Secretome translation shaped by lysosomes and lunapark-marked ER junctions

Abstract

The endoplasmic reticulum (ER) is a highly interconnected membrane network that serves as a central site for protein synthesis and maturation¹. A crucial subset of ER-associated transcripts, termed secretome mRNAs, encode secretory, lumenal and integral membrane proteins, representing nearly one-third of human protein-coding genes¹. Unlike cytosolic mRNAs, secretome mRNAs undergo co-translational translocation, and thus require precise coordination between translation and protein insertion^2,3. Disruption of this process, such as through altered elongation rates⁴, activates stress response pathways that impede cellular growth, raising the question of whether secretome translation is spatially organized to ensure fidelity. Here, using live-cell single-molecule imaging, we demonstrate that secretome mRNA translation is preferentially localized to ER junctions that are enriched with the structural protein lunapark and in close proximity to lysosomes. Lunapark depletion reduced ribosome density and translation efficiency of secretome mRNAs near lysosomes, an effect that was dependent on eIF2-mediated initiation and was reversed by the integrated stress response inhibitor ISRIB. Lysosome-associated translation was further modulated by nutrient status: amino acid deprivation enhanced lysosome-proximal translation, whereas lysosomal pH neutralization suppressed it. These findings identify a mechanism by which ER junctional proteins and lysosomal activity cooperatively pattern secretome mRNA translation, linking ER architecture and nutrient sensing to the production of secretory and membrane proteins.

Fig. 1. Secretome mRNAs are translated on ER and show confined motion.

a, Inset, design of secretome mRNA reporters (inset) containing a 5′ UTR, an open reading frame, 24 MS2 binding sites and a 3′ UTR. Main image, representative spinning-disk confocal microscopy of SiT-EGFP mRNA on ER labelled with SNAP–Sec61β. Scale bar, 5 µm. Right, magnified views of indicated regions showing trajectories of two confined mRNAs (top; yellow) or a rapidly moving mRNA (bottom; yellow). Scale bars, 1 µm. b, Violin plots of log(MSD_{τ = 1 s}) for SiT-EGFP mRNAs under control (n = 1,616) and puromycin (Puro)-treated (n = 1,608) conditions. Red dashed line indicates MSD = 0.055 µm². Dunnett’s test, ****P < 0.0001. c, Box plots showing fraction of slow-moving SiT-EGFP mRNAs in control (n = 17 cells) and puromycin-treated (n = 19) cells. Two-tailed unpaired t-test, ****P < 0.0001. d, Control (left) and puromycin-treated (right) U-2 OS cells expressing SNAP–Sec61β (grey) and SiT-EGFP mRNA. Puncta are pseudocoloured and designated as translating (blue, MSD < 0.055 µm²) or non-translating (MSD ≥ 0.055 µm²). Scale bars, 5 µm. e, ER-specific SunTag reporter design. f, Representative merged images of SunTag (green), mRNA (magenta) and trajectories (yellow). Left, SunTag(+) (translating) punctum with confined motion. Right, SunTag(–) (non-translating) punctum showing rapid motion. Scale bars, 1 µm. g, Violin plots of log(MSD_{τ = 1 s}) for cytERM–SunTag MS2 with SunTag(+) versus SunTag(–). Dunnett’s test, ****P < 0.0001. h, Violin plots of log(MSD_{τ = 1 s}) for SiT-EGFP–MS2 (n = 1,616), CD4-EGFP–MS2 (n = 251), CALR-mEmerald–MS2 (n = 250), ACTB-Halo–MS2 (n = 100) and SunTag-SEC61B–MS2 (n = 314). Red dashed line indicates MSD = 0.055 µm². Statistical comparisons versus SiT-EGFP–MS2, Dunnett’s test, ***P < 0.001, ****P < 0.0001; NS, not significant (P > 0.05). Source Data

Fig. 2. ER junctions are hotspots for secretome mRNA translation.

a, Time-lapse images of translating CD4-EGFP, SiT-EGFP and CALR-mEmerald mRNAs on ER (Sec61β, magenta) in U-2 OS cells. Scale bars, 1 µm. b, Box plots of mean distance to nearest ER junction per cell for translating (T) fractions of CALR (n = 29 cells), CD4 (n = 32) and SiT (n = 40) mRNA, non-translating (NT) SiT mRNA (n = 19), Halo–ACTB (n = 23), Halo–SEC61B (n = 36) and slow ribosomes (effective MSD_{τ = 1 s} < 0.04 µm², n = 37). Comparisons versus translating SiT-EGFP. Dunnett’s test, ****P < 0.0001, ***P < 0.001. c, Representative 3D structured illumination microscopy images of HCR-smFISH labelling of endogenous CD9 mRNA with the ER marker (mEmerald–Sec61β, magenta) in control and puromycin-treated U-2 OS cells. Scale bars, 1 µm. d, Quantification of distance of endogenous CD9 smFISH puncta to the nearest ER junction in control (n = 19 cells) and puromycin-treated (n = 18) cells. Each dot represents the mean per cell. Two-tailed unpaired t-test, ****P < 0.0001. e, Time-lapse images of a ribosome (cyan) with effective MSD_{τ = 1 s} < 0.04 µm² (slow, top) or MSD_{τ = 1 s} > 0.04 µm² (fast, bottom). Scale bars, 1 µm. Source Data

