Cell-autonomous innate immunity by proteasome-derived defence peptides

Abstract

For decades, antigen presentation on major histocompatibility complex class I for T cell-mediated immunity has been considered the primary function of proteasome-derived peptides^1,2. However, whether the products of proteasomal degradation play additional parts in mounting immune responses remains unknown. Antimicrobial peptides serve as a first line of defence against invading pathogens before the adaptive immune system responds. Although the protective function of antimicrobial peptides across numerous tissues is well established, the cellular mechanisms underlying their generation are not fully understood. Here we uncover a role for proteasomes in the constitutive and bacterial-induced generation of defence peptides that impede bacterial growth both in vitro and in vivo by disrupting bacterial membranes. In silico prediction of proteome-wide proteasomal cleavage identified hundreds of thousands of potential proteasome-derived defence peptides with cationic properties that may be generated en route to degradation to act as a first line of defence. Furthermore, bacterial infection induces changes in proteasome composition and function, including PSME3 recruitment and increased tryptic-like cleavage, enhancing antimicrobial activity. Beyond providing mechanistic insights into the role of proteasomes in cell-autonomous innate immunity, our study suggests that proteasome-cleaved peptides may have previously overlooked functions downstream of degradation. From a translational standpoint, identifying proteasome-derived defence peptides could provide an untapped source of natural antibiotics for biotechnological applications and therapeutic interventions in infectious diseases and immunocompromised conditions.

Fig. 1. Hundreds of potential antimicrobial peptides reside in the human proteome.

a, AMP sequences (green) in host proteins (grey). Conserved AMPs are indicated in orange, whereas those without conservation data are labelled as NA. b,c, Comparison of predicted amino-acid local-distance difference test (pLDDT) scores (b) and relative solvent-accessible surface area (rASA) (c) in putative AMPs compared with other regions of the same proteins. U-test, ****P < 0.0001. d, Distribution of peptides predicted by in silico proteasomal cleavage of the human proteome, showing only those scoring >5 on the basis of AMP biochemical characteristics. e, CFU count of intracellular S. typhimurium infection of A549 cells treated with bortezomib (50 nM) or untreated, normalized to the cell count. Data are mean ± s.e.m. (n = 6 biological replicates). Unpaired two-tailed Student’s t-test, **P = 0.004. f, Bacterial growth in conditioned medium (<10 kDa). OD, optical density. Diagram of the equipment in f was created with BioRender.com (https://BioRender.com/x10a923). g–i, Growth of S. enterica in medium from HCT116 (h) or A549 (g,i) cells treated with DMSO (control), epoxomicin (1 μM, concentrated medium ×2; g) or bortezomib (50 nM; h,i) for 6 h. Data are mean ± s.e.m. (n = 3 biological replicates). Unpaired two-tailed t-test: NS, P = 0.96, ***P = 0.0007, ****P < 0.0001 (g); **P = 0.0066 (h); *P = 0.0248 (i). j,k, Growth of S. enterica in medium from A549 (j) or HCT116 (k) cells treated with or without bortezomib and followed by proteinase K (1 μg ml^–1) for 6 h. Data are mean ± s.e.m. (n = 3 biological replicates). One-way ANOVA: NS, P = 0.38, *P = 0.033 (j); NS, P = 0.675; **P = 0.0038 (k).

Fig. 2. Identified PDDPs exhibit antibacterial activity.

a, Overview of the MAPP methodology. b, Sequence alignment of Histatin 3 (P15516) with previously reported AMPs (blue) and a peptide identified by MAPP (red). c, Overlap between MAPP-identified peptides and peptides detected in the A549 secretome. d, Workflow for scoring MAPP peptides to identify previously undescribed PDDPs. e, Distribution of AMP scores for MAPP-identified peptides. The dashed line shows the mean score of reported AMPs. f, CFU ml^–1 of Gram-negative (P. aeruginosa, S. enterica, E. coli) and Gram-positive (M. luteus, S. haemolyticus) bacteria after treatment with high-scoring peptides (named after the parental protein), LK20 or low-scoring peptides. Values normalized to control (DMSO). Data are mean ± s.e.m. (n = 3 biological replicates). g, Representative images of lysogeny broth agar plates showing tenfold serial dilutions of bacterial cultures treated with PPP1CB at MBC concentrations or with DMSO. h, FBCI values for combinations of 10 high-scoring peptides and a combination of five PDDPs (PPP1CB, DFNA5, DCTN4, PSMG2, CHMP2A) tested against M. luteus (Ml), P. aeruginosa (Pa), E. coli (Ec) and S. haemolyticus (Sh). FBCI ≤ 0.5 indicates synergy, 0.5 < FBCI ≤ 1 indicates additive, 1 < FBCI ≤ 4 indicates indifference and FBCI > 4 indicates antagonism. Low-scoring peptides (score of 0) and LK20 served as negative and positive controls, respectively. i, Bacterial permeabilization assays. j, Bacterial membrane permeabilization and propidium iodide uptake at 90 minutes after treatment with increasing concentrations of high-scored PDDPs. Values normalized to control (0.1% SDS). Data are mean ± s.e.m. (n = 3 biological replicates). k, Transmission electron microscope imaging of untreated bacteria or bacteria treated with PPP1CB at the minimal inhibitory concentration. Arrows indicate morphological disruptions and cytoplasmic release. Scale bars, 0.2 μm (S. enterica and E. coli) and 0.1 μm (M. luteus and S. haemolyticus). Schematics in panels a and i created with BioRender.com (https://BioRender.com/x10a923).

