Letter | Published:

Systematic analysis of ribophagy in human cells reveals bystander flux during selective autophagy

Nature Cell Biologyvolume 20pages135143 (2018) | Download Citation


Ribosomes are abundant cellular machines1,2 that are regulated by assembly, supernumerary subunit turnover and nascent chain quality control mechanisms1,2,3,4,5. Moreover, nitrogen starvation in yeast has been reported to promote selective ribosome delivery to the vacuole in an autophagy conjugation system dependent manner, a process called ‘ribophagy’6,7. However, whether ribophagy in mammals is selective or regulated is unclear. Using Ribo–Keima flux reporters, we find that starvation or mTOR inhibition promotes VPS34-dependent ribophagic flux, which, unlike yeast, is largely independent of ATG8 conjugation and occurs concomitantly with other cytosolic protein autophagic flux reporters8,9. Ribophagic flux was not induced upon inhibition of translational elongation or nascent chain uncoupling, but was induced in a comparatively selective manner under proteotoxic stress induced by arsenite10 or chromosome mis-segregation11, dependent upon VPS34 and ATG8 conjugation. Unexpectedly, agents typically used to induce selective autophagy also promoted increased ribosome and cytosolic protein reporter flux, suggesting significant bulk or ‘bystander’ autophagy during what is often considered selective autophagy12,13. These results emphasize the importance of monitoring non-specific cargo flux when assessing selective autophagy pathways.


Substantial cellular resources are used to produce the 106 to 107 ribosomes present in a typical mammalian cell (~10% of protein mass), raising the question of to what extent ribosome recycling into building blocks is regulated. Previous studies in Saccharomyces cerevisiae revealed approximately threefold faster delivery of ribosomes to the vacuole when compared with other cytosolic proteins under conditions of nitrogen starvation. Ribophagic flux in yeast required conjugation of the Atg8p Ub-like protein to phosphatidylethanolamine via Atg7p and the Atg1p-Atg17p kinase that promotes phagophore formation6,8. In mammals, starvation-induced autophagy requires the VPS34-BECN1 PI3 kinase and the ULK1 protein kinase for phagophore initiation. In contrast, the ATG8 conjugation pathway (employing the ATG7 E1 enzyme and the ATG5-ATG12 E3 complex) is not essential for starvation-induced flux, but its deletion reduces the success rate of autophagosomal closure by ~70% and productive lysosome fusion by ~30 min (ref. 9). Conservatively, approximately 103 ribosomes may be randomly engulfed into a 0.5 μm3 autophagosome14 during starvation-induced autophagy in a 3,000 μm3 mammalian cell. However, the extent to which mammalian ribosomes are degraded through autophagy in healthy or stressed cells is unknown, as is whether selective forms of ribophagy are used to degrade, for example, translationally defective ribosomes3,4,5. Conversely, are particular cytosolic constituents excluded from autophagosomal engulfment during selective autophagy? If not, to what extent are cytosolic contents degraded through bystander autophagy12,13 during specific turnover of damaged organelles? Stated differently—how selective is selective autophagy?

To address these questions, we established a Ribo–Keima system (Fig. 1a) in which chromosomal RPS3 (a 40S protein) or RPL28 (a 60S protein) in HCT116 and HEK293 cells was C-terminally tagged with Keima by gene editing in the heterozygous or homozygous form (Fig. 1b–d and Supplementary Fig. 1a). Keima is a pH-responsive reporter that undergoes a chromophore resting charge state change upon trafficking to the lysosome (pH of ~4.5), allowing flux measurements by flow cytometry or microscopy (Fig. 1a)15. Keima itself is also stable to lysosomal proteases in mammals, and the appearance of a processed Keima protein therefore reveals lysosomal trafficking15,16,17. Ribo–Keima proteins formed ribosomes and polysomes (Fig. 1e,f and Supplementary Fig. 1b) and efficiently incorporated puromycin into nascent chains, indicating little effect on global translation (Supplementary Fig. 1c). Excess ribosomal subunits are degraded via the proteasome4,5. Ribo–Keima cells treated with the proteasome inhibitor bortezomib (BTZ, 4 h) failed to accumulate free monomeric RPS3–Keima or RPL28–Keima protein (Fig. 1e,f and Supplementary Fig. 1b), indicating stoichiometric assembly into ribosomes. As with DsRed (ref. 18), heat-induced hydrolysis of the N-acyl imine in Keima during denaturation produced a ~9 kDa N-terminal Keima fragment (fused with either RPS3 or RPL28) and a 21 kDa C-terminal Keima–Flag fragment (Supplementary Fig. 1d–f), accounting for bands indicated by an asterisk in the immunoblots for RPS3 or RPL28 in Fig. 1d and Supplementary Fig. 1b,d.

Fig. 1: Construction of a system for monitoring ribophagy flux in human cells.
Fig. 1

a, Schematic of ribophagy detection in mammalian cells using a Ribo–Keima reporter system. FP, fluorescent protein. b, View of the C terminus of RPL28 and RPS3 in the structure of an 80S complex showing the solvent exposed site for Keima protein tagging. PDB: 5AJ0. c, Generation of homozygous RPL28–Keima (top) and heterozygous RPS3–Keima (bottom) reporter cell lines via CRISPR-Cas9-based gene editing. d, Confirmation of RPL28–Keima and RPS3–Keima expression in HEK293 and HCT116 cells by immunoblotting. Asterisks indicate Ribo–Keima protein fragments formed after the hydrolysis of the N-acyl imine in the Keima chromophore during heat denaturation (Supplementary Fig. 1d–f). Data shown represent more than three independent experiments with similar results. wt, wild type; KI, Keima-expressing cells. e,f, Sucrose density gradient centrifugation, showing proper assembly of RPL28–Keima (e) and RPS3–Keima (f) with 60S, 40S, 80S and polysomes with or without 4 h BTZ treatment (0.1 μM) to stabilize any unassembled subunits. Two independent experiments showed similar results. Unprocessed original scans of blots are provided in Supplementary Fig. 6. wcl, whole-cell lysate.

To examine ribophagic flux, we initially treated cells with the mTOR inhibitor Torin1 (Tor1). Inhibition of mTOR, a negative regulator of ULK1 (refs 19,20), results in increased bulk autophagic flux in a manner that partially mimics starvation. We found that all four Ribo–Keima cell lines displayed an increase in the 561/488 nm excitation ratio by flow cytometry that was reversed by addition of the lysosomal acidification inhibitor bafilomycin A1 (BafA) (Fig. 2a,b). Similarly, we observed the formation of a processed form of Keima (25 kDa) reflecting degradation of the RPS3 and RPL28 fragments of each fusion protein by lysosomal hydrolases (Fig. 2c), and this was also blocked by the VPS34 inhibitor SAR405 (Fig. 2c). In parallel, we found that Tor1 promoted the formation of VPS34-dependent cytoplasmic Keima-positive (red) puncta15 (Fig. 2d,e) that co-localize with lysosomal markers (LAMP1-eGFP and Lysotracker Green, Lyso-G) (Fig. 2f,g). A subset (~20%) of Keima puncta were coincident with the eGFP–LC3B-positive puncta observed on addition of Tor1 (24 h) (Fig. 2h), indicative of trafficking through the autophagosome. Control experiments in which the Keima signal was monitored by confocal microscopy at various times after washout of Tor1 (24 h treatment) revealed that Keima puncta are maintained for at least 8 h, and the number of puncta is not altered by treatment with lysosomal hydrolase inhibitors, indicating that Keima is very stable within the lysosome, as previously reported15 (Supplemental Fig. 1h,i). An estimate of ~10% of ribosomes trafficking to lysosomes over the 24 h period in Tor1-treated cells was obtained by comparing the Keima fluorescence of cells at pH 7.3 and 4.5 with that of Tor1-treated cells in the presence of BafA (see Methods). By contrast, <1% of ribosomes trafficked into lysosomes in complete growth media (based on flow cytometry; see Methods), and this level of ribophagy is generally undetectable by immunoblotting for processed Keima in Ribo–Keima reporter cells (Fig. 2c), indicating low basal ribophagic flux.

