were also seen, in that wild-type cells showed a significant 30% decrease (p Ͻ 0.0001) in total P i when grown in phosphate replete MM-KH 2 PO 4 medium compared with phosphate depleted MM-KCl. In contrast, hmt1⌬ cells exhibited no significant changes in P i levels between phosphate replete and depleted media. No differences in P i levels were found between wild-type and hmt1⌬ grown in YEPD (which has a higher concentration of phosphate than MM-KCl and has higher nutrients).
Finally, polyP abundance between wild-type and hmt1⌬ cells was examined when cells were subcultured from YEPD to different media; this is known to affect polyP synthesis and accumulation (86). Accordingly, we compared the polyP accumulation between wild-type and hmt1⌬ that were first grown in YEPD and then subcultured to either phosphate depleted, phosphate replete or YEPD media. polyP accumulation was examined at the lag, log, stationary and overnight stages of growth ( Fig. 3C and supplemental Fig. S8), by enzymatic hydrolysis of polyP to P i with purified exopolyphosphatase Ppx1p. Upon transfer of cells to phosphate depleted medium (MM-KCl), we observed low overall polyP levels and little accumulation of polyP from lag through to stationary phase in wild-type or hmt1⌬ (86). However, compared with the wild-type, hmt1⌬ showed a significant decrease in polyP levels at the log phase (p ϭ 0.0023). Upon transfer of cells from YEPD to phosphate replete medium (MM-KH 2 PO 4 ), successful accumulation of polyP was evident in wild-type and hmt1⌬ cells (as seen by overall higher P i as compared with MM-KCl). This is consistent with previous reports (87). Despite the downregulation of VTC genes and proteins in hmt1⌬ cells (Tables I and II), which are involved in polyP synthesis (19), a significant increase in polyP concentration in hmt1⌬ was observed compared with the wild-type after growth overnight (p ϭ 0.0464). Upon transfer of cells from YEPD to fresh YEPD, wild-type and hmt1⌬ cells from lag to stationary phase showed an expected increase in polyP levels (86). Contrary to VTC regulation (Tables I and II), however, a significant increase in polyP concentration was observed in hmt1⌬ compared with the wild-type in the lag (p Ͻ 0.0005) and log (p ϭ 0.0361) phases.
Pho4p Is An In Vitro Substrate of Hmt1p with A Single Arginine Methylation Site at Arg-241-Our studies of gene and protein expression in hmt1⌬, and the analysis of extracellular phosphatases and intracellular P i , highlighted a dysregulation in phosphate homeostasis. Given that many of the genes are regulated by transcription factor Pho4p, which undergoes nucleocytoplasmic shuttling, and given that Hmt1p methylation can mediate nuclear exit of proteins (88,89), we investigated whether Pho4p is a substrate of Hmt1p. We also examined whether Pho2p, which complexes with Pho4p for activation of gene expression, is a substrate of Hmt1p.
Recombinant Hmt1p, Pho2p and Pho4p were purified from E. coli and used in in vitro methylation assays. Purified Npl3p, a known substrate of Hmt1p (90), was used as a positive Assay results are at time 180 min, after adding cells to the pNPP reaction mix. B, The hmt1⌬ mutant shows lower total P i compared with the wild-type in phosphate depleted medium (MM-KCl). No differences in total P i were observed for hmt1⌬ between low (MM-KCl) and high phosphate (MM-KH 2 PO 4 ) media, however the wild-type did show a significant decrease between these two conditions. C, The hmt1⌬ mutant, compared with wild-type, shows significant changes in polyP levels at some phases of growth in MM-KCl, MM-KH 2 PO 4 and YEPD. Time 3 h is lag, 7 is log, 10 is stationary and 25 h is overnight. Data information: Data are presented as mean and error bars indicate standard deviation. Mean was obtained from at least two biological replicates. In A, B and C, * indicates p Ͻ 0.05, ** indicates p Ͻ 0.002, *** indicates p Ͻ 0.0005 and **** indicates p Ͻ 0.0001 (Student's unpaired t test). control, whereas reactions omitting the methyl group donor SAM were negative controls. The samples were then subjected to gel electrophoresis and immunoblotting with an anti-mono-methylarginine primary antibody. Interestingly, this indicated that Hmt1p could catalyze the methylation of Pho4p in vitro (Fig. 4). By contrast, Pho2p showed no methylation with or without the presence of SAM, indicating that it was not methylated by Hmt1p. As expected, the positive control Npl3p was successfully methylated by Hmt1p and Pho4p was not methylated in the negative control.
