CoChat
Sign Up
Cited by 173Antimicrobial Peptides and ActivitiesChemical Synthesis and AnalysisMicrobial Natural Products and Biosynthesis

The DOI matches its Crossref record; 70 of the 70 references it cites were found in a registry or an index.

Cell-free biosynthesis combined with deep learning accelerates de novo-development of antimicrobial peptides

Open access
Pandi, Amir, Adam, David, Zare, Amir, Trinh, Van Tuan, Schaefer, Stefan L., Burt, Marie, Klabunde, Björn, Bobkova, Elizaveta, Kushwaha, Manish, Foroughijabbari, Yeganeh, Braun, Peter, Spahn, Christoph, Preußer, Christian, Pogge von Strandmann, Elke, Bode, Helge B., von Buttlar, Heiner, Bertrams, Wilhelm, Jung, Anna Lena, Abendroth, Frank, Schmeck, Bernd, Hummer, Gerhard, Vázquez, Olalla, Erb, Tobias J.·2023·Nature Communications
DOI 10.1038/s41467-023-42434-9PMC PMC10632401

Published version of a preprint

Open this paper in CoChat — free accountOpen PDF

Abstract

Bioactive peptides are key molecules in health and medicine. Deep learning holds a big promise for the discovery and design of bioactive peptides. Yet, suitable experimental approaches are required to validate candidates in high throughput and at low cost. Here, we established a cell-free protein synthesis (CFPS) pipeline for the rapid and inexpensive production of antimicrobial peptides (AMPs) directly from DNA templates. To validate our platform, we used deep learning to design thousands of AMPs de novo. Using computational methods, we prioritized 500 candidates that we produced and screened with our CFPS pipeline. We identified 30 functional AMPs, which we characterized further through molecular dynamics simulations, antimicrobial activity and toxicity. Notably, six de novo-AMPs feature broad-spectrum activity against multidrug-resistant pathogens and do not develop bacterial resistance. Our work demonstrates the potential of CFPS for high throughput and low-cost production and testing of bioactive peptides within less than 24 h.