Fig. 2: mTOR inhibition promotes ribophagy flux in a VPS34-dependent manner.
Fig. 2

a, Ribo–Keima reporter cells were treated with BafA (50 nM, 1 h), Torin1 (150 nM, 24 h, Tor1), or a combination of the two, then analysed by flow cytometry. Frequency distributions of 561/488 nm ex. ratios are shown (n = 10,000 cells per condition). b, Histograms of mean values of the biological triplicate experiments for 561/488 nm ex. ratios (from a). Error bars represent standard error of the mean (s.e.m.); ****P< 0.0001, **P< 0.01, *P< 0.04, two-way ANOVA. c, Ribo–Keima reporter cell lines treated with Torin1 (150 nM, 24 h), SAR405 (an inhibitor of VPS34, 1 μM, 24 h), or a combination of the two were immunoblotted for Keima and LC3B. (Asterisk indicates Keima with different folding status, see Supplementary  Fig. 1d.) d, Confocal images of live HEK293 cells expressing RPS3–Keima after Torin1 (150 nM, 24 h) or Torin1 (150 nM, 24 h)/SAR405 (1 μM, 24 h) co-treatment. Scale bars, 20 μm. e, Unbiased quantitation of the live-cell images in d; the numbers of red Keima puncta per cell are shown. Data are shown as mean ± s.e.m. (n = 52, 60 and 60 cells from three independent experiments). f, HEK293 RPS3–Keima cells stably expressing LAMP1–eGFP were incubated in the presence or absence of Torin1 for 24 h before live-cell imaging. Scale bars, 20 μm. g, HEK293 RPS3–Keima cells treated as in f were stained with LysoTracker Green before live-cell fluorescence microscopy. Scale bars, 20 μm. h, HEK293 RPS3–Keima cells stably expressing eGFP–LC3 were treated as in f, then subjected to live-cell fluorescence microscopy. Scale bars, 20 μm. For fh bottom left, co-occurrence (%) of red Keima puncta with LAMP1–eGFP (n > 1,800 puncta), LysoTracker Green (n > 2,000 puncta) and eGFP–LC3 (n > 110 puncta) were calculated then plotted as black bars. Random co-occurrence (%) is shown as white bars (see Methods). For fh bottom right, fluorescence line scans along the white lines in the merged insets. Data are presented as mean ± s.e.m; ****P < 0.0001, ***P < 0.001, **P < 0.01, two-way ANOVA. Statistical source data for b and fh are provided in Supplementary Table 2. All experiments were performed three times with similar results. Unprocessed original scans of blots are provided in Supplementary Fig. 6.

Ribophagy in yeast following nitrogen deprivation requires Atg8p conjugation, the Ubp3p deubiquitylating enzyme and Cdc48p ATPase6,7. Surprisingly, ATG5 deletion in HEK293 Ribo–Keima cells had no effect on ribophagic flux in response to Tor1 or starvation (Fig. 3a–g and Supplementary Fig. 2a,c,e), while, as expected based on VPS34 inhibition, deletion of the VPS34 regulatory subunit BECN1 (ref. 20) greatly reduced ribophagic flux measured by flow cytometry, confocal imaging or immunoblotting for processed Keima (Fig. 3a–d and Supplemental Fig. 2a–d). Moreover, ribophagic flux in HEK293 RPS3–Keima;ATG5−/− cells induced by Tor1 (24 h) was reversed by treatment with SAR405 (Fig. 3c,d and Supplementary Fig. 2c), indicating that ATG5-independent flux nevertheless requires VPS34 (refs 20,21). HEK293 RPS3–Keima;ATG5−/− cells also generated lysosomally processed Keima at rates similar to those seen in ATG5-expressing cells (Supplementary Fig. 2c). Ribophagic flux in HEK293 RPS3–Keima cells was initially detectable 8 h post-starvation and increased slightly at 16 h (Fig. 3e–g). However, no discernable kinetic differences in Keima puncta per cell were found to depend upon ATG5 status (Fig. 3e–g). In HCT116 RPL28–Keima cells, starvation promoted a time-dependent increase in Keima puncta per cell, which was 24–44% lower in parallel ATG5−/− cells (Fig. 3h and Supplementary Fig. 2f–i). The reduced autophagic flux was rescued upon stable expression of ATG5 in ATG5−/− cells, as found by immunoblotting for lysosomally processed Keima (Supplementary Fig. 2i). Additional evidence that ribophagy in HEK293 cells upon mTOR inhibition is independent of the conjugation system includes (1) the presence of ribosomes in autophagosomal structures of ATG5−/− cells by electron microscopy (Fig. 3i), (2) co-localization of the Keima signal (561 nm excitation (ex.)) with LAMP1–eGFP-labelled lysosomes (Supplementary Fig. 2j), (3) no effect on Keima puncta number or intensity in ATG5−/− cells treated with lysosomal hydrolase inhibitors for 14 h during mTOR inhibition, indicating that the Keima turnover rate in the lysosome is similar in ATG5+/+ and ATG5−/− cells (Supplementary Fig. 2k), and (4) absence of eGFP–LC3 puncta associated with 561 nm excitation puncta (Supplementary Fig. 2l). Thus, neither Tor1 nor starvation-induced ribophagy absolutely requires ATG8 conjugation, but both require the VPS34 pathway. Similar results were obtained using HEK293 cells expressing levels of free Keima protein comparable to that in cells expressing RPS3–Keima (Supplementary Fig. 3a), consistent with the absence of selectivity in ribophagy flux observed following Tor1 treatment or starvation. Interestingly, Keima–LC3 protein expressed in ATG5−/− cells showed lysosomal flux comparable to ribosomes, further suggesting ATG8 conjugation-independent bulk autophagy.

Fig. 3: Ribophagy in response to mTOR inhibition in HEK293 cells is ATG5-independent but BECN1-dependent.
Fig. 3