To characterize the type and site(s) of methylation, we analyzed in vitro methylated Pho4p by ETD-MS/MS. This revealed mono-methylation on Arg-241, on a peptide of sequence R met SSGALVDDDKR (Fig. 5). This peptide, and its modification, was validated by chemically synthesizing the same peptide and comparing its fragmentation spectra with the native counterpart. Both peptides showed near-identical fragmentation and neutral losses characteristic of arginine monomethylation (Fig. 5B). To investigate whether only one amino acid of Pho4p is subject to methylation, site-directed mutagenesis was performed, whereby Arg-241 in recombinant Pho4p was substituted to lysine to generate Pho4p R241K. An in vitro methylation assay showed that this mutant protein could not be methylated by Hmt1p, as detected by immunoblotting (Fig. 4B).
We investigated the Arg-241 monomethylation site in vivo, for its presence and possible function when mutated to lysine. Overexpressed, His-tagged Pho4p from wild-type yeast was purified and analyzed by ETD-MS/MS. Unexpectedly, this did not reveal the presence of arginine methylation at Arg-241 (supplemental Fig. S9). A chromosomal GFP fusion of Pho4p (29) was also immunoprecipitated and analyzed by MS/MS. Although expressed at native abundance, this also did not reveal the presence of methylation at Arg-241, although it should be noted that the peptides covering this region of Pho4p were detected at low levels (supplemental Fig. S10). Phosphosites at Ser-242 and Set-243 were, however, confirmed (supplemental Figs. S9 and S10). Finally, we mutated Arg-241 to lysine in the chromosomal pho4 gene and assayed extracellular acid phosphatase levels compared with wild type. This was done in MM-KCl and in MM-KH 2 PO 4. This revealed a dysregulation of extracellular phosphatase activity (supplemental Fig. S11) although as a significant increase compared with wild type (p Ͻ 0.05) in MM-KCl. This was unexpected as a significant decrease in activity was seen in hmt1⌬ cells under the same conditions (Fig. 3A).
HMT1 Knockout Does Not Affect Nuclear Localization of Pho4p Under Phosphate Limitation-It is known that Pho4p multimerizes with Pho2p to activate the transcription of genes involved in the PHO regulatory pathway (91). Pho4p is localized to the nucleus and the cytoplasm in phosphate depleted and phosphate rich media, respectively (92). The shuttling of this transcription factor is dependent on its phosphorylation state, where the hypophosphorylation of Pho4p is critical to its nuclear localization (93). By contrast, Pho2p is localized in the nucleus and does not shuttle (94). As noted above, arginine methylation is known to be involved in the nucleocytoplasmic transport of Hmt1p substrates. The shuttling of Npl3p and Nab2p between the nucleus and the cytoplasm is methylation dependent (85,88,89,95). Hence, we investigated whether the localization of Pho4p, being an in vitro substrate of Hmt1p, is affected by the loss of arginine methylation.
A strain containing chromosomally GFP-tagged Pho4p, previously used for the study of Pho4p localization (29), and one containing GFP-tagged Pho2p (96) were subjected to deletion of HMT1. GFP-tagged Tef1p (96) was used as a cytoplasmic localization control. We studied the localization of Pho2p and Pho4p fusion proteins in phosphate replete and depleted media. In both the wild-type and hmt1⌬ mutant, there was no difference in Pho4p localization in either condition. Pho4p-GFP localized in the cytoplasm under replete phosphate and localized to the nucleus upon phosphate depletion (Fig. 6). Statistically, there was no significant difference in the level of Pho4p-GFP nuclear localization between the wild-type and knockout mutant (supplemental Fig. S12). As expected, the localization of Pho2p-GFP was nuclear in the wild-type and hmt1⌬ mutant, whereas the localization of the control protein translational elongation factor Tef1p-GFP was cytosolic.

DISCUSSION
Here we report that phosphate homeostasis in S. cerevisiae is disrupted through the deletion of HMT1. This was seen at both a transcriptomic and proteomic level; on deletion of HMT1, we observed a significant decrease in expression and abundance of many Pho4p target genes and their protein products. These included the genes PHO5, PHO11, PHO12, SPL2, PHO89, and VTC3 (Table I) and proteins Pho3p, Pho8p, Pho84p, Vtc1p, Vtc3p, and Vtc4p (Table II) (19). Functionally, we found that these changes were associated with a reduction in the activity of component(s) of PHO regulation, especially but not only the decrease in extracellular acid phosphatase activity (Fig. 3A). We showed that this phenotype was also seen in a G68R mutant of Hmt1p, showing that the loss of Hmt1p activity, and not just a loss of the entire protein, leads to dysregulation of phosphate homeostasis. A decrease in total P i levels was seen in hmt1⌬ in low phosphate conditions (Fig. 3B). By contrast, the concentration of total P i was similar between hmt1⌬ and wild type in high phosphate media MM-KH 2 PO 4 or in YEPD, which could reflect the accumulation of phosphate in cells through passive diffusion (97). Interestingly, although significant changes in polyP levels were observed in hmt1⌬ compared with wild type (Fig. 3C), these were contrary to those expected, being a decrease in polyP concentrations in phosphate replete media YEPD and MM-KH 2 PO 4 and increase in the phosphate depleted medium MM-KCl (86). This may be because of the VTC complex having functions separate to polyphosphate metabolism, such as vacuolar membrane fusion, V-ATPase activity and microautophagy (70,71,98).
