Catalytic residues, substrate specificity, and role in carbon starvation of the 2-hydroxy FA dioxygenase Mpo1 in yeast

C26:0 VLCFA, most of which is 2-hydrox-ylated (1, 15, 16). In yeast, Scs7 (a homolog of FA2H) is Abstract The yeast protein Mpo1 belongs to a protein family that is widely conserved in bacteria, fungi, protozoa, and plants, and is the only protein of this family whose function has so far been elucidated. Mpo1 is an Fe 2+ -dependent dioxygenase that catalyzes the  -oxidation reaction of 2-hydroxy (2-OH) long-chain FAs (LCFAs) produced in the degradation pathway of the long-chain base phytosphingosine. However, several biochemical characteristics of Mpo1, such as its catalytic residues, membrane topology, and substrate specificity, remain unclear. Here, we report that yeast Mpo1 contains two transmembrane domains and that both its N- and C-ter-minal regions are exposed to the cytosol. Mutational analyses revealed that three histidine residues conserved in the Mpo1 family are especially important for Mpo1 activity, suggesting that they may be responsible for the formation of coordinate bonds with Fe 2+ . We found that, in addition to activity toward 2-OH LCFAs, Mpo1 also exhibits activity toward 2-OH very-long-chain FAs derived from the FA moiety of sphingolipids. These results indicate that Mpo1 is involved in the metabolism of long-chain to very-long-chain 2-OH FAs produced in different pathways. We noted that the growth of mpo1  cells is delayed upon carbon deprivation, suggesting that the Mpo1-mediated conversion of 2-OH FAs to nonhydroxy FAs is important for utilizing 2-OH FAs as a carbon source under carbon starvation.

In addition to the direct oxidation of FAs, 2-OH FA is also generated in the degradation pathway of the LCB phytosphingosine (PHS) (17,18). PHS is present in the epidermis, small intestine, and kidney in mammals (19)(20)(21) and is a major LCB in yeast (15,16). LCBs commonly have hydroxyl groups at C1 and C3 and an amino group at C2. The simplest LCB is dihydrosphingosine. The most abundant LCB in mammals is sphingosine, which has a trans double bond between C4 and C5. PHS has a hydroxyl group at C4. LCBs undergo three common reactions in their degradation pathway: phosphorylation at C1, cleavage between the C2 and C3 positions, and oxidation (5,17,18,22,23). Through these reactions LCBs are converted to two-carbon shortened FAs, and PHS is converted to 2-OH FAs (17,18). Because 2-OH FAs cannot be used for glycerolipid synthesis or -oxidation, they need to be converted to nonhydroxy FAs. This conversion is achieved by oxidation, where 2-OH FAs are metabolized to nonhydroxy FAs with one less carbon (making them odd-chain nonhydroxy FAs) (17,18). The -oxidation reactions differ between mammals and yeast. Mammalian -oxidation involves three steps (CoA addition, long-chain aldehyde production by cleavage between the C1 and C2 positions, and oxidation to FAs), whereas yeast -oxidation is conducted in one step (17,18,24). Accordingly, the proteins involved in -oxidation differ between mammals and yeast. We previously identified MPO1 as the gene functioning in the degradation pathway of PHS in yeast (17). It subsequently became clear that Mpo1 is involved in the oxidation of 2-OH FAs as an Fe 2+ -dependent dioxygenase (EC 1.14.18.12), which directly converts 2-OH LCFAs to nonhydroxy FAs with one less carbon (24).
Mpo1 homologs are present in bacteria, fungi, protozoa, and plants, but not in animals. Of the members of Mpo1 family [previously known as domain of unknown function number 962 (DUF962)], Mpo1 is the only protein that has been analyzed. However, there remain many unsolved problems regarding the structural and enzymatic characterization of Mpo1. For instance, it is an ER protein that is predicted to have some transmembrane domains, but the exact number of these domains and its membrane topology are unknown; the Mpo1-catalyzed -oxidation reaction requires Fe 2+ , but the region and amino acid residues involved in Fe 2+ binding are unclear; and Mpo1 exhibits activity toward 2-OH LCFAs derived from PHS (24), but it is unclear whether Mpo1 is involved in -oxidation of 2-OH VLCFAs derived from the FA portion of sphingolipids. Although the 2-OH C26 VLCFA is abundant in yeast sphingolipids, its degradation process is unclear, including whether it is subject to -oxidation by Mpo1. Furthermore, the physiological significance of 2-OH FA degradation is also unknown. In the present study, we revealed the following unknown aspects of Mpo1: membrane topology, the residues important for its activity, its substrate specificity, and its role in the carbon starvation response. Our findings give clues to the elucidation of the functions and activities not only of Mpo1, but also of other members of the Mpo1 family.