a, Frequency distributions of 561/488 nm ex. ratios measured on HEK293 RPS3–Keima cells lacking ATG5 or BECN1 were compared after Torin1 or Torin1/BafA treatment using flow cytometry (n = 10,000 cells per condition). b, Average 561/488 nm ex. ratios calculated from the biological triplicate experiments from a (orange, blue, and grey dotted lines). Data are presented as mean ± s.e.m. ****P < 0.0001, **P < 0.01, two-way ANOVA. c, Confocal images of live HEK293 RPS3–Keima cells lacking ATG5 after Torin1 (150 nM, 24 h) or Torin1 (150 nM, 24 h)/SAR405 (1 μM, 24 h) co-treatment. Scale bars, 20 μm. d, Number of red Keima puncta per cell measured from live-cell images of HEK293 RPS3–Keima WT, ATG5−/− or BECN1−/− cells taken after Torin1 (150 nM, 24 h) or Torin1 (150 nM, 24 h)/SAR405 (1 μM, 24 h) co-treatment. Data are presented as mean ± s.e.m. The total numbers of cells from three biologically independent samples are indicated in the graph. e,f, HEK293 RPS3–Keima cells (with or without ATG5) were incubated with HBSS in the presence of lysosomal hydrolase inhibitors (LHI, E64d and pepstatin, 30 μM each) for the indicated times prior to live-cell imaging. Scale bars, 20 μm. g, Unbiased quantification of the live-cell images obtained as shown in e and f. h, Unbiased quantification of red Keima puncta obtained from live HCT116 RPL28–Keima cells (with and without ATG5) treated with HBSS/LHI, as shown in Supplementary Fig. 2g,h. In g and h, total numbers of cells from three biologically independent samples are indicated in the graph. Data are presented as mean ± s.e.m. i, Electron microscopy images of HEK293 RPS3–Keima WT, ATG5−/− and BECN1−/− cells 4.5 h after HBSS treatment in the presence of BafA (50 nM). Red arrow: ribosomes in autophagosomes or autophagolysosomes; yellow arrow: ribosomes bound to ER in cytosol. Scale bars, 500 μm. Data represent two independent experiments. The statistical source data for b is provided in Supplementary Table 2. All experiments were repeated at least three times unless otherwise indicated.

We examined the involvement of USP10 (Ubp3p in yeast) and p97/VCP (Cdc48p in yeast) in mammalian ribophagy7. The p97 inhibitor CB5083 (ref. 22) had no effect on Ribo–Keima processing during mTOR inhibition (Supplementary Fig. 3b), indicating that p97 is not required for this form of ribophagy. The USP10 inhibitor spautin23 reduced ribophagy of both 40S and 60S subunits (Supplementary Fig. 3c,d); however, this probably reflects the role of USP10 in maintaining BECN1–VPS34 activity in mammals23 rather than being directly involved in ribophagy. Consistent with recent work9, we did not observe a proposed role8 for the trans-Golgi as a source of membranes for ATG8 conjugation-independent autophagy, as brefeldin A had no effect on Tor1-induced ribophagy in Ribo–Keima ATG5−/− cells (Supplementary Fig. 3e). Moreover, we did not observe a correlation of Keima-positive lysosome localization with the Golgi network but did observe association with the endoplasmic reticulum (ER), a source of autophagic membranes21 (Supplementary Fig. 3f,g).

Maintaining healthy ribosomes is thought to be critical for cellular homeostasis, and agents that induce ribosomal stress could promote a need for the re-establishment of ribosome pools24. We therefore screened translation inhibitors and stress agents for increased ribophagic flux. Addition of the nascent chain terminator puromycin, the elongation inhibitor cycloheximide (CHX) and the p97 inhibitor CB5083 (ref. 22), which blocks extraction of ubiquitylated nascent chains from the ribosome, had no appreciable effect on ribophagic flux (Fig. 4a,b). Similarly, proteasome inhibitor BTZ or ER-stress agent tunicamycin (TM) did not promote ribophagic flux (Fig. 4a,b). By contrast, ribophagic flux similar to or greater than that seen with mTOR inhibition was induced by reversine, an MPS1 inhibitor that creates proteome imbalance through chromosome mis-segregation11,25,26 and by sodium arsenite (NaAsO2, AS), which induces stress granules10 and translational inhibition (Supplementary Fig. 4a). Oxidative stress by H2O2 tended toward an increase in ribophagic flux (P > 0.1) (Fig. 4a,b).

Fig. 4: A screen of ribosome stress agents identifies sodium arsenite and reversine as ribophagy inducers.
Fig. 4

a, HEK293 RPS3–Keima cells were exposed to Tor1, 150 nM, 24 h; reversine, 0.5 μM, 48 h; AS, 10 μM, 24 h; H2O2, 250 μM, 24 h; cycloheximide, 10 μM, 5 h; bortezomib, 250 nM, 5 h; puromycin, 2 μM, 5 h; p97 inhibitor CB5083, 100 nM, 24 h; or tunicamycin, 0.6 μM, 24 h, and the 561/488 ratio was measured (n = 10,000 cells per condition). b, HEK293 RPS3-Keima:WT, ATG5−/−, and BECN1−/− cells treated as in a (DB, diazaborine, 200 μM, 24 h). Mean 561/488 nm ex. ratios are plotted (n = 3 independent experiments, Supplementary Table 2). c, Frequency distributions of 561/488 nm ex. ratios measured in HEK293 RPS3–Keima:ATG5+/+, ATG5−/− cells treated with AS (n = 4200 cells per condition) ± BafA. d, The average 561/488 nm ex. ratios from triplicate experiments as in c. e, HEK293 RPS3–Keima:ATG5−/− cells transduced with a lentivirus expressing either ATG5WT or conjugation defective ATG5K130R mutant were treated with AS (10 μM, 24 h) ± BafA. The average 561/488 ratios of biological triplicate experiments are shown. Mean of two independent experiments is shown for ATG5K130R cells. f, Immunoblotting of indicated cells treated with AS (20 μM, 20 h) ± SAR405. g, Average 561/488 nm ex. ratios of the indicated cell lines treated with reversine (0.5 μM, 48 h) ± BafA (50 nM, 1 h) (n = 3 independent experiments). h, Frequency distributions of 561/488 nm ex. ratios for HCT116 RPL28–Keima cells with or without reversine treatment ± BafA (n = 10,000 cells per condition). i, Immunoblot of HCT116 RPL28–Keima cells treated with reversine ± SAR405. j, Imaging of HCT116 RPL28–Keima:ATG5+/+, ATG5−/− cells ± reversine (0.5 μM, 48 h). Scale bars, 20 μm. k, Quantification Keima puncta number per cell of the images in j. Mean ± s.e.m. Total numbers of cells are indicated as n. l, Average 561/488 nm ex. ratios of cells treated with reversine (0.5 μM, 48 h) ± BafA (n = 3 independent experiments). m, Immunoblots of indicated cell lines treated as in l. Mean ± s.e.m. is shown in b,d,e,g and l. ****P < 0.0001, ***P < 0.001, **P < 0.01. In b, one-way ANOVA. In d,e,g and l, two-way ANOVA. All experiments were repeated three times unless otherwise indicated. Statistical source data are provided in Supplementary Table 2. Unprocessed original scans are provided in Supplementary Fig. 6.

Stress granules are thought to undergo ATG8- and p97-dependent selective autophagy27,28. Consistent with previous reports27, AS induces stress granules marked with eIF4E–eGFP and 40S ribosomal protein RPS3–Keima, but not with RPL28–Keima (Supplementary Fig. 4b,c). However, stress granule formation was dependent on AS concentration (Supplementary Fig. 4d). At 20 μM AS, processed Keima in HEK293 RPS3–Keima cells was detected by ~3 h post AS treatment and was blocked by SAR405 up to 18 h post treatment without obvious stress granule or ribosome aggregate formation (Supplementary Fig. 4e,f). We found that AS-induced ribophagy as measured in HEK293 RPS3–Keima cells was largely dependent upon ATG5 and its ATG12 conjugate, and was reversed by BafA (Fig. 4c–f and Supplementary Fig. 4g). Consistent with flow cytometry (Fig. 4a,b), AS produced more processed Keima than Tor1 treatment, and also led to inhibition of pT389 S6K (Supplementary Fig. 4h), in agreement with reduced mTOR activity29. Unexpectedly, however, ribophagic flux was also increased by AS in HEK293 RPL28–Keima cells and was reduced by BafA (Fig. 4e), indicating that the 60S subunit also undergoes autophagy. Ribophagic flux for both HEK293 RPS3–Keima and RPL28–Keima was reduced upon addition of CB5083 (Supplementary Fig. 4i). Thus, although both Tor1 and AS inhibit S6K phosphorylation and induce ribophagic flux, AS-induced ribophagy largely requires ATG5 and p97 while Tor1 does not.