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1038/s41467-023-42434-9

Cites

70 of 70 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 The workflow for de novo-development of AMPs via deep learning and cell-free biosynthesis. a Generative variational autoencoders (VAE) for de novo-design of AMPs after being trained on known AMP sequences. b Predictive convolutional or recurrent neural networks as regressors for the MIC prediction after being trained on known AMPs and their MIC. c Trained generative and predictive models are used for sampling from the latent space (de novo-design of AMPs) and prioritization of AMPs (predicting their MIC), respectively. d Experimental pipeline for rapid cell-free biosynthesis of the designed AMPs from synthetic DNA fragments and direct testing of produced AMPs in the cell-free mix to bacterial cultures followed by overnight continuous growth assay. Created with BioRender.com.
Fig. 1 The workflow for de novo-development of AMPs via deep learning and cell-free biosynthesis. a Generative variational autoencoders (VAE) for de novo-design of AMPs after being trained on known AMP sequences. b Predictive convolutional or recurrent neural networks as regressors for the MIC prediction after being trained on known AMPs and their MIC. c Trained generative and predictive models are used for sampling from the latent space (de novo-design of AMPs) and prioritization of AMPs (predicting their MIC), respectively. d Experimental pipeline for rapid cell-free biosynthesis of the designed AMPs from synthetic DNA fragments and direct testing of produced AMPs in the cell-free mix to bacterial cultures followed by overnight continuous growth assay. Created with BioRender.com.CC BY · PubMed Central
Fig. 2 Cell-free production of de novo-generated and prioritized AMPs and activity screening against B. subtilis and E. coli. a We used different generative and regressor models (Supplementary Table 4) to design and prioritize AMPs in five rounds, produced and screened a total number of 500 AMPs from synthetic DNA fragments and found 30 functional candidates. b Charge and AlphaFold-predicted structure of the functional AMPs with associated slowed/stopped growth curves for B. subtilis and E. coli. All (including control) AMPs were produced using CFPS and no peptide purification was carried out prior to the activity test. Growth curves (OD600 0–0.45 over time 4–20 h for all) are the average of n = 3 independent experiments. Growth curves with error bars as standard deviation are provided in Supplementary Fig. 1. c 2D-projections of the 50-dimensional latent space were obtained by principal component analysis (PCA) for the two variational autoencoders (VAEs, without and with the KL-term annealing, VAE_v1 and VAE_v2, respectively, see Methods) that were used for de novo-design of AMPs. Blue color intensity represents the frequency of training AMPs in the latent space. Functional AMPs (red), BP100 and Cecropin B (black) annotated back into the latent space. Source data for b are provided as a Source Data file. Created with BioRender.com.
Fig. 2 Cell-free production of de novo-generated and prioritized AMPs and activity screening against B. subtilis and E. coli. a We used different generative and regressor models (Supplementary Table 4) to design and prioritize AMPs in five rounds, produced and screened a total number of 500 AMPs from synthetic DNA fragments and found 30 functional candidates. b Charge and AlphaFold-predicted structure of the functional AMPs with associated slowed/stopped growth curves for B. subtilis and E. coli. All (including control) AMPs were produced using CFPS and no peptide purification was carried out prior to the activity test. Growth curves (OD600 0–0.45 over time 4–20 h for all) are the average of n = 3 independent experiments. Growth curves with error bars as standard deviation are provided in Supplementary Fig. 1. c 2D-projections of the 50-dimensional latent space were obtained by principal component analysis (PCA) for the two variational autoencoders (VAEs, without and with the KL-term annealing, VAE_v1 and VAE_v2, respectively, see Methods) that were used for de novo-design of AMPs. Blue color intensity represents the frequency of training AMPs in the latent space. Functional AMPs (red), BP100 and Cecropin B (black) annotated back into the latent space. Source data for b are provided as a Source Data file. Created with BioRender.com.CC BY · PubMed Central
Fig. 3 Molecular dynamics (MD) simulations of AMP interactions with membranes. a Overview of simulation setup and exemplary snapshot of AMP #5 after 1 µs of simulated time on the IM. The peptide is shown in red cartoon representation, solvent (water + ions) is shown as transparent blue surface and the membrane is shown as gray licorice representation with lipid headgroup phosphates shown as spheres. b Distributions of distances between the centers of mass of the AMPs and the membrane midplane along the direction of the membrane normal (y-axis). Distributions are calculated from the last 940 ns of 1 µs long replicates, run with different lateral membrane tensions (x-axis) and with different membranes (blue: PM; red: IM). The thin dotted line indicates the headgroup phosphate positions. AMP #2 was simulated with and without disulfide bond (Methods). c Rendered simulation snapshots of potent AMPs (Fig. 4) on IM (top) and PM (bottom). See Fig. 5c, d for the mode of action experiments. The interacting leaflet of the membrane is depicted with gray lipid tails and orange (phosphates) and green (cholesterol oxygen) spheres. Created with BioRender.com.