Fig. 3. In vivo antimicrobial activity of the PPP1CB-derived peptide in models of pneumonia, bacteraemia and sepsis.

a, Pneumonia model. i.n., intranasal. b, Box plots of CFU count of P. aeruginosa (log₁₀) per lung in mice treated with PBS, PPP1CB (5 mg kg^–1 i.v.) or tobramycin (50 mg kg^–1 i.p.). Data are mean ± s.d. (n = 8). One-way ANOVA, ****P < 0.0001. c, Representative haematoxylin and eosin (H&E)-stained and Gram-stained lung tissue sections from uninfected mice or mice infected with P. aeruginosa (Pa) treated with vehicle (PBS) or PPP1CB peptide (5 mg kg^–1 i.v.). Scale bars, 50 µm (uninfected (H&E), Pa PPP1CB (H&E) and Pa PBS (H&E)) and 20 µm (Pa PBS (100×, Gram)). Red arrows indicate bacterial staining. d, Bar plot showing mean ± s.e.m. of tissue injury in the pneumonia model. One-way ANOVA, **P < 0.01, ***P < 0.001. e, Neutrophil infiltration analysis in lung tissue graded on the basis of severity: 1, minimal (<1%); 2, slight (1–25%); 3, moderate (26–50%); 4, moderate–severe (51–75%); 5, severe–high (76–100%). Data are mean ± s.e.m., one-way ANOVA, **P < 0.01. f, bacteraemia model. g, Box plots of CFU count of P. aeruginosa (log₁₀) per spleen in mice treated with PBS, PPP1CB (10 mg kg^–1 i.v.) or tobramycin (50 mg kg^–1 i.p.). Data are mean ± s.d. (n = 8). One-way ANOVA, ****P < 0.0001. h, Representative H&E-stained and Gram-stained tissue sections from spleen uninfected or treated with PBS or PPP1CB peptide (10 mg kg^–1 i.v.). Scale bars, 50 µm (H&E stained) and 20 µm (Gram stained). i–m, Bar plots representing the degree of lesions in different organs (spleen (i), heart (j), kidney (k), liver (l), lung (m)) in the bacteraemia model. Data are mean ± s.e.m. One-way ANOVA, *P = 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. n, Kaplan–Meier survival curve showing survival rates of mice treated with PBS, PPP1CB or tobramycin (10 mg kg^–1 retro-orbitally (r.o.) for both treatments) over six days after P. aeruginosa infection. Log-rank test, ****P < 0.0001, followed by pairwise comparisons using log-rank tests. NS, P > 0.05 (n = 15). Schematics in panels a and f created with BioRender.com (https://BioRender.com/x10a923).

Fig. 4. Targeted proteasomal degradation of PDDPs containing proteins and its effect on bacterial growth and infection.

