TORC1 inactivation stimulates autophagy of nucleoporin and nuclear pore complexes

Macroautophagy (hereafter autophagy) is an intracellular degradation pathway found in most eukaryotes in which different cellular components are sequestered within double-membrane vesicles called autophagosomes and transported to lytic compartments (lysosomes or vacuoles; Yang and Klionsky, 2010; Ohsumi, 2014). Although autophagic sequestration used to be deemed a nonselective process, an increasing number of studies have revealed that a wide variety of proteins and organelles are sequestered into autophagosomes in a selective manner (Kirkin, 2020; Gatica et al., 2018). In selective autophagy, proteins called autophagy receptors recognize specific cargo molecules or structures. In Saccharomyces cerevisiae, autophagy receptors interact with Atg11 to recruit the core Atg proteins, which initiate autophagosome formation on the cargos (Zientara-Rytter and Subramani, 2019). The receptors also bind to Atg8 on forming autophagosome membranes (isolation membranes or phagophores), thereby promoting cargo sequestration into autophagosomes (Noda et al., 2010; Johansen and Lamark, 2020).

The nuclear pore complex (NPC) is a huge protein assemblage embedded in the nuclear envelope and has vital roles in gene expression by mediating transport between the nucleus and cytoplasm as well as regulating genome architecture (Strambio-De-Castillia et al., 2010). The NPC is composed of seven substructures (the transmembrane ring, outer ring, inner ring, linker, pore filaments, nuclear basket, and cytoplasmic filaments), which are assembled by ∼30 different nucleoporins (Nups; Aitchison and Rout, 2012; Strambio-De-Castillia et al., 2010; Kim et al., 2018). As with other cellular components, to control the copy number, function, and quality of the NPC, appropriate degradation and turnover of the NPC or Nups must be important. Previous studies have revealed that the turnover rate differs substantially among Nups (Tran and Wente, 2006). In addition, a surveillance system for NPC assembly, involving the endosomal sorting complexes required for transport (ESCRT) machinery, has been reported (Webster et al., 2014). However, the cellular system that degrades the NPC or Nups is still largely unknown.

Here, we report that the NPC and Nups are selectively degraded by autophagy when Tor kinase complex 1 (TORC1) is inhibited in S. cerevisiae. While the whole NPC is targeted to two receptor-mediated selective autophagy pathways, unassembled Nup159, a component of the cytoplasmic filament of the NPC, is degraded by autophagy via its direct binding to Atg8.

We investigated whether the NPC is degraded by autophagy using a GFP processing assay. In this assay, autophagic degradation of GFP-fused proteins is assessed by monitoring the generation of GFP fragments, which are relatively resistant to vacuolar proteases. We chose a Nup from each of the seven substructures of the NPC (Fig. 1 A; Strambio-De-Castillia et al., 2010) based on the presence of human homologues and taggability and fused GFP to the C termini of these Nups. We confirmed the localization of the GFP-tagged Nups to the nuclear envelope by fluorescence microscopy (Fig. S1 A). We treated these cells with the TORC1 inhibitor rapamycin (Loewith et al., 2002; Noda and Ohsumi, 1998), which induces different types of autophagy. In immunoblotting analysis using anti-GFP antibodies, GFP fragments were detected in all cells expressing different Nups after treatment with rapamycin (Fig. 1 B). The accumulation of GFP fragments was also observed when cells were shifted to nitrogen starvation, in which TORC1 is inactivated (Fig. S1 B). GFP fragments began to appear at 8 h after rapamycin treatment (Fig. 1 C) and were not observed in ATG1 knockout (atg1Δ) cells, which are defective in autophagy (Matsuura et al., 1997; Fig. 1 B and Fig. S1 B). These results suggest that Nups are degraded by autophagy when TORC1 activity is attenuated. It should be noted that in these immunoblotting experiments, the levels of GFP-tagged Nups were largely different from each other and not necessarily consistent with their previously estimated copy numbers in the NPC (Kim et al., 2018; Rajoo et al., 2018). This would be due to different efficiencies in extraction of Nups from cells and/or their transfer from SDS-PAGE gels to polyvinylidene fluoride membranes.

