Two standard media, YPD medium (1% yeast extract (Bacto), 2% peptone (Bacto), and 2% glucose) and synthetic dextrose (SD) medium (Mai et al., 2018) were used to culture yeast cells. Unless otherwise noted, yeast cells were grown aerobically and exponentially at 30°C in liquid media. After the addition of chemicals to the culture media, the cells were further incubated under the same conditions. To support the growth of opi3Δ and/or cho3Δ strains, SD medium was supplemented with choline (1 mM final). For the agar plates, SD medium was soldified with 2% Bacto agar. Uracil/5-fluoroorotic acid (5-FOA) SD agar plates contained 100 mg/L uracil and 1.0 g/L 5-FOA.

Culture optical density was monitored using the spectrophotometer, BioRad Smartspec 3000.

Plasmids pRS316-IRE1 (the URA3 selectable marker; Promlek et al., 2011) and pRS313-IRE1 (the HIS3 selectable marker; Kimata et al., 2004) are yeast YCp-type single-copy plasmids that carry the IRE1 gene (open reading frame plus the 5′- and 3′-untranslated regions). Plasmid pYT-TDH3p-PMA1-mCherry (Mai et al., 2019) is also a yeast YCp-type single-copy plasmid carrying the URA3 selectable marker. Herein, the PMA1-coding sequence on pYT-TDH3p-PMA1-mCherry was replaced by a CHO2-coding sequence to generate pYT-TDH3p-CHO2-mCherry, which was used for expression of mCherry-tagged Cho2 under the control of the constitutive strong promoter of TDH3. Plasmid pRS316 (Sikorski and Hieter, 1989) was used as a control empty vector. Plasmid pPM28 was used to express eroGFP (Merksamer et al., 2008).

Unless otherwise noted, we used the standard yeast strain BY4741 (MAT a his3Δ1 leu2Δ0 met15Δ0 ura3Δ0) as the wild-type strain. Single-gene-deletion mutants of BY4741 (xxx::KanMX4) were obtained from EUROSCARF (http://www.euroscarf.de/). To obtain an opi3::HIS3MX strain, we transformed the EUROSCARF opi3::KanMX4 strain with the HIS3MX gene carried on pUG27 plasmid (Gueldener et al., 2002). The opi3::HIS3MX construct was PCR-amplified from the resultant transformant and used for transformation of the EUROSCARF cho2::KanMX4 strain, which yielded a cho2::KanMX4/opi3::HIS3MX strain.

Another yeast strain KMY1516 (MATα ura3 his3 trp1 ire1::TRP1 UPRE-lacZ::LYS2 UPRE-GFP::LEU2; Kimata et al., 2004) was also employed in this study. After transformation with pRS316-IRE1, KMY1516 was further transformed with the xxx::KanMX4 constructs and/or the opi3::HIS3MX construct (xxx represents opi3 or cho2), which had been PCR-amplified from the EUROSCARF strains or their derivative to generate gene-deletion mutants. The resulting ire1::TRP1/xxx::KanMX4 strains (or the ire1:TRP1/cho2::KanMX4/opi3::HIS3MX strain) containing pRS316-IRE1 were further transformed with pRS313-IRE1 or its mutants, and cultured on uracil/5-FOA-containing agar plates to counter-select pRS316-IRE1. The IRE1 mutants were generated as previously described (Kimata et al., 2007; Tran et al., 2019).

HeLa cells (0.15 × 10⁵ cells) were inoculated in six-well dishes (Corning) with 3 ml Dulbecco’s modified Eagle’s medium supplemented with 10% FCS. The dishes were incubated at 37°C (5% CO₂) for 2 days.

Total RNA samples, which were extracted from yeast cells using the hot-phenol method (Collart and Oliviero, 2001), were subjected to RT-PCR in which the poly(dT) RT primers and the HAC1-specific PCR primers were employed (Promlek et al., 2011; Mai et al., 2018). As the forward and reverse PCR primers interposed the HAC1-intron sequence, RT-PCR yielded different-sized products from unspliced (HAC1u) and spliced (HAC1i) HAC1 mRNAs. The RT-PCR products were electrophoresed on a 2% agarose gel, and the ethidium bromide-fluorescent image was captured with a UV-transilluminating imager E-Box (Vilber Lourmat). The gel images were analyzed using ImageJ software (https://imagej.nih.gov/ij/), and the HAC1 mRNA-splicing efficiency was calculated using the following formula: The   HAC1   mRNA − splicing   efficiency = 100 × ( HAC1i   band   intensity ) { ( HAC1i   band   intensity ) + ( HACL1u   band   intensity ) }

To check the XBP1 mRNA splicing pattern in HeLa cells, we extracted total RNA and performed RT-PCR analysis as described by Tsuchiya et al. (2018). The sequences of the human XBP1 specific PCR primers were 5′-TTA​CGA​GAG​AAA​ACT​CAT​GGC​C-3′ and 5′-GGG​TCC​AAG​TTG​TCC​AGA​ATG​C-3′.