a, dTAG system of PPP1CB protein. b, Western blot analysis of degradation of dTAG–GFP and dTAG–PPP1CB. Cells were treated with dTAG^V-1 (500 nM) or DMSO, with or without proteasome inhibitor (1 µM epoxomicin) for 6 h (n = 3 biological repeats). c, Bacterial growth assay using conditioned medium. MWCO, molecular weight cut-off. d, S. enterica growth in conditioned medium from A549 cells degrading GFP (left), PPP1CB (middle) and the ratio of dTAG^V-1 to DMSO treatment (right). Data are mean ± s.e.m. (n = 4 biological replicates). Unpaired two-tailed Student’s t-test: left, NS, P = 0.63; middle, *P = 0.007; right, *P = 0.036. e, Experiment using the dTAG system. Cells expressing dTAG were incubated for 1 h with dTAG^V-1 (500 ng ml^–1) to activate PPP1CB degradation by the proteasome, followed by S. typhimurium infection (a multiplicity of infection of 100 was used for CFU counts and of 5 was used for imaging). Bacteria that did not infect the cells were washed with gentamicin and intercellular bacteria were either plated for CFU count or analysed by imaging. f,g, Bar plots showing the ratio of intracellular CFU ml^–1 counts of M. luteus (f) and S. typhimurium (g) in cells expressing dTAG–PPP1CB or dTAG–GFP, treated with dTAG^V-1 or DMSO. Data are mean ± s.e.m. n = 6 and n = 3 biological replicates (f and g, respectively). Unpaired two-tailed Student’s t-test: *P = 0.0106 (f); **P = 0.0022 (g). h–j, Imaging of A549 cells expressing dTAG–GFP (green) or PPP1CB, treated with DMSO or dTAG^V-1, infected with red fluorescent protein (RFP) S. typhimurium (magenta) or DAPI (blue). h, Representative images. Scale bars, 50 µm. i,j, Infection with RFP S. typhimurium per number of nuclei. Data are mean ± s.e.m. (n = 4 biological replicates, five images each) for dTAG–GFP (i) and dTAG–PPP1CB (j). Unpaired two-tailed Student’s t-test: NS, P = 0.9244 (i); *P = 0.0249 (j). Schematics in panels a, c and e created with BioRender.com (https://BioRender.com/x10a923).

Fig. 5. Bacterial infection promotes the recruitment of PSME3 and induces cationic cleavage activity.

a, Infection MAPP experiment. b, Frequency of unique peptides, red (infection), blue (control) and grey (shared). c, Distribution of scores for infection or control unique peptides. Mann–Whitney U-test, P < 0.0001. d, Normalized percentage of C-terminal amino acids of MAPP peptides after infection. e,f, Percentage of unique peptides. C-terminal Lys, Arg or His (e) or Trp, Phe or Tyr (f) per protein in infection (blue) or control (grey). Mann–Whitney U-test, ****P < 0.0001. g, Model of pull-down proteasome activity assay. h–j, Proteasome activity of S. enterica infected A549 cells (4 h, dark blue; 1 h, light blue) or uninfected (4 h, dark grey; 1 h, light grey). Bar plots at the analysis end point for Arg-Leu-Arg (i) and Leu-Leu-Val-Tyr (j) peptides. Data are mean ± s.e.m. (n = 6 biological replicates). One-way ANOVA, NS, P = 0.075, ***P = 0.0003 (i); ***P = 0.0008, ****P < 0.0001. (j). RFU, relative fluorescence unit. k, Volcano plot of proteasome co-immunoprecipitated proteins from infected A549 cells. l,m, Relative PSME3 intensity by proteomics or proteasome pull-down after 1 h (l) or 4 h (m) of infection. Data are mean ± s.e.m. (n = 3 biological replicates). One-way ANOVA, NS, P = 0.64, **P = 0.0012 (l); NS, P = 0.74, ****P < 0.0001 (m). n,o, A549 cells expressing shCtrl or shPSME3 infected with S. typhimurium. n, Representative image of colonies; o, CFU quantification. Data are mean ± s.e.m. (n = 6 biological replicates). Two-tailed Student’s t-test, ***P = 0.0006. p, Bacterial growth (1 h) in conditioned medium from A549 cells expressing control short hairpin RNA (shCtrl) or short hairpin RNA against PSME3 (shPSME3). Data are mean ± s.e.m. (n = 6 biological replicates). Two-tailed Student’s t-test, ***P = 0.0007, ***P = 0.0006, *P = 0.026, NS, P = 0.23. Schematics in panels a and g created with BioRender.com (https://BioRender.com/x10a923).

Extended Data Fig. 1. Detection of AMPs across the human proteome.

a, Pie chart illustrating the total number of experimentally validated AMPs sourced from four comprehensive databases: DBAASP (Database of Antimicrobial Activity and Structure of Peptides), dbAMP, DRAMP (Data Repository of Antimicrobial Peptides), and CAMP (Collection of Antimicrobial Peptides). b, Examples of AMP-containing proteins (cyan regions) mapped onto their respective protein structures (light gray). c, Gene ontology (GO) analysis of AMP-containing proteins categorizing immune-related and non-immune functions. d, Heatmap representing tissue-specific expression levels of all 273 identified AMP-containing proteins and antimicrobial proteins derived from diverse organisms within the human proteome. e, Heatmap highlighting expression profiles of seven well-characterized human antimicrobial proteins (subset of the 273 AMP-containing proteins). f, Schematic illustration delineating protein regions into surface-accessible (rASA > 25), buried (rASA <25), and intrinsically disordered regions (IDRs; predicted by pLDDT scores). g, Bar plot showing the percentage of proteasome-derived peptides predicted in silico across peptidomics datasets from various biological fluids. h, Distribution of previously reported AMPs, classified according to antimicrobial biochemical properties (cationicity and hydrophobicity). Illustrations created in BioRender. Merbl, Y. (2025) https://BioRender.com/x10a923.