TORC1 inactivation induces both nonselective and selective autophagy. To determine whether autophagic degradation of Nups is selective, we examined this process in cells lacking Atg11, which, in association with autophagy receptors, plays a central role in target recognition in most selective autophagy pathways (Zientara-Rytter and Subramani, 2019). ATG11 knockout caused marked defects in degradation of all of the GFP-tagged Nups we examined (Fig. 2 A). Autophagy receptors interact with both degradation targets and Atg8 family proteins to link the targets to forming autophagosomal membranes (Gatica et al., 2018; Kirkin, 2020). These receptors contain the Atg8-family interacting motif (AIM; or the LC3 interacting region), which binds to the AIM-binding pocket (AIMBP) of Atg8 family proteins (Noda et al., 2010; Johansen and Lamark, 2020). The P52 and R67 residues are located around the AIMBP, and an alanine substitution at these residues decreases the receptor-binding ability of Atg8 (Noda et al., 2008). Similar to ATG11 knockout, this mutation also severely impaired autophagic degradation of GFP-tagged Nups (Fig. 2 B). These results suggest that a receptor-dependent mechanism mediates autophagic degradation of Nups.

Nups assembled into seven different NPC substructures were all degraded by autophagy (Fig. 1 B). In addition, their degradation was similarly affected by ATG11 knockout and mutations in the AIMBP of Atg8 (Fig. 2, A and B). These results suggest that Nups are mainly sequestered into autophagosomes in the state of the NPC, while some Nups may also be degraded by autophagy in a form of assembly intermediates or unassembled proteins. Consistent with this prediction, immunoelectron microscopy detected signals of the pore filament component Nsp1 and the cytoplasmic filament component Nup159 (Nup159-HA) at electron-dense structures embedded in nuclear envelope–derived, double-membrane vesicles (see next section) within autophagic bodies, which are autophagosomal inner vesicles that accumulate within the vacuolar lumen in vacuolar protease-deficient cells (Takeshige et al., 1992; Fig. 2, C and D; and Fig. S1 C).

We previously reported that double-membrane vesicles form by budding from the nucleus and are selectively sequestered into autophagosomes in yeast cells treated with rapamycin (Mochida et al., 2015). This pathway is called “nucleophagy” and requires the outer nuclear membrane protein Atg39 as a specific receptor. Because Nups were found in double-membrane vesicles within autophagic bodies (Fig. 2, C and D), we predicted that nucleophagy was responsible for NPC degradation. Indeed, ATG39 disruption decreased degradation of GFP-tagged Nups (Fig. 2 A). However, substantial amounts of GFP fragments were still observed in these cells. These results suggest that a proportion of the NPC is degraded via Atg39-dependent nucleophagy, but there exists another autophagy pathway for NPC degradation under TORC1-inactivating conditions. We also found autophagic bodies encapsulating Nsp1-positive double-membrane vesicles in ATG39 knockout cells (Fig. S1 D), suggesting that an Atg39-independent mechanism also sequesters the NPC into autophagosomes in this manner.

We examined degradation of Nups in the absence of the other autophagy receptors Atg19, Atg32, Atg34, Atg36, Atg40, and Cue5, which are known to target vacuolar proteins, mitochondria, α-mannosidase, peroxisomes, the endoplasmic reticulum, and polyQ proteins, respectively (Fig. S2 A; Shintani et al., 2002; Kanki et al., 2009b; Okamoto et al., 2009; Suzuki et al., 2010; Motley et al., 2012; Lu et al., 2014; Mochida et al., 2015). Only ATG40 knockout decreased Nup degradation. We previously reported that although Atg40 is a receptor for ER-phagy, a proportion of this protein resides in the nuclear envelope and that its absence partially impairs nucleophagy (Mochida et al., 2015). We also showed that ATG40 knockout exacerbated defects in Nup degradation in ATG39 knockout cells (Fig. S2 B). These results allowed us to speculate that Atg40 in the nuclear envelope, in addition to Atg39, participates in Nup degradation as a receptor. However, the AIM of Atg40 was dispensable for degradation of nuclear components, although it was strictly required for ER-phagy; its reticulon-like function probably promotes the formation of double-membrane vesicles in the nuclear envelope (our unpublished results). Thus, Atg40 is unlikely to act as a receptor in NPC degradation. We also compared degradation of Nups (Ndc1 and Nup85) with the inner nuclear membrane protein Src1, which was used to examine nucleophagy activity in our previous study (Mochida et al., 2015). ATG39 knockout only partially affected degradation of the Nups but largely impaired that of Src1 (Fig. S2 B). Conversely, ATG40 knockout had stronger effects on Nup degradation. These results suggest that the NPC and Nups are not merely degraded by nucleophagy as a bystander. In addition, our data showed that ATG11 knockout and Atg8 AIMBP mutant cells exhibited more severe defects in NPC degradation than ATG39 knockout cells (Fig. 2, A and B; and Fig. S2 C). These results suggest the existence of an unknown receptor for an Atg39-independent degradation of the NPC.