Yeast strains and media
The budding yeast Saccharomyces cerevisiae was used in this study. The strains we used were BY4741 (MATa his31 leu20 met150 ura30) and its derivatives 4378 (mpo1::KanMX4), 261 (pxp1:: KanMX), and KSKY4 (BY4741, pxp1::KanMX mpo1::NatNT2) (25,26). KSKY4 cells were constructed by introducing mpo1::NatNT2 into strain 261. Cells without plasmid were grown in synthetic complete (SC) medium (2% d-glucose, 0.5% casamino acids, and 0.67% yeast nitrogen base without amino acids) supplemented with 20 mg/l adenine, 20 mg/l uracil, and 20 mg/l tryptophan. Cells harboring the pAKNF316 (URA3 marker) plasmid or its derivatives were grown in the above medium without uracil (SC medium supplemented with 20 mg/l adenine and 20 mg/l tryptophan). Cells bearing the plasmid encoding MPO1 fused with the enhanced GFP (EGFP) gene, under control of the MET25 promoter, were induced by removing methionine from the medium. For this purpose, medium lacking uracil and methionine (2% d-glucose, amino acid supplements without methionine, 20 mg/l adenine, and 0.67% yeast nitrogen base) was used. The following media were used to examine growth under carbon or nitrogen starvation conditions. SC medium, SC medium without a carbon source (SCC; SC medium without glucose), and synthetic dextrose medium without a nitrogen source (SDN; 2% d-glucose and 0.17% yeast nitrogen base without amino acids or ammonium sulfate). All yeasts were cultured at 30°C with rotation.

Plasmids
The pAKNF316 vector (CEN) is designed to express a protein tagged with 3×FLAG at the N terminus under the control of the glycerol-3-phosphate dehydrogenase gene (TDH3) promoter (27). The pNK46 plasmid encoding 3×FLAG-MPO1 is a derivative of the pAKNF316 vector (18) pKSK35 (3×FLAG-MPO1 H115Y), and pKSK36 (3×FLAG-MPO1 E119D) plasmids were constructed using a QuikChange sitedirected mutagenesis kit (Agilent Technologies, Santa Clara, CA) with the pNK46 plasmid and appropriate primers. The pMT66 (N terminus), pMT67 (Lys47), pMT72 (Leu69), pMT68 (Val93), pMT69 (Arg121), and pMT70 (C terminus) plasmids encode 3×FLAG-MPO1 fused with N-glycosylation reporter at the indicated positions. These are derivatives of the pNK46 plasmid. For their construction, the BamHI site was first introduced into each position using a QuikChange site-directed mutagenesis kit. The resulting plasmids were digested with BamHI and ligated with the N-glycosylation reporter fragment, which had been prepared by digesting the pMT50 plasmid encoding amino acid residues 40-221 of the invertase Suc2, using BamHI.

Protein deglycosylation
Deglycosylation of Mpo1 carrying the N-glycosylation reporter was performed using endoglycosidase H (New England Biolabs, Beverly, MA), according to the manufacturer's instructions.

PHS labeling assay
The PHS labeling assay using [11, H]PHS was performed as described previously (17).

