Defective chaperone-mediated autophagy in the retinal pigment epithelium of age-related macular degeneration patients

Analysis of mRNA levels of different LAMPs in publicly available healthy human retina scRNA-seq datasets (GSE142449↗) (Voigt et al, 2020) revealed that, similarly to its murine counterpart (Notomi et al, 2019), LAMP2 expression in human retina is markedly higher in the RPE and glial cells (Fig. EV1 and Dataset EV1). The network of effectors, positive and negative modulators of CMA, has been extensively characterized and their expression levels can be used to infer CMA activity (CMA score) by performing a directed, weighted average (Bourdenx et al, 2021). CMA score showed a trend towards a reduction selectively in the RPE (RPE65⁺BEST1⁺ cells) of patients with AMD (Fig. 1B) compared to other cell types within the subretinal space (Fig. EV2A) in a publicly available RPE/choroid scRNA-seq dataset (GSE135922↗). Supporting a putative impairment of CMA, proteomic analyses revealed that abundance of KFERQ-containing proteins was increased in both the RPE of patients with early AMD (MGS2) as well as within drusen (Crabb et al, 2002), when compared to healthy patients (MGS1) and the reference human proteome (Fig. 1C). Furthermore, there was a significant enrichment on literature-validated CMA protein substrates (Dataset EV2) in the RPE of AMD donors (Fig. 1D and Table EV1). Finally, we also observed a significant decrease in the protein levels of both LAMP-2A and HSC70 in the RPE of histological slides from AMD patients (Figs. 1E and EV3).

CMA impairment was limited to the RPE as no major changes were observed in the neuroretina of AMD patients (Fig. EV2B; GSE135092↗) or in donors stratified according to MGS guidelines (Fig. EV2C; GSE115828↗). We have previously described that in several tissues and cell types, including the retina, CMA is modulated at the transcriptional level by the RARα transcription factor, and that the inhibitory effect of RARα on CMA can be reversed by enhancing its interaction with the co-repressor N-CoR1 (Gomez-Sintes et al, 2022). However, in the case of the neuroretina, we did not observe any changes in the protein levels of LAMP-2A or N-CoR1/RARα ratio, by immunoblotting (Fig. EV2D,E) or immunofluorescence (Fig. EV2F). While dysfunctional macroautophagy has been described in both neuroretina and RPE (Golestaneh et al, 2017; Mitter et al, 2014), herein we demonstrate for the first time selective impairment of CMA in the RPE that could be responsible for the selective vulnerability of this cell type throughout AMD progression and the cause of other cellular alterations unique to these cells such as impaired mitochondrial function (Ferrington et al, 2017) or aberrant ROS production (Datta et al, 2017).

In order to validate these findings and investigate their degradation by CMA, we compared the levels of representative proteins in the presence of a specific macroautophagy inhibitor (10 μM MRT68921) or total lysosomal proteolysis inhibition (N/L) as previously described (Juste and Cuervo, 2019) (Figs. 3D,E and EV4F). HK2 (Hexokinase 2) initiates glycolysis by phosphorylating glucose to generate glucose-6-phosphate, it has previously been shown that it can undergo CMA (Xia et al, 2015) and this was also the case in healthy iPSC-RPE (Fig. 3E,F). Surprisingly, we did not observe N/L-induced accumulation of GAPDH (glyceraldehyde 3-phosphate dehydrogenase), traditionally used to monitor CMA, highlighting the cell-type specificity of CMA substrates (Kaushik and Cuervo, 2018). As previously mentioned, the healthy and AMD donor-derived cells shared a pool of CMA substrates with a high representation of cholesterol metabolism (Fig. 3C). We observed similar levels of CMA of SQLE (Squalene monooxygenase), that catalyzes cholesterol biosynthesis, in both groups, although the fraction of lysosomal degradation insensitive to MRT inhibition was markedly lower in AMD, supporting reduced CMA in this group (Fig. 3E,F). Interestingly, using the recently developed KFERQ-Finder tool (Kirchner et al, 2019) we identified two putative acetylation-generated KFERQ-like motifs in the sequence of SQLE (Fig. 3E). Considering the high levels of lipid peroxidation (Fig. 2K) and representation of ferroptosis-related proteins observed in ORA analysis (Fig. 3C), we assessed CMA of ACSL4 (Acyl-CoA Synthetase Long chain family member 4). Lipid remodeling mediated by ACSL4 can propel ferroptosis execution (Doll et al, 2017). It has been previously validated as a CMA substrate (Liu et al, 2022), and in accordance, we found that cellular levels of ACSL4 were markedly elevated and lysosomal degradation of ACSL4 was stalled in iPSC-RPE from donors with AMD (Fig. 3E,F).