We also showed that the sorting nexin Atg24, which is involved in selective autophagy pathways via an unknown mechanism (Nice et al., 2002; Kanki et al., 2009a; Shpilka et al., 2015; Nemec et al., 2017), is required for degradation of the NPC (Fig. S2 D). On the other hand, the absence of Nvj1, which is important for piecemeal microautophagy of the nucleus (Cadou and Mayer, 2015; Kvam and Goldfarb, 2007; Otto and Thumm, 2020), did not compromise NPC degradation (Fig. S2 E). This result is consistent with the observation that the NPC was excluded from the nucleus–vacuole junction, where piecemeal microautophagy of the nucleus occurs, in cells treated with rapamycin (Fig. S2 F and Fig. S1 A).

A number of selective autophagy substrates, such as fatty acid synthase, the clock repressor CRY1, and the nuclear receptor corepressor NCoR1, directly interact with Atg8 family proteins (Shpilka et al., 2015; Toledo et al., 2018; Saito et al., 2019). We found that the immunoprecipitation of FLAG-tagged Atg8 coprecipitated all of the HA-tagged Nups analyzed (Fig. 3 A). Moreover, almost all Nups were detected in immunoprecipitates of Atg8 by mass spectrometry (Table S1). These results suggest that the NPC interacts with Atg8. We showed that the NPC–Atg8 interaction depended on rapamycin treatment (Fig. 3 B). Moreover, the interaction was abolished by the mutation in the AIMBP of Atg8 (Fig. 3 C). These results suggest that an unknown AIM-containing protein mediates the interaction between the NPC and Atg8 upon TORC1 inactivation.

We hypothesized that the NPC binds to Atg8 via a Nup containing the AIM. Among putative AIM-containing proteins listed in the iLIR database (https://ilir.warwick.ac.uk; Jacomin et al., 2016), Nup159, Nup116, and Nup1 contain two, two, and one putative AIM, respectively (Fig. 4 A and Fig. S3 A). However, since coimmunoprecipitation of HA-tagged Nup1 with FLAG-Atg8 was not detected by immunoblotting (Fig. S3 B), Nup1 was excluded from further analysis. The core of the AIM is composed of four amino acid residues, in which aromatic and aliphatic residues at the first and fourth positions, respectively, are crucial for Atg8 binding (Noda et al., 2010; Johansen and Lamark, 2020). We replaced the first and fourth residues in the putative AIMs of Nup159 and Nup116 with alanine, expressed these mutants from plasmids in the presence of the chromosomally encoded WT proteins, and examined interactions of the mutant proteins with Atg8 by coimmunoprecipitation analysis (Fig. 4 A; and Fig. S3, A and B). The F443A L446A mutation in Nup159 and the two mutations in Nup116 did not impair the coprecipitation of these proteins with Atg8. By contrast, the Y1078A L1081A mutation in Nup159 reduced its coprecipitation with Atg8 to an undetectable level. (Although W303-derived strains were used in this study, the residue numbering of Nup159 is based on that in S288C-derived strains, because the NUP159 gene from BY4741 was cloned and used for mutational analysis. Y1078 and L1081 correspond to Y1052 and L1055 of Nup159 in W303-derived strains, which lacks a part of the FG repeats.) We also constructed yeast strains expressing Nup159-GFP mutants in the absence of chromosomally encoded Nup159 and obtained similar results (Fig. 4 B). In addition, we showed that the coprecipitation of other Nups (Pom152 and Nsp1) with Atg8 was also impaired in Nup159^{Y1078A L1081A} mutant cells (Fig. 4 B). Thus, the Y1078-D-K-L1081 sequence in Nup159 indeed functions as an AIM, and the interaction between Nup159 and Atg8 through this motif contributes considerably to the NPC–Atg8 interaction. We showed that knockout of ATG39 and ATG40 did not affect the interaction between Nup159 and Atg8 (Fig. S3 C).