In vitro 2-OH FA -oxidation assays
The membrane fraction of the yeast cells was prepared by ultracentrifugation after the cells had been crushed with zirconia beads, as described previously (24). A proteoliposome containing 3×FLAG-Mpo1 was prepared using the ProteoLiposome Expression Kit (CellFree Sciences, Ehime, Japan), essentially as described previously (24). Briefly, 3×FLAG-Mpo1 was translated by the cell-free system using wheat germ lysate from the mRNA transcribed from the pZKN29 plasmid (24) in the presence of liposomes. The pEU-E01-T1R1 plasmid included in the kit was used as a control. Liposomes were prepared as follows. After the reactions, the lipids were extracted by successive addition and mixing with 3.75 vol of chloroform/methanol (1:2, v/v), 1.25 vol of chloroform, and 1.25 vol of water. After centrifugation (9,000 g, room temperature, 1 min), the organic phase was recovered and dried. The reaction product FAs were derivatized to N-(4-aminomethylphenyl)pyridine (AMPP) using an AMP+ MS kit (Cayman Chemical) according to the manufacturer's instructions and quantified by LC/MS/MS as described below.

LC/MS/MS analyses
To quantify PCs, the lipids were extracted from the yeast and purified by TLC as described previously (24). During lipid extraction, 1 nmol C12:0/C12:0 PC (Avanti Polar Lipids) was added as an internal standard. To quantify 2-OH ceramides, the lipids were extracted from the yeast by incubating cell pellets containing 1. The PCs and ceramides extracted from yeast cells and the FAs derivatized to AMPP after the in vitro assay were resolved by UPLC on a reverse-phase column (Acquity UPLC CSH C18 column; particle size, 1.7 m; column size, 2.1 × 100 mm; Waters, Milford, MA) at 55°C and detected using an electrospray ionization tandem triple quadrupole spectrometer (Xevo TQ-S; Waters) in MRM mode. LC conditions for the PCs and ceramides were as follows. The flow rate was 0.4 ml/min in the binary gradient system using a mobile phase A [acetonitrile/water (3:2, v/v) containing 10 mM ammonium formate] and a mobile phase B 2), with the collision energy set at 35 eV. The Q1 and Q3 values and collision energies for PCs and FAs derivatized to AMPP were as previously described (24). Data were analyzed and quantified using MassLynx software (Waters).

Mpo1 is a membrane protein with two transmembrane domains and N and C termini exposed to the cytosol
Although we had previously revealed that Mpo1 is an ER protein (17), the number and regions of its transmembrane domains and its membrane topology remained unknown. Our hydrophobicity plot revealed that Mpo1 has two highly hydrophobic regions (H2 and H3) and three moderately hydrophobic regions (H1, H4, and H5) (Fig.  1A). To determine its membrane topology, we performed an N-glycosylation reporter assay. In this assay, an N-glycosylation cassette containing multiple N-glycosylation sites is inserted into certain regions of the target protein. Because N-linked sugar chains are added to proteins only on the luminal side of the ER, the orientation (cytosolic or luminal side of the ER) of the region into which the cassette was inserted can be determined by examining the N-glycosylation status of the cassette (29,30). We inserted an N-glycosylation cassette into the middle of each hydrophilic region that was flanked by two hydrophobic regions (after Lys47, Leu69, Val93, or Arg121), as well as at the N terminus and the C terminus of Mpo1 (Fig. 1A). These inserts were expressed as 3×FLAG-tagged proteins in the mpo1 cells and detected by immunoblotting. Of the six proteins with an N-linked glycosylation cassette insert, only the one with the cassette inserted after Leu69 exhibited slower mobility on SDS-PAGE than expected for its molecular weight (Fig.  1B). This mobility shift was canceled upon treatment with endoglycosidase H, which removes the N-linked sugar chain. These results indicate that a hydrophilic region containing Leu69 is present in the ER lumen. The fact that the other inserts were not N-glycosylated suggests that they and their adjacent hydrophilic regions are oriented to the cytosolic side of the ER. From these results, we concluded that Mpo1 contains two transmembrane domains and that both its N and its C termini face the cytosol (Fig. 1C).