In addition, we interrogated whether macroautophagy and the ubiquitin-proteasome system were being upregulated in response to CMA deficiency in AMD iPSC-RPE, as previously reported in other cell types (Massey et al, 2006). First, we measured macroautophagy flux (LC3-II_N/L/LC3-II_Control) and observed no differences between healthy and AMD iPSC-RPE (Figs. 3E and EV6A). No alterations were observed neither in lysosomal function, evaluated using fluorogenic Cathepsin B substrate (Fig. EV6B), nor in acidic lysosomal mass (Fig. EV6C). To obtain subcellular resolution and detect more subtle alterations, we performed immunofluorescence staining of autophagic cargo (p62, magenta), autophagic vacuoles (LC3, cyan), or endolysosomes (LAMP-1, yellow) and analyzed single-, double-, or triple-positive vesicles (Fig. EV6D). No differences were detected in any of the parameters analyzed (Fig. EV6E–K). In addition, we also performed a recently developed targeted proteomics approach that offers robust quantification of autophagy-related proteins, including several selective autophagy receptors (SAR) (Leytens et al, 2025). Healthy and AMD iPSC-RPE showed similar flux of ATG8-family proteins (GABARAPL1), mitophagy receptors (BNIP3) and selective autophagy adaptors (CALCOCO2/NDP52, SQSTM1/p62, TAX1BP1) upon lysosomal degradation inhibition with N/L (Fig. EV6L), indicating that macroautophagy is indeed fully functional in both backgrounds. Finally, we also measured the levels of K48-linked poly-ubiquitinated proteins, the canonical targeting signal for degradation via ubiquitin–proteasome system, but observed no differences between healthy and AMD iPSC-RPE (Fig. EV6M).

Our findings suggest a selective impairment of CMA in iPSC-RPE derived from donors with AMD, phenocopying our initial observations in AMD donor tissue (Fig. 1). Accumulation of specific proteins or rerouting of others to alternative protein disposal pathways also play a role in RPE dysfunction during AMD etiopathogenesis. Notably, no compensatory upregulation or crucial alterations were observed in macroautophagy or ubiquitin-proteasome, highlighting CMA deficiency as the main driver of the observed changes in lysosomal protein degradation and of the subsequent proteotoxicity in AMD iPSC-RPE.

In conclusion, pharmacological activation of CMA with CA77.1 restores proteostasis, improves mitochondrial function, and decreases oxidative stress in iPSC-RPE derived from donors with AMD.

Donor data for publicly available transcriptomic and proteomics data are provided in Dataset . Formalin-fixed and paraffin-embedded (FFPE) macular sections from anonymous donors were obtained from Lion's Gift of Sight Eye Bank (Minnesota, USA). De-identified samples were obtained with informed consent of the donor or donor's family and in accordance with the WMA's Declaration of Helsinki and the Department of Health and Human Services Belmont Report. The Lion's Gift of Sight Eye Bank is licensed by the Eye Bank Association of America (Accreditation #0015204) and accredited by the FDA (FDA Established Identifier 3000718538). De-identified donor tissue is exempt from the process of Institutional Review Board approval. Donors were classified as Healthy or AMD by a trained ophthalmologist, and additional clinical observations are included in Table . EV1 EV1

Conjunctiva from de-identified donors was similarly obtained from Lion's Gift of Sight Eye Bank (Minnesota, USA). Donors were classified as Healthy (MGS1) or AMD (MGS2-3) by a Board-Certified Ophthalmologist, and additional clinical observations, including the Minnesota Grading System (MGS) score, are included in Table EV2. Reprogramming of conjunctival epithelial cells into hiPSCs was performed using the CytoTune™ 2.0 Sendai Reprogramming Kit (A16517, Thermo Fisher) following the manufacturer's instructions as previously described (Geng et al, 2017).

