Endosomal microautophagy is activated by specific cellular stresses in trout hepatocytes

These results suggest that mild-oxidative stress in RT cells induces the targeting of KFERQ-containing proteins to LE in a process similar to eMI in flies and mammals⁶.

To next evaluate whether the KFERQ- and ESCRT-I-dependent but CMA-independent eMI-like puncta might also rely on MA, we designed a Mo to inhibit the translation of mRNAs encoding the E1-like enzyme Atg7 in the RT. This approach aims to impair MA initiation⁴⁰. First, we found that the reduction of Atg7 by Moatg7 led to a significant reduction in the conjugation of Atg5 to Atg12 (Supplementary Fig. 2G and H), a key process in MA initiation recognized to be dependent on the activity of Atg7⁴¹. Besides, we assessed LC3 lipidation as a readout of the MA flux after exposing cells to H₂O₂, with or without the autophagosome-lysosome fusion inhibitor chloroquine (CQ, at 10 µM)⁴². Moatg7 markedly reduced LC3-II levels compared to cells transfected with the negative control oligo (MoSTD–) (Fig. 3G; quantification in Fig. 3H). Together with the loss of Atg12–Atg5 complex formation, this confirmed effective MA inhibition. We then examined Venus-puncta formation under mild-oxidative stress and observed that MA inhibition did not decrease the number of eMI-sensor puncta (Fig. 3I, quantification in Fig. 3J). Instead, it increased the puncta count compared to MoSTD-, similar to findings in Drosophila fat body³⁶. This suggests that increased cellular stress might enhance eMI-like process in RT as well.

Together, our results demonstrate that the formation of the KFERQ-Venus puncta upon H₂O₂ incubation specifically relies on ESCRT-I machinery but is indeed independent of both CMA and MA activities, supporting -for the first time- the existence of an eMI-like process in fish. Interestingly, we have observed that the activity of such process in RT increases when either CMA or MA are impaired, which points to an established and probably finely tuned crosstalk between these autophagy pathways to maintain cellular proteome homeostasis. This interaction certainly warrants further research.

We then examined the effect of genotoxic insults by inducing DNA double-strand breaks with etoposide (50 µM)⁴⁵. Our results showed that etoposide clearly increased Venus puncta compared to CT (Fig. 4A, quantification in Fig. 4B), similarly to findings in Drosophila³⁶ and mammals³⁷. In mammals, CMA participates in the selective degradation of the checkpoint kinase 1 (Chk1), a KFERQ-motif containing protein, to ensure cell cycle progression after DNA repair⁴⁶. Conversely, CMA impairment was associated with defects in nuclear proteostasis, genomic instability and reduced cell survival⁴⁶. Although it remains to be determined whether Chk11 can also be targeted by eMI, these findings suggest a conserved protective mechanism that maintains genome integrity through the degradation of KFERQ-substrates after DNA damage, which could be, at least partly, fulfilled by eMI in RT.

Finally, we observed that serum starvation (-FBS), known to induce both CMA^3,39,47 and MA in various species, including RT^48–50, did not alter Venus-puncta formation. In contrast, combined serum and amino acids deprivation (-FBS -AA) with HBSS medium for 16 h substantially increased eMI-like puncta compared to CT (Fig. 4A, quantification in Fig. 4B). This aligns with increased eMI during prolonged nutrient deprivation in the CMA-deficient species Drosophila⁴³. In mammals, where both CMA and eMI pathways are present, the regulation of eMI during nutrient deprivation is more complex²⁸. Short-time (1 h) HBSS exposure triggers an ESCRT-III-dependent, but ESCRT-I and Hsc70-independent, eMI variant in human cell lines. This variant likely degrades selective autophagy receptors (e.g., LC3B and GABARAPL2), preventing unwarranted selective MA in the short-term, while enhancing bulk MA during prolonged starvation⁵¹. Conversely, long-term nutrient starvation, a well-known CMA activator³¹, inhibits eMI in rodents, likely due to the CMA-mediated degradation of the eMI component BAG6⁹. This research highlights the possible role of BAG6 as a switcher, directing KFERQ-containing proteins to either eMI or CMA and enabling these pathways to compensate for each other's deficiencies. In RT, although the silencing of bag6 appears to inhibit the eMI process similar to what was observed in mice⁹, both autophagic pathways function simultaneously during stress, such as mild-oxidative stress and nutrient deprivation (16 h), targeting KFERQ-containing proteins to lysosomes and LE/MVB. This raises the question of whether a Bag6-dependent mechanism coordinates both eMI and CMA pathways in RT as observed in mammals. However, the differences between our results and those in mammals may stem from the experimental models used. In our in vitro model, -FBS -AA represents an acute condition with negligible extracellular levels of these compounds, while mammalian studies involved in vivo nutrient starvation experiments⁹, where circulating nutrient and growth factor levels likely remain physiological. To the best of our knowledge, no data exists on the in vitro activity of eMI in mammals during both long-term and acute nutrient deprivation. Therefore, we cannot ruled out that the simultaneous activation of CMA and eMI observed in the present study may be a conserved physiological response regarding extreme cellular stress across vertebrates.

