High-dimensional Characterization of Genome–Environment Fitness Landscapes in Klebsiella pneumoniae
Curation statements for this article:-
Curated by eLife
eLife Assessment
This study provides a valuable mapping of how genomic variation from a large collection of Klebsiella pneumoniae clinical isolates shapes bacterial fitness across 214 conditions. Solid data describe that fitness is strongly condition-dependent and that features can be advantageous under selection. Though the predictive, plasmid causality and GWAS association findings would be further strengthened by additional analyses ruling out population-structure confounding, these findings are of value for our understanding of the power of genotype to predict phenotype for this important human pathogen.
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (eLife)
Abstract
Background Bacterial fitness is shaped by interactions between genome variation and environmental context, yet how these interactions determine its predictability and heritability remains unclear. In the clinically important pathogens of Klebsiella pneumoniae , a leading cause of hospital-acquired infections, this question is particularly pressing. Despite extensive genomic characterization, we still lack a systematic understanding of how genome-wide variation translates into fitness across diverse environments in K. pneumoniae . Methods We filled this gap by profiling a systematic collection of 1,462 clinical K. pneumoniae isolates across 214 diverse environmental and pharmacological stress conditions using high-throughput chemical genomics. Fitness was quantified from colony growth and integrated with whole-genome sequencing data. Genome-wide association analyses identified genetic determinants of fitness, and machine learning models incorporating genomic features were used to predict fitness. Results Fitness exhibited a strongly environment-dependent genetic architecture, with modest but significant concordance between genetic background and phenotypic variation. Under antibiotic and stress-combination conditions, fitness was driven by discrete, high-effect determinants, including known resistance genes, resulting in stronger signals and improved predictability. In contrast, non-antibiotic environments showed more polygenic and distributed architectures with weaker associations. Genome-wide analyses identified both established and previously uncharacterized genes linked with fitness across conditions. Resistance and virulence determinants exhibited clear context-dependent trade-offs, conferring fitness advantages under selection but imposing costs in non-selective environments. Consistent with this, plasmid carriage showed environment- and genotype-dependent fitness effects, with benefits under antibiotic pressure and measurable costs otherwise. Genomic variant–based models for fitness prediction achieved moderate performance (Mean Spearman correlation (ρ) = 0.36 (95% CI: 0.18–0.67) for predicted versus observed values in unseen data) across conditions, with improved accuracy under strong antibiotic selective pressures, and produced well-calibrated prediction intervals with high coverage. Despite strong population structure effect on predictions, models captured predictive gene and SNP biomarkers for fitness. Conclusion These findings highlight that bacterial fitness is an emergent property of genome–environment interactions rather than a fixed attribute of genotype. This work establishes a unified high-dimensional genotype–phenotype framework linking genomic variation to fitness across diverse conditions in a major pathogen, with broader implications for other pathogenic bacterial species.
Article activity feed
-
eLife Assessment
This study provides a valuable mapping of how genomic variation from a large collection of Klebsiella pneumoniae clinical isolates shapes bacterial fitness across 214 conditions. Solid data describe that fitness is strongly condition-dependent and that features can be advantageous under selection. Though the predictive, plasmid causality and GWAS association findings would be further strengthened by additional analyses ruling out population-structure confounding, these findings are of value for our understanding of the power of genotype to predict phenotype for this important human pathogen.
-
Reviewer #1 (Public review):
This manuscript by Zhou, Williams, Millner, and colleagues reports findings from a large-scale, multicenter study of Klebsiella pneumoniae (Kp) isolates. These isolates were sequenced using short-read sequencing and extensively characterized using a variety of field-appropriate tools, such as Kleborate, AMRFinderPlus, and PlasmidFinder. Then, the authors measured the growth of these Kp isolates across many (>200) environmental conditions they pre-determined as important for making human-centric fitness conclusions. These genotype and phenotype data were integrated, and the authors deployed an association-based strategy to identify genetic determinants of the observed fitness effects. Finally, the authors built models to predict fitness outcomes using the associated genetic determinants. From these data, the …
Reviewer #1 (Public review):
This manuscript by Zhou, Williams, Millner, and colleagues reports findings from a large-scale, multicenter study of Klebsiella pneumoniae (Kp) isolates. These isolates were sequenced using short-read sequencing and extensively characterized using a variety of field-appropriate tools, such as Kleborate, AMRFinderPlus, and PlasmidFinder. Then, the authors measured the growth of these Kp isolates across many (>200) environmental conditions they pre-determined as important for making human-centric fitness conclusions. These genotype and phenotype data were integrated, and the authors deployed an association-based strategy to identify genetic determinants of the observed fitness effects. Finally, the authors built models to predict fitness outcomes using the associated genetic determinants. From these data, the authors conclude that fitness is "an emergent property" of a given Kp isolate's genetic background and environmental constraints, with a nod toward the future development of biomarkers to predict fitness outcomes using genotype alone.