Fig. 3. LNPK marks ER junctions that enhance secretome mRNA translation.

a, Representative image of LNPK–GFP and ER (Halo–Sec61β, cyan). Inset, enlarged view. Scale bar, 5 µm. b, Top, translating SunTag(+) cytERM-SunTag–MS2 (grey) with overlapping SunTag (green) and LNPK (magenta). Bottom, non-translating SunTag(–) cytERM–SunTag–MS2, showing no colocalization with LNPK. Scale bars, 1 µm. c, Percentage of cytERM–SunTag–MS2 puncta that are SunTag(+) or SunTag(–) within 300 nm of LNPK (SunTag(+)) or outside this range (SunTag(–)). n = 14 cells. Paired two-tailed t-test, ****P < 0.0001. d, Representative images of SiT-EGFP–MS2 and particle trajectories (yellow) on ER (red) in control, LNPK-KD or CLIMP63-KD cells. Scale bars, 1 µm. e, Translating fractions of SiT-EGFP mRNA in control (n = 17 cells), CLIMP63-KD (n = 14), LNPK-KD (n = 13) and puromycin-treated (n = 17) cells. Statistical comparisons were performed against control using Dunnett’s test, ****P < 0.0001. f, Protein yield from whole-cell extracts (top) and membrane fractions (bottom) of 10⁶ control and LNPK-KO U-2 OS cells. Each extract was normalized to its control. a.u., arbitrary units. Dunnett’s test, **P < 0.005. Source Data

Fig. 4. LNPK-enriched ER junctions recruit lysosomes and boost translation.

a, Schematic of PLA. Anti-LNPK antibody (via oligo-labelled anti-rabbit, blue) and organelle marker antibodies (via oligo-labelled anti-mouse, green) generated proximity signals amplified by rolling circle amplification and visualized with a fluorescent probe (orange). b, Representative images from PLA experiment, showing DNA, PLA puncta (orange) and merged views of LNPK antibody with EEA1 (early endosome), LAMP1 (lysosome) or TOMM20 (mitochondria). Scale bars, 5 µm. c, Quantification of PLA spots per cell for LNPK with EEA1, LAMP1 and TOMM20, and for LAMP1 with REEP5 (control). Data from 10 fields of view, 3 replicates. Statistical comparisons versus LNPK–LAMP1. Dunnett’s test, ****P < 0.0001. d, Schematic of the optogenetic ER–lysosome recruitment tool comprising iLID–Halo–LAMP1 (lysosome) and sspB–mCherry–Sec61β (ER). mCh, mCherry. e, Representative images of iLID–Halo–LAMP1 (magenta) and sspB–mCherry–Sec61β (green) during 488 nm activation in control and LNPK-KO cells. Scale bars, 1 µm. f, Quantification of ER signal within lysosomal mask over time following activation. Control cells: t_1/2 = 7.9 s; LNPK-KO cells: t_1/2 = 26 s. Shaded areas represent 95% confidence intervals. Orange dashed line indicates a normalized intensity of 1. g, Representative images of SunTag, LAMP1, cytERM-SunTag–MS2 (mRNA) and merged views. Arrows mark translating mRNAs. Scale bar, 1 µm. h, Top, relative SunTag intensity versus distance to lysosomes (binned at 500 nm) for cytERM–SunTag–MS2 in control and LNPK-KD cells. Normalized to intensity at 4.25 µm. Shaded areas represent 95% confidence intervals. Orange dashed line indicates a relative SunTag intensity of 1. Multiple unpaired t-tests were done, ****P < 0.0001, **P < 0.01. Bottom, schematic showing how lysosome proximity increases SunTag intensity, reflecting higher ribosome occupancy. Source Data

Fig. 5. Lysosome activity and LNPK regulate secretome translation via eIF2.