Cell culture plates were coated with Matrigel (356234, Corning). From P0-P2, cells were maintained in TheraPEAK X-VIVO-10 serum-free cell culture medium (BP04-743Q, Lonza) containing 1× Pen/strep antibiotics (15070-063, Gibco) and supplemented with 10 μM Y-27632 ROCK inhibitor (S1049, SelleckChem) for the first week or with 10 mM Nicotinamide (72340, Merck) for the rest. For >P2, cells were maintained in MEMα+GlutaMAX (32561, Gibco) containing 1× Pen/Strep, 1× N1 (N6530, Merck), 1× non-essential amino acids (11140050, Gibco), 250 mg/L taurine (T0625, Merck), 55 nM hydrocortisone (H6909, Merck) and 6.5 ng/L triiodo-L-thyronine (T5516, Merck) supplemented with 5% FBS (S11150H, R&D Systems) and 10 μM Y-27632 for the first week or with 2% FBS and 10 mM Nicotinamide for the remaining time in culture. Cells were passaged every 4 weeks using Accumax (00-4666-56, Thermo Fisher), and the medium was changed twice a week. All experiments were performed at P3-P4 and at least 2 weeks after the last passage. For assays analyzing mitochondrial function or metabolism, cells were cultured without nicotinamide.

To physiologically induce autophagy, cells were subjected to starvation in MEMα+GlutaMAX without any supplements or FBS for 24 h. To analyze the contribution of different pathways to substrate protein degradation, cells were treated with 10 μM MRT68921 (S7949, SelleckChem) or 100 μM Leupeptin (L2884, Merck) and 20 mM NH₄Cl (A9434, Merck) for 6 h. To pharmacologically induce CMA, cells were treated with 10 μM CA77.1 (SML3197, Merck) for 24 h as previously reported (Gomez-Sintes et al, 2022).

Cells were routinely tested for mycoplasma contamination, and iPSC-RPE cell identity was confirmed based on cobblestone-like morphology and expression of prototypic RPE proteins (Fig. EV4A,B).

FFPE slides (4 μm thickness) were generated by a trained technician at Lion's Gift of Sight Eye Bank (Minnesota, USA). Hematoxylin/eosin staining was performed, and histopathological observations and diagnosis were carried out by a certified ophtalmologist (Table EV1). FFPE slides were deparaffinized and rehydrated using standard protocols (3×Xylene, 100% EtOH, 96% EtOH, 70% EtOH, dH₂O; 5 min each), then subjected to Tris-based antigen retrieval performed according to manufacturer's instructions (H-3301-250, Vector Labs). Samples were permeabilized with 0.3% Triton X-100 (BP151-100, Fisher Scientific) for 20 min, then blocked with NGS buffer (10% NGS (G9032, Merck), 0.1% Triton X-100 in PBS) for one hour at RT. Primary antibodies (Table EV3) were diluted in NGS buffer and incubated overnight at 4 °C. After three 5-min washes with PBS, secondary antibodies (Table EV3) were similarly diluted in NGS buffer and incubated for 1 h at RT. To counterstain nuclei, 1 μg/mL DAPI (D9542, Merck) was added and incubated together with the secondary antibodies. After three 5-min washes with PBS, samples were mounted using Vectashield (H-1000-10, Vector Labs). Confocal imaging was performed using a Zeiss LSM 510 (Zeiss) confocal microscope equipped with a ×63 immersion objective. Negative (secondary antibody-only) and positive controls (ARPE-19 cells) are included in Fig. EV2.

For immunofluorescence staining of iPSC-RPE, 2 × 10⁴ cells per well were seeded over Matrigel-coated cell culture slides. Cells were carefully washed twice with PBS and fixed using 4% paraformaldehyde for 15 min at RT. Samples were permeabilized/blocked using BGT buffer (BSA-Glycine-Triton; 3 mg/mL BSA, 0.25% Triton X-100, and 100 mM glycine (1610718, Bio-Rad) in PBS) for 20 min. Primary antibodies (Table EV3) were diluted in BGT buffer and incubated for 1 h at RT. After three 5-min washes with PBS, secondary antibodies (Table EV3) were similarly diluted in BGT buffer and incubated for 1 h at RT. To counterstain nuclei, 1 μg/mL DAPI was added and incubated together with the secondary antibodies. To measure protein aggregate accumulation, the ProteoStat assay (ENZ-51035-K100, Enzo) was performed following the manufacturer's instructions. After three 5-min washes with PBS, samples were mounted using ProLong Diamond (P36970, Merck) and cured overnight. Confocal imaging was performed using a Zeiss LSM 510 (Zeiss) confocal microscope equipped with a ×63 immersion objective.