Reversine-induced chromosome mis-segregation produces supernumerary protein subunits, protein complex imbalance, proteotoxic stress and activation of autophagy to presumably facilitate removal of misfolded or supernumerary proteins11. To examine reversine-induced ribophagic flux, we employed HCT116 Ribo–Keima cells, which have a stable and near diploid karyotype. We observed increased ribophagy flux with both RPS3– and RPL28–Keima upon reversine treatment, but the 60S subunit RPL28–Keima reporter showed approximately twofold higher Keima ratio change responsive to BafA (Fig. 4g). In RPL28–Keima cells, processed Keima was detected as early as 24 h post reversine treatment (Fig. 4h,i). Reversine-dependent ribophagic flux was reversed by BafA and SAR405 (Fig. 4g–i) and was strikingly ATG5-dependent (Fig. 4j–m).

The distinct ATG5 dependence observed between Tor1-induced versus AS or reversine-induced ribophagic flux led us to examine the relative selectivity of ribophagy in comparison to additional Keima flux reporters, including autophagosomes themselves (Keima–LC3), cytosol (LDHB–Keima), cytoskeleton (Keima–ACTB–), proteasome (PSMD12–Keima) or mitochondria (TOMM20–Keima). Stable Keima reporter cells were sorted to achieve comparable levels of expression, though lower than that for the Ribo–Keima reporters (Fig. 5a–d and Supplementary Fig. 5a). To assess relative cargo flux, HEK293 reporter cell lines were treated with either Tor1, AS or SAR405 to block basal autophagy, and extracts were examined by immunoblotting for processed Keima (Fig. 5b) with signals then normalized to Tor1-treated cells (see Methods) (Fig. 5e). AS induces lysosomal delivery of Keima–LC3 to a level comparable to that induced by Tor1, indicating elevated overall autophagy flux. Similarly, non-ribosomal Keima reporter proteins displayed a similar level of increase in autophagic flux after AS and Tor1 treatment (Fig. 5b,e). Moreover, in contrast to other reporter proteins, ribophagic flux measured with both RPS3–Keima and RPL28–Keima was three- to fourfold higher than with Tor1, indicating a level of selectivity for this cargo. The increased selectivity for ribophagic flux with AS reflects a much lower rate of basal flux when compared to the other cargo and a larger induced flux relative to Tor1 treatment (Fig. 5b and Supplementay Fig. 5b,c). Analogous experiments were performed with HCT116 cells in the context of reversine treatment, again revealing selectivity in reversine-dependent turnover of RPL28–Keima when compared with Tor1 treatment, while with the other reporters, the increased flux was again comparable to that seen with Tor1 treatment (Fig. 5c,d,f). Unlike AS, however, reversine did not inhibit S6K phosphorylation (Fig. 5b,d) and did not promote selective turnover of RPS3–Keima (Fig. 5f), suggesting distinct flux parameters for large and small ribosome subunits in this setting.

Fig. 5: Quantitative western blot analyses of various Keima reporter cell lines reveal selective capture of ribosomes during AS and reversine treatment and the relative quantity of bystander autophagy during selective autophagy.
Fig. 5

a, Confocal images of live HEK293 cells expressing seven different Keima reporter proteins show proper intracellular localization. b, Extracts from the indicated cell lines treated with SAR405 (1 μM, 20 h), AS (20 μM, 20 h), or Tor1 (150 nM, 20 h) were subjected to immunoblotting. The representative western blot images were developed using HRP-conjugated secondary antibodies, whereas the quantitative western blot analysis of biological triplicate experiments was performed using near-infrared fluorescent secondary antibodies. Blot images were processed using an Odyssey CLx imager for quantitation. c, Proper localization of Keima reporter proteins expressed in HCT116 cells was confirmed by live-cell imaging. Scale bars (a,c), 5 μm. d, Extracts from the indicated HCT116 Keima reporter cell lines treated with SAR405 (1 μM, 24 h), Tor1 (150 nM, 24 h) or reversine (Rev, 0.5 μM, 48 h) were subjected to immunoblotting. e,f, Quantitation of immunoblots obtained as in b (e) and d (f). The processed Keima band intensity of each lane was normalized by the intact Keima band intensity in the same lane. The relative abundance of processed Keima in AS or reversine treated cells was then normalized with Torin1-treated cells (internal standard for each cell line). Data are presented as mean ± s.e.m. Total numbers of biologically independent experiments are indicated. gi, Bystander autophagy during mitophagy (g), pexophagy (h) and lysophagy (i) was analysed using HEK293 Keima reporter cell lines. Corresponding cells were treated either with SAR405, antimycin A (10 μM, 15 h) and oligomycin A (5 μM, 15 h) (AO) or with clofibrate (500 μM) for the indicated time to induce mitophagy (g) and pexophagy (h), repectively. Keima reporter cell lines used for mitophagy analysis stably express exogenous PARKIN. To induce lysophagy, cells were treated with either SAR405 or l-leucyl–l-leucine methyl ester (LLOMe, 1 mM) for 1 h, then further incubated with fresh medium for 24 h. Red arrow indicates lysozyme-resistant Keima (25 kDa) fragment. All experiments were repeated more than three times unless otherwise indicated. Statistical source data are provided in Supplementary Table 2. Unprocessed original scans of blots are shown in Supplementary Fig. 6.

Increased flux for various cytosolic cargo (for example, LDHB) in response to AS treatment, which ostensibly promotes selective targeting of stress granules for autophagy28, led us to ask the extent to which conditions that are thought to promote selective forms of autophagy induce bystander autophagy12,13. The extent to which cytosolic proteins are captured during what would otherwise be considered selective cargo capture has not been examined systematically. Relative to basal conditions, PARKIN-dependent mitophagy via mitochondrial depolarization with antimycin A and oligomycin A (AO) induced autophagic flux for all reporters tested, to an extent similar to that seen with Tor1, with LDHB–Keima, for example, increasing two- to fourfold above basal conditions (Fig. 5g). Similar results were obtained for cells treated with pexophagy-inducing clofibrate (Fig. 5h) and lysophagy-inducing LLOMe (Fig. 5i). Thus, agents used for the induction of selective autophagy of cytosolic membrane-bound organelles also promoted capture of additional cytosolic proteins with diverse subcellular localization and assembly patterns.

Here, we set out to systematically assess ribophagic flux in mammalian cells. Under basal conditions, ribophagic flux is extremely low (<1% over an ~24 h period) in HEK293 cells. As with yeast, nutrient deprivation (as well as mTOR inhibition) increases ribophagic flux (to ~10% of ribosomes over a 24 h period in human cells), but in contrast with yeast, neither ATG8 conjugation nor the p97 ATPase is required. Moreover, ribophagic flux was comparable to free Keima flux, indicating that ribophagy under these conditions is not selective, as classically defined30. Surprisingly, agents that induce defects in ribosomal quality control did not promote robust ribophagy, suggesting that autophagy is not generally used for turnover of translationally impaired ribosomes. However, we found that proteotoxic stress agents such as AS and reversine promote ATG5-dependent ribophagic flux and, in the case of AS, flux depended substantially on p97. The comparative selectivity of ribophagic flux relative to other cytosolic cargo primarily reflected a lower basal rate of ribophagic flux. Interestingly, autophagic flux of the cytosolic reporter LDHB–Keima in response to either AS or reversine was largely dependent upon ATG5−/− (Supplementary Fig. 5c,d). In contrast, Tor1-dependent turnover of cytosolic forms of Keima is independent of ATG5, but with similar overall flux rates as observed with ribosomes (Supplementary Fig. 3a). This suggests that bulk autophagy induced by mTOR inhibition is mechanistically distinct from that induced by some, but not all, forms of proteotoxic stress. It remains to be determined whether the key parameters for determining the dependency on ATG8 conjugation involve either the types of cargo being produced or, alternatively, the overall abundance of the cargo needing to be degraded through autophagy.