a, Translating fraction of SiT-EGFP mRNA under control conditions, with chloroquine (CQ; 10 µM, 4 h; n = 16 cells), lysosomal protease inhibitor (Lys. Pro. Inhibitors; 1 h; n = 16) or amino acid starvation (−AA; 16 h; n = 15). Comparisons versus control; Dunnett’s test, ***P = 0.0003, ****P < 0.0001. b, Relative SunTag intensity versus distance from lysosome (500 nm bins) for cytERM–SunTag–MS2 under control conditions, with LNPK knockdown, amino acid depletion or CrPV IRES expression. Shaded areas represent 95% confidence intervals. Comparisons versus LNPK KD; multiple unpaired t-tests, ****P < 0.0001. c, Left, translating fractions of cytERM–SunTag in control (Ctrl; n = 20 cells), LNPK-KD cells (n = 38) and LNPK-KD cells with 200 nM ISRIB (n = 24). Comparisons versus control; Dunnett’s test, *P = 0.0081 (LNPK KD), P = 0.7951 (LNPK KD plus ISRIB). Right, translation of cytERM–SunTag with CrPV IRES 5′ UTR in control (n = 34) and LNPK-KD (n = 34) cells. Unpaired two-tailed t-test, P = 0.3777. d, Translating fraction of SiT-EGFP mRNA under control conditions, with rapamycin (100 nM, 1 h, n = 16), torin-1 (1 µM, 1 h, n = 17), LNPK KD (n = 13), 1 µM thapsigargin (Thaps; 1 h, n = 20), 1 µM thapsigargin plus 200 nM ISRIB (1 h, n = 32), amino acid starvation (16 h, n = 15), or amino acid starvation plus 200 nM ISRIB (1 h, n = 36). Comparisons versus control. Dunnett’s test, ****P < 0.0001. e, Western blots of total eIF2α, phosphorylated eIF2α (p-eIF2α), LNPK and tubulin (as loading control) from control and LNPK-KO cells (3 replicates). f, Quantification of total eIF2α (blue) and p-eIF2α immunofluorescence in control and LNPK-KO cells. Unpaired two-tailed t-test versus control: eIF2α, *P = 0.0346; p-eIF2α, P = 0.14. g, Ratio of p-eIF2α/eIF2α from immunofluorescence (5 replicates). Unpaired two-tailed t-test, **P = 0.0095. h, Time-lapse images of FRAP recovery of SunTag (green) at cytERM–SunTag MS2 (magenta) in control cells, LNPK-KD cells and LNPK-KD cells treated with ISRIB. Arrows indicate photobleaching sites. i, FRAP recovery curves for control (n = 31), LNPK-KD (n = 31), control plus ISRIB (n = 10) and LNPK-KD plus ISRIB (n = 32) conditions. The vertical dashed line indicates time of photobleaching and the horizontal dashed line indicates a normalized intensity of 1. Shaded areas represent 95% confidence intervals. j, Model showing how LNPK and lysosomes form ER junctional hubs to enhance secretome translation via eIF2-dependent initiation. AAs, amino acids. Source Data

Extended Data Fig. 1. Motion distinguishes translating from nontranslating secretome mRNAs.

a, Golgi-like distribution of SiT-EGFP MS2 mRNA in U-2 OS cells. SiT-EGFP (green), MS2 (cyan), ER (orange). b, Mean squared displacement (MSD, µm²) vs lag time (τ, sec) for SiT-EGFP MS2 mRNAs under control (red) or puromycin (blue). Fitted curves: D_control = 0.004 µm²/s, D_puro = 0.033 µm²/s. c, MSD vs lag time of SiT-EGFP MS2 trajectories classified as translating (MSD_τ=1sec < 0.055 µm²). d, Schematic of cytERM-SUNTAG MS2 construct (upper) and CrPV IRES cytERM-SUNTAG MS2 (lower). e, Boxplot of translating fractions of unmodified cytERM-SUNTAG MS2 in control (N = 20 cells), LNPK KD (N = 38), and CLIMP63 KD (N = 39). f, MSD vs lag time of SUNTAG(+) cytERM-SUNTAG MS2 mRNAs, fitted Dapp=0.007 µm²/s. g, Violin plots of log(MSD_τ=1sec) for SUNTAG(+) cytERM-SUNTAG MS2 (n = 72), CD4-EGFP MS2 (n = 251), SiT-EGFP MS2 (n = 1616), Calreticulin-mEmerald MS2 (n = 250), ER5-mEmerald MS2 (n = 1155), ER3-mEmerald MS2 (n = 2475), β-actin-Halo MS2 (n = 100), SUNTAG-Sec61β MS2 (n = 314), TOMM20-Halo MS2 (n = 233), SiT+puromycin (n = 1608), and SUNTAG(–) cytERM-SUNTAG MS2 (n = 680). Red dotted line, MSD = 0.055 µm². h, Time-lapse images of Actin-Halo MS2 mRNA (green) and ER (magenta). Scale bars, 1 µm. Source Data