In summary, we identified and characterized an eMI-like pathway in a RT hepatocyte cell line. This process is induced by oxidative stress, DNA damage, or nutrient deprivation, mirroring findings in Drosophila^36,43, but not in response to FBS removal, supporting that eMI induction in RT is stimulus-specific. These findings suggest a significant role for this eMI-like process in fine-tuning the cellular proteome during stress in RT. Moreover, they open new avenues for exploring this function using complementary model organisms, such as zebrafish and RT.

Minimum Essential Medium (61,100-053), essential (11,130-036) and non-essential amino acids (11,140–050), L-glutamine (25,030-024), sodium pyruvate (11,360-070), penicillin–streptomycin (15,140-122), geneticin (11,811,031), neomycin (11,598,906), puromycin (10,296,974), fetal bovine serum (10,270-106), and Hank's Balanced Salt Solution (14,025-092), were all purchased from Gibco. HEPES (BP299-1) and PBS (BP2944) were procured by Fisher BioReagents. Hydrogen peroxide solution (H1009), D-(+)-Glucose (G7021) and chloroquine (C6628) were all purchased from Sigma-Aldrich. Etoposide (HY-13629) was purchased from MedChemExpress, paraformaldehyde 4% fixative (22,023) from Biotium, and the antifade mounting medium containing DAPI (H-2000) from Vector Laboratories, Inc. The TRIzol Reagent (15,596,018), RIPA buffer (89,901), the cocktail of protease and phosphatases inhibitors (78,422), the Qubit protein assay kit (Q33211), SuperBlock Blocking Buffer in PBS (37,515), SuperSignal West Pico Plus Chemiluminescent Substrate (34,578), and No-Stain Protein Labeling Reagent (A44449), were all acquired from ThermoFisher Scientific. All laboratory plastic ware, unless otherwise stated, were procured by Sarstedt AG & Co. KG.

Ensembl Genome database (Ensembl release 112; http://www.ensembl.org↗) was used to identify orthologous genes to the human genes coding for the main eMI players in the genome of RT (Genome assembly: USDA_OmykA_1.1), or of Danio rerio (Genome assembly: GRCz11), and their identity was later confirmed using Genomicus version 110.01⁵². To screen whether the identified genes are expressed across different tissues of RT and zebrafish, we interrogated the PhyloFish RNA-seq database³⁵. Instead, to evaluate their expression in RTH-149 cells, we performed a Bulk RNA barcoding and sequencing (BRB-seq) analysis⁵³ using the services of Alithea Genomics S.A as explained below.

Regarding oligos, up to four new siRNAs against the RT tsg101 mRNA sequence (Ensembl ID-gene: ENSOMYG00000006944) were designed using the siDESIGN Center. After testing their efficiency in reducing both mRNA and protein levels of Tsg101 by qPCR and western blot (data not shown), the most effective was selected for following experiments (sitsg101; 5′-UGAUUAGGUGAGCGAAUUA-3′). The different siRNAs using in the present work, including the siRNAs against the two RT lamp2a's (i.e., si14; 5′-UAAAGAAAUUGCUCGGCUC-3′, and si31; 5′- UGGAGAAGCGGCUGUGUUA-3′, previously validated for RT²⁰, and a negative control siRNA against LucGL2 (siLUC; 5′-CGUACGCGGAAUACUUCGA-3′), not expressed in RT cells, were purchased from Horizon Discovery. On the other hand, a morpholino oligonucleotide (Moatg7; 5′-CAGAAGCCATCTCTATCACCG-3′) was designed in collaboration with GeneTools customer support to targeting the start codon of the RT atg7 mRNA sequence (Ensembl ID-gene: ENSOMYG00000075941) for blocking its translation. Besides, two different splice junction-targeting morpholinos were designed, each directed against pairs of RT bag6 paralogs. The first targeted the paralogs located on chromosomes 2 and 3 (Ensembl gene IDs: ENSOMYG00000013461 and ENSOMYG00000018793; designated Mobag6_chr2^&_chr3; sequence: 5′-CAGACAGCATTACTCACTGTAC-3′). The second targeted the paralogs located on chromosomes 18 and 32 (Ensembl gene IDs: ENSOMYG00000034772 and ENSOMYG00000023282; designated Mobag6_chr18^&_chr32; sequence: 5′- TAGTTGCTGGCTTTTTTACTCAC-3). A standard negative control morpholino (MoSTD-; sequence: 5′-CCTCTTACCTCAGTTACAATTTATA-3′), targeting against an intronic region of the human β-globin gene associated with β-thalassemia, was also designed. All morpholinos were purchased from GeneTools.