Three histidine residues are important for the activity of Mpo1
Mpo1 is an Fe 2+ -dependent dioxygenase (24), but the amino acid residues important for its activity, and in particular those involved in the binding of Fe 2+ , had remained unclear. To identify these, the amino acid sequences of 18 bacterial, fungal, protozoan, and plant Mpo1 family members were aligned. Six amino acid residues (Tyr15, His19, His28, His115, Glu119, and Pro123) were highly conserved ( Fig. 2A). To examine their importance for Mpo1 activity, each amino acid residue was substituted to Ala, and the resulting mutants (Y15A, H19A, H28A, H115A, E119A, and P123A) were expressed in the mpo1 cells. Although the expression levels of the Y15A mutant were slightly lower than those of the WT protein, the other mutants were detected at almost the same levels as the WT (Fig. 2B).
Next, we performed [ 3 H]PHS labeling experiments. PHS is metabolized either to sphingolipids or glycerophospholipids ( Fig. 2C) (17). In the latter, which is a degradation pathway, the PHS metabolite 2-OH FAs are converted to nonhydroxy odd-chain FAs by -oxidation, followed by incorporation into glycerophospholipids. When the mpo1 cells harboring the vector were labeled with [ 3 H]PHS, PHS was metabolized only to sphingolipids (Fig. 2D, blue) and not to glycerophospholipids (Fig. 2D, green). Expression of WT Mpo1 in mpo1 cells restored this impaired PHSto-glycerophospholipid metabolism. PHS was also not metabolized to glycerophospholipids in the mpo1 cells expressing the H19A, H28A, or H115A mutant, but it was in those expressing the Y15A, E119A, or P123A mutant. These results indicate that three histidine residues, His19, His28, and His115, are important for Mpo1 function. In several Fe 2+ -binding proteins, histidine residues are involved in the binding action (31)(32)(33)(34)(35), implying that His19, His28, and His115 are responsible for Fe 2+ binding.
In the yeast S. cerevisiae, C18 PHS is the most abundant PHS species, followed by C20 PHS (30,36). These are metabolized to C15:0 FA (C15:0-COOH) and C17:0-COOH, respectively, via the degradation pathway involving Mpo1. A portion of the C15:0-COOH thus generated is converted to C17:0-COOH via FA elongation, and some of both of these are converted to C15:1-COOH and C17:1-COOH, respectively, by desaturation (24). These odd-chain LCFAs are incorporated into glycerophospholipids after conversion to acyl-CoAs (17). To quantitatively evaluate the activity of each Mpo1 mutant, the amounts of PC, the most abundant glycerophospholipid class, were measured by LC/MS/MS. The measurement results for five of the most abundant of the PC species containing odd-chain LCFAs that are products of Mpo1 (or their derivatives) are shown in Fig. 3A: C16:1-C15:0, C16:1-C17:1, C18:1-C15:0, C18:1-C15:1, and C16:1-C17:0. The quantities of these odd-chain PC species in mpo1 cells harboring vector were much lower than in those expressing WT Mpo1, but were similar to those in the H19A-or H115A-expressing cells. The levels of odd-chain PCs in H28A-expressing cells were lower than in WT-expressing cells, but higher than in cells harboring vector. The quantities of odd-chain PCs in the mpo1 cells expressing the Y15A, E119A, or P123A mutant were similar to those of the WT Mpo1-expressing cells.
Next, the quantities of odd-chain PCs were measured in the presence of 10 M 2-OH C16:0-COOH in the medium; that is, they were measured under conditions in which the substrate of Mpo1 was abundant. Upon addition of 2-OH C16:0-COOH, odd-chain PC levels in mpo1 cells expressing WT Mpo1 increased approximately 11fold (compare Fig. 3B with Fig. 3A: the total percentage of odd-chain PCs is 9.8% in the former and 0.9% in the latter). However, the increase was very slight in cells harboring vector or expressing the H19A, H28A, or H115A mutants, and although the increase was greater in those expressing the Y15A or E119A mutants, it