Fig. 3 Molecular dynamics (MD) simulations of AMP interactions with membranes. a Overview of simulation setup and exemplary snapshot of AMP #5 after 1 µs of simulated time on the IM. The peptide is shown in red cartoon representation, solvent (water + ions) is shown as transparent blue surface and the membrane is shown as gray licorice representation with lipid headgroup phosphates shown as spheres. b Distributions of distances between the centers of mass of the AMPs and the membrane midplane along the direction of the membrane normal (y-axis). Distributions are calculated from the last 940 ns of 1 µs long replicates, run with different lateral membrane tensions (x-axis) and with different membranes (blue: PM; red: IM). The thin dotted line indicates the headgroup phosphate positions. AMP #2 was simulated with and without disulfide bond (Methods). c Rendered simulation snapshots of potent AMPs (Fig. 4) on IM (top) and PM (bottom). See Fig. 5c, d for the mode of action experiments. The interacting leaflet of the membrane is depicted with gray lipid tails and orange (phosphates) and green (cholesterol oxygen) spheres. Created with BioRender.com.CC BY · PubMed Central
Fig. 4 Bioactivity characterization of chemically synthesized functional AMPs. a Minimum inhibitory concentration (MIC) of the AMPs against B. subtilis (gray) and E. coli (black), average of n = 3 independent experiment. CC50 (50% cytotoxicity, blue) and HC50 (50% hemolysis, red) values of the AMPs on HCT116 human colon cells and human red blood cells, respectively (average of n = 2 independent experiments). Arrows show values above the measured limits, black, blue and red for MIC > 100 µM, CC 50 > 250 µM and HC50 > 250 µM, respectively. MIC, HC50 and CC50 values are provided in Supplementary Table 10. b MIC values of the AMPs tested against ESKAPE pathogens including E. faecium, Methicillin-resistant S. aureus (MRSA), K. pneumoniae, A. baumannii, P. aeruginosa, Enterobacter spp measured in duplicates (n = 2 independent experiments) and MIC of the six potent, broad-band AMPs on Y. pestis and B. anthracis and S. pneumoniae as the average of triplicates (n = 3 independent experiments). Created with BioRender.com.
Fig. 4 Bioactivity characterization of chemically synthesized functional AMPs. a Minimum inhibitory concentration (MIC) of the AMPs against B. subtilis (gray) and E. coli (black), average of n = 3 independent experiment. CC50 (50% cytotoxicity, blue) and HC50 (50% hemolysis, red) values of the AMPs on HCT116 human colon cells and human red blood cells, respectively (average of n = 2 independent experiments). Arrows show values above the measured limits, black, blue and red for MIC > 100 µM, CC 50 > 250 µM and HC50 > 250 µM, respectively. MIC, HC50 and CC50 values are provided in Supplementary Table 10. b MIC values of the AMPs tested against ESKAPE pathogens including E. faecium, Methicillin-resistant S. aureus (MRSA), K. pneumoniae, A. baumannii, P. aeruginosa, Enterobacter spp measured in duplicates (n = 2 independent experiments) and MIC of the six potent, broad-band AMPs on Y. pestis and B. anthracis and S. pneumoniae as the average of triplicates (n = 3 independent experiments). Created with BioRender.com.CC BY · PubMed Central
Fig. 5 Resistance and mode of action test of AMPs on E. coli and the mode of action assays. a Twenty one days MIC measurement of daily culture passage each day from the cells grown in half-MIC concentrations of AMPs. The MIC values were normalized by the minimum of 21 daily averages for each AMP. Bars are the average of n = 3 independent experiments. See Supplementary Fig. 7 for raw MIC data. b Impact of different AMPs at the concentration of ¼ MIC on the number of E. coli outer membrane vesicles (OMVs) for untreated E. coli (control), (AMP) solvent, and six potent AMPs, and polymyxin B. Bars are the average of independent experiments n = 8 for the control, n = 6 for AMP#3, and n = 3 the rest. Statistics: ordinary one-way ANOVA p < 0.0001 (4.86×10−6), with Polymyxin B as the only significantly different group. c Plate reader measured fluorescence (in arbitrary unit) of propidium iodide with E. coli cells untreated (control) or treated with AMPs. Bars are the average of n = 3 independent experiments. P values are for unpaired t-test of each AMP compared to untreated E. coli control. d Phase-contrast microscopy of E. coli cells untreated or treated with six potent AMPs and Cecropin B (a membrane disruptive AMP) and stained with propidium iodide (red). Source data for a–c and raw images in d (plus six more acquisitions for each, two images of n = 3 biological replicates with similar results) are provided as a Source Data file. Created with BioRender.com.
Fig. 5 Resistance and mode of action test of AMPs on E. coli and the mode of action assays. a Twenty one days MIC measurement of daily culture passage each day from the cells grown in half-MIC concentrations of AMPs. The MIC values were normalized by the minimum of 21 daily averages for each AMP. Bars are the average of n = 3 independent experiments. See Supplementary Fig. 7 for raw MIC data. b Impact of different AMPs at the concentration of ¼ MIC on the number of E. coli outer membrane vesicles (OMVs) for untreated E. coli (control), (AMP) solvent, and six potent AMPs, and polymyxin B. Bars are the average of independent experiments n = 8 for the control, n = 6 for AMP#3, and n = 3 the rest. Statistics: ordinary one-way ANOVA p < 0.0001 (4.86×10−6), with Polymyxin B as the only significantly different group. c Plate reader measured fluorescence (in arbitrary unit) of propidium iodide with E. coli cells untreated (control) or treated with AMPs. Bars are the average of n = 3 independent experiments. P values are for unpaired t-test of each AMP compared to untreated E. coli control. d Phase-contrast microscopy of E. coli cells untreated or treated with six potent AMPs and Cecropin B (a membrane disruptive AMP) and stained with propidium iodide (red). Source data for a–c and raw images in d (plus six more acquisitions for each, two images of n = 3 biological replicates with similar results) are provided as a Source Data file. Created with BioRender.com.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 69 references

