Endoplasmic reticulum–mitochondria junction is required for iron homeostasis

The endoplasmic reticulum (ER)–mitochondria encounter structure (ERMES) is a protein complex that physically tethers the two organelles to each other and creates the physical basis for communication between them. ERMES functions in lipid exchange between the ER and mitochondria, protein import into mitochondria, and maintenance of mitochondrial morphology and genome. Here, we report that ERMES is also required for iron homeostasis. Loss of ERMES components activates an Aft1-dependent iron deficiency response even in iron-replete conditions, leading to accumulation of excess iron inside the cell. This function is independent of known ERMES roles in calcium regulation, phospholipid biosynthesis, or effects on mitochondrial morphology. A mutation in the vacuolar protein sorting 13 (VPS13) gene that rescues the glycolytic phenotype of ERMES mutants suppresses the iron deficiency response and iron accumulation. Our findings reveal that proper communication between the ER and mitochondria is required for appropriate maintenance of cellular iron levels.

; overexpression of Mdm10 complementing proteins 1 and 2 (20), which are mitochondrial proteins of unknown function involved in lipid homeostasis; and deletion of Ups2p, a mitochondrial intermembrane space protein that is involved in phospholipid metabolism (21). Among these, gain-of-function mutations in VPS13 arise naturally with high frequency and suppress the phenotypic consequences of ERMES deletion but fail to restore the ER-mitochondria junctions (18). The VPS13 family proteins are highly conversed, and mutations in human VPS13 orthologues are associated with Cohen syndrome, a genetic disorder that affects motor skills, mental development, and chorea-acanthocytosis, a neurodegenerative disorder (22)(23)(24). Orthologues of the defining components of the yeast ERMES complex have not been found in metazoans, suggesting that the functions of ERMES may be achieved through other mechanisms in higher eukaryotes.
Iron is an essential element for almost all living organisms and participates in a wide variety of biological processes as a critical cofactor for numerous enzymes and proteins. However, iron can also be toxic due to its ability to generate reactive oxygen species in aerobic conditions. Cells therefore maintain an iron quota that is determined by its utility versus toxicity. In yeast, cells acquire iron through non-reductive siderophoreiron transporters as well as reductive uptake systems that includes a family of ferric reductases and low-and high-affinity iron transporters (25). In iron-limited conditions, the major iron-dependent transcription factor activator of ferrous transport protein 1 (Aft1p) shuttles and accumulates in the nucleus to activate a transcriptional program for both iron uptake and acclimation to iron-limited metabolism (26 -28). The mRNA-binding protein cysteine-three histidine 2 (Cth2p) is also activated to post-transcriptionally regulate many genes related to iron homeostasis through mediating RNA degradation, which is functionally similar to apoaconitase iron regulatory proteins 1 and 2 in vertebrate cells (29,30). Interestingly, disruption of iron-sulfur (Fe-S) cluster assembly, which occurs in the mitochondrial matrix, triggers the iron deficiency response. For instance, cells lacking yeast frataxin homolog 1 (Yfh1p), a mitochondrial component of the ironsulfur cluster assembly pathway, activate the Aft1-dependent transcriptional program and accumulate excess iron in the mitochondria (31).
Considering that iron-sulfur cluster biosynthesis occurs in mitochondria and impaired mitochondrial iron-sulfur biosynthesis activates the iron deficiency response, we investigated whether iron homeostasis is perturbed in ERMES mutants. Here, we show that the loss of ER-mitochondria junctions induces an Aft1p-dependent iron deficiency response, leading to iron accumulation in the cell and mitochondria. Genetic disruption of the iron regulatory system exacerbated the respiration defect by ERMES deficiency. Furthermore, a dominant mutation in Vps13p suppressed the iron deficiency response and the iron accumulation phenotype of the ERMES mutants. Our findings indicate that ER-mitochondrial junctions are required for proper iron usage by the cell, expanding the functional repertoire of this important protein complex.