Selective autophagy often involves the Ub system30. Whether and how Ub supports ribophagy in mammals is unknown, although 40S ribosomal subunits are ubiquitylated upon stress31. We found that the Ub activating enzyme UBA1 inhibitor MLN7243 reduces AS-induced ribophagic flux by three- to fourfold and LDHB flux to a lesser degree (a factor of 1.6; Supplementary Fig. 5e–h), but did not affect Tor1 or reversine-induced ribophagic flux (Supplementary Fig. 5e–i). AS promotes phosphorylation of the Ub-binding autophagy receptor p62 (also called SQSTM1), stimulating its ability to bind and inhibit the NRF2 Ub ligase KEAP1 as well as promoting p62 recruitment to autophagic cargo29. However, p62 deletion in Ribo–Keima reporter cells had no effect on AS or reversine-induced ribophagic flux (Supplementary Fig. 5j–l), indicating that p62 is not essential for ribophagic flux under these conditions. Further studies are necessary to understand the mechanism of Ub usage in cargo selection during stress-induced autophagy. Finally, we found that diverse cytosolic proteins are captured under conditions ostensibly used to promote selective forms of organellar autophagy at levels comparable to that seen upon induction of bulk autophagy with mTOR inhibition. Although cytosolic proteins may be captured within autophagosomes during organellar capture, referred to as bystander autophagy12,13, the magnitude of non-selective cargo capture observed here was often comparable to that seen upon induction of bulk autophagy by Tor1. This raises the possibility that some agents used to promote selective organellar autophagy may also induce proteotoxic stress, which then simultaneously promotes bulk autophagy. In this regard, formation of ectopic autophagosomes unlinked to mitochondrial cargo has been observed upon treatment with depolarization agents32. Whether or not stress granules are selectively degraded as opposed to being cleared via stress-induced bulk autophagy that is ATG8-dependent remains unclear, but is suggested by our finding that levels of AS that do not generate obvious stress granules nevertheless induce widespread autophagic degradation of ribosomes and LDHB (Supplementary Fig. 5c). This work suggests the importance of testing alternative cargo when examining the selectivity of cargo capture and degradation, emphasizes the emerging theme8,9 that some forms of induced autophagy need not be ATG8-conjugation-dependent, and provides a framework and tools for analysis of selective autophagy.


Antibodies and chemicals

The following antibodies and reagents were used in this study: Keima (M126-3M, MBL), RPS3 (9538s, Cell Signal Technology), RPL28 (ab138125, Abcam), LC3B (2775S, Cell Signal Technology), ATG5 (12994s, Cell Signal Technology), ATG12 (4180T, Cell Signal Technology), BECN1 (3495, Cell Signal Technology), p70 S6K (2708T, Cell Signal Technology), p70 S6K phospho-T389 (9234S, Cell Signal Technology), actin (sc-69879, Santa-Cruz), tubulin (ab7291, Abcam), PCNA (SC56, Santacruz), KEAP1 (10503-2-AP, Proteintech), p62 (H00008878-M01, Novus Biologicals), Flag (M8823, Sigma), puromycin antibody (MABE343, EMD Millipore), p62 p-S351 (AP20690b, Abgent), ubiquitin (z045801-2, Dako), anti-mouse IgG (H+L) HRP conjugate (w4021, Promega), Western Lightning Plus-ECL (NEL105001, Perkin Elmer), anti-rabbit IgG (H+L), HRP conjugate (w4011, Promega), bortezomib (A2614, APExBIO), IRDye 800CW goat anti-mouse IgG H+L (925-32210, LI-COR), Torin1 (14379, Cell Signal Technology), SAR405 (A8883, APExBio), bafilomycin A (B1793, Sigma), chloroquine diphosphate salt (C6628, Sigma), Lysotracker Green (L7526, ThermoFisher Scientific), ER-Tracker Green (E34251, ThermoFisher Scientific), CellLight Golgi-GFP (C10592, ThermoFisher Scientific), HBSS (14025092, Invitrogen), E64d (A1903, APExBIO), pepstatin A (P5318, Sigma), reversine (R3904, Sigma), sodium (meta) arsenite (S7400, Sigma), H2O2 solution (H1009, Sigma), cycloheximide (C1988, Sigma), diazaborine (530729, EMDmillipore), puromycin dihydrochloride (P-600-500, Gold Biotechnology), brefeldin A (9972s, Cell Signal Technology), CB5083 (CB-5083, Selleckchem), tunicamycin (T7765, Sigma), poly-l-lysine solution (P4832, Sigma), lipofectamine 3000 (L3000008, Invitrogen), FluoroBrite DMEM (A1896701, ThermoFisher Scientific), benzonase nuclease HC (71205-3, Millipore), the Surveyor mutation detection kit (706025, Integrated DNA Technologies), GeneArt precision gRNA synthesis kit (A29377, ThermoFisher Scientific), RNasin (N2111, Promega) and the Neon transfection system (MPK1025, ThermoFisher Scientific). Dilutions and clone numbers of all antibodies used in this study are provided in Supplementary Table 3.

Cell culture

HEK293, HCT116 and HEK293T cells were grown in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal calf serum and maintained in a 5% CO2 incubator at 37 °C.

Generation of Keima knock-in cell lines using CRISPR-Cas9 gene editing

gRNAs targeting the C-terminus region of human RPS3 and RPL28 genes were designed using the CHOPCHOP website (http://chopchop.cbu.uib.no/). The guide sequence for the RPS3 (5′- GACATACCTGTTATGCTGTG-3′) and RPL28 (5′-CTTTCTCACCTGCCTCGAC-3′) genes were assembled into pX330 plasmid33. Donor vector was constructed by assembling Keima–Flag transgene with upstream and downstream homology arms (650 nucleotides each) into digested pSMART plasmid by Gibson assembly. HEK293 and HCT116 cells were transfected with donor and gRNA vectors (1:1 ratio) by lipofectamine 3000 (Invitrogen). Five days after transfection, Keima fluorescent-positive cells were sorted into 96-well plates by flow cytometry (MoFlo Astrios EQ, Beckman Coulter). Three weeks later, the expanded single-cell colonies were screened for integration of the Keima transgene by genotyping, followed by immunoblotting with anti-Keima, anti-RPS3 or anti-RPL28 and anti-Flag antibodies. Finally, genomic DNA of colonies showing integration of Keima by genotyping and immunoblotting methods were isolated and the targeted regions were PCR amplified for sequencing confirmation. The sequences of primers used for genotyping are provided in Supplementary Table 1.