In vitro investigation into Pho4p revealed that it could be monomethylated at arginine 241 by Hmt1p, in an RSS motif ( Fig. 4 and 5). The motif is different to the canonical RGG motif for Hmt1p; noncanonical sites of Hmt1p methylation have been reported elsewhere (10,11) although these too have arisen from in vitro experiments. Our analysis of Pho4p methylation in vivo was inconclusive as we did not detect methylation at Arg-241 or any other site, whether on a natively expressed GFP Pho4p fusion or on overexpressed Pho4p. However, it is possible that Arg-241 methylation only occurs under specific conditions, or is of low stoichiometry, and thus was undetected here. It is also possible that other regulatory relationships exist between arginine methylation and phosphorylation homeostasis in S. cerevisiae cell; a recent study reported the presence of polyphosphate on protein Nsr1p (99) which is a known in vivo substrate of Hmt1p (5).
The localization of some Hmt1p substrate proteins is known to be affected by methylation, including Npl3p, Hrb1p, Nab2p and Gbp2p (16,74,85,89,100). Knowing this, we investigated whether the loss of HMT1 could affect the localization of Pho4p, interfering with the Pho4p nucleocytoplasmic shuttling that is dependent on its degree of phosphorylation (101). Using chromosomal GFP-fusions, we detected no change in the subcellular localization of Pho4p in the absence of HMT1 (Fig. 6). This suggests that methylation of Pho4p, if confirmed in vivo, must affect phosphate regulation through mechanism(s) that affect the transcriptional activity but not localization of Pho4p.
The location of the putative arginine methylation site in Pho4p may provide clues to its function. There are two possible models for this. The first is that arginine methylation acts to increase the homodimerization of Pho4p or its binding affinity to DNA. To activate transcription of its target genes, Pho4p binds DNA as a homodimer with its basic helix-loophelix (bHLH) motif, located at the C terminus (102). The methylation site at Arg-241 is located between the oligomerization and bHLH DNA-binding domains of Pho4p (Fig. 7). Based on the C-terminal crystal structure of Pho4p (103), the methylarginine is unlikely to be within the DNA-binding region. However, the proximity of Arg-241 to the oligomerization B, The annotated ETD-MS/MS spectra for the methylated Pho4p peptide R met SSGALVDDDKR from the in vitro methylation assay (top) and the synthetic methylated peptide (bottom) show near-identical fragmentation patterns, validating the identification of this methylarginine site on Pho4p. Summarized ion fragment coverages, where c-and z-ions and their derivatives are shown in each spectra. Precursor and chargereduced precursor ions -c-and z-ions, which are prominent ions resulting from -NH 3 and methylarginine-associated losses -and their measured masses are labeled in the spectra. Methylarginine-associated neutral losses are abbreviated as follows: monomethylamine (MMA), monomethylguanidine (MMG). Losses from NH 3 are shown as '*'.
FIG. 6. Deletion of HMT1 has no effect on the nuclear localization of Pho4p-GFP under phosphate limited conditions compared with the wild-type. Pho2p-GFP was localized to the nucleus, even with the deletion of HMT1, whereas Tef1p-GFP was localized in the cytoplasm. Images were taken at 100x objective (oil immersion). and bHLH domains means its methylation could be involved in mediating the protein-protein interactions necessary for the activation of the target genes of Pho4p. There is a dimerization precedent in hnRNP protein Npl3p, where nuclear Hmt1p-mediated arginine methylation increases selfdimerization by ϳ2-fold (52,104). The dimerization domain of Npl3p is in the C-terminal SR-domain, from positions 276 -364, and loss of Npl3p dimerization results in defects in translation (105). We hypothesize that the function of the putative arginine methylation site on Pho4p could be like Npl3p, potentially increasing its homodimerization. If this is the case Pho4p could still dimerize in the absence of arginine methylation, albeit with lower affinity. This model could explain the significant reduction, but not complete loss of expression, of Pho4p target genes and proteins in hmt1⌬.