Loss of ERMES function induces the iron deficiency response
We first performed genome-wide gene expression analysis using mRNA sequencing in the wild-type (WT) and ERMES subunit deletion mmm1⌬ and mdm34⌬ strains. To determine whether loss of ERMES induces the iron deficiency response, we examined the genes that comprise the iron regulon including the master transcription factors Aft1p and Aft2p, siderophore-iron transporters Aft1 regulon 1 and 2 and FITs, ferric reductases, and high-affinity iron transporters. The iron regulon genes are up-regulated in mmm1⌬ and mdm34⌬ strains compared with WT, which is similar to when WT cells are grown in iron-poor medium through treatment of cells with bathophenanthroline disulfonate (BPS), a strong iron chelator (Fig. 1B). Analysis of all the iron homeostasis genes annotated in the Saccharomyces Genome Database (n ϭ 28 genes) revealed that the median expression level of iron homeostasis genes is significantly higher in mmm1⌬ and mdm34⌬ strains as well as in BPS-treated WT cell compared with WT cells grown in rich medium ( Fig. 1, C and D). Gene ontology (GO) analysis of the up-regulated genes in both mmm1⌬ and mdm34⌬ strains confirmed significant enrichment for iron transport-related genes (Fig. 1E), which is similar to the enrichment of GO terms of the induced genes with BPS treatment (data not shown).
The transcriptional response to iron deficiency also involves activation of genes such as CTH2, which is involved in mediating the metabolic acclimation to iron deficiency by sparing iron use, and genes involved in iron-independent pathways such as glutamate dehydrogenase GDH3 (30,32,33). These genes are up-regulated in both mmm1⌬ and mdm34⌬ strains as well as in BPS-treated WT cells (Fig. 1F). The iron deficiency response also involves the repression of iron-dependent pathways including especially abundant iron-sulfur-containing proteins such as glutamate synthase GLT1, isopropylmalate isomerase LEU1, biotin synthase BIO2, and electron transport chain genes such as succinate dehydrogenases to spare iron. These genes are also down-regulated in mmm1⌬ and mdm34⌬ strains as well as in BPS-treated WT cells (Fig. 1F). GO analysis of the co-down-regulated genes in mmm1⌬ and mdm34⌬ strains revealed enrichment in electron transport chain genes (Fig.  1G), which is similar to WT cells grown in iron-depleted condition, but also significant enrichment in genes that function in protein translation (Fig. 1G).
To determine whether the iron deficiency response in ERMES mutants causes iron accumulation, we collected cells in the log phase of growth and divided the same batch of cells to measure the total cellular and mitochondrial iron levels in WT, mmm1⌬, mdm34⌬, and yfh1⌬ as a positive control. As shown in Fig. 1H, iron is significantly accumulated in the mmm1⌬ and mdm34⌬ with ϳ3-4-fold more intracellular iron than WT. The yfh1⌬ strain also contained more intracellular iron as expected. However, unlike the yfh1⌬ strain that preferentially accumulates iron in the mitochondria, we did not find preferential accumulation of iron in mitochondria isolated from mmm1⌬ and mdm34⌬ (Fig. 1I). Thus, iron accumulates equally in the cell and the mitochondria of ERMES mutants, distinguishing them from the yfh1⌬ strain.

Impairment of ERMES disrupts iron homeostasis
Iron accumulation in the ERMES mutants is dependent on the high-affinity iron uptake system because deleting FET3, which encodes an oxidoreductase required for high-affinity iron uptake, in mmm1⌬ (mmm1⌬fet3⌬) prevented the accumulation of iron ( Fig. 2A). Interestingly, whereas mmm1⌬ has a longer doubling time compared with WT, inhibition of iron uptake in mmm1⌬ by deletion of FET3 (mmm1⌬fet3⌬) exacerbates the growth defect (Fig. 2B). This suggests that ERMES mutants induce the iron deficiency response likely due to an inability to fully utilize intracellular iron and depend on the extra iron to meet their demands. Taken together, our transcriptome and iron measurement analyses indicate that when cells lose the junctions between ER and mitochondria the iron deficiency response is induced, causing intracellular iron overload.

ERMES and the iron regulon cooperate to ensure optimal cellular respiration
The ERMES mutants have defects in mtDNA maintenance and cannot grow in non-fermentative conditions (6 -9).
Because iron deficiency causes diminished respiration (34), we next investigated whether the role of ERMES in maintaining iron homeostasis complements the iron regulon to ensure proper cellular respiration. We replaced the promoter of Mitochondrial iron (nmol/mg of protein) * *** * *** P < 0.001 * P < 0.05