Concerning plasmids, two constructions originally designed for tracking eMI in Drosophila³³ and based on the backbone pUAST-attB (RRID:DGRC_1419) plus the 21^st amino acids of RNASe A, including its KFERQ motif, fused to either the N- or C-Split Venus⁵⁴, were gifted by Patrik Verstreken (KU Leuven). Those two plasmids were digested to isolate the N- and C-Split Venus regions tagged to the KFERQ motif, for their posterior sub-cloning in the backbone of the pEGFP-N1-ro1GFP [a gift from S. James Remington (Addgene plasmid #82369; http://n2t.net/addgene:82369↗; RRID:Addgene_82369)] for a suitable expression in RT cells, but only after the removal of the EGFP sequence. The KFEAA-N-Venus was synthetized by Genewiz (Azenta Life Sciences) and inserted in the same pEGFP-N1-ro1GFP backbone detailed before. The KFEAA-C-Venus was also synthetized by Genewiz but inserted in the pCMV6-G2-ATP8-V5 plasmid [a gift from Matthew O'Connor (Addgene plasmid # 86852; http://n2t.net/addgene:86852↗ ; RRID:Addgene_86852). Moreover, mRFP-Rab7 plasmid was a gift from Ari Helenius (Addgene plasmid #14436; http://n2t.net/addgene:14436↗; RRID:Addgene_14436).

The RT hepatoma-149 cell line (RTH-149 cells; supplied by ATCC #CRL-1710; RRID:CVCL_3467), passage number below 20, and routinely screened to confirm absence of mycoplasma contamination by using the Myco-Sniff Mycoplasma PCR Detection Kit (MP Biomedicals #093050201) following the manufacturer's recommendations, was used for all cell culture experiments in the current work. This cell line has been previously validated as a reliable model for nutrition and autophagy research in the RT^20,49. Cells were grown at 18 °C without CO₂ in complete medium (hereafter referred as control condition, CT) consisting in Minimum Essential Medium containing essential amino acids, 1X L-glutamine and glucose (5 mM), and supplemented with 1X non-essential amino acids, sodium pyruvate (1 mM), HEPES (25 mM), 1X Penicillin–Streptomycin, and 10% fetal bovine serum.

RTH-149 cells stably transfected with the two N- and C-Split Venus regions tagged to the KFERQ motif were selected for the resistance to the selective antibiotic geneticin, followed by cell sorting using a FACS Aria 2-Blue 6-Violet 3-Red 5-YelGr, 2 UV laser configuration (BD Biosciences, Le Pont de Claix Cedex, France) in biosafety cabinet. RTH-149 cells transfected to stably express the mutated KFERQ motifs (i.e., KFEAA-N-Venus and KFEAA-C-Venus) were selected for their resistance to both the selective antibiotics neomycin and puromycin. On the other hand, RTH-149 cells stably expressing a KFERQ-PA-mCherry1 generated and selected in our previous work²⁰ were additionally used for tracking CMA. Those eMI- or CMA-reporter cells, respectively, were maintained in a version of the CT medium detailed above in which the antibiotic used was geneticin. All the cells transfections with either plasmids (1–2.5 µg), siRNAs (1 µM) or morpholino oligos (5 µM), were performed with the Transfection System device (Invitrogen MPK5000) and Neon 100 µL samples Kit (Invitrogen MPK10096).