For steady-state radiolabeling, cells were grown in YPD or SD medium (supplemented with 1 mM choline) containing 370 kBq/ml ³²P-orthophosphate (NEX053, PerkinElmer) for more than 4.5 h (2–3 generations). After centrifugal harvest, cells (approximately equivalent to OD_{600 =} 1.0) were washed once with water, were suspended in 100 µl of chloroform:methanol (1:1), and were lysed by bead-beating with 100 µl of glass beads (425–600 µm; Sigma-Aldrich). The cell lysates were mixed with 200 µl of chloroform:methanol (2:1) and clarified by centrifugation (15,000 × g, 1 min). The supernatants were further mixed with 50 µl of chloroform-methanol (2:1), 50 µl of chloroform, and 150 µl of water. After agitation, the mixtures were centrifuged at 15,000 × g for 5 min, and the organic bottom layers were dried in vacuo and analyzed by thin-layer chromatography (TLC) as described by Sakakibara et al. (2015). Autoradiographs of the TLC plates were captured using the phosphor imaging system Amersham Typhoon and analyzed using ImageJ software.

After harvest by centrifugation, cells (equivalent to OD₆₀₀ = 40.0) were subjected to lipid extraction as described for PMME detection. The vacuum-dried lipid samples were solubilized in PBS containing 1% Triton X-100 and enzymatically assayed for PC concentration using the colorimetric PC detection kit, LabAssay^TM Phospholipid (Fujifilm).

In accordance with the Microbial Identification (MIDI) protocol (Sasser, 2001), cellular fatty acids were converted into fatty acid methyl esters (FAMEs) through saponification and methylation. The resulting FAME samples were analyzed by gas chromatography, which was performed by TechnoSuruga Laboratory Co., Ltd. (Shizuoka, Japan).

As described in our previous report (Mai et al., 2018), yeast cells were disrupted by glass bead beating in a Triton X-100-containng buffer. The cell lysates were separated by ultracentrifugation (160,000 × g for 3 h) and analyzed by anti-BiP Western blotting.

Cells carrying the eroGFP-expression plasmid, pPM28, were observed under SP8 FALCON (Leica) with the 100×/1.40 HC PL APO CS2 objective lens. For excitation, a 405 nm diode laser (UV/violet-light excitation, 67% output) and a 496 nm white-light laser (blue-light excitation, 100% output) were employed. For detection, a hybrid detector (gating 492–571 nm) was employed. The pinhole size was 1.70 AU. The resulting images were processed as previously described (Phuong et al., 2021).

The culture optical density, the HAC1 mRNA-splicing efficiency, the Western-blot band densities, and the PMME and PC contents were determined from triplicate cultures, and were subjected to calculation of averages and standard deviations. To obtain p-values, two-tailed unpaired t-tests were performed using Microsoft Excel.

In yeast cells, Cho2 acts as the PE methyltransferase for the conversion of PE to PMME, which is further converted to PC by another phospholipid methyltransferase, Opi3 (de Kroon, 2007; Supplementary Figure S1A). As previously reported (Thibault et al., 2012), Ire1 is activated to induce HAC1-mRNA splicing in yeast cho2Δ cells and opi3Δ cells cultured in SD medium, probably because a shortage of PC causes LBS (Figure 1A, the “No choline” condition). As PC is also produced through the Kennedy pathway, through which free choline is conjugated to diacylglycerol (Gibellini and Smith, 2010; Supplementary Figure S1A), growth retardation of cho2Δ cells and opi3Δ cells was rescued by the addition of choline into the medium (Figure 1B). Nevertheless, opi3Δ cells, but not cho2Δ cells, were found to exhibit considerable HAC1-mRNA splicing even in the presence of choline (Figure 1A, the “1 mM choline” condition). In the experiment shown in Supplementary Figures S1B,C, we monitored phospholipid composition through steady-state ³²P labeling of cells and thin-layer chromatography. Consistent with a previous report by Sakakibara et al. (2015), PMME accumulation in wild-type cells and opi3Δ cells was completely abolished by the introduction of the cho2Δ mutation. Importantly, the HAC1-mRNA splicing in Δopi3Δ cells was also abolished by the introduction of the cho2Δ mutation (Figure 1C).