In all experiments, 2.5 × 10⁵ cells per well were seeded over Matrigel-coated 24-well standard culture plates. Cells were washed twice with sterile PBS and total RNA was isolated using QIAGEN RNeasy Mini kit (74104, QIAGEN) according to the manufacturer's instructions. To obtain cDNA, 1 μg of RNA was retrotranscribed using SuperScript III First-Strand Synthesis System (18080051, Thermo Fisher). Transcript mRNA levels were determined by RT-qPCR using either PowerUp SYBR Green (A25780, Thermo Fisher) with in-house primers (Table EV4) or TaqMan probes (Table EV5) with NZY Speedy qPCR Probe Master Mix (MB23003, NZYtech). Ribosomal 18S was included as a housekeeping gene in all experiments.

Using either RT-qPCR or publicly available RNA-seq data, the CMA score was calculated by averaging the weighted and directed transcript levels of known CMA effectors and positive and negative regulators as previously described (Bourdenx et al, 2021). Briefly, counts were log2-normalized, LAMP2, since it is the limiting effector in the pathway, was given a weight of 2, and the rest of the components −1 or 1 based on their effect on CMA activity.

In all experiments, 5 × 10⁵ cells per well were seeded over Matrigel-coated 12-well standard culture plates. Total protein was collected using ice-cold RIPA lysis buffer (R0278-50, Merck) supplemented with protease (P8340, Merck) and phosphatase (P5726, Merck) inhibitors. Protein concentration was measured using Pierce BCA Protein assay (A55865, Thermo Fisher). Samples were diluted, supplemented with 4× Laemmli Sample buffer (1610747, Bio-Rad) and briefly boiled before loading them on precast 4–15% Mini-PROTEAN® TGX (4561086, Bio-Rad) or AnykD Criterion TGX Midi gels (5671125, Bio-Rad). Protein was transferred onto 0.2 μm PVDF membranes (1704272, Bio-Rad) using a TransBlot semi-dry transfer system (Bio-Rad). Transfer efficiency was assessed using Ponceau S staining (78376, Merck). Membranes were blocked with 5% BSA (A30075-100.0, RPI) or milk in TBS-T (0.1% Tween-20 (1706531, Bio-Rad) in PBS) for 1 h and incubated overnight at 4 °C with primary antibodies diluted 1:1000 (Table EV3) in 5% BSA in PBS. Secondary antibodies were diluted 1:2000 in TBS-T (Table EV3) and incubated for 1 h at RT. Membranes were developed using Pierce ECL Western Blotting substrate (32106, Thermo Fisher) or SuperSignal West Dura (34076, Merck) and x-ray film (AGFA) using a CURIX 60 Processor (AGFA). High-abundance proteins were detected using a Chemidoc imaging system (Bio-Rad). After detecting all the pertinent antibodies, total protein within the membranes was stained again using Bio-Safe Coomassie Stain (1610786, Bio-Rad) and air-dried.

HEK293T cells (CRL-3216; ATCC) were transfected with VSV-G, pMDLg/pRRE, REV, and KFERQ-PS-Dendra plasmids using CaCl₂ method. After 24 h, supernatant was collected, centrifuged at 2000 × g for 3 min, and filtered using a 0.45 μm syringe filter. For transient lentiviral transduction, 2 × 10⁴ iPSC-RPE cells per well were seeded over Matrigel-coated cell culture slides (81816, Ibidi). Filtered supernatant containing packaged KFERQ-PS-Dendra vector (Dong et al, 2020) was diluted 1:1 in medium containing 10 μg/mL polybrene (107689, Merck). After 24 h, an additional 1:1 volume was added per well. After 48 h, transduction efficiency was assessed using an inverted widefield fluorescence microscope, the medium was replaced and, if necessary, treatments were added. Cells were carefully washed twice with PBS and fixed using 4% paraformaldehyde (AA433689M, Thermo Fisher) for 15 min at RT, and nuclei were counterstained with 1 μg/mL DAPI in PBS for 5 min. After three 5-minute washes with PBS, samples were mounted using ProLong Diamond and cured overnight. Confocal imaging was performed using a Zeiss LSM 510 (Zeiss) confocal microscope equipped with a ×63 immersion objective.