The primary strength of this study is the integration of two large datasets, one genotypic and the other phenotypic. This approach is novel in its scale (genotypic, geographic, etc.), which adds to the strength of the overall conclusions. These conclusions, for the most part, are consistent with the data. As such, this study can act as an example to other studies tying genotype to phenotype, especially in the context of careful genetic analysis and rational phenotypic design. A secondary strength is that the findings are supported by other, smaller-scale studies such that there is a growing consensus as to the general fitness determinants employed by Kp that are potentially driving its increased global importance as a bacterial pathogen.
The primary weakness of this study is that some confounding factors are not taken into account or are deemphasized. The primary confounding factor is that the associations between genetic signatures of fitness are not fully considered. This is especially important given that an important driver of Kp fitness is its accessory genome, which is heavily influenced by mobile genetic elements and plasmids. The use of short-read sequencing leads to an inherent inability to tie fitness effects to such genetic elements, which may lead to masking or overinterpretation of particular genotype-phenotype relationships. For example, nearly all of the gene-level determinants of condition-specific fitness and pleiotropic genes are known plasmid-associated genes; however, it is unclear if these specific genes are driving fitness or if they are merely signatures of other associated genetic elements. This broadly weakens the conclusions and translatability, and in the absence of long-read sequencing to resolve genetic architecture, it is an unresolvable weakness. A second is the over-interpretation of the variables used to assess bacterial growth, which is the primary fitness metric. Colony opacity can be influenced by Kp capsule production, which is a highly variable trait that is associated with environmental conditions. Finally, it is unclear whether clonal isolates are excluded to avoid overinterpretation of fitness outcomes.
-
Reviewer #2 (Public review):
Summary:
The authors mapped context-dependent fitness of more than 1,400 clinical K. pneumoniae isolates across 214 antibiotic, chemical, environmental, metal, envelope and nutrient conditions, integrating fitness measurements with whole-genome sequencing to build a genotype-phenotype-fitness framework and a machine learning pipeline for predicting bacterial fitness across environments.
Strengths:
The scale of the dataset is remarkable both in terms of isolates and conditions tested. The central findings that fitness is strongly condition-dependent and that resistance-associated features help under selection are well supported. Resource accessibility (ColonyExplorer portal, code, trained models) is excellent, allowing future integration of different genetic backgrounds and environmental contexts.
Weaknesses:
Reviewer #2 (Public review):
Summary:
The authors mapped context-dependent fitness of more than 1,400 clinical K. pneumoniae isolates across 214 antibiotic, chemical, environmental, metal, envelope and nutrient conditions, integrating fitness measurements with whole-genome sequencing to build a genotype-phenotype-fitness framework and a machine learning pipeline for predicting bacterial fitness across environments.
Strengths:
The scale of the dataset is remarkable both in terms of isolates and conditions tested. The central findings that fitness is strongly condition-dependent and that resistance-associated features help under selection are well supported. Resource accessibility (ColonyExplorer portal, code, trained models) is excellent, allowing future integration of different genetic backgrounds and environmental contexts.
Weaknesses:
(1) Normalisation uses a plate-level random effect and plate-median scaling, whereas within-plate (position) effects and their correction are not described, although typically required in such high-density arrays.
(2) Hypermucoid or heavily encapsulated isolates may form smaller but denser colonies. The area-opacity correlation (Figure S3A) does not exclude this, so the virulence-associated costs in selected conditions (Figures 2F-G, S5A) could be partly measurement-driven.
(3) Plasmid effects are attributed causally from observational data. Fitness effects are inferred from short-read replicon calls (Figure 3) and may be confounded by lineage and co-occurring resistance determinants.
(4) The study does not apply linear mixed model GWAS on the continuous fitness phenotype, relying instead on median dichotomization with a non-mixed model that is vulnerable to population structure (especially in a highly clonal species) and could exacerbate apparent separation, inflating association counts.
(5) GWAS-derived features appear to have been selected using all isolates rather than within each training fold. Test-set labels thus influenced which features entered the models, biasing correlation, coverage and interval estimates upward. Although mitigating overfitting, nested hyperparameter tuning and lineage-stratified tests do not remove this bias. Because of this, I do not find the claim that genomic variants predict fitness beyond population structure adequately supported yet.
With the points above addressed, I believe this work would be highly valuable for prioritising genes and variants for functional validation and for generally studying fitness trade-offs in hospital pathogens.
-