We next employed yeast cells producing Cho2 tagged with mCherry at the C-terminus (wild-type Cho2-mCherry) under the control of the strong TDH3 promoter. While wild-type Cho2-mCherry supported the growth of cho2Δ cells in SD medium that did not contain choline, such support was not observed when Cho2-mCherry carried the G102A/G104A point mutation in the putative enzymatic reaction center of the Cho2 moiety (Shields et al., 2003; Supplementary Figure S2A). Therefore, as expected, the G102A/G104A mutation is likely to inactivate Cho2. Consistent with this observation, the high expression of wild-type Cho2-mCherry resulted in an increase in the cellular abundance of PMME, which was abolished by the G102A/G104A mutation (Supplementary Figure S3). Importantly, high expression of wild-type Cho2-mCherry, but not of its G102A/G104A variant, induced the HAC1-mRNA splicing in wild-type (CHO2OPI3) yeast cells (Figure 1D). Therefore, we deduce that, in addition to PC shortage, PMME accumulation causes ER stress.

The deletion of the IRE1 gene is known to considerably worsen the growth of yeast cells upon ER stress (Cox et al., 1993). In the experiment shown Figure 1E, we employed ura3Δire1Δ yeast cells carrying a URA3/IRE1 plasmid, which was counter-selected by 5-FOA (Boeke et al., 1984). This strain could not grow on agar plates containing 5-FOA and choline when carrying the opi3Δ mutation, but grew well when carrying the cho2Δ mutation, the cho2Δopi3Δ mutation, or the intact CHO2OPI3 genes (wild-type; WT). Thus, the ire1Δ mutation abolished the growth of opi3Δ cells, but not that of wild-type, cho2Δ, or cho2Δopi3Δ cells in the presence of choline. Therefore, PMME accumulation is likely to harm cells through the induction of ER stress.

The ER stress induced by PMME accumulation is unlikely to be sufficient to cause severe growth retardation. As shown in Figure 1B, the growth rate of opi3Δ cells was similar to that of wild-type cells in the presence of choline. Moreover, the growth of wild-type cells was not retarded by the high expression of wild-type Cho2-mCherry (Supplementary Figure S2B).

To further confirm our proposition that PMME per se causes ER stress, we added free monomethylethanolamine (MME; 10 mM) to yeast cultures, resulting in an modest increase in cellular PMME abundance (Figure 2A; Supplementary Figure S3). MME is likely to be converted to PMME via the Kennedy pathway, as a similar phenomenon was observed even when cells carried the cho2Δ mutation (Figure 2B). As shown in Figure 2C, we induced ER stress by adding an ER-stressing reagent, dithiothreitol (DTT) or tunicamycin, into yeast cultures containing or not containing MME or choline. In the absence of ER stressors (Figure 2D; DTT 0 mM), exogenously added MME did not induce HAC1-mRNA splicing in wild-type cells. This is probably because, unlike the opi3Δ mutation or the high expression of wild-type Cho2-mCherry, the extracellular addition of MME increased the cellular PMME abundance only moderately (Supplementary Figure S3). Moreover, 3 mM DTT led to high-level HAC1-mRNA splicing independently of MME (Figure 2D). Nevertheless, MME intensified the HAC1-mRNA splicing induced by low-dose (1 mM) DTT (Figure 2D). Unlike MME, choline did not enhance the HAC1 mRNA-splicing level even when DTT (1 mM) was added to the cultures (Figure 2E). MME also boosted the HAC1-mRNA splicing that was induced by tunicamycin, although not strongly (Figure 2F).

Even when MME was added to yeast culture at a higher concentration (30 mM), HAC1-mRNA splicing was only marginal in the absence of another ER stress stimulus (Figure 2G). To monitor HAC1-mRNA splicing, we did not treated yeast cells with a higher concentration of PMME, which considerably inhibited cellular growth (Supplementary Figure S4). We then asked what occurs in the case of mammalian cells. IRE1α is the major paralog of mammalian Ire1 family proteins and promotes the splicing of XBP1 mRNA upon ER stress (Walter and Ron, 2011). MME induced the XBP1-mRNA splicing in mammalian HeLa cells in a dose-dependent manner even without other ER-stress stimuli (Figure 2H).

Sakakibara et al. (2015) reported that the cellular accumulation of PMME caused by the opi3Δ mutation inhibits mitophagy in yeast cells. If PMME accumulation in opi3Δ cells induces ER stress and triggers UPR simply via the inhibition of mitophagy, HAC1-mRNA splicing may be provoked in cells carrying genetic mutations that inhibit mitophagy. However, as shown in Supplementary Figure S5, such was not observed in the case of the atg32Δ mutation or the atg5Δ mutation, which is known to abolish mitophagy (Kanki et al., 2009; Okamoto et al., 2009; Kanki et al., 2015).