was still smaller than that in the WT-expressing cells (Fig. 3B). The difference from the WT-expressing cells was smallest in the cells expressing the E119A mutant. These results indicate that the activity of all of these mutants (H19A, H28A, H115A, Y15A, and E119A) was depressed. Although there was no observable effect of the mutation on the activity of cells expressing the Y15A or E119A mutant at low concentrations of the substrate (physiological conditions), the effects became apparent under an increased substrate concentration (addition of 2-OH C16:0-COOH to the medium). Together with the results shown in  (31,33). To investigate the functional relationship between the Mpo1 motif and those of class I diiron-oxo proteins, the tyrosine (Tyr15) and glutamate (Glu119) residues of Mpo1 were replaced with the corresponding amino acid residues from the latter (histidine and aspartate, respectively). The resulting Y15H and E119D mutants were expressed in mpo1 cells, and their activity was examined by [ 3 H]PHS labeling. The activity and expression levels of the two mutants were comparable to the WT protein (Fig. 4A, B). Tyrosine can potentially form a coordinate bond with Fe 2+ , as can histidine. Both glutamate and aspartate are acidic amino acids. We speculated that these similarities would make it possible to replace the amino acid residues in the Mpo1 motif with the corresponding amino acid residues from the class I diironoxo proteins.
We also created mutants in which each of the histidine residues in the Mpo1 motif was replaced with a tyrosine residue (H19Y, H28Y, and H115Y). None of these mutants exhibited any activity (Fig. 4B), similar to the alanine mutants (H19A, H28A, and H115A; Fig. 2D). Thus, tyrosine residues cannot substitute for the function of the histidine residues in Mpo1, probably due to the differences in the orientation and/or position of the coordinate bond ligands.
To exclude the possibility that the cause of the impaired PHS/2-OH FA metabolism in the Mpo1 mutant-expressing cells was due to mislocalization, the intracellular localization of each alanine mutant was examined using EGFPfusion. All Mpo1 mutants exhibited a double ring structure characteristic of yeast ER, which is composed of perinuclear ER and cortical ER, as does the WT Mpo1 (Fig. 5). Therefore, none of the mutations affected the intracellular localization of Mpo1.
Next, to investigate the possibility that the cause of the impaired PHS/2-OH FA metabolism in the Mpo1 mutantexpressing cells was decreased Mpo1 activity, the enzyme activity of each alanine mutant was examined by in vitro assay. The membrane fraction of the mpo1 cells expressing each Mpo1 mutant was prepared and incubated with 1 mM 2-OH C16:0-COOH in the presence of 1 mM Fe 2+ , and the amount of C15:0-COOH produced was determined by LC/MS/MS analysis. The -oxidation activity in cells expressing WT Mpo1 was approximately five and one-half times higher than that in cells harboring vector (Fig. 6A). Those expressing the Y15A, H19A, H28A, or H115A mutants exhibited very low levels of activity, similar to the cells harboring vector. The activity in cells expressing the E119A mutant was only slightly lower than that in the WT cells, and in cells expressing the P123A mutant, it was comparable to that of the WT cells.
WT Mpo1 was active not only toward 2-OH C16:0-COOH but also toward 2-OH C14:0-COOH, producing C13:0-COOH (Fig. 6B). When the -oxidation activity of Mpo1 toward 2-OH C14:0-COOH was measured using different concentrations of the Fe 2+ , the activity reached a plateau at an Fe 2+ concentration of 20 M. The K D value of Mpo1 for Fe 2+ was calculated to be 9.7 M. However, no activity was observed for the Y15A, H19A, H28A, or H115A mutants, even when the Fe 2+ concentration was increased to 1 mM (Fig. 6C), suggesting that the affinity of these mutants for Fe 2+ was much lower and that their K D values were above 1 mM.