At the beginning of each experiment, similar to what was previously detailed in Vélez et al.²⁰, cells were counted using a Cellometer K2 device (Bioscience LLC, Lawrence, MA, USA) and plated at a density of 60.000 cells/well onto 4-well culture slides (Ibidi 80426 or FALCON 354114) for Venus or mCherry microscope imaging experiments, respectively. In the case ofRNA extraction and both protein collection and IEM experiments, cells were plated in 6-cm dishes at 400.000 or 500.000 cells/dish, respectively. Before experimental treatments, the cells were gently washed twice with 1X PBS, and then incubated with CT medium alone, or supplemented with either hydrogen peroxide (25 μM, H₂O₂ medium, mild-oxidative stress), High-Glucose (D-glucose at 25 mM), or etoposide (50 µM), as appropriate and for the times detailed elsewhere. Otherwise, cells were incubated with H₂O₂ medium supplemented with chloroquine (CQ condition, 10 µM), or instead, with 1X Hank's Balanced Salt Solution supplemented exclusively with HEPES (25 mM, HBSS condition, serum and nutrient starvation medium), or also with 1X essential amino acids, 1X non-essential amino acids, and 1X L-glutamine (-FBS condition, serum starvation medium). To asses CMA activity, the CMA-reporter expressing cells were photoactivated with a 405-nm light source for 10 min prior to starting the incubations²⁰.

Cells were fixed for 20 min using paraformaldehyde 4% before mounting and counterstaining nuclei with DAPI, and then the images were acquired in the Bordeaux Imaging Center (CNRS-INSERM and Bordeaux University, member of the national infrastructure France BioImaging). On the one hand, all cells images of KFERQ-Venus and mRFP-Rab7 fluorescence were acquired in widefield using a 20X/0.95 dry objective and a 1X intermediate lens with the high-content screening microscope Zeiss Cell Discoverer 7 (Zeiss Microscopy, Germany), which is equipped with Dual Zeiss Colibri LED-based illumination (from 385 to 625 nm), the dichroic mirrors 405/493/575/653, 450/538/610, 405/493/610, the emission filters 412-438/501-527/583-601/662-756, 458-474/546-564/618-756, 412-433/501-547/617-758, the color camera Zeiss Axiocam 512 as a detector, and its controlled by the Zen Blue 3.1 software (Zeiss). On the other hand, as reported before²⁰, for imaging CMA-reporter cells we used a widefield Microscope Leica DM 5000 (Leica Microsystems, Germany) equipped with a 40X/1.25 oil objective, DAPI and TRITC filter cubes, a Greyscale sCMOS camera (Hamamatsu Flash 4.0 V2, Japan), and controlled by MetaMorph v7.8.10.0 software (RRID:SCR_002368).

eMI-reporter puncta per image were automatically detected in the Venus-channel and counted with Fiji (RRID:SCR_002285) by using the ComDet v.0.5.5 plugin (developed in Cell Biology group of Utrecht University and available at https://github.com/UU-cellbiology/ComDet↗), with an approximate particle size of 4 pixels and an intensity threshold ranging from 5 to 20 SD as settings. Next, the number of nucleus (> 100 pixels) in each image were counted by using the Fiji Analyze Particles function in the DAPI channel after IsoData thresholding and watershed segmentation. In the case of PA-mCherry1 fluorescence experiments, the CMA puncta were counted on images in gray (red channel) and blue (DAPI) by using the Cell Counter plugin of Fiji (RRID:SCR_002285) as previously detailed²⁰, followed by nuclei counting. All images were counted in blind and random order and excluding cells on the edges or very superposed. For each image, the number of either eMI- or CMA-puncta was normalized by the number of nuclei within the image, and the results showed represent the number of puncta per cell in a minimum of 60 images coming from 3 independent experiments. In all cases, more than 600 cells/condition were counted.

The BIOP version (https://github.com/BIOP/ijp-jacop-b↗) of the original JACoP plugin (RRID:SCR_025164) for Fiji was used to evaluate the co-ocurrence between Venus-puncta and late endosomes labeled with mRFP-Rab7 in cells incubated with H₂O₂ medium. We analyzed in 85 single cells the PCC, which ranges from − 1 to 1, where − 1 indicates a negative correlation and + 1 highlights a complete positive correlation, between the pixel-intensity of the two contrasted channels³⁸.

