Virus Engineering: ORACLE reveals a bright future to fight bacteria

A new way to alter the genome of bacteriophages helps produce large libraries of variants, allowing these bacteria-killing viruses to be designed to target species harmful to human health.
  1. Willow Coyote-Maestas
  2. James S Fraser  Is a corresponding author
  1. Department of Bioengineering and Therapeutic Sciences, UCSF, United States
  2. Quantitative Biosciences Institute, UCSF, United States

Antibiotics are usually quite effective at killing bacteria that cause disease, but they often end up eliminating huge swaths of microorganisms beneficial to health . Limiting this collateral damage by solely targeting pathogenic bacteria remains challenging, as only slight differences separate harmful and beneficial bacterial species.

An alternative treatment to chemical antibiotics could be to harness viruses called bacteriophages (or phages), which have evolved to recognize and prey on highly specific strains of bacteria (Abedon et al., 2011). Yet engineering phages to target harmful bacterial species requires scalable genetic tools that can precisely alter the genomes of these viruses. Now, in eLife, Srivatsan Raman and colleagues at University of Wisconsin-Madison – including Phil Huss as first author – report a new method that can create thousands of mutations in a host-specifying phage protein, showing how these changes alter which bacteria the virus can target (Huss et al., 2021).

Before this study, most protein engineering tools that ‘tweaked’ the genome of a bacteriophage to create desirable traits relied on either rational or directed evolution methods (Favor et al., 2020; Yehl et al., 2019). In rational approaches, the detailed knowledge of the three-dimensional structure of a protein – and its associated function – helps to guide mutations that will lead to desired properties. Directed evolution approaches, in contrast, bypass the need for such detailed understanding as they involve making random mutations which are then filtered for beneficial effects through multiple rounds of selection. However, these methods tend to only sample a small proportion of possible mutations.

Another approach to identify beneficial genetic variation is deep mutational scanning (DMS for short), where every single amino acid in a protein is changed to all other amino acids to see how it affects the function of a protein. This creates both a mechanistic structure-function understanding of the protein, and a starting point for protein engineering. DMS guided-methods are becoming widely used in protein engineering to develop better antibodies, enzymes and virus-based gene delivery systems (Adams et al., 2016; Ogden et al., 2019; Romero et al., 2015). Until the work by Huss et al., however, DMS approaches in bacteriophages were limited by a lack of methods to build sizeable and comprehensive mutational libraries in bacteriophage genomes.

To address this problem, the team developed the ORACLE method (short for Optimized Recombination Accumulation and Library Expression), a new phage genome engineering approach which creates large and unbiased mutational libraries of a phage gene. ORACLE can produce complete libraries of thousands of mutations, enabling DMS-based approaches in phages. The approach was used to create a library of mutant bacteriophages that each carried a single amino acid mutation in the tip of their receptor binding protein (known as RBD), the structure the viruses use to recognize their target species.

Massively parallel DNA sequencing was then applied to assess which mutant bacteriophages multiplied after being exposed to various strains of bacteria, revealing how each amino acid mutation in the RBD tip contributed to host specificity. In particular, changes in amino acids exposed to the surface and in contact with the host gave rise to phage variants highly specific to different bacterial strains.

Additional experiments further showcased this remarkable specificity, and the power of ORACLE. In particular, the phage library obtained through the DMS approach was tested on bacterial strains that carried on their surface slightly modified versions of the sugar molecules (or lipopolysaccharide) recognized by the viruses. This allowed Huss et al. to find phage variants whose ability to infect their prey depended on the type of lipopolysaccharides present on the bacteria. Finally, the team demonstrated the potential therapeutic utility of this approach, selecting for and characterizing RBD tip mutations that made phages target a strain of bacteria which causes urinary tract infections. This work by Huss et al. demonstrates how DMS-based approaches can help to understand the basic principles underlying bacteriophage specificity, and dramatically improve their ability to prey on pathogenic bacteria.

New functions often emerge when multiple genetic changes accumulate and interact, creating larger (or smaller) effects on protein function than would be expected from the sum of individual mutations. DMS experiments alone are ill suited to examine those interactions, as it is impractical to make all double, triple, and other higher-order mutations. However, machine-learning models trained on DMS datasets can help to explore these higher order interactions and to predict useful multi-mutational variants. This approach was applied, for instance, to help turn a small virus into a tool to deliver genes of interest to specific mammalian cells (Ogden et al., 2019). Such DMS-trained models could be employed for bacteriophages in order to expand or specify the host range beyond what is possible with single mutations alone.