Mpo1 is active toward 2-OH VLCFAs
In yeast, 2-OH FAs are generated not only via the PHS degradation pathway but also through the oxidation of the C2 position of FAs in ceramide by the sphingolipid 2-hydroxylase Scs7 (1,17). The 2-OH FAs produced by degradation of PHS are LCFAs (mainly C16 and C18) and are mostly incorporated into glycerophospholipids after being converted to nonhydroxy odd-chain LCFAs by Mpo1 (Fig.  3) (24). The substrates of Scs7 are thought to be VLCFAs (mainly C26) in ceramides (1,38), although the possibility remains that free VLCFAs are substrates of Scs7. So far, the metabolism of 2-OH VLCFAs in sphingolipids is completely unknown. We examined the possibility that Mpo1 converts 2-OH C26:0-COOH into C25:0-COOH after the release of 2-OH C26:0-COOH from sphingolipids by inositol phosphosphingolipid phospholipase C and ceramidase. The levels of ceramide containing 2-OH C26:0-COOH were comparable between WT and mpo1 cells (Fig. 8A), but those of ceramide containing 2-OH C25:0-COOH were approximately 38-fold lower in mpo1 than in WT cells (Fig. 8B). The latter is predicted to be produced from the Mpo1 product C25:0-COOH by CoA addition, amide bond  formation with LCB (ceramide production), and 2-hydroxylation (Fig. 8C). We then performed an in vitro assay using purified Mpo1 and 2-OH C24:0-COOH as the 2-OH VLCFA substrate. Mpo1 indeed exhibited activity toward 2-OH C24:0-COOH (Fig. 8D). These results indicate that Mpo1 is involved not only in -oxidation of 2-OH LCFAs, which are the metabolites of PHS, but also in -oxidation of 2-OH VLCFAs produced by the Scs7-mediated pathway (Fig. 8E).