Generation of gene knockout cell lines using CRISPR-Cas9 gene editing

ATG5 and BECN1 knockout on HEK293 cell lines was carried out by the electroporation method. Initially, six guide RNAs for ATG5 and BECN1 knockout were designed using the CHOPCHOP website. In vitro transcription of gRNA templates was carried out using a GeneArt Precision gRNA synthesis kit, followed by purification. Cas9 protein and gRNA complex were electroporated into HEK293 cells using a Neon transfection system kit (Thermo Fisher Scientific). At 48 h after electroporation, the gene cleavage efficiency of each guide RNA was measured by surveyor assay. The following gRNA was shown to have the best cutting efficiency among the tested guides: ATG5 (5′-GATCACAAGCAACTCTGGAT-3′, exon5) and BECN1 (5′-GTAGTAGGCACCTGGCTGGG-3′, exon4). Single cells were sorted into 96-well plates using a limiting dilution method, and expanded clonal cells were screened by immunoblotting with anti-ATG5 or anti-BECN1 antibodies. ATG5 and p62 knockout on HCT116 cell lines and p62 knockout on HEK293 cell lines were carried out by plasmid-based transfection of Cas9/gRNA using pX459 plasmid, as described previously33 using the following guide sequence: 5′-GATCACAAGCAACTCTGGAT-3′ (exon5) for ATG5, 5′-GCGAGGGAAAGGGCTTGCAC-3′ (exon3) and 5′-TGTAGCGTCTGCGAGGGAAA-3′ (exon3) for p62. After initial screening by immunoblotting, clear knockout was further confirmed by next-generation sequencing. The sequences of primers used for the surveyor assay and next-generation sequencing are provided in Supplementary Table 1.

Cloning and generation of stable cell lines


Full-length clones for human Lamp1, LC3, eIF4E, LDHB, TOMM20, ACTB, PSMD12, PEX3 and ATG5 were obtained from the human orfeome collection (version 8). Lysine to arginine substitution in ATG5 was carried out using standard site-directed mutagenesis. The destination pHAGE vectors containing N-terminal or C-terminal eGFP and Keima tags were prepared by Gibson assembly. Using the Gateway cloning system, lentiviral-based expression vectors containing N-terminal tandem eGFP-tagged LC3, C-terminal tandem eGFP-tagged Lamp1 and eIF4E, and N- or C-terminal tandem Keima-tagged constructs were generated. In the case of the Lamp1–eGFP construct, the linker sequence between Lamp1 and the eGFP was replaced with GGSGGS sequence by Gibson assembly to prevent possible mis-localization of the protein because the linker immediately follows the YXXI sorting motif on the C-terminal site of Lamp1.

Stable cell line generation

The 293T cells were transfected with the corresponding lentiviral expressing vectors using lipofectamine 3000 (Invitrogen). Virus-containing supernatant was collected 48 h post-transfection. HEK293 and HCT116 cell lines were infected with the virus to create stable cell lines. In the case of eGFP–LC3 and Lamp1–eGFP expressing cell lines, eGFP-positive cells were single cell sorted by flow cytometry to generate clonal cell lines. Free Keima, Keima–LC3, mito–Keima, LDHB–Keima, TOMM20–Keima, Keima–ACTB, PSMD12–Keima and PEX3–Keima expressing cell lines were first selected with puromycin (Invitrogen) for at least one week, then sorted by flow cytometry (MoFlo Astrios EQ, Beckman Coulter) twice, with an interval of two weeks, to equalize the Keima expression levels. The Keima signal level of RPS3–Keima expressing cells was used as the standard for sorting cells with similar Keima expression level.

Cell lysis and immunoblotting assay

Cells were cultured in the presence of the corresponding chemicals to ~50–60% confluency in a six-well plate, then trypsinized, pelleted by centrifugation, washed with Dulbecco's Phosphate-Buffered Saline (DPBS) and pelleted again. The cell pellet was lysed with radioimmunoprecipitation assay buffer (RIPA buffer) containing mammalian protease inhibitor cocktail (Sigma) and 20 units ml−1 benzonase (Millipore), followed by centrifugation (13,000 r.p.m.). The concentration of the supernatant was measured by Bradford assay and the whole-cell lysate was denatured by the addition of NuPAGE LDS (Lithium dodecyl sulfate) Sample Buffer (Thermo Fisher Scientific) supplemented with 100 mM dithiothreitol, followed by boiling at 75 °C for 5 min. Each lysate (40 μg) was loaded onto the 4–20% Tris-Gly gel (Thermo Fisher Scientific) or 4–12% NuPAGE Bis-Tris gel (Thermo Fisher Scientific), followed by SDS–PAGE with Tris-Glycin SDS running buffer (Thermo Fisher Scientific) or NuPAGE MES (2-(N-morpholino)ethanesulfonic acid) SDS running buffer (Thermo Fisher Scientific), respectively. The proteins were electrotransferred to polyvinylidene difluoride (PVDF) membranes (0.45 μm, Millipore) and then immunoblotted using antibodies as indicated. For quantitative immunoblotting of Keima reporter cells, 80–100 μg of total lysate was loaded onto SDS–PAGE gel due to the low expression level of the Keima construct and small amount of processed Keima, except for Keima–LC3 (35 μg). A 4–20% Tris-Gly gel was used to resolve the proteins. After transfer, the membrane was blocked with 5% non-fat milk, incubated with anti-Keima mouse antibody (4 °C, overnight), washed three times with tris-buffered saline containing 0.25% tween 20 (TBST) (total 30 min) and further incubated with fluorescent IRDye 800CW goat anti-mouse IgG H+L secondary antibody (1:20,000) for 2 h. After a severe wash with TBST for 30 min, the processed Keima band intensity (25 kDa) was measured using an OdysseyCLx imager and divided by the normalized signal of intact Keima protein. The processed Keima to intact Keima protein ratio was compared within the gel and, where quantitations obtained from independent immunoblots were compared, the Torin1 treatment condition was used as internal control. In the case of Keima–LC3 cell lines, tubulin was used to normalize the processed Keima signal because LC3 is an autophagy machinery rather than a cargo. Thus, the accumulated processed Keima level normalized by tubulin level reflects total autophagy flux.

Sucrose density gradient fractionation

Cells were grown to 60% confluency in 15 cm dishes with or without bortezomib for 4 h. The cells were then treated with cycloheximide (100 μg ml–1) for 10 min and collected after trypsin treatment. 1.5 ml of ribosome purification buffer (200 mM HEPES, 300 mM NaCl, 50 mM MgCl2, 100 μg ml–1 cycloheximide, 0.5% Triton-X, protease inhibitor tablet (Roche), 50 unit ml−1 RNasin) was added to resuspend the cells and the suspension was passed through a 23G needle over 10 times. After centrifugation for 5 min at 13,000 r.p.m., the supernatant was loaded on top of the 5–50% sucrose gradient column, prepared in house. Using an SW40 motor (Beckman Coulter), the sucrose gradient was centrifuged for 2 h at 35,000 r.p.m. (4 °C), followed by 1 h of no-break deceleration. A small hole was made at the bottom of the tube and the sample was fractioned into 1.5 ml tubes (750 μl each). RNA level was measured by Nano-drop (260 nm ultraviolet (UV) absorption) and each fraction was immunoblotted using the corresponding antibodies.

In vitro translation reactions

Cells were grown to 60% confluency in six-well plates. The cells were treated with puromycin (1 μg ml–1) for 15 min, then collected after trypsin treatment. After cell lysis using RIPA buffer, an immunoblotting assay was performed with anti-puromycin antibody.

Flow cytometry analysis

Cells were grown in the presence of the corresponding chemicals to 60% confluency in six-well plates, then collected after trypsin treatment. After resuspending the cells in FACS buffer (1× DPBS, 1 mM EDTA, 1% FBS, 25 mM HEPES, final pH 7.3–7.5), the cells were analysed by flow cytometry (MoFlo Astrios EQ, Beckman Coulter). Data were processed with FlowJo software. The 561/488 nm ratio distribution graph was processed in Prism software after exporting the 488 and 561 nm ex./em. intensities of individual cells (10,000 cells) were exported. An example of the gating strategy used for flow cytometry analysis is shown in Supplementary Fig. 5.