Research works printed but not linked: 1

  1. Konur, O. Handbook of Algal Science, Technology and Medicine (Academic Press, 2020).

Cited by

Sorted by citations

Items 101–172 of 172

Related

See all

Metadata

Authors
Pandi, A., Adam, D., Zare, A., Trinh, V. T., Schaefer, S. L., Burt, M., Klabunde, B., Bobkova, E., Kushwaha, M., Foroughijabbari, Y., Braun, P., Spahn, C., Preußer, C., Pogge von Strandmann, E., Bode, H. B., von Buttlar, H., Bertrams, W., Jung, A. L., Abendroth, F., Schmeck, B., Hummer, G., Vázquez, O., & Erb, T. J.
Affiliations
  • Max Planck Institute for Terrestrial Microbiology (DE)
  • Philipps University of Marburg (DE)
  • Max Planck Institute of Biophysics (DE)
  • AgroParisTech (FR)
  • Fraunhofer Institute for Translational Medicine and Pharmacology (DE)
  • Goethe University Frankfurt (DE)
  • German Center for Infection Research (DE)
Corresponding author
  • Pandi, Amir
  • Erb, Tobias J.
ORCID
Venue
Nature Communications
Publisher
Springer Science and Business Media LLC
Published
2023-11-08
Volume
14
Issue
1
Pages
7197-7197
ISSN
2041-1723 (ISSN-L)
DOI
10.1038/s41467-023-42434-9
PMC
PMC10632401
OpenAlex
W4388492286
Cited by
173
References
69
Citations by year
2026: 54 · 2025: 67 · 2024: 50 · 2023: 2
FWCI
31.50
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Communications (publishedVersion, cc-by)
  • bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • HAL (submittedVersion, other-oa)
  • RePEc: Research Papers in Economics (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • Fraunhofer-Publica (Fraunhofer-Gesellschaft) (submittedVersion)
Funders
  • Max-Planck-Gesellschaft
  • European Molecular Biology Organization (ALTG 165-2020)
Topics
  • Life Sciences › Immunology and Microbiology › Microbiology › Antimicrobial Peptides and Activities
  • Physical Sciences › Chemistry › Organic Chemistry › Chemical Synthesis and Analysis
  • Life Sciences › Immunology and Microbiology › Applied Microbiology and Biotechnology › Microbial Natural Products and Biosynthesis
Keywords
antimicrobial peptides, cell-free protein synthesis, de novo peptide design, deep learning, bioactive peptides, multidrug-resistant pathogens, broad-spectrum antimicrobial activity, peptide toxicity, molecular dynamics simulation
MeSH terms
Deep Learning, Antimicrobial Peptides, DNA Replication, Protein Biosynthesis, Molecular Dynamics Simulation

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review