As described in the Introduction section, it is widely accepted that Ire1 directly senses ER accumulation of unfolded proteins (Kimata et al., 2007; Gardner and Walter, 2011). In the case of yeast Ire1, this ability is compromised by a luminal-domain partial deletion, namely the ΔIII mutation (Kimata et al., 2007; Promlek et al., 2011; Tran et al., 2019). On the other hand, the transmembrane domain of Ire1 serves as a sensor for LBS (Volmer et al., 2013; Halbleib et al., 2017). Halbleib et al. (2017) and Tran et al. (2019) previously indicated that this ability of Ire1 is compromised by a point mutation, V535R, which is located on the amphipathic helix of the transmembrane domain. Therefore, by using these Ire1 mutants, we can categorize stimuli that cause ER stress into two types (Tran et al., 2019). As shown in Figure 3A, the V535R mutation, but not the ΔIII mutation, of Ire1 attenuated the UPR induced by the OPI3-gene deletion. Thus, we deduce that high accumulation of PMME due to the opi3Δ mutation activates Ire1 by triggering LBS.

We then checked if the opi3Δ mutation affects the fatty-acid composition of lipidic molecules. In the experiment shown in Figure 3B, yeast cells were subjected to alkaline saponification, and the resulting fatty acids were methylated and quantitatively detected by gas chromatography. This revealed that the opi3Δ mutation, but not the cho2Δ mutation, increased the proportion of lauric acid (C12:0) (Figure 3B).

On the other hand, the strong HAC1-mRNA splicing induced by co-treatment of cells with MME and low-dose (1 mM) DTT was compromised by the ΔIII mutation (Figure 4A). Unlike the case of wild-type Ire1, MME did not boost the low-level HAC1-mRNA splicing triggered by 1 mM DTT when cells carried ΔIII Ire1 (Figure 4B). Thus, we presume that this stress stimulus impairs protein folding in the ER. To support this idea, we performed BiP sedimentation analysis (Figure 4C) using a method developed in our previous studies (Promlek et al., 2011; Mai et al., 2018). Yeast cells were lysed in the presence of a mild detergent, Triton X-100, and subjected to ultracentrifugation. ER-accumulated unfolded proteins tend to form BiP-bound aggregates, the amount of which is estimated by anti-BiP Western blot analysis of the pellet fraction. As shown in Figure 4C, BiP was abundantly carried in the pellet fraction obtained from wild-type cells dually treated with low-dose DTT and MME. This result suggests that PMME aggravates DTT-induced disturbance of ER protein folding.

We next checked the disulfide bond-forming ability of the ER using the cells producing eroGFP, an ER-located GFP variant that changes its excitation spectrum depending on intracellular disulfide-bond formation. After chemical treatment, the cells were illuminated by two different wavelength laser beams, and the ratio of the two fluorescent signals was expressed as the eroGFP value. When eroGFP is reduced in the ER, cells show a higher eroGFP value (Merksamer et al., 2008). As shown in Figure 4D, treatment of cells with 1 mM DTT increased the eroGFP value, which was not further elevated by MME. Therefore, we presume that PMME does not aggravate the DTT-induced ER stress directly by stimulating the activity of DTT to reduce protein disulfide bonds.

Lastly, we investigated whether the endogenous level of PMME affects ER-stress status in yeast cells. A low dose of DTT (1 mM) induced HAC1-mRNA splicing in wild-type cells more strongly than in cho2Δ cells when they were cultured in standard nutrient-rich YPD medium (Figure 5A). As YPD medium contains free choline, the PC abundance did not differ between wild-type and cho2Δ cells (Figure 5B). Thus, we assume that wild-type cells are prone to ER stress owing to endogenously accumulated PMME. Cells were then cultured in SD medium containing choline, which was added to support the growth of cho2Δ cells, and were stressed by 1mM DTT. Figure 5C indicates that, in this case, wild-type cells and cho2Δ cells showed similar HAC1 mRNA-splicing profiles, suggesting that endogenously accumulated PMME does not potentiate ER stress in cells cultured in SD medium. The different outcomes between YPD and SD media can be explained by the result shown in Figure 5D, which indicates that the PMME abundance was considerably lower in wild-type cells cultured in SD medium than those cultured in YPD medium. On the other hand, it is unlikely that ER stress affected cellular PMME abundance (Figure 5E).

We next examined if PMME aggravates ER stress that is induced by a stress stimulus besides DTT. We previously reported that ethanol stress impairs ER protein folding and activates Ire1 (Miyagawa et al., 2014). As shown in Figure 5F, the cho2Δ mutation attenuated the HAC1-mRNA splicing induced by ethanol in cells cultured in YPD medium.