Calculation of approximate percentage of ribosomes undergoing ribophagy before and after Torin1 treatment

HEK293 RPS3-Keima cells were collected in pH 7.3 or 4.5 buffer and incubated for 30 min. The 561/488 nm excitation fold increase from cells in pH 7.3 buffer to pH 4.5 buffer was calculated to be 2.020 (Value_a). This condition mimics the situation of nearly 100% ribosomes undergoing ribophagy, because all the cytosolic ribosomes are at pH 4.5. The 561/488 nm excitation fold increases of control cells and Torin1 (24 h) treated cells responsive to bafilomycin A were 0.027 and 0.194, respectively (Value_b and Value_c, respectively). These values represent the lysosomal population of Ribo–Keima before and after 24 h of Torin1 treatment. The per cent ratio of (Value_b) to (Value_a) is 1.3% and the ratio of (Value_c) to (Value_a) is 10.4%, which represents the percentage of ribosomes undergoing ribophagy before and after Torin1 treatment compared to the total ribosomes.

Live-cell confocal microscopy

Cells were plated onto a 33-mm glass-bottomed dish (no. 1.5, glass diameter of 14 mm, MatTek) pretreated with poly-l-lysine, then incubated in phenol-red free medium (FluoroBrite DMEM, Thermo Fisher) containing l-glutamine (2 mM) and sodium pyruvate (1 mM) for 48 h. Followed by treatment with corresponding chemicals or HBSS, the cells were stained with 2.5 μg ml–1 Hoechst 33342 (Thermo Fisher Scientific) for 10 min. The medium was replaced with fresh FluoroBrite DMEM (37 °C) and cells were imaged using a Yokogawa CSU-X1 spinning disk confocal microscope with a Spectral Applied Research Aurora Borealis modification on a Nikon Ti motorized microscope equipped with a Nikon Plan Apo ×60/1.40 NA objective lens. Pairs of images for ratiometric analysis of mKeima fluorescence were collected sequentially using 100 mW/442 nm and 100 mW/561 nm solid-state lasers attenuated and controlled with an AOTF (Spectral Applied Research LMM-5) and emission (em.) collected with a 620/60 nm filter (Chroma Technologies). Wide-field fluorescence images of Hoeschst were collected using a Lumencor SOLA light source, 395/35 nm ex. and 480/40 nm em. filters (Chroma Technologies). Both confocal and wide-field images were acquired with the same Hamamatsu ORCA-ER cooled charge-coupled device camera and MetaMorph software. Seven z-series optical sections were collected with a step size of 0.7 μm, using the Nikon Ti internal focus motor. The z series are displayed as maximum z projections, and gamma, brightness and contrast were adjusted equally for each image using FiJi software. For the live-cell imaging of cells expressing Golgi-mEGFP, HEK293 RPS3-Keima cells were transiently transfected with CellLight Golgi–GFP (C10592, ThermoFisher Scientific) for 24 h. Cells were then incubated with Torin (150 nM) or the combination of Torin and SAR405 (150 nM and 1 μM, respectively) for 24 h before imaging. Confocal images of mEGFP fluorescence were collected using a 100 mW/488 nm solid-state laser and a 525/50 nm em. filter (Chroma Technologies).

Quantitative image analysis

Serial z-stack image sections taken with a ×60 objective lens were combined using the maximum intensity projection function in FiJi software. Without any cutting, the original images were processed as in the following sections.

Step 1: Cell segmentation

The Hoechst channel showing stained nuclei was processed as follows: (1) background subtraction, (2) autothreshold, make binary, fill holes, dilate and water shed, (3) connected components labelling using ‘find connected regions’ plug-in to make ‘marker’. The 442 nm ex./620 nm em. Keima channel was processed as follows: (1) Gaussian blur, (2) subtract background, (3) autothreshold, make binary and dilate. Using the marker controlled segmentation function in MorphoLibJ plug-in, the two processed image channels were combined to generate ‘mask’.

Step 2: Track-mate

The 442 nm ex./620 nm em. and 561 nm ex./620 nm em. channels were stacked, and Keima puncta co-localized with in 3 pixel distance were analysed using TreackMate function. This step generated a.xml file containing co-localization and puncta intensity information of the whole image plain.

Step 3. MatLab coding

Using Matlab software, the.xml file was processed in a way that any puncta with a 561/448 intensity ratio over 0.5 were assigned to the corresponding cells segmented in Step 1. This resulted in two new images showing segmented cells and numerically assigned puncta in each cell. Co-localization of Lamp1–eGFP, eGFP–LC3 and Lysotracker Green with red Keima puncta was performed by FiJi software. Any puncta with a pixel size of 7 and intensity of 7 (arbitrary number) were found in both 488 nm ex./620 nm em. and 561 nm ex./620 nm em. channels. Among the puncta that passed the intensity and size thresholds, the total puncta number that appeared in both channels within a pixel distance of 5 was divided by the total number of puncta analysed in the 488 nm ex./620 nm em. channel image. For the random co-occurrence measurement, the 488 nm ex./561 nm em. channel image was shifted 10 μm in a random direction. Co-localization analysis between the two channels was performed following the same procedure.

Electron microscopy

HEK293; wt, ATG5−/−, BECN1−/− cells were grown to 60% confluency in a six-well plate and incubated with HBSS + BafA (50 nM) for 4.5 h. The cells were then fixed with 1.25% paraformaldehyde, 2.5% glutaraldehyde and 0.03% picric acid, followed by osmication and uranyl acetate staining, dehydration in alcohols, and then embedded in Taab 812 Resin (Marivac). Sections were cut with a Leica ultracut microtome and picked up on formvar/carbon-coated copper slot grids. Immediately before imaging, the grids were stained with 2% uranyl acetate for 5 min to increase ribosome contrast, followed by 0.2% lead citrate staining. The samples were imaged under a Phillips Tecnai BioTwin Spirit transmission electron microscope.

Statistics and reproducibility

All statistical data were calculated using GraphPad Prism 7. Comparisons of data in Fig. 4b were performed by one-way analysis of variance (ANOVA) with Sidak’s multiple comparisons test. Comparisons of all other data were performed by two-way ANOVA with Tukey’s multiple comparisons test. P values <0.01 were considered significant. All experiments were repeated at least three times unless otherwise indicated.

Life Sciences Reporting Summary

Further information on experimental design is available in the Life Sciences Reporting Summary.

Code availability

The MatLab code used for unbiased image quantitation in this study is available from the corresponding author upon reasonable request.

Data availability

The source data for statistical analyses of Figs. 2b, f–h, 3b, 4b,d,e,g,l and 5e–f and Supplementary Figs. 2j, 3a and 5c,h are provided in Supplementary Table 2. All other data supporting the findings of this study are available from the corresponding author upon reasonable request.

Additional information

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.


  1. 1.

    Warner, J. R. The economics of ribosome biosynthesis in yeast. Trends Biochem. Sci. 24, 437–440 (1999).

  2. 2.

    Zhang, T., Shen, S., Qu, J. & Ghaemmaghami, S. Global analysis of cellular protein flux quantifies the selectivity of basal autophagy. Cell Rep. 14, 2426–2439 (2016).

  3. 3.