Mpo1 is important for carbon utilization from 2-OH FAs under carbon starvation conditions
Although we revealed that Mpo1 is involved in the metabolism of a variety of chain lengths of 2-OH FAs, the physiological significance of Mpo1 remains largely unknown. Deletion of the MPO1 gene did not affect growth under nutrient-rich growth conditions (24) (Fig. 9A). We examined the effects of MPO1 deficiency on growth under starvation conditions. The growth of mpo1 cells was slower than that of WT cells under carbon-deprived conditions (Fig. 9B). On the other hand, the growth of mpo1 cells was similar to that of WT cells under nitrogen starvation conditions (Fig. 9C). Addition of nonhydroxy FAs (C16:0-COOH), but not 2-OH FAs (2-OH C16:0-COOH), to the carbon-deprived medium restored the growth of mpo1 cells (Fig. 9D, E). These results suggest that yeast utilizes 2-OH FAs as a carbon source under carbon-starved conditions, whereas this utilization is abolished in mpo1 cells due to impaired -oxidation of 2-OH FAs.

DISCUSSION
Mpo1 is an Fe 2+ -dependent dioxygenase, and oxygen bound to Fe 2+ is involved in the -oxidation of 2-OH FAs (24). However, the amino acid residues involved in Fe 2+ binding were unclear. In the present study, we have revealed that three histidine residues (His19, His28, and His115) are particularly important for Mpo1 activity (Figs.  2-4, 6). The K D value of WT Mpo1 toward Fe 2+ was 10 M (Fig. 6B), but no activity was detected for the mutants in which one of these histidine residues had been substituted (H19A, H28A, and H115A), even in the presence of 1 mM Fe 2+ . In class I diiron-oxo proteins, each of the two histidine-containing motifs (motif 1, H....HXXXE; motif 2, HXXXH....HXXXXD) forms a coordinate bond with Fe 2+ (31,33). On the other hand, the amino acid sequence that we have revealed here, which is important for Mpo1 activity, is YXXXH....H....HXXXE (the Mpo1 motif). In this sequence, the tyrosine and glutamate residues can be  replaced by histidine and aspartate residues, respectively (Fig. 4B). The H....HXXXE portion of the Mpo1 motif is identical to motif 1 of class I diiron-oxo proteins, and the combination of YXXXH and HXXXE in the Mpo1 motif is similar to motif 2. At present, it is not clear which residues in the Mpo1 motif actually form the coordinate bonds with Fe 2+ . However, we speculate that the His19 and His115 residues, which are indispensable for Mpo1 activity in any of our assays, are at least involved in the bond formation process.
Depressed activity of the H19A, H28A, and H115A mutants was observed in our [ 3 H]PHS labeling experiment (Fig. 2D) and our measurements of odd-chain PC levels (Fig. 3A). In contrast, the activity of the Y15A and E119A mutants was not depressed in those experiments, although this was the case to some extent when 2-OH C16:0-COOH was added to the culture medium (Fig. 3B), and also in the in vitro assay (Fig. 6A). In the first two in vivo experiments, the Mpo1-mediated reaction was carried out under conditions that mimicked physiological conditions in the living organism, with low levels of endogenous substrates. Under these conditions, a slight depression of enzyme activity was difficult to detect. On the other hand, the latter two experiments were conducted under comparatively unnatural conditions: the in vivo experiment was conducted in the presence of high levels of substrate, and the in vitro conditions were not necessarily optimized for the enzyme, enabling us to detect the depression of its activity with high sensitivity. Our results indicate that Tyr15 and Glu119 also have some role in the enzyme activity of Mpo1.
To date, 3,062 Mpo1 homologs have been found in a wide range of species (1,986 species) from bacteria, fungi, protozoa, and plants (http://pfam.xfam.org/family/ PF06127#tabview=tab1). The function of yeast Mpo1 (2-OH FA dioxygenase) was the first function to be elucidated from this family (17,24). Sequence alignment of representative Mpo1 family members revealed that the Mpo1 family contains three highly conserved regions consisting of 32, 7, and 40 amino acid residues [collectively called the Mpo1 domain (previously the DUF962 domain) hereafter], respectively ( Fig. 2A). Mpo1 (S. cerevisiae) has gaps of 25 and 29 amino acids separating these highly conserved regions. We also revealed that Mpo1 spans the ER membrane via the highly hydrophobic regions H2 and H3. These regions correspond to the gaps between the highly conserved regions in the Mpo1 family. On the other hand, in some Mpo1 family members (e.g., Alcanivorax borkumensis and Bacillus sp.), the three highly conserved regions (Mpo1 domain) are nearly contiguous, with almost no gaps between them, suggesting that they do not have transmembrane domains. Considering the high sequence conservation in this family of enzymes, it is highly probable that all family members are dioxygenases, like Mpo1. However, the substrates of the soluble family members may not be lipids (2-OH FAs) but rather hydrophilic 2-OH-containing molecules. In addition, some Mpo1 family members contain domains such as cytochrome P450-like, protein kinase, or peptidase (http://pfam.xfam.org/family/PF06127#tabview=tab1), in addition to the Mpo1 domain, suggesting that this family has a variety of cellular functions.
The regions other than the transmembrane domains of Mpo1 are oriented toward the cytosol; in other words, the active sites of Mpo1 exist in the cytosol. This means that 2-OH FAs, which are the substrates of Mpo1, exist in the cytosolic leaflet of the ER membrane with their carboxy group exposed to the cytosol, and receive oxygen attack from the cytosol. The first irreversible reaction in the degradation pathway of LCBs is catalyzed by LCB 1-phosphate lyase. The active site of this enzyme is also oriented to the cytosolic side of the ER membrane (39). Therefore, the metabolism of LCBs proceeds in the cytosolic leaflet of the ER membrane.
Deletion of MPO1 caused growth retardation under carbon-starved conditions (Fig. 9B). The cells recovered from the growth retardation after addition of nonhydroxy FA to the medium (Fig. 9D). On the other hand, deletion of MPO1 did not affect growth under nitrogen starvation conditions (Fig. 9C). These results suggest that Mpo1 is not involved in the induction of autophagy, but rather functions to supply carbon sources from 2-OH FAs. Both 2-OH FAs and PHS, a precursor of 2-OH FAs, are components of sphingolipids. Although little is known about sphingolipid metabolism under carbon-starved conditions, our results suggest that sphingolipids are degraded and utilized as a carbon source to produce energy. 2-OH FAs cannot be used as substrates for -oxidation. Therefore, they must be converted to nonhydroxy FAs by Mpo1 to become substrates.
In conclusion, here, we have revealed the membrane topology, substrate specificity, amino acid residues important for activity, and physiological role of Mpo1. Of the Mpo1 family members, only Mpo1 itself has been analyzed. However, our findings give clues as to the enzymatic characterization and roles of the entire Mpo1 family.

Data availability
All data are contained within the article.