Skip to PREreview

PREreview of Targeted phage hunting to specificKlebsiella pneumoniaeclinical isolates is an efficient antibiotic resistance and infection control strategy

CC BY 4.0

We, the students of MICI5029/5049, a Graduate Level Molecular Pathogenesis Journal Club at Dalhousie University in Halifax, NS, Canada, hereby submit a review of the following BioRxiv preprint: 

Targeted phage hunting to specific Klebsiella pneumoniae clinical isolates is an

efficient antibiotic resistance and infection control strategy

Celia Ferriol-González, Robby Concha-Eloko, Mireia Bernabéu-Gimeno, FelipeFernández-Cuenca, Javier E. Cañada-García, Silvia García-Cobos, Rafael Sanjuán, Pilar Domingo-Calap


We will adhere to the Universal Principled (UP) Review guidelines proposed in: 

Universal Principled Review: A Community-Driven Method to Improve Peer Review. Krummel M, Blish C, Kuhns M, Cadwell K, Oberst A, Goldrath A, Ansel KM, Chi H, O'Connell R, Wherry EJ, Pepper M; Future Immunology Consortium. Cell. 2019 Dec 12;179(7):1441-1445. doi: 10.1016/j.cell.2019.11.029 

SUMMARY:  Ferriol-González et al. present their efforts to identify phages to target specific Klebsiella pneumoniae clinical isolates. The authors isolated 83 phages from environmental samples and sequenced the genomes of 71 of these phages, which they distributed into different viral clusters. They annotated the opening reading frames (ORFs) of each sequenced phage, assigning proposed functional attributes when possible. To better understand the potential for phage-host binding and recognition of isolated phages, the authors identified viral attachment protein structures for the different phages, with a particular focus on conserved depolymerase domains involved in binding and degrading host cell surface sugars. Analysis of host-range was conducted using the data collected from infecting 77 Klebsiella spp. reference strains with the isolated phages. A cross-infection matrix was constructed to determine trends in host-range for each phage-host interaction. The authors developed a phage cocktail based on the previously noted phage-host interactions and tested the efficacy of the cocktail against the Klebsiella spp. reference strains by conducting spot tests using serial dilutions of the phage cocktail. They found that 42 of the 77 strains were susceptible to the phage cocktail; however, this was lower than their expected number of susceptible strains based on individual phage-host interactions. They obtained 58 carbapenem-resistant K. pneumoniae high risk clones circulating in Spain to test their phage cocktail against. Prior to testing the phage cocktail, they looked for correlations between capsule types, carbapenemase-encoding genes, and anti-phage defense systems in the high-risk clones, noting that none of the spacer sequences found in CRISPR systems aligned with any of the phages present in their cocktail. They found the phage cocktail could infect 18 of the 58 high risk clones and 14 of these had capsular locus types present in the reference strains. This shows a significant relationship between the phage cocktail successfully infecting a high-risk strain, and the strain capsule locus having been present in the reference strain group. Considering the low infectivity rate of their phage cocktail, they repeated previous experiments via their targeted phage hunting approach using the clinical isolate KL64, which had a capsule locus type that was well represented within the high-risk strains. Using a method focused on identifying phages against a specific clinical isolate, they increased their rate of infectivity across 20 newly identified phages. Overall, this study demonstrates that taking a targeted approach to identify phages to infect specific bacterial strains will lead to identification of phages that infect the specific bacterial strain used for the reference and screening process.


STRENGTHS: The authors describe the clinical relevance of phage therapy in combating antimicrobial resistance in a clear and digestable way. They also assess the limitations of their study in the context of phage therapy literature.

WEAKNESSES: The conclusion that phage isolation is an efficient way to implement phage therapeutics as personalized tools is not supported by the data or in the literature. Several deficiencies in data presentation and incomplete analysis (described in detail below) should be addressed to improve readability. 



1.   Quality: Experiments (1–3 scale; note: 1 is best on this scale) SCORE = 3

·  Figure by figure, do experiments, as performed, have the proper controls? [note: we use this ‘figure-by-figure' section for broader detailed critiques, rather than only focusing on controls]. 

·  Fig. 1 – This figure is not described in the body text in a way that promotes understanding of the relatedness or significance of the different phage clusters. This data could be more accessible to the reader if the clusters were represented as phylogenetic trees with helpful supporting text. Moreover, there are discrepancies between the number of phages isolated, sequenced, and presented in this figure without explanation or justification for these differences in numbers. 

·  Fig. 2 – There is a lot of information in this figure full of annotated genomes that is not discussed by the authors. Perhaps the reader would be better served if this large figure were moved to the supplement, while genome annotations or other relevant information that the authors wish to discuss in the paper could be condensed and presented as a main figure.