Next, we examined the impact of the Nup159–Atg8 interaction on autophagic degradation of the NPC. Fluorescence microscopy confirmed that the Nup159-GFP mutants localized to the nuclear envelope, similar to WT Nup159-GFP (Fig. 4 C and Fig. S3 D). The GFP processing assay revealed that whereas Nup159^{F443A L446A}-GFP was degraded by autophagy comparably to WT Nup159-GFP, Nup159^{Y1078A L1081A}-GFP exhibited a significant defect in its degradation (Fig. 4 D and Fig. S3 E). However, other Nups we tested were all degraded normally in Nup159^{Y1078A L1081A} mutant cells (Fig. 4 E). These results suggest that while Nup159 interacts with Atg8 via the AIM both inside and outside of the NPC, the interaction of Nup159 with Atg8 is important for degradation of this Nup itself, but not for that of the whole NPC. Nonetheless, it is still possible that the Nup159–Atg8 interaction is also involved in degradation of the whole NPC along with other interactions that link the NPC to Atg8, including an interaction mediated by an unknown receptor.

In this study, we provided the first evidence for autophagic degradation of the NPC and Nups, which we term “NPC-phagy” and “nucleoporinophagy,” respectively. Our results suggest that multiple selective autophagy pathways mediate NPC/Nup degradation in S. cerevisiae cells when TORC1 activity is attenuated. Immunoelectron microscopy revealed that NPCs in the state of being embedded in nuclear envelope–derived, double-membrane vesicles are enclosed within autophagosomes (Fig. 5 A). Atg39-dependent nucleophagy is likely to participate in NPC degradation in this manner.

In addition, our results suggested the existence of an unknown receptor, which might be specific to NPC-phagy and function in collaboration with Atg11. Although we have been unable to identify such a receptor to date, we found that the cytoplasmic filament component Nup159 interacts with Atg8 via the AIM and that this interaction is pivotal for the NPC association with Atg8. Given these results, we predicted that autophagic degradation of the NPC was affected by the mutation in the AIM of Nup159. However, although this mutation significantly retarded degradation of the Nup159 mutant itself, it did not influence degradation of other Nups. These results were contrary to our expectation but clearly showed that Nups not assembled into the NPC can also be targets for selective autophagy (nucleoporinophagy; Fig. 5 B). In the case of Nup159, binding to Atg8 via the AIM contributes to its efficient degradation. We observed cytoplasmic puncta of Nup159-GFP (Fig. 4 C and Fig. S3 D), allowing us to speculate that aggregates of Nup159 or Nup159-containing subcomplexes are taken up by autophagosomes via the interaction with Atg8. Unassembled Nup159 or Nup159 dissociated from the NPC may be directed to autophagic elimination to control the integrity or function of the NPC under TORC1-inactivating conditions. A previous study reported that monoubiquitylated Nup159 recruits the dynein light chain to the NPC, which is important for nuclear segregation during mitosis (Hayakawa et al., 2012). Therefore, it is possible that autophagic degradation of Nup159 is involved in the control of cell cycle progression. Other Nups may also be subjected to nucleoporinophagy by binding to Atg8 or unknown receptors under certain conditions.

NPC-phagy can contribute to the control of the quality and quantity of the whole NPC. TORC1 inhibition evokes different cellular responses, including autophagy, and causes cell cycle arrest at G1 phase with delayed G2/M transition (Loewith and Hall, 2011; Barbet et al., 1996; Nakashima et al., 2008). In S. cerevisiae, the nuclear envelope and NPCs are not disassembled during mitosis, and NPC assembly occurs continuously throughout the cell cycle (Winey et al., 1997). NPC-phagy may promote NPC turnover or remove misassembled/malfunctioned NPCs and thereby participate in the maintenance of the NPC under TORC1-inactivating conditions. Consistent with this idea, disruption of NUP116, which destabilizes the NPC (Onischenko et al., 2017), stimulated autophagic degradation of the NPC (Fig. S3 F). In other organisms in which NPC disassembly occurs and Nups are dispersed in the cytoplasm during cell division, some Nups may be degraded by nucleoporinophagy.

During the course of this study, another group also reported that NPCs are degraded by selective autophagy under nitrogen starvation in S. cerevisiae and that this NPC degradation was enhanced by the absence of Nups (Nup120 and Nup133) that cause aberrant NPC clustering in the nuclear envelope (Lee et al., 2020). However, there are some contradictory results between the two studies. The same AIM in Nup159 has also been identified in that study, but unlike our results, the AIM mutation significantly reduced degradation of other Nups, and accordingly, Nup159 has been proposed as an autophagy receptor for NPC degradation. In addition, deletion of NUP116 rather impeded autophagy of the NPC (Lee et al., 2020). These discrepancies might be due to different strain backgrounds used in the two studies, but close examination will be required to resolve this issue.