Extended Data Fig. 2. Examining the antimicrobial activity of putative PDDPs.

a, Sequence alignments of Histone 2JA with previously reported AMP peptides (light blue) and peptides identified through MAPP (red). Sequence alignments of Fau protein with previously reported AMP peptides (light blue) and peptides identified through MAPP (red). b, Histogram showing peptide length distribution, highlighting density differences between known AMPs (dark blue) and MAPP-derived AMPs (light blue). c, Bar plot indicating the percentage of proteasome-derived peptides (MAPP peptides) identified across various peptidomics datasets from biological fluids. d, Protein structural models generated by AlphaFold 2 for 10 selected high-scoring peptides (AMP score > 5), with the locations of proteasome-derived defense peptides (PDDPs) within the protein highlighted in cyan.

Extended Data Fig. 3. Assessing permeabilization and morphological effects of PDDPs.

a-b, Heatmaps representing minimal inhibitory concentration (MIC) experiments against five bacterial species. Data show bacterial growth inhibition in response to increasing concentrations of 10 selected peptides with high scores, while b shows bacterial growth with 5 low-scoring peptides (AMP score = 0), percentage from (DMSO). LK20 was used as a positive control. Data are presented as mean (n = 3 biological replicates). c, Bacterial membrane permeabilization assay showing the percentage of propidium iodide uptake over time of 5 bacteria treated with increasing concentrations of PPP1CB-derived peptide. mean ± s.e.m. (n = 3 biological replicates). d Mammalian cell permeabilization assay showing the percentage of propidium iodide uptake in A549 and HCT116 cells over time with increasing concentrations of PPP1CB-derived peptide. Data are presented as mean ± s.e.m. (n = 3 biological replicates). e, Heatmap showing propidium iodide uptake in A549 cells treated with all selected high-scoring PDDPs and a low-scoring peptide (AMP score = 0) at the 90-minute time point. Data are presented as mean (n = 3 biological replicates). f, Representative transmission electron micrographs (TEM). Bacteria were treated with PPP1CB-derived peptide at minimal bactericidal concentration (MBC) or left untreated. Arrows indicate disruptions in bacterial membranes and other morphological changes.

Extended Data Fig. 4. Testing the antimicrobial function of the PPP1CB-dervied derived peptide in vivo.

a, Predicted half-life distribution of MAPP-derived peptides classified as putative AMPs (cyan) compared to peptides with AMP scores below the threshold (<5; gray). Dashed lines indicate the stability of 10 high-scored proteasome-derived peptides. b, Pearson correlation between the predicted half-life of MAPP peptides in blood and their AMP score. c, Transepithelial/transendothelial electrical resistance (TEER) assay showing the barrier index of Caco-2 cells treated with EGTA (control), Pseudomonas aeruginosa, or PPP1CB peptide at varying concentrations for 6 h. Data are presented as mean ± s.e.m. (n = 3 biological replicates). One-way ANOVA, *** P = 0.0006; **** P < 0.0001. d, Representative images from an in vivo bacteremia model with Pseudomonas aeruginosa (PAO1) infection. H&E-stained and gram-stained tissues are shown for four organs: liver, spleen, kidney, and heart. Columns display uninfected controls and tissues treated with PBS or PPP1CB peptide (intravenous, 10 mg/kg).

Extended Data Fig. 5. Validation of two model targets of proteasomally-cleaved peptides by the dTAG system.