·  Fig. 3 – The authors should provide rationale for the data analysis presented in Fig. 3; this figure is not referenced at all in the body text of the Results section, making it difficult for the reader to understand the authors’ intent and the significance of the data. While it makes sense that receptor binding proteins, types and abundance for each phage may impact the ability for a particular phage to infect a particular strain of Klebsiella, the authors do not explore this further beyond the quantification provided here. Further explanation of the significance of these protein structures is required throughout the text. 

·  Fig. 4 – A) When evaluating phage host-range, using a non-binary scoring system allows for separate categorization of partial lysis of bacterial cells from complete or no lysis. This approach would promote understanding of the degree of efficacy and number of phages that were completely or partially lytic against different strains of Klebsiella spp. Additionally, based on the information provided, it does not appear that potentially lysogenic phages containing integrase genes were removed from the assay prior to testing. This is a critical step as only lytic phages would be approved for use as treatments. The modularity calculation performed on the crossed-infection matrix does not seem necessary. B) The information provided in this figure panel does not seems to directly relate to the research question. C) Because this is one of the most important figures in the preprint, the authors should revise the axis label for greater clarity and give it more attention in the main body text. Some additional exposition about this data in the Results and Discussion sections could be very impactful. 

·  Fig. 5 – A table with KL types corresponding to sequence types would be a useful reference table to have in the supplementary data for the readers to reference. A mosaic plot may be more appropriate to convey the relationship between these categorical variables in a clearer way without splitting KL types. There should also be a grey box included indicating this group even though it is explained in the figure legend. Additionally, the significance of this data is somewhat lost on the reader as there has not been enough information provided on Klebsiella spp. biology, such as different capsule types, for the reader to appreciate the correlation between the KL and ST types. 

·  Fig. 6 – We found this figure to be the most informative in terms of data presentation. It still could benefit from greater elaboration by the authors in the main text to highlight the significance of the data presented here.                                

·  Are specific analyses performed using methods that are consistent with answering the specific question? Is there appropriate technical expertise in the collection and analysis of data presented? 

·  Standard approaches of phage cocktail design and evaluation such as isolating phages using the host organisms of interest (i.e. clinically relevant Klebsiella), ensuring removal of lysogenic phages, accurate host range evaluation, and quantitative evaluation of cocktail killing ability need to be improved/included in this study to address the research aim of creating a bacteriophage cocktail to treat a Klebsiella infection. Furthermore, genomic data analyses should directly address a specific research question, such as RBP and capsule binding interactions, rather than only serving to broadly identification of genomic characteristics. 

·  Do analyses use the best-possible (most unambiguous) available methods quantified via appropriate statistical comparisons?  

·  The use of a modularity calculation for the cross-infection matrix is not standard in the phage field and seems to misrepresent the relationship of the data based on how the phages and strains are arbitrarily laid out, seemingly creating clusters artificially. This does not contribute to the readers understanding of the host-range of a particular phage and we would recommend excluding this from the dataset. 

·  We found the information provided by the authors as justification for their bioinformatics tools and how they utilized them to be insufficient. Unicycler is a tool that acts as a SPAdes optimizer; therefore, it is unclear why the authors would choose to use a non-optimized tool for their analysis. Additionally, the statement “SPAdes includes a step for read error correction and quality trimming of Illumina reads, adding an extra layer of quality control” (pg. 7), misleads the reader because Unicycler is actually using SPAdes. Additionally, bioinformatic analyses did not include software and database version numbers in the methods section, making reproducibility of this work impossible. Many phage isolation and characterization methods were not the standard for the field, and the failure to remove lysogenic phages from the pool of phage candidates excludes this phage cocktail and this bank of phages from being used in a cocktail for future clinical use.

·  Are controls or experimental foundations consistent with established findings in the field? A review that raises concerns regarding inconsistency with widely reproduced observations should list at least two examples in the literature of such results. Addressing this question may occasionally require a supplemental figure that, for example, re-graphs multi-axis data from the primary figure using established axes or gating strategies to demonstrate how results in this paper line up with established understandings. It should not be necessary to defend exactly why these may be different from established truths, although doing so may increase the impact of the study and discussion of discrepancies is an important aspect of scholarship.  

·  The assessment of phage host range using a binary scoring system is not the standard in the field and a range-based approach would allow for a better appreciation of partial vs. complete vs. no lysis outcomes for different phage-host infections.

2.   Quality: Completeness (1–3 scale) SCORE = 2.5

·     Does the collection of experiments and associated analysis of data support the proposed title- and abstract-level conclusions? Typically, the major (title- or abstract-level) conclusions are expected to be supported by at least two experimental systems. 

·  Some of the data presented requires further experimentation to reach a convincing conclusion. For example, because the RBPs were identified bioinformatically, without characterization of binding interactions between RBPs and different bacterial strains, the significance of these interactions remain unclear. Additionally, it is unclear how the domains identified in the RBPs contribute to host-range of any individual phage. Furthermore, a secondary method to demonstrate phage-host killing in a quantifiable way would better support the spot plaquing assays conducted by the authors. 