The NPC has been implicated in the replicative life span of S. cerevisiae (Lord et al., 2015). In addition, a deficiency in the NPC leads to various human diseases and aging (Nofrini et al., 2016; Sakuma and D’Angelo, 2017). The basic framework for understanding NPC-phagy and nucleoporinophagy determined in this study provides a foundation for studies of the molecular mechanisms and physiological/pathological significance of these new autophagy pathways in various organisms.

Yeast strains used in this study were derived from W303-1A or BJ2168 (Thomas and Rothstein, 1989; Jones, 1991) and are listed in Table S2. Gene disruption and tagging were performed by a standard method (Janke et al., 2004). Yeast cells were grown at 30°C in YPD medium (1% yeast extract, 2% peptone, and 2% glucose) or SD+CA+ATU medium (0.17% yeast nitrogen base without amino acids and ammonium sulfate, 0.5% casamino acid, 2% glucose, 0.002% adenine sulfate, 0.002% tryptophan, and 0.002% uracil). SD+CA+ATU without uracil (SD+CA+AT) was used for the culture of strains carrying pRS316-derived plasmids. To induce autophagy, cells grown to mid-log phase were treated with 200 ng/ml rapamycin or incubated in SD-N medium (0.17% yeast nitrogen base without amino acids and ammonium sulfate and 2% glucose).

pRS316-based centromeric plasmids for the expression of Atg8^WT and Atg8^{P52A R67A}, Atg8^G116A–GFP, and Atg8^{P52A R67A G116A}–GFP were described previously (Mochida et al., 2015; used in Fig. 2 B, Fig. 3 C, and Fig. S2 C). pRS316-NUP159-GFP, pRS316-NUP1-6×HA, and pRS316-NUP116-6×HA (used in Fig. 4 B and Fig. S3, B, D, and E) were constructed as follows. The coding sequences of NUP159-GFP, NUP1-6×HA, and NUP116-6×HA with their original promoters were amplified by PCR using genomic DNA prepared from YTY171, YTY304, and YTY309 (Table S2). The PCR products were cloned into pRS316 using the Gibson Assembly Kit (New England Biolabs). Point mutations were introduced into these plasmids by a PCR-based method. To introduce mutations into NUP159 on genomic DNA, 2,415–4,302 bp of the NUP159 ORF followed by the GFP or 6×HA sequence as well as this DNA fragment containing the Y1078A L1081A mutation were cloned into pRS303 together with the PGK1 terminator using the Gibson Assembly Kit, yielding pRS303-NUP159^2415-4302bp-GFP/6×HA and pRS303-NUP159^{2415-4302bpY1078A L1081A}-GFP/6×HA. These plasmids were digested within the NUP159 sequence by SpeI and introduced into yeast cells for integration into the NUP159 locus by homologous recombination to obtain the strains used for the analyses shown in Fig. 4, C, D, and E.

For immunoblotting, antibodies against GFP (632381; Clontech), Atg8 (Nakatogawa et al., 2012), HA (11867423001; Roche), Nsp1, and Pom152 (gifts from Dr. Tohru Yoshihisa, University of Hyogo, Hyogo, Japan) were used. To prepare the samples for the Nup–GFP processing assay, yeast cell pellets were suspended in 0.1 M NaOH, kept on ice for 5 min, and centrifuged at 5,000 rpm for 5 min at 4°C. The pellets were suspended in Urea SDS sample buffer (37.5 mM MOPS-NaOH, pH 6.8, 2% SDS, 100 mM DTT, 4 M urea, and a trace amount of bromophenol blue) and incubated at 65°C for 10 min. These samples were centrifuged at 15,000 g for 1 min, and the supernatants were subjected to SDS-PAGE and immunoblotting.