    Shao, S. & Hegde, R. S. Target selection during protein quality control. Trends Biochem. Sci. 41, 124–137 (2016).

  4. 4.

    Sung, M. K., Reitsma, J. M., Sweredoski, M. J., Hess, S. & Deshaies, R. J. Ribosomal proteins produced in excess are degraded by the ubiquitin–proteasome system. Mol. Biol. Cell 27, 2642–2652 (2016).

  5. 5.

    Sung, M. K. et al. A conserved quality-control pathway that mediates degradation of unassembled ribosomal proteins. eL ife 5, e19105 (2016).

  6. 6.

    Kraft, C., Deplazes, A., Sohrmann, M. & Peter, M. Mature ribosomes are selectively degraded upon starvation by an autophagy pathway requiring the Ubp3p/Bre5p ubiquitin protease. Nat. Cell Biol. 10, 602–610 (2008).

  7. 7.

    Ossareh-Nazari, B. et al. Cdc48 and Ufd3, new partners of the ubiquitin protease Ubp3, are required for ribophagy. EMBO Rep. 11, 548–554 (2010).

  8. 8.

    Nishida, Y. et al. Discovery of Atg5/Atg7-independent alternative macroautophagy. Nature 461, 654–658 (2009).

  9. 9.

    Tsuboyama, K. et al. The ATG conjugation systems are important for degradation of the inner autophagosomal membrane. Science 354, 1036–1041 (2016).

  10. 10.

    Anderson, P. & Kedersha, N. RNA granules. J. Cell Biol. 172, 803–808 (2006).

  11. 11.

    Santaguida, S., Vasile, E., White, E. & Amon, A. Aneuploidy-induced cellular stresses limit autophagic degradation. Gene. Dev. 29, 2010–2021 (2015).

  12. 12.

    Galluzzi, L. et al. Molecular definitions of autophagy and related processes. EMBO J. 36, 1811–1836 (2017).

  13. 13.

    Anding, A. L. & Baehrecke, E. H. Cleaning house: selective autophagy of organelles. Dev. Cell 41, 10–22 (2017).

  14. 14.

    Kishi-Itakura, C., Koyama-Honda, I., Itakura, E. & Mizushima, N. Ultrastructural analysis of autophagosome organization using mammalian autophagy-deficient cells. J. Cell Sci. 127, 4089–4102 (2014).

  15. 15.

    Katayama, H., Kogure, T., Mizushima, N., Yoshimori, T. & Miyawaki, A. A sensitive and quantitative technique for detecting autophagic events based on lysosomal delivery. Chem. Biol. 18, 1042–1052 (2011).

  16. 16.

    Mizushima, N., Yoshimori, T. & Levine, B. Methods in mammalian autophagy research. Cell 140, 313–326 (2010).

  17. 17.

    Ni, H. M. et al. Dissecting the dynamic turnover of GFP-LC3 in the autolysosome. Autophagy 7, 188–204 (2011).

  18. 18.

    Gross, L. A., Baird, G. S., Hoffman, R. C., Baldridge, K. K. & Tsien, R. Y. The structure of the chromophore within DsRed, a red fluorescent protein from coral. Proc. Natl Acad. Sci. USA 97, 11990–11995 (2000).

  19. 19.

    Chan, E. Y., Kir, S. & Tooze, S. A. siRNA screening of the kinome identifies ULK1 as a multidomain modulator of autophagy. J. Biol. Chem. 282, 25464–25474 (2007).

  20. 20.

    Hurley, J. H. & Young, L. N. Mechanisms of autophagy initiation. Annu. Rev. Biochem. 86, 225–244 (2017).

  21. 21.

    Ktistakis, N. T. & Tooze, S. A. Digesting the expanding mechanisms of autophagy. Trends Cell Biol. 26, 624–635 (2016).

  22. 22.

    Anderson, D. J. et al. Targeting the AAA ATPase p97 as an approach to treat cancer through disruption of protein homeostasis. Cancer Cell 28, 653–665 (2015).

  23. 23.

    Liu, J. et al. Beclin1 controls the levels of p53 by regulating the deubiquitination activity of USP10 and USP13. Cell 147, 223–234 (2011).

  24. 24.

    Harper, J. W. & Bennett, E. J. Proteome complexity and the forces that drive proteome imbalance. Nature 537, 328–338 (2016).

  25. 25.

    Hewitt, L. et al. Sustained Mps1 activity is required in mitosis to recruit O-Mad2 to the Mad1-C-Mad2 core complex. J. Cell Biol. 190, 25–34 (2010).

  26. 26.

    Santaguida, S., Tighe, A., D’Alise, A. M., Taylor, S. S. & Musacchio, A. Dissecting the role of MPS1 in chromosome biorientation and the spindle checkpoint through the small molecule inhibitor reversine. J. Cell Biol. 190, 73–87 (2010).

  27. 27.

    Panas, M. D., Ivanov, P. & Anderson, P. Mechanistic insights into mammalian stress granule dynamics. J. Cell Biol. 215, 313–323 (2016).

  28. 28.

    Buchan, J. R., Kolaitis, R. M., Taylor, J. P. & Parker, R. Eukaryotic stress granules are cleared by autophagy and Cdc48/VCP function. Cell 153, 1461–1474 (2013).

  29. 29.

    Ichimura, Y. et al. Phosphorylation of p62 activates the Keap1-Nrf2 pathway during selective autophagy. Mol. Cell 51, 618–631 (2013).

  30. 30.

    Stolz, A., Ernst, A. & Dikic, I. Cargo recognition and trafficking in selective autophagy. Nat. Cell Biol. 16, 495–501 (2014).

  31. 31.

    Higgins, R. et al. The unfolded protein response triggers site-specific regulatory ubiquitylation of 40S ribosomal proteins. Mol. Cell 59, 35–49 (2015).

  32. 32.

    Yoshii, S. R., Kishi, C., Ishihara, N. & Mizushima, N. Parkin mediates proteasome-dependent protein degradation and rupture of the outer mitochondrial membrane. J. Biol. Chem. 286, 19630–19640 (2011).

  33. 33.

    Ran, F. A. et al. Genome engineering using the CRISPR-Cas9 system. Nat. Protoc. 8, 2281–2308 (2013).

Download references


This work was supported by the National Institutes of Health (grants R37NS083524 and RO1GM095567 to J.W.H.). The authors acknowledge the Nikon Imaging Center and the Imaging and Data Analysis Core (Harvard Medical School) for imaging assistance.

Author information


  1. Department of Cell Biology, Harvard Medical School, Boston, MA, USA

    • Heeseon An
    •  & J. Wade Harper


  1. Search for Heeseon An in:

  2. Search for J. Wade Harper in:


J.W.H. and H.A. conceived the study. H.A. performed all experiments. H.A. and J.W.H. analysed the data and wrote the paper.

Competing Interests

The authors declare no competing financial interests.

Corresponding author

Correspondence to J. Wade Harper.

Supplementary information

  1. Supplementary Information

    Supplementary Figures 1–6, Legends

  2. Life Sciences Reporting Summary

  3. Supplementary Table 1

    Sequences of guide RNAs used for CRISPR tagging and knock-out, and primers for genotyping and next-generation sequencing.

  4. Supplementary Table 2

    Statistics Source Data. The source data for statistical analyses of Figs. 2b, 2f-h, 3b, 4b, 4d-e, 4g, 4l, 5e-f, and Supplemental Figs. 2j, 3a, 5c, and 5h.

  5. Supplementary Table 3

    Information of antibodies used in the study.

About this article

Publication history






Further reading