Extended Data Fig. 2. Translating mRNAs and ribosomes localize at ER junctions with LNPK.

a, Wide-field images of SiT-EGFP MS2 mRNAs (cyan) and ER (red) in U-2 OS, COS7, HeLa, and HT1080 cells. Large image scale bars, 10 µm; inset, 1 µm. b, Box plots of average distance of translating SiT-EGFP MS2 mRNAs (MSDτ = 1 sec<0.055 µm²) to nearest ER junctions in U-2 OS, HeLa, COS7, and HT1080. Each dot, single cell mean. c, Cumulative distribution function (CDF) fitting of L10a-HaloTag ribosome displacements in control cells. Fits: one coefficient (1D, purple), two (2D, yellow), three (3D, red). d, Residual plots of observed vs fitted data (from c). e, Circular graph of ribosome mobility states in control vs puromycin-treated cells. Fractions: slow D = 0.004 µm²/s (blue), medium D = 0.04 µm²/s (yellow), fast D = 0.3 µm²/s (purple). f, Wide-field image of cytERM-SUNTAG MS2 mRNAs (gray), SUNTAG signal (green), and anti-LNPK staining (magenta). Translating puncta circled (green dotted); non-translating puncta circled (magenta). Scale bars, 1 µm. g, Cumulative distribution function of log₁₀ SUNTAG intensity of cytERM-SUNTAG MS2 mRNAs in control vs LNPK KD cells. h, Distribution of log₁₀ SUNTAG intensities from cytERM-SUNTAG MS2 puncta in control vs LNPK KD cells. Source Data

Extended Data Fig. 3. LNPK loss reduces membrane protein synthesis and translation near lysosomes.

a, Schematic of L-homopropargylglycine (HPG) incorporation assay to label newly synthesized proteins in whole-cell and membrane fractions. b, Click-labeled Alexa-488-HPG signal in whole-cell and digitonin-extracted membrane fractions of control (green) and LNPK KO U-2 OS cells (magenta). Whole-cell: no significant difference (unpaired two-tailed t-test, P = 0.8492). Membrane fraction: significant reduction in LNPK KO (****P < 0.0001). c, Tables of proteins and mRNAs differentially regulated in LNPK KO cells ( ≥ 2-fold change in protein or RNA, adjusted P < 0.05). d, Scatter plot of protein fold-change (Protein 2FC, LNPK KO/WT) vs RNA fold-change (RNA 2FC, LNPK KO/WT). Pink, proteins with adjusted P < 0.05; blue, mRNAs with adjusted P < 0.05. Dotted line, 2-fold threshold. e, Log₂ protein/mRNA ratios in KO vs WT. Central dotted line, 1:1 ratio; flanking dotted lines, 2-fold boundaries. Pink dots, higher in control; blue, higher in LNPK KO. f, Histogram of log₂ difference in protein/mRNA ratio between KO and control. Pink bars, genes with higher ratios in control; blue bars, higher ratios in LNPK KO. g, The average distance between SiT-EGFP mRNA and the nearest lysosome (µm) of translating (T, green) and non-translating (NT, red) mRNA. Each line represents the same cell. N = 15 cells. Source Data

Extended Data Fig. 4. LNPK depletion alters eIF2 signaling and translation recovery.

a, Relative eIF2α levels (normalized to tubulin) in control and LNPK KO U-2 OS cells. Unpaired two-tailed t-test, P = 0.037. b, Relative phosphorylated eIF2α (p-eIF2α) levels in control and LNPK KO cells. Unpaired two-tailed t-test, P = 0.14. c, Immunoblots of ATF4, Tubulin, eIF2α, and p-eIF2α from control (WT), LNPK KO, and U-2 OS cells treated with bortezomib (100 nM, 4 h). All blots imaged under identical conditions. d, Western blots of LNPK, eIF2α, p-eIF2α, ATF4, and Tubulin from control siRNA, LNPK siRNA (LNPK KD), and LNPK siRNA+bortezomib (100 nM, 4 h). e, qPCR quantification of spliced/unspliced XBP-1 ratio in control, LNPK KO, and LNPK KD cells, with or without thapsigargin (1 µM, 1 h). Data show no induction of unfolded protein response by LNPK loss. f, FRAP recovery of cytERM-SUNTAG MS2 puncta in control cells (black, n = 32) vs cycloheximide-treated cells (red, n = 10). Recovery is blocked by cycloheximide, confirming fluorescence recovery reflects new peptide synthesis rather than antibody exchange. Source Data