Yeast cells grown to mid-log phase were converted to spheroplasts by incubating them in 0.5× YPD or 0.5× SD+CA+AT medium containing 1 M sorbitol and 10 U/OD₆₀₀-unit-cells zymolyase 100T (Nacalai Tesque). The spheroplasts were washed with 20 mM Hepes-KOH (pH 7.2) and 1.2 M sorbitol, treated with 200 ng/ml rapamycin (LC Laboratories) in 0.5× YPD or 0.5× SD+CA+AT medium containing 1 M sorbitol for 2 h at 30°C, and lysed with IP buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 10% glycerol, and 5 mM EDTA) containing 2× cOmplete Protease Inhibitor Cocktail (Roche), 2 mM phenylmethylsulfonyl fluoride, and 1.0% n-dodecyl-β-D-maltoside for 30 min at 4°C. The cell lysates were centrifuged at 15,000 g for 15 min, and the supernatants were incubated with NHS FG-beads (Tamagawa Seiki) conjugated with anti-FLAG M2 antibody (F1804; Sigma-Aldrich) or GFP Nanobody (Harada et al., 2019) at 4°C for 2 h. The beads were washed with IP buffer containing 0.1% n-dodecyl-β-D-maltoside three times and then incubated in Urea SDS sample buffer at 65°C for 10 min to elute bound proteins.

Fluorescence microscopy to confirm the nuclear localization of GFP-tagged Nups was performed using an inverted fluorescence microscope (IX83; Olympus) equipped with a 488-nm laser, electron multiplying charge-coupled device camera, and a 150× objective lens as described previously (Kotani et al., 2018). Images were acquired using MetaMorph (Molecular Devices) and processed using Fiji (ImageJ; Schneider et al., 2012; Schindelin et al., 2012).

Electron microscopy was performed by Tokai-EMA based on rapid freezing and freeze-fixation methods. Cells were sandwiched with copper disks and quickly frozen in liquid propane at −175°C. The frozen samples were substituted with 1% tannic acid in ethanol and 2% distilled water at −80°C for 48 h and then and kept at −20°C for 4 h and at 4°C for 1 h. These samples were dehydrated in anhydrous ethanol for 30 min three times and infiltrated with a 50:50 mixture of ethanol and LR white resin (London Resin) at 4°C for 1 h. After infiltration, the samples were washed with LR white three times by incubation at 4°C for 30 min for each wash process and transferred to fresh LR white, followed by polymerization at 50°C overnight. The polymerized resins were ultrathin sectioned, and the sections were placed on nickel grids and immunostained with anti-Nsp1 antibodies and secondary antibodies conjugated to 10-nm gold particles as reported previously (Shima et al., 2019). Electron microscopy was performed using a transmission electron microscope (JEM-1400Plus; JEOL), and digital images (3,296 × 2,472 pixels) were obtained using a charge-coupled device camera (EM-14830RUBY2; JEOL).

Yeast cells expressing 3×FLAG-tagged Atg8 (FLAG-Atg8) or nontagged Atg8 were spheroplasted, treated with rapamycin, and subjected to immunoprecipitation using an anti-FLAG antibody as described in Materials and methods. The resulting immunoprecipitates were analyzed by mass spectrometry. Identified proteins with a normalized abundance in samples from FLAG-Atg8 cells that is fivefold higher than that in samples from nontagged Atg8 cells were selected, and Nups, proteins involved in autophagosome formation (autophagy machinery), known substrates, or receptors for selective autophagy (autophagy substrate/receptor) are listed in Table S1.

Fig. S1 shows the results related to autophagic degradation of Nups. Fig. S2 shows the results of Nup degradation in various autophagy-related mutants. Fig. S3 shows the results of the analysis of Nup mutants. Table S1 shows the results of mass spectrometry of Atg8 immunoprecipitates. Table S2 lists the yeast strains used in this study.

We thank the members of our laboratories for materials, discussions, and technical and secretarial support. We also thank Dr. Tohru Yoshihisa for antibodies against Nsp1 and Pom152 and the Biomaterial Analysis Center, Technical Department at Tokyo Institute of Technology for DNA sequencing services.

This work was supported in part by the Ministry of Education, Culture, Sports, Science, and Technology (KAKENHI Grants-in-Aid for Scientific Research 25111003, 17H01430, and 19H05708 to H. Nakatogawa), the Japan Science and Technology Agency (CREST grant JPMJCR13M7 to H. Nakatogawa), the Tokyo Tech Fund (STAR grant to H. Nakatogawa), and the NOVARTIS Foundation (Japan) for the Promotion of Science (to H. Nakatogawa).

The authors declare no competing financial interests.

Author contributions: Y. Tomioka, T. Kotani, Y. Ohsumi, and H. Nakatogawa designed the project. Y. Tomioka performed most of the experiments with the help of H. Kirisako. Y. Oikawa, Y. Kimura, and H. Hirano performed mass spectrometry analysis. Y. Tomioka and H. Nakatogawa wrote the manuscript. All authors analyzed and discussed the results and commented on the manuscript.