In all experiments, 4 × 10⁴ cells per well were seeded over Matrigel-coated Seahorse XFe96 plates (103792-100, Agilent). Sensor cartridges were hydrated overnight using Seahorse XF Calibrant (103792-100, Agilent) inside an incubator at 37 °C without CO₂. To assess mitochondrial function, cells were washed once and medium wash replaced with XF Base Media (103575-100, Agilent) supplemented with 2 mM glutamine (G8540, Merck), 5.5 mM glucose (G7528, Merck) and 1 mM sodium pyruvate (P5280, Merck) and subjected to sequential injections of 2 μM Oligomycin (11341, Cayman), 1 μM FCCP (15128, Cayman) and 1 μM Antimycin A+Rotenone (34799, 13995, Cayman). To assess glycolytic capacity, cells were washed once and medium was replaced with XF Base Media supplemented with 2 mM glutamine and 1 mM sodium pyruvate, incubated for 1 h at 37 °C without CO₂ and subjected to sequential injections of 20 mM glucose, 2 μM oligomycin, and 100 mM 2-deoxyglucose (14325, Cayman). Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured using standard MitoStress and GlycoStress protocols using Wave software (v2.6.3.8, Agilent). Recordings were normalized to cell number after addition of 1 μg/mL Hoechst 33342 (H3570, Thermo Fisher) and automated cell number quantification using a Cytation microplate imager (Agilent).

In all experiments, 4×10⁴ cells per well were seeded over Matrigel-coated 96-well black clear-bottom plates. A combination of 1 μM Lysotracker Green (lysosomal acidity; L7526, Merck), 1× Magic Red Cathepsin B substrate (acid protease activity; 947, ImmunoChemistry), and 1 μg/mL Hoechst 33342 was added, and cells were incubated for 15 min at 37 °C, washed, and imaged using a Cytation microplate imager (Agilent).

For proteomics analyses, 1 × 10⁶ cells were seeded in Matrigel-coated 6-well plates and maintained as previously described. Treatments were added for 24 h after the final culture medium change. Cells were detached using Accumax, pelleted by centrifugation at 300 × g for 5' and washed twice with sterile ice-cold PBS. After the final centrifugation, PBS wash carefully aspirated using a p1000 micropipette, and the cell pellets were flash-frozen using dry ice. Both targeted and bulk proteomics were performed by the Metabolomics and Proteomics Platform (MAPP) at the Université de Fribourg (Switzerland) as previously described (Leytens et al, 2025), and raw data were processed using Skyline (Pino et al, 2020) and Spectronaut (Biognosys), respectively. Statistical analysis to find differentially enriched proteins was performed using the matrixTests package in R, and GSEA software (Subramanian et al, 2005) was used for pathway enrichment analysis, including 1000 permutations based on gene-set size. CMA literature-validated substrate list was manually curated (Dataset EV2). Canonical and non-canonical (phosphorylation, acetylation) KFERQ-like motif abundance was determined using the KFERQ finder (v0.8) tool (Kirchner et al, 2019). ORA of putative CMA degradation substrates (FC > 1.25;p value < 0.01; KFERQ-motif⁺ in N/L-treated iPSC-RPE) was performed using g:Profiler (Kolberg et al, 2023) including KEGG, Reactome, and WikiPathways databases. p value and fold-change cut-offs were established based on previous proteomics studies from the lab assessing protein abundance differences in RPE and lysosomal degradation of KFERQ proteins (Shang et al, 2024; Bourdenx et al, 2021). Resulting data was exported to CytoScape (v3.10.2.) (Shannon et al, 2003), filtered based on a Jaccard index of 0.25, and clustered using EnrichmentMap (Merico et al, 2010) and AutoAnnotate plugins (Kucera et al, 2016). Heatmaps were generated using the tidyverse, ggplot2, viridis, RColorBrewer, and pheatmap packages in R (v4.2.0.).

Experimental design diagrams and a graphical abstract were created with BioRender.com.

Data points represent individual donors or donor-derived iPSC-RPE cell lines. Sample size was determined based on previous studies from the lab, including enough cell lines to account for donor heterogeneity and be able to determine significance between healthy/AMD donors. Bioinformatic, biochemical, and microscopy image analyses were performed by a researcher blinded to the condition. All data were evaluated for normality and heteroscedasticity. Normally distributed data were analyzed using a one-way or two-way analysis of variance with appropriate post hoc comparisons (more than two groups) or a two-tailed Student's t test (two groups). Non-normally distributed data were analyzed using Mann–Whitney's U-test. All statistical tests were performed using GraphPad Prism (v10.0) or R (v4.2.0), and data are presented as the mean ± standard error of the mean (s.e.m.).