·     Are there experiments or analyses that have not been performed but if ‘‘true’’ would disprove the conclusion (sometimes considered a fatal flaw in the study)? In some cases, a reviewer may propose an alternative conclusion and abstract that is clearly defensible with the experiments as presented, and one solution to ‘‘completeness’’ here should always be to temper an abstract or remove a conclusion and to discuss this alternative in the discussion section. 

·  Authors conclude that reference Klebsiella strains would be good vehicles of phage production; however, no experimentation was conducted to evaluate phage titers achieved through production in a reference strain compared to a clinical isolate. Additionally, the conclusion that phage isolation is a faster and more efficient way to implement phage therapeutics as personalized tools is not supported by the authors data and has not been shown to be fast and efficient throughout the literature.

3. Quality: Reproducibility (1–3 scale) SCORE = 3

·     Figure by figure, were experiments repeated per a standard of 3 repeats or 5 mice per cohort, etc.?

·  While authors stated three replicates were performed using plaque assays to assess phage-strain infectivity, they considered the strain to be susceptible if plaques were found 2/3 times. Without additional supporting data measuring infectivity by another means, we don’t believe this that the plaque assay data can stand alone without additional replicates to attain significance.  

·     Is there sufficient raw data presented to assess the rigor of the analysis? 

·  No, there is no image presentation of plaque data. 

·     Are methods for experimentation and analysis adequately outlined to permit reproducibility? 

·  Authors must include bioinformatic software version numbers; otherwise, these analyses cannot be reproduced and may be misleading to others in the field referencing this article for bioinformatic methods. Additionally, further explanation of how authors assessed and defined certain parameters, such as narrow-mid-broad host-range, should be included for better understanding.

·     If a ‘‘discovery’ dataset is used, has a ‘‘validation’ cohort been assessed and/or has the issue of false discovery been addressed? N/A

4. Quality: Scholarship (1–4 scale but generally not the basis for acceptance or rejection) SCORE = 2

·     Has the author cited and discussed the merits of the relevant data that would argue against their conclusion? 

·   Good. The authors discuss some of the limitations of the study, including reference bias due to the host isolates used, lack of quantitative data, and many phages used in their cocktail.

·     Has the author cited and/or discussed the important works that are consistent with their conclusion and that a reader should be especially familiar when considering the work? 

·     The authors should provide more context and information from the broader field regarding the role of Klebsiella capsule types and how they influence susceptibility to phage infection. In general, we think that providing more information about Klebsiella biology would help the reader better understand the dataset and wider implications of the study.

·     Specific (helpful) comments on grammar, diction, paper structure, or data presentation (e.g., change a graph style or color scheme) go in this section, but scores in this area should not be significant basis for decisions.

·  We believe that some of the supplementary figures from the paper should be included in the main text based on how frequently they are referenced by the authors and how important they are for following the narrative of the manuscript. Because many of the figures presented in the main text are single panel, there are good opportunities to add these supplementary data figures to these main figures to create multi-panel figures. 


1.   Impact: Novelty/Fundamental and Broad Interest (1–4 scale) SCORE= 3.5

A score here should be accompanied by a statement delineating the most interesting and/or important conceptual finding(s), as they stand right now with the current scope of the paper. A ‘‘1’’ would be expected to be understood for the importance by a layperson but would also be of top interest (have lasting impact) on the field.

How big of an advance would you consider the findings to be if fully supported but not extended? 

·  This manuscript does not advance novel concepts. Instead, it successfully demonstrates one of the known drawbacks to phage therapy, which is that most phages have narrow host-range and phage therapy needs to be assessed on a case-specific basis. Therefore, the conclusions drawn by the authors do not support the title of the article claiming targeted phage hunting is an efficient antibiotic resistance and infection control strategy. 

2.   Impact: Extensibility (1–4 or N/A scale) SCORE =  N/A

Has an initial result (e.g., of a paradigm in a cell line) been extended to be shown (or implicated) to be important in a bigger scheme (e.g., in animals or in a human cohort)?  This criterion is only valuable as a scoring parameter if it is present, indicated by the N/A option if it simply doesn’t apply. The extent to which this is necessary for a result to be considered of value is important. It should be explicitly discussed by a reviewer why it would be required. What work (scope and expected time) and/or discussion would improve this score, and what would this improvement add to the conclusions of the study? Care should be taken to avoid casually suggesting experiments of great cost (e.g., ‘‘repeat a mouse-based experiment in humans’’) and difficulty that merely confirm but do not extend (see Bad Behaviors, Box 2)

Competing interests

The authors declare that they have no competing interests.