Impairment of ERMES disrupts iron homeostasis
MMM1 with the GAL1 promoter (pGAL-MMM1) to enable inducible MMM1 expression and analysis of genetic interactions. Growth in the presence of glucose, which represses the expression of MMM1, resulted in a slight growth defect. As expected, pGAL-MMM1 cells showed a respiratory defect when grown on non-fermentable carbon sources such as ethanol and glycerol (YPEG), which repress MMM1 expression (Fig. 3A). Longer incubation on YPEG (3 days) allows for growth of pGAL-MMM1 cells likely due to leaky expression of MMM1. Importantly, deletion of ATX1, a cytosolic copper chaperone that is required for effective iron uptake (35,36), causes a moderate respiratory defect. However, the respiratory defect is much more severe when atx1⌬ is combined with MMM1 shutoff (Fig. 3A). Thus, the impairments of ERMES and iron uptake exhibit a synthetic growth defect on non-fermentable carbon sources. This is even the case when iron distribution inside the cell is disrupted. When we combined shutoff of MMM1 with constitutive overexpression of a mitochondrial iron transporter, mitochondrial RNA splicing 3 (Mrs3p), which significantly increased cellular and mitochondrial iron content (supplemental Fig. S1A), or the vacuolar iron importer Ccc1p by replacing their promoters with the high-expressing glyceraldehyde-3-phosphate dehydrogenase isozyme 3 (TDH3) promoter (37), the respiratory defect of the double mutants was much more severe than MMM1 shutoff alone (Fig. 3B). However, deletions of MRS3, MRS4, or both mitochondrial iron transporters had no effect on the respiratory deficiency of the MMM1 shutoff strain (supplemental Fig. S1B). Altogether, these findings further confirm that ERMES and a working iron uptake and distribution system function complementarily to ensure optimal respiratory growth. We next reasoned that further increases in intracellular iron levels may rescue the respiratory defect of the MMM1 shutoff strain. We therefore engineered a strain harboring a constitutively active Aft1 (Aft1-1 up ) by introducing a previously described mutation, aft1-C291F (26). Surprisingly, strains carrying the Aft1-1 up allele also exacerbated the respiration defect in MMM1 shutoff (Fig. 3C). Altogether, these data suggest that disruption of ERMES causes a defect in intracellular iron usage that extends beyond simply the levels of iron. Too little or too much iron, both of which can be problematic (see "Discussion"), have adverse effects when the function of ERMES is impaired.

A dominant mutation in VPS13 partially suppresses the iron deficiency response and iron accumulation of ERMES mutants
Dominant mutations in VPS13 suppress the phenotypic consequences of ERMES deficiency including respiration defects, although the mechanism of suppression is unknown (18). To determine whether a VPS13 dominant mutation (vps13-D716H) can also suppress the iron deficiency response and iron accumulation in ERMES mutants, we deleted the MMM1 gene in a vps13-D716H strain and determined gene expression and intracellular iron levels. As shown in Fig. 4A, the vps13-D716H point mutation suppressed the growth defect of mmm1⌬ on both YPD (glycolytic) and YPEG (respiratory) plates. Global gene expression analysis in medium with glucose revealed that the median expression of iron regulon genes, which were upregulated in mmm1⌬, was lower in vps13-D716H mmm1⌬ double mutants (Fig. 4B). However, the vps13-D716H mmm1⌬ does not fully restore the expression of certain iron uptake genes such as FIT2 and FIT3 to levels observed in WT or vps13-D716H itself (Fig. 4C). More interestingly, the genes required for respiration have higher expression in vps133-D716H mmm1⌬ compared with the WT or vps13-D716H strain alone even in the presence of glucose (i.e. YPD) (Fig. 4D). Consistent with the gene expression changes, intracellular and mitochondrial iron levels were decreased in vps13-D716H mmm1⌬ compared with mmm1⌬ alone (Fig. 4, E and F). These data indicate that a VPS13 dominant mutation significantly suppresses the iron deficiency response and alleviates the iron overload phenotype of a mutation in ERMES.

Requirement of ERMES for iron homeostasis may be related to its mitochondrial protein import but not other known functions
To determine which known functions of ERMES may be related to the iron deficiency response, we determined whether iron accumulates to similar extents in strains with deletion of genes that function in various ERMES-related roles. Deletions of the phosphatidylcholine biosynthesis genes CHO1 and PSD1 or the cardiolipin biosynthesis gene CRD1, representing the loss of ERMES' phospholipid exchange function, had no effect on intracellular iron levels (Fig. 5A). Loss of mtDNA caused by deletion of a mitochondrial RNA helicase, mitochondrial RNA helicase 4 (MRH4), also did not lead to iron overload (Fig. 5A), and neither did deletion of GTPase EF-hand protein of mitochondria 1 (GEM1), a calcium-binding Rho-like GTPase and potential regulatory component of ERMES complex involved in calcium-dependent mitochondrial movement and inheritance (15,38) (Fig. 5A). In addition, deletion of FIS1, which is required for mitochondrial fission (39), did not increase cellular iron content inside of cells, suggesting that any alteration to mito-