A second model for the putative Arg-241 methylation in Pho4p involves the interplay between phosphorylation and methylation. Interplay between modifications is emerging as a widespread means for the regulation of protein function, including in protein interaction codes (106). Pho4p is phosphorylated at eight different serine residues, five of which are phosphorylated by the Pho80p-Pho85p cyclin-CDK complex under high-phosphate conditions (92,93). Four of these five phosphosites are well characterized in their respective roles: phosphorylation of Ser-114 and -128 signals nuclear export; phosphorylation of Ser-152 decreases nuclear import; and phosphorylation of Ser-223 decreases the binding affinity for Pho2p. The remaining three phosphoserines at positions 204, 242, and 243 are not phosphorylated by Pho85p, and their kinase(s) and molecular functions are unknown (107)(108)(109). It is noteworthy that the methylarginine site on Pho4p, reported in this study, is directly adjacent to the phosphoserines at positions 242 and 243 (Fig. 7) (109). Given its proximity to the DNA-binding domain, arginine methylation of Pho4p could potentially serve as a recruitment signal for transcriptional machinery, or to recruit kinases for the phosphorylation of Ser-242 and -243. A similar example involves the methylation of transcription factor STAT6 at Arg-27 (110), where the loss of arginine methylation leads to a decrease in IL4-dependent phosphorylation. Alternatively, given that all of the characterized phosphoserines in Pho4p inhibit its transcrip-tional activity (93), Pho4p methylation could serve to block this inhibition. In human cells, the methylation of Arg-296 and -299 by PRMT1 inhibited the phosphorylation of Ser-302 on hnRNP K, involved in chromatin remodelling, transcription, RNA splicing, mRNA stability and translation (111). Loss of methylation on these arginine residues led to an increase of p53-independent apoptosis upon DNA damage.
Post-translational modification of proteins, or combinations thereof, are important ways by which nutrient sensing can be controlled. Examples include the PHO pathway itself and, in mammalian cells, the interplay between O-linked ␤-N-acetyl glucosamine (O-GlcNAc) and phosphorylation (112). Our research has highlighted a possible new link between phosphate regulation and SAM levels, where SAM is the major methyl donor inside the cell. SAM is synthesized alongside phosphate and diphosphate from methionine and ATP by the S-adenosylmethionine synthetases Sam1p and Sam2p. Interestingly, the intracellular concentrations of SAM and phosphate are known to be correlated. The accumulation of SAM in the adenosine kinase mutant ado1⌬ was accompanied by an upregulation in phosphate related transcripts and increased cellular concentrations of phosphate and polyP (113). Furthermore, polyP contributes to stability of SAM (113). The downregulation of the PHO regulon upon deletion of HMT1, revealed in our study, suggests an interdependency between the regulation of SAM and phosphate levels. Simplistically, this may be a means by which the cell does not expend resources on phosphate sequestration if there is not sufficient SAM to undertake a large range of other metabolic activities. This donor sensing and regulation of protein activity is like the dependence of acetyl-CoA donor levels on the activity of lysine acetyltransferases (KATs), which acetylate enzymes that regulate various metabolic processes in the cell (114). In conclusion, this study has shown that loss of Hmt1p-mediated arginine methylation leads to the dysregulation of phosphate homeostasis. Although the monomethylation of Pho4p at Arg-241 may play a role in this process, this was not validated in vivo and thus the exact molecular mechanisms underlying dysregulation remain to be confirmed. Acknowledgments-We thank the Ramaciotti Centre for Genomics and the Bioanalytical Mass Spectrometry Facility, both at the University of New South Wales, for technical support in gene expression analysis and proteomics, respectively. We acknowledge Erin O'Shea (Harvard University) for the gift of the Pho4p-GFP strain.
DATA AVAILABILITY Data from SILAC experiments has been submitted to the ProteomeXchange Consortium (http://proteomecentral. proteomexchange.org) via the PRIDE partner repository (57) with the data set identifier PXD004054. MaxQuant output for SILAC experiments are available as supplemental Data S1 and S2. Annotated MS/MS spectra can be visualised by using MS-Viewer (http://msviewer.ucsf.edu/prospector/cgi-bin/msform. cgi?formϭmsviewer) with the search keys: 1olc8v6ekm (dataset A) and 0rgyw6hg1i (dataset B). RNA-Seq reads have been deposited in the NCBI sequence read archive (SRA) under the accession number SRP072252. Microarray data have been deposited in the NCBI Gene Expression Omibus under the accession number GSE99869 (58). * MRW acknowledges financial support from the Australian Research Council (DP130100349 and DP170100108), the Australian Government NCRIS scheme, the New South Wales State Government RAAP scheme and the University of New South Wales. GHS acknowledges financial support from the Australian Research Council (DE150100019) and GHS and MAE acknowledge support from the University of New South Wales. SZC and DY acknowledge the support of Australian Postgraduate Awards.