For IEM, cells were fixed for 1 h using a solution of paraformaldehyde 2% and glutaraldehyde 0.1% in PBS. After chemical fixation, cells were scrapped and pelleted in 12% gelatin. Small pieces were cut and incubated in 2.3M sucrose overnight at room temperature. Each sample were immersed in liquid nitrogen, and then sliced in ultrathin Sect. (70 nm) at − 120 °C with a 35° angle diamond knife (Diatome®) and collected on nickel grid (Electron Microscopy Sciences). Immunogold labeling of Venus proteins were achieved on section with a goat anti-Venus polyclonal first antibody (Antibodies.com, A121679), followed by secondary antibody donkey-anti-goat IgG (H&L) coupled with a 10 nm gold particules (Aurion). At the end, contrasts were performed with a solution of 1% aqueous uranyl acetate in 2% methylcellulose for 10 min in ice in the dark. Sections were imaged using a Hitachi H7650 transmission electron microscope equipped with a Gatan Orius CCD camera, as previously described⁵⁵. MVBs were identified using previously established criteria^56,57.

RNA samples were collected and processed with TRIzol reagent following the manufacturer's protocol. Then, RNA concentration was determined using a NanoDrop 2000 (Thermo Scientific) and integrity confirmed in an agarose gel, as previously reported⁵⁸. In the case of Bulk RNA barcoding and sequencing (BRB-seq) analysis⁵³, 8 independent RNA samples (50 ng of RNA/each) were collected and isolated from RTH-149 cells cultured in growth medium and successively sent to Alithea Genomics S.A for its processing as explained in detail elsewhere²². On the other hand, for quantitative RT-PCR the samples were prepared as described before^49,55 and cDNA synthesis was performed in duplicate, followed by duplicated PCR analysis²⁰. The relative quantification of the target gene tsg101 and of all bag6 paralogs was calculated using the ΔCT method⁵⁹ with eef1a1 (eukaryotic translation elongation factor 1 alpha 1) as a reference gene. Primers used for tsg101 (forward: 5′-CCAGTCAGCTACAGAGGAAACA-3′/reverse: 5′-AGATCTTGCCGTTGGCATCA-3′) and for all bag6 paralogs (forward: 5′-CTGCAGGATGAGAGGAC-3′/reverse: 5′-TGCTGGTGCTGACTCTG-3′) were designed in the current work by using the Primer-Blast tool of NCBI, and the amplicons were sent to Genewiz (Azenta Life Sciences) to confirm primer specificity by Sanger Sequencing. Primers used for eef1a1 (forward: 5′-TCCTCTTGGTCGTTTCGCTG-3′/reverse: 5′-ACCCGAGGGACATCCTGTG-3′) were previously validated⁶⁰.

Proteins were collected with RIPA buffer supplemented with protease and phosphatases inhibitor⁴⁹, and then quantified using a Qubit 2.0 fluorometer (ThermoFisher Scientific). Later, 5 µg of protein were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and then proteins were transferred to polyvinylidene fluoride (PVDF) membranes (Merk-Millipore IPFL00010). Next, total protein labeling was performed with No-Stain Protein Labeling Reagent following the manufacturer's protocol for posterior accurate total protein normalization. After blocking membranes with SuperBlock Blocking Buffer in PBS, they were incubated with the corresponding primary antibody rabbit anti-TSG101 (Sigma-Aldrich HPA006161, at 1/500), rabbit anti-LC3B (Cell Signaling Technologies 2775, at 1/1000), or rabbit anti-ATG5 (Novus, NB110-53818), followed by incubation with a goat anti-rabbit secondary antibody HRP (Invitrogen 31460, at 1/10,000). Immunoreactive bands were developed thanks to the SuperSignal West Pico Plus Chemiluminescent Substrate, and imaged using Smart Exposure to assure signal linearity in an iBright FL1500 Imaging System (ThermoFisher Scientific). Finally, the images were quantified using the iBright Analysis Software (ThermoFisher Scientific, RRID:SCR_017632). Full-length blots are available in the supplementary Fig. 3.

All statistical analyses in the current study were performed using GraphPad Prism (RRID:SCR_002798) version 8.0.1 for Windows (GraphPad Software, Inc., www.graphpad.com↗) considering a p value < 0.05 as a level of significance, and all data presented are reported as means ± SEM or fold induction respect the control group ± SEM, from a minimum of 3 independent experiments. Before performing parametric or non-parametric tests, Normal distribution was assessed. The differences between two groups were evaluated through a parametric two-tailed unpaired Student's T-test, or a non-parametric Mann Whitney test. Instead, for the comparison between more than two groups in Fig. 2B, Kruskal–Wallis followed by Dunn's multiple comparisons test was utilized.