Impairment of ERMES disrupts iron homeostasis
chondrial morphology does not necessarily lead to iron accumulation. However, deletion of the SAM complex subunit SAM37, which impairs mitochondrial protein import similarly to ERMES mutants, led to accumulation of iron to a similar extent as mdm10⌬. Expression analysis in sam37⌬ confirmed that loss of SAM37 also induces the iron deficiency response (Fig. 5, B and C). These data indicate that disruption of phospholipid or calcium exchange or mitochondrial morphology does not necessarily affect intracellular levels, but interruption of mitochondrial protein import can induce the iron deficiency response.
Both ERMES mutants and sam37⌬ induce the iron deficiency response and iron overload, which raises the question of whether SAM or ERMES complex regulates iron homeostasis via the same pathway. Our previous data showed that vps13-D716H can suppress the growth defect and iron deficiency response of ERMES mutants. To determine whether vps13-D716H can also suppress the growth defect and iron deficiency response in sam37⌬, we deleted the SAM37 gene in vps13-D716H strain and determined the effects on growth, gene expression, and intracellular iron levels. As shown in Fig. 5D, the VPS13 dominant mutation did not rescue the growth defect in sam37⌬ on a YPD plate. Genome-wide gene expression analysis and intracellular iron measurement also revealed that the neither the iron deficiency response (Fig. 5, B and C) nor the cellular (Fig. 5E) or mitochondrial iron accumulation (Fig. 5F) in sam37⌬ is suppressed by vps13-D716H. Considering that vps13-D716H suppresses the iron deficiency response and iron accumulation of ERMES mutants, our data suggest that the iron overload phenotype of ERMES mutants and sam37⌬ may be through different cellular mechanisms.

Discussion
Physical contacts between organelles through membrane contact sites establish a route for interorganelle communication that is necessary for homeostatic control of cellular processes. This is particularly imperative for mitochondria because they do not communicate with other organelles via the vesicular trafficking pathway. Mitochondria contain essential pathways for the generation of energy, metabolism of amino acids and phospholipids, and synthesis of iron-sulfur clusters among others. Two protein complexes, ERMES and vCLAMP, physically connect the mitochondria to the ER (3) and vacuole (4), respectively. Together, ERMES and vCLAMP are required for phospholipid transport between mitochondria and the ER (14). ERMES has also been implicated in calcium exchange between the ER and mitochondria, mitochondrial protein import, mitochondrial attachment to actin and segregation into daughter cells during cell division, and maintenance of mitochondrial genome (17). Whether ERMES directly participates in these functions or they are secondary effects due to the specific abnormal mitochondrial morphology in ERMES mutants

Impairment of ERMES disrupts iron homeostasis
remains to be established. Our study now uncovers an additional role for ERMES in regulation of iron homeostasis. Disruption of ERMES function, from either the ER or the mitochondrial surfaces, causes inappropriate induction of the iron deficiency response and accumulation of excess iron inside the cell. This new role of ERMES is independent of its function in phospholipid and calcium exchange, indicating that communication between the ER and mitochondria is also required for proper control of cellular iron levels. Iron is an essential element for numerous biochemical pathways operating in every cellular compartment. Iron needs to be transported to the mitochondria for biosynthesis of iron-sulfur clusters, which are among the most ancient protein cofactors and fulfill many vital functions in fundamental pathways such as respiration, DNA replication, DNA repair, transcription, telomere maintenance, and translation (40). Defects in mitochondrial iron-sulfur cluster biosynthesis induce iron uptake and iron redistribution inside the cell (31) but also cause respiratory deficiency, slow growth, and loss of mtDNA (41), which are phenotypes similar to those described for ERMES mutants. This raises the possibility that defects in mtDNA maintenance and respiration in ERMES mutants may be related to abnormal iron usage in the cell. This notion is supported by the negative genetic interactions observed between MMM1 and essential components of the iron uptake pathway (Fig. 2). As with other functions of ERMES, further studies are needed to determine whether ERMES effects on iron homeostasis are direct or indirect. Considering the fact that a SAM37 deletion mutant accumulated excess iron (Fig. 5A), has condensed giant mitochondria (11), and has a respiration defect (42), it is possible that loss of ERMES affects mitochondrial protein important, which in turn impairs iron-sulfur cluster biosynthesis, inducing the iron deficiency response, slow growth, and loss of mtDNA. Our data indicate that the genetic architecture underlying the function of ERMES in iron homeostasis differs from the SAM complex because a mutation that suppresses the ERMES phenotype has no effect on accumulation of iron in sam37⌬. However, it is still possible that iron overloading is secondary to inhibition of mitochondrial protein uptake because the ERMES mutant and sam37⌬ mutant likely block protein uptake by different mechanisms.
Defects in mitochondrial iron-sulfur biosynthesis but not cytosol iron-sulfur processes induce iron uptake and iron redistribution inside the cell, indicating that the signal for such a response must originate in the mitochondria (27). However, how the mitochondria communicate with the cell to regulate the iron deficiency response is not clear. The fact that ERMES mutant induced the iron deficiency response raises the possibility that cells may regulate iron homeostasis partly through the membrane contact sites between the ER and mitochondria.
We also found that the vps13-D716H mutation suppressed the iron deficiency response in an ERMES mutant. This sug-