A lack of genetic tools and the sheer number of bacteriophage species have turned these viruses into the understudied dark matter of the viral universe. Ongoing ‘metagenomic’ studies that sample the genetic material of phages in the natural world are revealing many strains whose biology is unknown, and which carry massively diverse sequences (Jordan et al., 2014). ORACLE is unlocking new ways to characterize bacteriophages, paving the way to one day harness these sequences to understand and engineer the microbial world within and around us.


Article and author information

Author details

  1. Willow Coyote-Maestas

    Willow Coyote-Maestas is in the Department of Bioengineering and Therapeutic Sciences and the Quantitative Biosciences Institute, UCSF, San Francisco, United States

    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-9614-5340
  2. James S Fraser

    James S Fraser is in the Department of Bioengineering and Therapeutic Sciences, and the Quantitative Biosciences Institute, UCSF, San Francisco, United States

    For correspondence
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5080-2859

Publication history

  1. Version of Record published: April 15, 2021 (version 1)


© 2021, Coyote-Maestas and Fraser

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.


  • 1,635
  • 118
  • 0

Views, downloads and citations are aggregated across all versions of this paper published by eLife.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Willow Coyote-Maestas
  2. James S Fraser
Virus Engineering: ORACLE reveals a bright future to fight bacteria
eLife 10:e68277.

Further reading

    1. Biochemistry and Chemical Biology
    2. Plant Biology
    Henning Mühlenbeck, Yuko Tsutsui ... Cyril Zipfel
    Research Article

    Transmembrane signaling by plant receptor kinases (RKs) has long been thought to involve reciprocal trans-phosphorylation of their intracellular kinase domains. The fact that many of these are pseudokinase domains, however, suggests that additional mechanisms must govern RK signaling activation. Non-catalytic signaling mechanisms of protein kinase domains have been described in metazoans, but information is scarce for plants. Recently, a non-catalytic function was reported for the leucine-rich repeat (LRR)-RK subfamily XIIa member EFR (elongation factor Tu receptor) and phosphorylation-dependent conformational changes were proposed to regulate signaling of RKs with non-RD kinase domains. Here, using EFR as a model, we describe a non-catalytic activation mechanism for LRR-RKs with non-RD kinase domains. EFR is an active kinase, but a kinase-dead variant retains the ability to enhance catalytic activity of its co-receptor kinase BAK1/SERK3 (brassinosteroid insensitive 1-associated kinase 1/somatic embryogenesis receptor kinase 3). Applying hydrogen-deuterium exchange mass spectrometry (HDX-MS) analysis and designing homology-based intragenic suppressor mutations, we provide evidence that the EFR kinase domain must adopt its active conformation in order to activate BAK1 allosterically, likely by supporting αC-helix positioning in BAK1. Our results suggest a conformational toggle model for signaling, in which BAK1 first phosphorylates EFR in the activation loop to stabilize its active conformation, allowing EFR in turn to allosterically activate BAK1.

    1. Biochemistry and Chemical Biology
    2. Neuroscience
    Katarzyna Marta Zoltowska, Utpal Das ... Lucía Chávez-Gutiérrez
    Research Article

    Amyloid β (Aβ) peptides accumulating in the brain are proposed to trigger Alzheimer’s disease (AD). However, molecular cascades underlying their toxicity are poorly defined. Here, we explored a novel hypothesis for Aβ42 toxicity that arises from its proven affinity for γ-secretases. We hypothesized that the reported increases in Aβ42, particularly in the endolysosomal compartment, promote the establishment of a product feedback inhibitory mechanism on γ-secretases, and thereby impair downstream signaling events. We conducted kinetic analyses of γ-secretase activity in cell-free systems in the presence of Aβ, as well as cell-based and ex vivo assays in neuronal cell lines, neurons, and brain synaptosomes to assess the impact of Aβ on γ-secretases. We show that human Aβ42 peptides, but neither murine Aβ42 nor human Aβ17–42 (p3), inhibit γ-secretases and trigger accumulation of unprocessed substrates in neurons, including C-terminal fragments (CTFs) of APP, p75, and pan-cadherin. Moreover, Aβ42 treatment dysregulated cellular homeostasis, as shown by the induction of p75-dependent neuronal death in two distinct cellular systems. Our findings raise the possibility that pathological elevations in Aβ42 contribute to cellular toxicity via the γ-secretase inhibition, and provide a novel conceptual framework to address Aβ toxicity in the context of γ-secretase-dependent homeostatic signaling.