a, Bar graph showing S. enterica growth in conditioned medium from A549 cells expressing shCTRL or shPPP1CB at the final time point (4.5 h). mean ± s.e.m. (n = 6 biological replicates). Unpaired two-tailed t-test, ns P = 0.8778. b, Western blot analysis of PPP1CB levels in A549 cells expressing shCTRL or shPPP1CB, normalized to Actin. c, Western blot of PPP1CB degradation using the dTAG system across time points (1–6 h). d, Sequence alignments of PPP1CB peptides identified by MAPP. Below (scores <2), above (scores > 2). Zoomed-in view highlights peptides with scores > 5. e, Bar graphs showing S. enterica growth in conditioned medium from A549 dTAG cells at the final time point (4.5 h) for GFP (left) and PPP1CB (right). mean ± s.e.m. (n = 4 biological replicates, 2 or 3 technical replicates). Unpaired two-tailed t-test, ns P = 0.6338, ** P = 0.0073. f, Schematic representation of the dTAG peptidomics experiment. g, Spectra and MS/MS ions by which the PPP1CB-derived peptide was identified in the A549 secretome following dTAG-PPP1CB degradtation. h, Western blot analysis of HA-PSMG2 levels after 6 h with DMSO or dTAGV−1. i, Bar graph of S. enterica growth in conditioned medium from A549 cells expressing dTAG-GFP or dTAG-PSMG2 treated with DMSO or dTAGV−1. mean ± s.e.m. (n = 6; 2 biological replicates). Unpaired two-tailed t-test, ** P = 0.0068. j-l Imaging of A549 cells expressing dTAG-PSMG2 treated with DMSO or dTAGV-1, infected with RFP- S. typhimurium (magenta), and stained with DAPI (blue). j, Representative images k, Bar plot showing quantification of RFP- S. typhimurium per nucleus. mean ± s.e.m. (n = 4 biological replicates, five images per replicate). Unpaired two-tailed t-test, *** P = 0.0002. l, Bar plots showing quantification of RFP/Salmonella per nucleus in A549 cells dTAG PSMG2. Five images per biological replicate ± s.e.m. (n = 4 independent samples) Unpaired two-tailed t-test, *** P = 0.0002. Illustrations created in BioRender. Merbl, Y. (2025) https://BioRender.com/x10a923.

Extended Data Fig. 6. Changes in MAPP-identified peptides during bacterial infection.

a, Bar graphs of IL-8 secretion from A549 cells infected with S. enterica at 1 or 4 h, or uninfected. mean ± s.e.m. (n = 3 biological replicates). One-way ANOVA, ** P < 0.0280. b, Heat map of correlation between biological sample replicates. c, PCA analysis comparing samples under different conditions. d-e, Volcano plots of degradome analysis with highlighted AMP-containing proteins at 1 h (d) and 4 h (e) of infection relative to uninfected control. f, Sequence alignment of UBE2M peptides identified by MAPP under different conditions: exclusive to control (light blue), infection (red), or shared (light gray). g, Absolute counts of peptides with cationic termini in control and infected samples. Data are mean ± s.e.m. (n = 3 biological replicates). One-way ANOVA, ** P < 0.01, **** P < 0.0001. h, Percentage of peptides with cationic or hydrophobic/aromatic C-terminal amino acids identified by MAPP in control or infected cells. i, Percentage of unique peptides with cationic or hydrophobic/aromatic N-terminal amino acids identified by MAPP exclusively in control or infected cells. j, Percentage of N-terminal amino acids of peptides identified by MAPP in control or infected. h-j, Two-sided Mann-Whitney U-test.

Extended Data Fig. 7. Investigating proteasome inhibition and peptide cleavage sites.

a, S. enterica growth in conditioned medium from cells treated with Epoxomicin (1 µM) or Leupeptin (20 µM) for 6 h. b, Bar plot of bacterial growth at 6 h. mean ± s.e.m. (n = 3 biological replicates). One-way ANOVA, ** P < 0.01, *** P < 0.001. c, Graphic representation of proteasome cleavage sites, showing differences between constitutive and immunoproteasome activity. d, Heat map of C-terminal cleavages in peptides from MDA-MB231 cells treated with DMSO, Bortezomib (Bort, 50 nM), or the immunoproteasome inhibitor ONX-0914 (ONX, 1 μM). mean ± s.e.m. (n = 3 biological replicates). e, Western blot analysis of PSME3 levels in A549 cells expressing shCTRL or shPSME4, compared to vinculin. f, CFU counts of S. typhimurium from A549 cells infected after pre-treatment with JSH-23 (40 µM) for 2 h. mean ± s.e.m. (n = 3 biological replicates). One-way ANOVA, ns P > 0.05, **** P < 0.0001. g, Representative images of A549 cells expressing shCTRL or shPSME3, treated with IKK16 (1 µM), and infected with RFP- S. typhimurium (magenta). Nuclei are stained with DAPI (blue). h, Bar plots quantifying RFP- S. typhimurium per nucleus. Twelve images per biological replicate. Data are mean ± s.e.m. (n = 3 biological replicates). One-way ANOVA, ns P > 0.05, **** P < 0.0001. Illustrations created in BioRender. Merbl, Y. (2025) https://BioRender.com/x10a923.