Impairment of ERMES disrupts iron homeostasis
gests that the iron regulation function of ERMES can be bypassed without restoring the ER and mitochondria junctions (18), indicating that functional communication is more important than the physical connection between the ER and mitochondria. Orthologues of the ERMES complex in metazoans have not been found, but the VPS13 family is highly conserved between yeast and higher eukaryotes and may compensate for the absence of ERMES. Interestingly, mutations in the human VPS13 orthologues VPS13A, -B, and -C (43) are associated with neurodegenerative disorders (22)(23)(24). Because iron overload is also related to neurodegenerative diseases (44), it will be interesting to investigate whether the VPS13-related human diseases may have defects in iron homeostasis.

Yeast strains and media
Yeast strains used in this work are described in supplemental Table S1. Standard yeast media and manipulations were used. For BPS treatment, YPD was prepared using acid-washed glassware and sterilized by filtration (referred to as clean YPD). Cells in the log phase of growth were treated with 100 M BPS for 4 h in clean YPD.

RNA sequencing
Cells grown in YPD in the log phase (A 600 around 1) were collected. Total RNA was extracted using the hot acid phenol extraction method. The extracted RNA samples were treated with DNase I (Ambion TURBO DNA-free kit) and further purified with TRIzol regent (Ambion). Libraries of mRNA were prepared with Illumina TruSeq RNA sample preparation kit version 2 or KAPA stranded mRNA sample preparation kit. Libraries were sequenced, and reads were aligned using TopHat 2.0.8 with default setting (45). Transcript abundances were normalized to fragments per kilobase of exon per million fragments mapped using Cuffdiff 2.0.2 (46). For log 2 ratio calculations, all transcripts with fragments per kilobase of exon per million fragments mapped lower than 0.1 were replaced with 0.1.

Iron measurement
Cells grown in YPD in the log phase (A 600 around 1) were collected. For total cellular iron measurement, cells corresponding to 50 OD units were washed once with 1 mM EDTA and twice with purified water (Milli-Q), and the cell pellet was used for iron measurement. For mitochondrial iron measurement, mitochondria were purified using a sucrose gradient as described (47). The protein concentration of the purified mitochondria was determined with a BCA assay (Thermo Scientific Pierce BCA Protein Assay kit). The total cell or purified mitochondrial pellet, after a centrifugation step to compact the pellet, was overlaid with 286 l of 70% nitric acid and digested at room temperature for 24 h and at 65°C for an additional 2 h before being diluted to a final nitric acid concentration of 2% (v/v) with purified water (Milli-Q). 1:10 dilutions of the growth medium and water were treated with nitric acid to a final concentration of 2% (v/v) and measured directly. Metal contents were determined by inductively coupled plasma mass spectrometry on an Agilent 8800 triple quadrupole ICP-MS/MS instrument, in comparison with an environmental calibration standard, using 89 Y as an internal standard. The levels of 56 Fe were determined in MS/MS mode using H 2 as a cell gas. The average of three to six biological replicates and five technical replicate measurements was used for each individual strain. For a given biological replicate, the same batch of medium was used to grow all WT and mutant strains to minimize batch effect. The variation between the technical replicate measurements never exceeded 5% for an individual sample. The iron content was normalized either to OD or mitochondrial protein level.

Spot test
Cells were first inoculated in YPD and grown overnight, diluted to an A 600 of 0.3 in YPD, and grown for 4 -5 h prior to collection. 5-Fold serial dilutions of WT and mutants were spotted onto either a YPD (2% glucose) or YPEG (2% ethanol and 3% glycerol) plate and incubated at 30°C.

Doubling time determination
Cells were inoculated in clean YPD overnight and diluted to an A 600 of 0.2 in clean YPD. A 600 was measured at regular intervals, and doubling time was determined with GraphPad Prism 5.0 software.